March 17, 1995 Development of Novel Planar Transmission Lines for High-Frequency Passive Circuit Component and Array Applications Linda Katehi and Gabriel Rebeiz Final Report CONTRA CT IDENTIFIERS ONR Contract Number: N00014-92-J-1070 Period Covered by Report: 9/92 to 1/95 Name of Institution: University of Michigan Scientific Personnel Partially or Fully Supported by this Contract: Faculty Students Linda Katehi Rhonda Drayton (Supported 60% by ONR Gabriel Rebeiz and 40% by a Minority Fellowship) Tom Weller (Supported 50% by ONR and 50% by a NASA Fellowship) Steve Robertson (Supported 30% by ONR and 70% by a NASA Fellowship) 29226-1-F = RL-2404 1

STUDENT PROFILES STUDENT PROFILES This contract supported students part-time as indicated above. The rest of the support came through fellowships through GEM and the NASA Center of Space Terahertz Technology. A short description of this partial support and the relevance of the students work with the subject is described below: Rhonda Drayton, expected to graduate in May 1995. She has been supported part time by this contract and the rest of her support has been provided by a minority fellowship. Specifically, she developed micromachined circuits for high-frequency applications and she fabricated and successfully measured the first micromachined circuits with an integrated package. At the present time, Rhonda is completing the development of micromachined antennas. This work will be presented in the 1995 IEEE Antennas and Propagation Symposium. Tom Weller, expected to graduate in May 1995. He has been supported part time from this contract until March 1994 and full-time since then. The rest of his support came from the NASA Center of Space Terahertz Technology. He has developed the first membrane components which very effectively operated in frequencies up to 40 GHz. Tom's work has demonstrated the capability to effectively use membranes for the development of very efficient matching networks, filters and power dividers. Steve Robertson, expected to graduate in Winter 1997. He has been supported part time from this contract since March 1994. The rest of his support came from the NASA Center of Space Terahertz Technology. He has developed the first membrane components which very effectively operated in W band. Steve has developed the first planar filters operating in W band with less than 1 dB insertion loss. This work will be presented in the 1995 IEEE Microwave Theory and Techniques Symposium. HONORS and AWARDS During the duration of this Contracts we received the following Awards and Recognitions: * In September 1993, Prof. Rebeiz was awarded with the Isaac Goga Gold Medal Award by URSI, during the 24th General Assembly Meeting in Kyoto, Japan. This award is given only once every three years by URSI to the most prominent young Senior Member of IEEE. * In May 1994, Prof. Linda Katehi at the University of Michigan has been awarded the Humboldt Research Award for Senior U.S. Scientists in recognition of her accomplishments in research and teaching. According to Prof. P. Russer of the FerdinandBraun-Inst. f. Hochstfrequenztchnik, Berlin, this is the first Humboldt Award given to a U.S. Scientist with specialization in Electromagnetic Modeling, Development and Characterization of Monolithic Circuits and Antennas. Development of Novel Planar Transmission Lines for High-Frequency Passive Circuit

PATENTS * In June 1994, Ms. Rhonda F. Drayton received the first prize in the student paper competition by the IEEE Microwave Theory and Techniques Society during the 1994 International MTT-S Symposium in San Diego, CA, for the paper coauthored with her advisor Prof. Linda Katehi "Development of Miniature Microwave Circuit Components Using Micromachining Techniques". * In Fall 1994, Prof. Katehi became a Fellow of IEEE. * In October 1994, Professor Katehi received The University of Michigan Faculty Recognition Award, in recognition of outstanding contributions to the intellectual environment of the University. PATENTS The work performed under this contract is pending for a U.S patent entitled "Micromachined Monolithic Circuits with Integrated Packaging". The patent application was filed in September 9, 1995 with the U.S. patent office in Washington DC. PUBLICATIONS PH.D DISSERTATIONS 1. Rhonda Drayton, Tom Weller and Steve Robertson have not graduated yet. The first two students are graduating in May 1995 and the third in Winter 1997. Their dissertations will become available to ONR upon their graduation. JOURNAL PUBLICATIONS 1. N. I. Dib, R. F. Drayton, and L. P. B. Katehi, "A Theoretical and Experimental Study of Microshield Circuits", Microwave and Optical Technology Letters, Vol. 6, No. 6, pp. 333-339, May 1993. 2. Rhonda F. Drayton, N. I. Dib, and L. P. B. Katehi, "Design of Micromachined High Frequency Circuit Components", to be published January 1995 in The International Journal of Microcircuits and Electronic Packaging. 3. R. F. Drayton and L. P. B. Katehi, "Development of Self-Packaged High Frequency Circuits Using Micromachining Techniques", submitted for publication in the IEEE Transactions on Microwave Theory and Techniques. 4. L. P. B. Katehi, et. al., "Micromachined Circuits for Millimeter- and Sub-millimeterWave Applications," IEEE Antennas and Propagation Magazine, Vol.35, No. 5, pp. 9-17, October 1993. 5. T.M. Weller, L.P.B. Katehi and G.M. Rebeiz, "High-Performance Microshield-Line Components," in press, IEEE Transactions of Microwave Theory and Techniques (expected date June 1995). 6. T.M. Weller, L.P.B. Katehi and G.M. Rebeiz, "A 250 GHz Microshield Bandpass Filter," in press, IEEE Microwave and Guided Wave Letters. Development of Novel Planar Transmission Lines for High-Frequency Passive Circuit

PUBUCATIONS 7. T.M. Weller, L.P.B. Katehi and G.M. Rebeiz, "Single and Double Folded-Slot Antennas on Semi-Infinite Substrates," submitted to the IEEE Transactions on Antennas and Propagation. CONFERENCE PROCEEDINGS 1. L. P. B. Katehi, N. I. Dib, and R. F. Drayton, "Theoretical and Experimental Characterization of Microshield Circuits," International Symposium on Signals, Systems, and Electronic Proceedings, pp. 77-80, 1992. 2. R. F. Drayton and L. P. B. Katehi, "Microwave Characterization of Microshield Lines," 40th ARFTG Conference Digest, pp. 171-176, Dec 1992. 3. R. F. Drayton and L. P. B. Katehi, "Experimental Study of Micromachined Circuits," 4th International Symposium on Space Terahertz Technology, Los Angeles, CA, pp. 238-248, March 1993. 4. T.M. Weller, L.P.B. Katehi, G.M. Rebeiz, H.J. Cheng and J.F. Whitaker, "Fabrication and Characterization of Microshield Circuits," presented in the 1993 International Symposium on Space Terahertz Technology, Los Angeles, California, March 1993. 5. R. F. Drayton and L. P. B. Katehi, "Study of Micromachined Quasi-Planar Lines," 1993 Antenna and Propagation/URSI Radio Science Symposium, Ann Arbor, MI, p.30, June 1993. 6. R. F. Drayton and L. P. B. Katehi, "Micromachined Circuits for Mm-Wave Applications," 23rd European Microwave Conference, Madrid, Spain, pp. 587-588, September 1993. 7. T.M. Weller, G.M. Rebeiz and L.P.B. Katehi, "Experimental Results on Microshield Transmission Line Circuits," presented in the 1993 International Symposium of Microwave Theory and Techniques Society, Atlanta, Georgia, June 1993. 8. R. F. Drayton, C. Kidner, J. East, L. P. B. Katehi, "Micromachined Detector Mounts for Millimeter-wave Applications," 5th International Symposium on Space Terahertz Technology, Ann Arbor, MI, May 1994. 9. Tom Weller, S.V. Robertson, L.P.B. Katehi and G.M. Rebeiz, "Millimeter and Submillimeter-Wave Microshield Line Components," presented in the 1994 International Terahertz Symposium, Ann Arbor, MI. 10. R. F. Drayton and L. P. B. Katehi, "Development of Miniature Microwave Circuit Components Using Micromachining Techniques," 1994 IEEE MTT-S International Microwave Symposium, pp. 225-228. 11. R. F. Drayton, T. M. Weller, and L. P. B. Katehi, "Development and Characterization of Miniaturized Circuits for High Frequency Applications Using Micromachining Techniques," presented at the 1994 The International Society for Hybrid Microelectronics Symposium. 12. R. F. Drayton and L. P. B. Katehi, "Micromachined Conformal Packages for Microwave and Millimeter Wave Applications," to be presented at 1995 IEEE MTT-S International Microwave Symposium. 4 Development of Novel Planar Transmission Linesfor High-Frequency Passive Circuit

A BRIEF ANALYSIS OF THE RESEARCH FINDINGS 13. R. F. Drayton and L. P. B. Katehi, "Microstrip Patch Antennas on Micromachined Low-Index Materials," submitted to the 1995 Antennas and Propagation Society International Symposium. 14. Steve V. Robertson, Linda P.B. Katehi and Gabriel M. Rebeiz, "Micromachined Self-Packaged w-Band Bandpass Filters," accepted for presentation in the 1995 International Symposium of the IEEE Microwave Theory and Techniques Society, Orlando, FL, May 1995. 15. I. Papapolymerou, Rhonda F. Drayton and Linda P.B. katehi, "Surface Wave Mode Reduction for Rectangular Microstrip Antennas," submitted for presentation in the 1995 International Symposium of Antennas and Propagation Society, Seattle, WA, June 1995. A BRIEF ANALYSIS OF THE RESEARCH FINDINGS The work performed under this contact has resulted in what we consider to be major breakthroughs in planar technology. For the first time, it was demonstrated successfully that silicon micromachining can be used for a variety of high-frequency applications including packaging and high-performance interconnects. The success of our work can be summarized in the following: * The work performed under this contract has resulted in 8 Journal publications and 15 publications in Symposia proceedings. Copies of these papers are attached in Appendices B and C respectively. * The work has been submitted for a U.S. patent successfully and is pending decision. * For her work under this contract, Rhonda Drayton received the best paper award in the Student Paper Competition during the 1994 IEEE Microwave Theory and Techniques Symposium (see attachment in Appendix A). * All of the three students supported by this contract have been listed in the 1994 Technical Program Chairman's Honor Roll (see Appendix A). * The work performed by this contract has attracted the interest of small companies which plan to commercialize some of our components. * The work has also attracted the interest of MA/COM and Martin Marietta which plan to use some of our packaging ideas on their own planar circuits. This contract allowed us to perform very innovative and original work which, we feel, will become a landmark in the years to come. Development of Novel Planar Transmission Lines for High-Frequency Passive Circuit

APPENDIX A APPENDIXA S Development of Novel Planar Transmission Lines for High-Frequency Passive Circuit

,1 TRANSACTIONS (O ',1I. 'y\Vi: 1lt-0:() ANI T'r-CIINIQUtIS. V()l 42. NO! 2. D[:CEMB:-k i"4 1994 IMS Technical Program Presentation Recognitions Bob Eisenhart AT this year's Symposium there were two types of presentation competition. The first was the Student Paper Competition which followed an established format. This was judged by a select panel on all aspects of each presentation at a separate session on Tuesday evening. Each participant was then given a complimentary invitation to the Awards Banquet on Wednesday. The 1st, 2nd, and 3rd place winners were, respectively, Rhonda Drayton, University of Michigan; Steven Gearhart, University of Michigan and Peter Liao, University of California at Santa Barbara. They received their prizes at the Awards Banquet. A second, and new competition this year, resulted in a 'Technical Program Chairman's Honor Roll." This focused entirely on the quality of the presentation rather than the technical content and recognition was not limited to one or two speakers per session. All presenters were notified of the procedure and rating criteria, hopefully motivating them to put in that little extra effort. It is felt that this was very successful since there were many comments by registrants who specifically pointed out that the overall level of presentation was greatly improved over past years. This high quality was also evident in the rating scores received. In a rating system of 0 to 20 where 12 was defined as an average presentation, the average rating received was 14. The threshold for placement on the Honor Roll was for all scores greater or equal to 14, and resulted in the 165 presenters shown on the accompanying listing. These presenters have also received a Certificate of Recognition. When duplicate speakers and papers not receiving scores are removed, this represents more than 50% of the rated papers. Aside from the Presentation Recognition, all presenters of contributed papers received a Symposium Medallion in appreciation for their part in the Symposium. Thanks again to all involved in the Technical Program, and particularly those on the Technical Program Committee who took the time to provide reviews of the presentations. The results easily justified our efforts. 1994 IMS Technical Program Chairman's HONOR ROLL F. Alessandri R. Anholt D. Antsos M. A. Aust J. W. Bandler A. Bangert M. Becker E. M. Biebl H. Blinck C. Bohm J. Bornemann W. Boulais S. Bouthillette C. R. Boyd R. L Brown T. Bnrkiema S. C. Bundy J. Burns K. Chang C. H. Chen C. Y. Chi Y. Cho ED. Cohen M.Cohn C.E Collins A. Constanzo L. D'Evelyn A. S. Daryoush J. Detlefson N. Dib J. DiLotnzo G. F. Dionne R. F. Dravton H. Eisele M. Engles R. Faraji-Dana L. M. Felton A. Fenrro H. R. Fettennan J. Frank A. Galli S. S. Gearhart J. Goel M. A. Gouker R. Goyal P. A. Grobelny R. Hajji P. Halloran R. Heidemann W. Heinrich D. Heston H. Heuermann B. Hewitt S. Hofschen P. K. Ikalainen T. Ishizaki T. Ivanov P. Jukkala U. Karacaoglu R. Kaunisto T. Kawai J. S. Kenney K. W. Kobayashi R. Koic B. Kraemer M. Kubo S. Lardizabal J. Laskar B. Lax J. G. Leckey L. Lefrancois J. Lester J. Y. Liang U. Lott J. F. Luy M. R. Lyons S. Maas J. A. MacDonald R. R. Mansour M. J. Marcus H. Matsumoto S. R. Mazunder R. McGowan D. McQuiddy L Mehdi W. Menzel W. Mertin J. E. Miller K Minot S. MolekonL M. Mongiardo A. Morini L. Mullen T. Nakagawa S. Nakajima M. Nakhia T. D. Ni T. N. Nielsen H. Okazaki G. Olbrich P. Onno J. Owens D. Parker S. T. Peng A. Plattner M. Pospieszalski N. M. Pothecary S. J. Prasad S. Pritchett M. M. Radmanesh A. V. Raisanen C. Rauscher V. S. Reinhardt A. C. Reyes M. Righi M. Rittweger S. V. Robertson P. Russer K. Sabetfakhri H. Sakai B. C. Schmukler J. S. H. Schumacher R. Schwindt F. Sechi A. K. Sharma H. Shigesawa C. G. Shih T. Shimura R. N. Simons C. Sinclair P. M. Smith P. M. So F. Sterzer M. L. Stevens P. A. Stimson G. Strauss W. Struble D. G. Swanson S. Talisa D. Teeter J. C. Tippet T. Tokumitsu M. C. Tsai H. S. Tsai R. Vahldieck T. Van den Broeck F. Vergeyst M. P. Versleijen E. Vourch H. Wang M. R. Webb C. J. Wei R. Weigel T. M. Weller D. A. Willems J. T. Williams R. M Wohlert I. Wolff E. Yamashita H. W. Yao S. Ye M. Yousefi A. H. Zaabab 0018-9480/94$04.00 1994 IEEE

APPENDIX B APPENDIX B Development of Novel Planar Transmission Lines for High-Frequency Passive Circuit

STUDY OF MICROSHIELD CIRCUITS Nihad I Dib. Rhonda. Drayton. and Linda P. B Katehi. v ~..,.-"..2. kt. I: RMS /,," ~,. '', r..r.,,.\ * '.t:t,,'tet' "I if'd (/! (t, 'fi!t'-: l 'Pt prd4,'UtZi -.,, *<,,*.,;r;,;' k',;J[' rt. ( i:;l,.' Irn r)\ ~ Steo - S:e: 2 a Metalize Step 3 S.e. 4 HAlm Metallize Shlec o Pe OaCK 0o fe Ca Figure 1 Fjrrij.lilon procc.:'.*r 'ha rr.. -.. BSTRA( I ~4~ ',; i;U II, rd(. t: ~ " ' " " 'i ud t,j '!n' p. t' f rPi)t lhic' [p i pil r.;raJnuz i n:.'/;t n':'e presented This line ain operate althoul:hte 'teif 'or i ta hoet'.s,r Dhseuse of air bridees tor ground equali:a. iton. Furthermore. it has shoHn the tendency to radiate less than the,toil ettional coplanar Iwaveguide (CPWI and can provide a wide rullnc i impedances due to the many available parameters for deslwn The space-domatn integral equation method is used to analz:e let cral discontinuities of the proposed line. n order to validate the theorv measurements are performed on scaled models of a specific discontinurrtv The experimental and theoretical results are in good agreement. In addition. it is demonstrated that the proposed line disconwinuites radiate considerably less than the corresponding coplanar waveguide ones C /993 John Wile c& Sons. Inc 1. INTRODUCTION The first planar transmission line. the stripline. was introduced almost 40 years ago and created the basis of a new and revolutionary hybrid technology. Since then. hybrid technology has evolved to monolithic technology, drastically increasing operating frequencies and consequently reducing weight and volume. In conventional planar transmission lines the power is propagating by creating an rf voltage difference between two planar conductors printed on the same (coplanar waveguides. coplanar strips) or opposite surfaces (stripline. microstrip. coupled strips) of a dielectric slab structure. In most cases. the geometry of the conventional lines permitted great design flexibility. tremendous reduction of the space occupied by the circuit. and realization of very large scale, very high frequency applications. The planarization of the conductors in the above transmission lines provided the capability of integration, which led to unwanted mechanisms such as radiation and dispersion resulting in enhanced ohmic losses and electromagnetic coupling. These mechanisms are frequency dependent and impose serious limitations as we approach the submillimeterfrequency range. The ability to find new geometries [1-91 which reduce or eliminate the above loss or coupling mechanisms without affecting the monolithic character of the line will extend the operating frequencies long into the terahertz region and will improve circuit performance in existing applications. As a common practice. elimination of radiation losses and reduction of electromagnetic coupling has been achieved by enclosing the planar circuits in shielding cavities. In most cases. the cavities have to be placed away from the circuit to avoid proximity effects. thus introducing cavity resonances which interfere with the electrical performance of the circuits. While shielding is possible in many circuit applications, in monolithic arrays where the environment has to remain open. radiation from the feeding structure and parasitic coupling to the radiating elements have been major issues. Fhi, article presents, thcirc:ica;.,iJd \;c cn.:. '.. o, the microshield line ( Figure 1). a nc-. ".pe rt ni'nnoi'. line appropriate for circuit and arra\ appiications i l" Fni, geometry may be considered as an e(.iution ot the convcntional microstrip line or coplanar waaeguide and is char.:terized b. low radiation loss. reduced etir.rmalne:ic Intcrference and compatibility to antennas of mnicrotrip,r aperture type. In addition. the microshield geometr, cin cnr ate the basis of a new technology and can lead to novei Idc.i. in antenna and array design (111. In this new line. the ground plane has been deformed from Its original planar form to totally or partially surround the inner strip conductor as i shielding microcavity. Such a structure can be made monolithically using etching and metal deposition techniques Furthermore. the inner conductor can be suspended in air btr using membrane technology (121. This eliminates dielectric losses which could be high at millimeter-wave frequencies. Membranes have been effectively employed to provide highefficiency monolithic antennas and arrays operating at frequencies as high as 2.5 THz [131. One of the advantages of the microshield lint. In comparison with the more conventional ones such as the microstrip or the coplanar waveguide. is the abilit% to operate without the need for via holes or the use of air bridges for around equalization. Furthermore, due to the many available parameters in design, a wide range of impedances may be achieved. Specifically. by varying the size of the shielding microcavit, (see Figure 1) the per-unit length capacitance of the line can increase or decrease from the value of the corresponding microstrip or coplanar waveguide resulting in lower or higher values of the characteristic impedance. One loss mechanism still found in this novel structure is conductor loss Due to the geometry of the ground plane. conductor loss is expected to be slightly higher than microstrip and lower than conventional CPW. The presence of conductor losses limits the use of microshield lines and circuits to the lower end of the submillimeter-wave spectrum (up to 500 GHz). In this article the process for the fabrication of the microshield line is described in Section 2. and its feasibility is demonstrated by building a microshield line low-pass hiter teeding an aperture-type antenna. To provide an indication of the tendency of the microshield line to radiate less. cs ceral threedimensional discontinuities are analyzed and their response is compared to the same discontinuity geometries in conventional coplanar waveguide form. In all cases. the microshield discontinuities show much lower radiation than the conventional CPW ohes. The theoretical approach ulsed tI.Inalyze these discontinuities is presented briefly in Scctil'n A.nd numerical results are given in Section 5. In;iddJitr,, 'he characteristic impedance of the line is evaluated ii,,,- II \1.nalvsis. Moreover, measurements on scaled!1r', '. >'n -: MICROWAVE AND OPTICAL TECHNOLOGY LETTERS vo; 6: - 333

2. FABRICATION Ilhe:.thrit..Irion t miLrot.hic.ld clireuis is dependent r.:hi tnlli dielectric membrane t';.hnoios and; the.ininitrp',pK ni.-: oft the,uppo(rting vtters -A i memhrtnc an.1 thr., ',() vl- \,O() 0 truLrlt. 'r lltlh IU t c ll. nhrlt l ' ih. ' S.id tlla and rigid Sit —u;'n,rtinn,.har, ccri-tc, I he 5i ):.i\cr are rnade b, high-temperature therm.l xo\ldtiatn '!t,)') xhlle the S-,N. I.l\cr Is nmade h. chemical,vipr Jcp-,-ltion at S.()' 1121 Atter the dec elopment ot the threc-ljer.triicture. the membrane. aire tlhric.tted in tio.steps Fir.t.in fpening in the sIllcun-nlitridet lers i, denned on the.hack rt the w.iter. and then the sIlicon is etched until a transparent membrane appears. The etching solution is anisotropic and ields pyramidal cavities which can he exploited for the deeloipment of microshield transmission lines. Figure 1 giles the most important steps of the procedure In a simplified a-a. Once the membranes are fabricated. it is easy to lithographica;ll define several different microshield geometries for impedance and propagation constant measurements. Figure 2 shows a picture of a low-pass microshield filter exciting an aperture-type antenna. The eight-section filter consists of alternating 50-f and 125-f microshield sections and has Chebyshev low-pass characteristics with 0.5-dB passband ripple and a cutoff frequency of 76 GHz. The filter also performs an impedance transformation from 100 f) at the input to 50 fl at the output. This circuit has been made as a proof of fabrication feasibility and it does not present an optimized design. In fact. in order to successfully design this antenna configuration many important issues associated with the microshield geometry have to be studied and understood. Some of these issues are (i) parasitic radiation. (ii) transitions from conventional planar lines to microshield structures. and (iii) electromagnetic coupling to neighboring circuit elements. 3. THEORY The theoretical method used to analyze microshield line discontinuities is based on a space-domain integral equation (SDIE) which is solved using the method of moments. The SDIE approach has been previously applied to study several microstrip and coplanar waveguide (CPW) discontinuities and houndjrv prohlem s rpi:nt, s; npler, ne', ---.an cqui.llcnt maine!il curren-; i' - he,;. ',urtloc nrLltncn!i 'urrcnt r.:, ~:..,.:'d.,.- '.' tht'e rit;i o Mtns.il'i.,t I j ali<.;s~e * tf,:.'^ 'e,;, s.>., ltl.f!'. i! 1 h. v. X..:..;& ^,.....ipplied,i the.,on:tilnuist at the t.h n.: cl:.:t!:..''. the surface ot the s.iot apertures,..hich iid r'.. inte,:rai equation wi l j f| ( r r - (, r r i ] \" (F',/, where C(',, are the magnetic field d\adic Grecn', tunct;v the regions aboe and below the slots. respecltiels. ank an assumed ideal current source feeding the micro,,hicl [141. The free-space Green's function is expressed int:he of Sommerfeld integrals in space domain 115. ltj. %,hii Green's function in the shielding microcavits consist.,t ble summations over the waveguides cigensalues [14. Since the cavity dimensions are very small. higher-i microshield modes are not excited in the frequenc. ran interest. The integral equation (1) is solved using the methe moments where the unknown magnetic current is expa in terms of rooftop basis functions. Then Galerkin' me is applied to reduce the above equation to a linear ssste equations: Y,, Y,.. VI" Y. Y.:(V) (, ) - where Y, (i = y. z: j = y. ) represent blocks of the at tance matrix. V, is the vector of unknown y and: mag current amplitudes, and I, is the excitation vector whi identically zero everywhere except at the position of the sources. Finally, the equivalent magnetic current distrlbi and consequently the electric field in the slots are obta by matrix inversion. Away from the discontinuity. the fields form standing waves of the fundamental propagi mode. Using the derived electric field. an ideal transmit line method is applied to determine the scattering paramt of the discontinuity. From the scattering parameters. r; tion loss can be evaluated as a function of the frequency other geometrical parameters. 4. SCALED MODEL MEASUREMENTS Experimental characterization of circuits requires a.ccL measurement of the scattering parameters. the effectilv electric constant, and radiation and conductor losses S two-port scattering parameter measurements are limite, 100 GHz from state-of-the-art vector systems. experime characterization of the microshield line can be best perforl on scaled models even if the structure itself is intended submillimeter-wave applications. Due to the paslsie na of the microshield. most of the information obtained trorr scattering parameter measurements. excluding condu losses, can be scaled with frequency to the submillmllctrregion. The rf measurements were performed in the!it!-( frequency range on the HP 8510B automat.!c. '\,. Figure 2 A microshield line low-pass filter 334 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS Vo 6 No 6 May 1993

Ul) U in (n) C.)0n cn - 0n L......... U) -lo -....... C 03 7(3.......... E -~........... ~ -~ -.....-.E.E....... CT EJ'UC m9 U V U cc Lu 0 0 T 0 C" oM < o a-.0 U, -2 0 0 I-,U E 0 U U 0 0 <6E h-. -0

0 -10 -20 -30 -40 1 100 0 -100 0 11 12 13 14 15 16 17 18 Frequency (GHz).. *. I *' ' 1...,, ' ' I....* * I....* * I...., mu Angle(S12) == — =-^ Angle(Sl1) 10 11 12 13 14 15 16 17 18 Frequency (GHz) Figure 4 Scattering parameters (magnitude and phase) of a two open-end coupled dielectric microshield line discontinuit! s g = 0.5 mm). Solid lines: theory, dots: experiment validate the numerical method, scaled measurements were performed on the two coupled open-ends discontinuity. In this experiment, the microshield line has an inner conductor width of 1.5 mm, a slot width of 0.5 mm, and a microcavity width and height of 2.5 and 1.27 mm, respectively. Figure 4 shows the scattering parameters (magnitude and phase) as were evaluated theoretically and measured experimentally. The ripple seen in the measured data is mainly due to the very small coupling between the two lines. which intensifies the effect of connector repeatability. In the numerical results shown in Figures 5-7. the m shield lines and the conventional coplanar waveguides an inner conductor width of 0.5 mm and slot widths of mm. The substrate used in the dielectric microshield at the coplanar waveguide is 0.635 mm thick and has. diek constant e, = 2.1. The microcavity size [height (h)i and \ (a)] is specified in the figure captions. It is expected reducing the cavity size up to the point where the side touch the slot edges will improve field continemet,t rFh ity size does not present an optimum desiron.m,1t 'hc rt 336 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS Vol 6. No 6. May 1993

E 10 b0 j la 01 V, 0 9 10 w -16.17 E - - 4 -18 - -11 -- I ____X^J" 171 T - 1. I. - I.1 I. I. I -19 L 0.0 --— I - - - - - 0 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 Gap width, g (mm) Figure 5 Radiation loss factor of an open-end line (I/ = 3 mm. u mm. = 24 GHz).In 0o la I.. -12 -14 -16 -18 *Ic^ '. t i\ rTT'n i( *?':^;T 1i:,riii i ts.; v t. dimensions %were merelc ucd j. si,;i.. * '. -. cip.ahtilt\ of the proposed mincr,sh:c..,,. radiate considerahlt less th.Ir th e <,rh X.i cetllide ones O* pen end. -il,:urc ' e l'^ ' <, r.I.:;.:'...'. li)tl - ' of.n open end a.,. tun.:,;. %,ldth for the proposed microshield inc.a:'^i. r:.:, to the coplanar waieiuide open end As ca,.., '.".: the dielectric microshield open end h,,s the '\<:. i: tion loss. v, hich is approximately halt ot th.tt c",. ti:s.:.. in the con entional CPW, open end Lo, r aidn,:t, rc n.' from good held confinement in the slots. shiL.h is Juc. the presence of the dielectric substrate and the shicldin, _ ca. itv. Coupled Short Ends. Figure 6 shov,, the loss factor. 11, log(1 - 'SI - -IS,.'). for two coupled shortcd end. is a function of frequency. Again. the dielectric rnTicr,hieid has smaller radiation loss factor than the nmemt'rane microshield and the conventional CPW. Despite these differences. in all cases, the radiation loss is ser\ silmall due to the type of the planar discontinuity Coupled Open Ends. Figure 7 shows the loss factor for this discontinuity as a function of frequenc\y Here the cavity size is chosen such that the side walls touch the slot edges and with height equal to one slot,ildth It can be seen that radiation in the coupled microshield open ends is approximately one-third the radiation from coupled open ends made of conventional CPW. In addition. it can be seen that the dielectric microshield coupled open ends radiate more than the membrane microshield in contrast to the coupled short-end discontinuity (Figure 6). This may be due to the excitation of the surface swaves from the open ends along the direction of propagation in the dielectric microshield line. As mentioned earlier. in addition to low radiation. the microshield line has one more major advantage the ability to achieve a wide range of impedances by appropriately varying the geometry of the line and the size of its microcavitv. This range spans over the impedance ranges of the conventional microstrip line and coplanar waveguide. Figure S gives the characteristic impedance of a microshield line as a function of SI(S + 2W) for a given cavity size. where S is the center conductor width and W is the slot width. The characteristic impedance has been obtained using a TEM analsis 8IS. 191. This impedance should be the exact one as long as no higherorder modes are excited. It can be seen that as s (S - 2W) gets smaller, Z1, approaches that of a microstrip line 6. CONCLUSIONS A new type of monolithic planar transmission linc the microshield line, has been proposed. This line has the lfllowing advantages: * There is no need for via holes or air brides!,,r ground equalization. * A wide range of impedances can be oht.inedl * Wavelength is equal to the free-space,d,,lcithn s.hich results in larger circuit dimensions. * Radiates much less than the conventio., t '"A -2 I 18 20 22 24 26 28 30 32 34 36 Frequency (GHz) Figure 6 Radiation loss factor of two coupled short ends (h = 3 mm. a = 3 mm. S, = 0.25 mm) b.0o 0 U a -12 -13 -14 -15 -16 -18 -19 -20 -21 -22 -24 -)., I -. 7 17 m. -.::-; X.:... -, 7,F " I- - -, - II I I 7 1 I II I: I I I I I I III I i I I I I I - a I - Is * a - a.......... 18 20 22 24 26 28 30 32 34 36 Frequency (GHz) Figure 7 Radiation loss factor of two coupled open ends (h = 0.25 mm. a = I mm. S, = (.5 mm. g = 01.25 mm) MICROWAVE AND OPTICAL TECHNOLOGY LETTERS Vol 6 Nc '3 337

i -. - I ~.e '. I -.. i - a -(or GaAs) 300 -' ~ 200.*S 100 5 0.0 0.. 0.2 0.3... 0.5 0.6 0.7 0.8 0.9 1.0. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 S/(S+2W) Figure 8 Characteristic impedance of a microshield line (a = 1.2 mm. h = 0.2 mml The loss factor of several discontinuities has been evaluated and compared with similar discontinuities in conventional CPW. It has been found that the microshield line has lower radiation which is approximately half of that encountered in the conventional CPW. Preliminary experiments on scaled models have been performed where good agreement with theoretical results has been obtained. In addition, it has been shown that the microshield line has the ability to achieve a wide range of characteristic impedances. ACKNOWLEDGMENTS This work has been supported by the Office of Naval Research under Contract No. N00014-92-J-1070, the NASA Center for Space Terahertz Technology, and NSF under Contract No. ECS 8657951. The authors would like to thank Dr. W. Harokopus for supplying computer programs and Professor G. Rebeiz and his group for their scientific support. REFERENCES 1. R. Lagerlof. "Ridged Waveguide for Planar Microwave Circuits." IEEE Trans. Microwave Theory Tech.. Vol. MTT-2 1. July 1973. pp. 499-501. 2. R. Simons. "Propagation Characteristics of Some Novel Coplanar Waveguide Transmission Lines on GaAs at MM-Wave Frequencies." 1986 Conference on Millimeter Wave/Microwave Measurements and Standards for Miniaturized SYstenms. Aia Nov. 6-7. 1986. 3. F. Alessandri. U. Goebel. F. Melai. and R. Sorrentino. retical and Experimental Characterization of Nonsvmme Shielded Coplanar Waveguides for Millimeter-Wave Cir IEEE Trans. Microwave Theory Tech.. Vol MTT-37. Dec pp. 2020-2027. 4. Y. Wang and J. Okoro. "Impedance Calculations for M Coplanar Waveguides," /nt. J. Electron.. Ma p Ilo. p1 875. 5. K. Kawasaki and T. Itoh. "Analysis of Stitch Line for Mo Microwave Integrated Circuits." 1991 IEEE M1 TT.S Iierc Microwave Symposium Digest, Boston. pp 7S 1- 7.S4 6. K. Goossen et al.. "Nearly Dispersionless Micrcsrlrip for I Pulses Utilizing a Buried Silicide Groundplanc. I9, MTT-S International Microwave Symposmum DLe, \. Bos 1053-1056. 7. H. Ogawa. T. Hasegawa. S. Banba. and H Nakamoto. Transmission Lines for Multi-Layered NM1(.s1. /W MTT-S International Microwave Symposrumtt i)vetr. Bos 1067-1070. 8. T. Rozzi. G. Gerini. A. Morini. and MI Dc s.inils NM Buried Microstrip Inset Guide." in Pro(rcchl:,\, thl'c ropean Microwave Conference, pp. 673 -h's sV.pit 1i91 9. A. Engel and L. P. Katehi. "Low-Lo'. Mol.,tih., I rin Lines for Submillimeter and Terahertz F'rt,,,, i \pplk 338 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS Vol 6. No 6 May 1993

\ r lfarlkupu, P' Kchi( *.~ nru - '.,' **I'In I I I 2, (1 Kcli'i. ), ni~lircltinr1. \ ( 'III ' P ilirr1In. and [) R~urt~~c~.. sti i, in,jn i R \rr.hmc Icr-%kAnii''; kk,'fDimcni! n Ie I''fl.)r pp t-im 1 ( carh~irl a. 'I. I Da%cc. irid ( Chin In[L'ur~itcd I IIIA. Iin: c~ir ('o crnt *r( ~irnc -\rr~i. II t F_ 51 i, (,mdld i Ili' Lw ia~"t ppI), 1-4 N D'. P IKatehi. (9 Ptonchrak. and R Simons. l 'hcorctic:,jl eind E\perimenral Chlaracterizaitin int CoGplanair \Wdvcguied Di,Continuities Ior Filter Applications." IEEE Tran.r %licro a' r Thetrw Tech.. Mja 1991. pp. 873-SS2 I' liarokopus and P Katehl. "Radiation Los% l or Open (Coplanar 'A r ec~wuide Di,-tcontinuities.- /191 IEEE 51FTT-S Int'rnwnoal.I1,'rouiaie S.'wvmpostimii Die.ves. Bosron. pp 743-746. lb 6' W farokopus and P. Kiitehi. "Chatracterization ot Mficro-trip Discontinuitaes on Mulultilaver Dielectric Suhstrates Includine Radiatinon Los~ses." IEEE Tranis. Mlrowuave Theory Tech.. Dec 1989. pp. 21058-2 06 5. 17 N Dib and P. Katehi. "Modeling of Shielded CPW Discontinulties Using the Space Domain Integral Equation Method ISDIE)." J. Electroagii. WUvesc Appi.. Vol. 5. No. 4.5. 1991. pp 513-523. IS D Rowe aind B. Lao. "Numerical Analv-is of Shielded Coplanar Waveuimdes.- IEEE Trans. Micromave T/eorv Tech.. Nov. 1983. pp. 911-915. 19) D Marcuse. "Electrostatic Field of Coplanar Lines Computed wccith the Point Matching Miethod." IEEEJ. Quiamumtin Electron.. Mac 1989. pp. 939-947. Received 3-20-92 Mficroki-cie 4and Optical Technolog, Letters. 6/6. 333-339 1993 John Willev & Sons. Inc. CCC 0895-2477 93 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS Vol 6 No 6. 3 339

.,,,uJucu t t"tcu nermnaoaai journal oJ MacroCrcuus and iectronu Packagung Design of Micromachined High Frequency Circuit Components Rhonda Franklin Drayton, Nihad I. Dib, and Linda P. B. Katehi NASA Center for Space Terahertz Technology The University of Michigan Ann Arbor, MI 48109-2122 USA ABSTRACT This paper describes the development of high frequency micromachined circuit components with emphasis on the design and fabrication steps. These passive components have been fabricated using silicon micromachining processes which allow for the development of high-frequency circuits that have a monolithically integrated package. Some advantages of such configurations include reduction in the overall size and weight, improved isolation between neighboring circuits, minimized parasitic radiation effects, as well as increased circuit density due to the individual shielding of circuit geometries. Presented are circuit design and fabrication issues with extensive discussions of the implemented design procedure and of the adopted micromachining processes. Success of the circuit development is verified by comparison of on-wafer measurements to modeled results in the 10-40 GHz range. I. INTRODUCTION Planar transmission lines such as microstrip, stripline, and coplanar waveguide (CPW) have become conventional structures to use in the design of microwave and millimeter wave circuits due to the flexibility provided in fabricating passive components with predefined electrical functions as well as the enhanced ease in mounting active devices. Although microstrip and stripline have been utilized the most in passive circuits, limitations in mounting active devices have made the use of coplanar waveguide more popular since its physical geometry provides inherit advantages. A commonly observed problem in many of these lines, however, is degradation in circuit performance that results when coupling mechanisms associated with parasitics are excited along with radiation effects that arise in dense circuit environments. In order to address the above concerns, the development of a Page 1

transmission line geometry that offers electrical performance comparable to conventional ones, maintains ease in device mounting and reduces undesirable coupling effects is required. This can be attained with independent shielding of specific circuit components that can be achieved using micromachining techniques. Since most high frequency circuit applications address development, modeling, fabrication and experimental characterization of systems prior to packaging, the effect of the housing on the electrical performance is very difficult to predict. As a result, the electrical response of many packaged circuits suffers significant performance degradation, mainly attributed to the introduction of unwanted parasitics along with the excitation of multiple shielding resonances resulting from the interaction between the circuit board and metallic housing. To address the issue of proximity coupling and cavity resonances, monolithically integrated cavities can be developed which provide effective shielding to individual components while maintaining an overall geometry that is small enough to avoid the multiple resonance excitation in the range of operating frequencies. From a cost perspective, since conventional housing elements can be rather expensive and impractical to optimize for each generic circuit board, a solution toward cost minimization is easily achieved by using micromachining techniques. Consequently, for system level designs where weight and volume reduction as well as controllable parasitics are critical issues, overall system costs are directly reduced using these techniques. Industries related to satellite and mobile communications can benefit greatly from the technological development of miniature micromachined microwave components. Although well-established in sensor applications, silicon micromachining is currently being investigated for the development of miniature circuit components used for high frequency systems. A completely shielded micromachined circuit can be excited by a planar transmission line of coplanar type that is surrounded by an air-filled cavity in the upper region and a substrate-filled one beneath the line as shown in Figure 1. To address the specific design issues of shielding and isolation, a variety of micromachining processes have been explored and studied to identify approaches which provide circuits with the best electrical performance at these frequencies. Demonstration of the concept of micromachined circuits for RF applications is given through simple circuit components such as a tuning stub and lowpass filter. These simple components have been developed and their performance has Page 2

been measured and compared to conventional transmission lines [ I ], in this case coplanar waveguide. More specifically, this paper presents the development of micromachined miniature circuit components for high frequency applications in two sections. The first section, design approach for micromachined circuits, provides a general overview of the approach required to develop the circuits along with a brief summary of the theoretical models used. It also extensively discusses the fabrication procedure developed and implemented through this effort. The second section, experimental validation and discussion, describes the measurement considerations needed for circuit characterization and discusses the results obtained both experimentally and theoretically for various circuit components. I. DESIGN APPROACH FOR MICROMACHINED CIRCUITS High frequency micromachined circuits are developed in two phases: (a) design and analysis and (b) fabrication and measurement. In the design and analysis phase, determination of an appropriate commercially available CAD tool is a critical issue to address. Theoretically, while full-wave techniques offer the most accurate model prediction, they are inappropriate to use as design tools due to their complexity and computer time intensive requirements. Consequently, initial circuit design is best achieved using simpler quasi-static models that are based on TEM or quasi-TEM approximations. After an entry level circuit design has been completed, the specific circuit geometry is determined and analyzed using full-waVe analysis techniques to verify performance. Since these models require specification of the exact geometry, the modeled circuit provides a highly accurate prediction of the response. From this point, performance improvements can be subsequently obtained using an iterative method to fine tune the circuit dimensions for given design requirements. After an optimum design has been determined, fabrication is implemented using the procedures presented in section I-C below, and circuits are then tested to evaluate the overall circuit performance. A. Design Procedure To illustrate the procedures needed for realistic circuit development, a flow chart is presented in Figure 2. Using quasi-static or semi-empirical CAD tools, the basic circuit is designed electrically to meet the desired specifications. Once the appropriate response is obtained, the circuit geometry is determined and an analysis is performed using full-wave techniques such as the finite-difference time Page 3

domain (FDTD) method. During the analysis phase, the circuit response is evaluated and iterative changes are made to the geometry to improve the response. The final design of the micromachined circuit is then fabricated whereafter the response is measured using the experimental set-up described in section III-A. Having both theoretical and experimental results, comparisons are made to validate the circuit response. If the electrical response is within an acceptable performance margin, the procedure is complete; however, if the response is unacceptable, design objectives must then be reconsidered. In the case of high performance applications, discrepancies between theory and experiment mainly arise from fabrication tolerances. When design errors must be minimized, two options exist. The first option involves modifying the steps of the fabrication procedure that reduce the geometrical deviations between the theoretically modeled and the measured circuit. In instances where prototype cost is a critical issue, the second option involves modifying the electrical designs to make them less sensitive to fabrication tolerances while preserving desired electrical performance. B. Theoretical Models Utilized In the above design procedure, several comments can be made regarding the theoretical techniques implemented for the modeling of micromachined circuits. The characteristic impedance for basic transmission line structures such as the completely shielded feed line, used herein, has been computed using quasi-static models [2], while dispersion characteristics have been evaluated using the spectral domain technique [3]. This development is required in order to design appropriate feeding lines required for specific calibration procedures that operate in the 10 - 40 GHz range. Once the electrical behavior of the transmission lines is characterized, simple circuits are developed and modeled using a quasi-static software tool (PUFF [4]). This tool uses simple ideal transmission line theory and allows for easy design as a first iteration to the whole procedure. For more complicated circuit geometries, however, quasi-static models are inadequate and the use of full-wave analysis techniques is required. Upon completion of the design, analysis for verification is performed using full-wave models such as the finite difference time domain (FDTD) technique [5] since the fabricated geometries require versatility in modeling the exact shapes and dimensions of the realized planar lines and cavities (see Figure 1). Page 4

The finite difference time domain technique offers flexibility since an arbitrary geometry can be evaluated primarily by discretization. In this method, Maxwell's curl equations are expressed in discretized space and time domains and, then, are used to simulate the propagation of the initial excitation in the "leapfrog" manner [5]. To characterize any planar discontinuity, propagation of a specific time-dependent function through the structure is simulated using the FDTD technique. The time dependence of the excitation can be chosen arbitrarily; however, a Gaussian pulse is often used because it varies smoothly in time and its Fourier transform is also a Gaussian function centered at zero frequency. Following the time and space discretizations of the electric and magnetic field components, the FDTD equivalents of Maxwell's equations are then used to update the spatial distributions of these components at alternating half time steps [5]. The space steps, Ax, Ay, and Az are carefully chosen such that integral numbers of them can approximate the various dimensions of the structure. As a rule of thumb and in order to reduce the truncation and grid dispersion errors, the maximum step size is chosen to be less than 1/20 of the smallest wavelength, which corresponds to the highest frequency represented in the pulse, existing in the computational domain. The Courant stability criterion is then used to select the time step to insure numerical stability. In this paper, the electric conductors are assumed to be perfectly conducting with zero thickness and are treated by setting the tangential electric field components to zero at all time steps. For the electric field components lying on the dielectric-dielectric interface, the average of the two permittivities is used in the FDTD equations. To terminate the FDTD lattice at the front and back planes, the superabsorbing, first-order Mur boundary condition [6-8] is utilized to simulate infinite lines. The interested reader may consult [5,9,10] for a detailed description of the FDTD method. Compared to other methods, the FDTD method is relatively easy to implement for complicated geometries. It requires only simple arithmetic operations in the solution process and also has the advantage of being flexible for time and frequency domain analyses. Frequency characteristics can subsequently be obtained by applying the Fourier transform on the time response of the discontinuity. The response in a wide frequency range starting from dc can be obtained by only one pulse simulation in contrast to frequency domain techniques which provide the response at a specific frequency and are hard to apply at very low frequencies due to long feeding line requirement. Page 5

C. Fabrication Procedures The circuits mentioned above are a two silicon wafer system with <100> crystal orientation and rely primarily on micromachining of silicon. For high frequency circuit applications the fabrication processes are relatively new although they are fairly standard in microstructures. To understand micromachining technology for high frequency circuits, this section presents the fabrication as it applies to monolithically shielded geometries. The micromachined circuits presented have a shielding environment that has been monolithically integrated into a two-wafer system and is made of cavities in both upper and lower regions. As shown in Figure 1, the upper region consists of a metallized air-filled cavity while the lower region has a substrate-filled cavity of high resistivity silicon, er = 11.7, which is also metallized on the lower side. To measure the response of these circuits, grounded coplanar waveguide (GCPW) feeding lines are used followed by upper and lower cavities that are grounded through direct contact with the ground planes of the coplanar lines. Other important design issues are the development of accurate alignment marks between the various wafer layers that ensure repeatable alignment, a mechanism for exciting the individual circuits and the incorporation of a shielding environment for the individual circuits. Alignment is achieved by incorporating alignment and probing windows that are entirely etched through to expose the lower wafer surface. Shielding is achieved by developing a substrate-filled cavity beneath the line on the lower wafer while an air-filled cavity is formed over the line in the upper wafer. Descriptions of the individual wafer layers used to construct the completely shielded micromachined circuit are discussed in the next paragraph. The lower wafer shown in Figure 3 is a high resistivity, single-side polished silicon having a dielectric mask of silicon dioxide and wafer thickness of 350 pm. To develop the circuits, the planar lines are printed using standard photolithographic techniques and a seed layer of titanium/gold/titanium is initially evaporated and then electroplated to achieve the appropriate metal thickness, in this case 3 microns. After the circuits and alignment marks are printed, the lower side of the wafer is patterned using an infrared alignment to expose the regions that form the lower cavity. Since the cavities must be underneath the individual circuits, the alignment windows and the lower cavities are defined on the Page 6

unpolished side of the wafer. Prior to anisotropically etching the silicon, the oxide is removed from these areas using buffered hydrofluoric acid (BHF). The anisotropic etchant, ethylene diamine pyrocatechol (EDP), uses the <111> crystal plane of <100> silicon wafers as an etch stop causing an etch angle profile of 54.750 to form non-vertical sidewalls in the cavities as seen in Figure 3c with the lower cavity having its widest dimension near the upper surface. Prior to evaporating the final metal, the oxide in the groove underneath the ground plane is removed and then metallized and plated to metal thickness of 3 microns to form direct contact. The upper wafer, shown in Figure 4, requires patterning of both sides of the wafer prior to etching the cavity and probe windows. In this study, a low resistivity silicon wafer is used having a thickness of 500 pgm with 7,500 A of thermally grown oxide on both sides. After defining the probe window and alignment marks to be etched using photolithography, a thin metal layer of Ti/Au is evaporated for a lift-off procedure to mask the remaining silicon. This extra masking layer serves two purposes; it offers additional protection to the upper cavity roof and is used as a mask during infrared alignment to pattern the lower side of the windows as well as the upper cavities. Prior to etching, the oxide is removed from the areas where silicon is eventually etched as shown in Figure 4 and from the area where etch rulers are located that have widths corresponding to specific depths. This double sided etching process provides mechanical strength in multiple cavity wafers, Figure 5, by including a structural beam in the probe windows which increases handling ease during mounting of the upper wafer. Integration of the upper shield to the lower shielded planar circuits is completed after alignment and attachment of the two wafers via the microscope using regular adhesion methods. The resulting configuration is shown in Figure 6 where the heights of the upper and lower cavities are 280 and 350 microns respectively, while the widths are 1180 and 980 microns respectively. In applications where only upper half shielding is required, fabrication is easily achieved by excluding the lower cavity formation. Page 7

III. EXPERIMENTAL VALIDATION AND DISCUSSION A. Measurement Technique In order to characterize the circuits up to 40 GHz, current measurement systems require the use of a probe station to measure planar geometries. The set-up used for this work consisted of an HP 851 OB Network Analyzer, an Alessi probe station and Cascade Microtech ground-signal- ground probes with a probe pitch1 of 150 jim. To accurately measure the circuit performance of the micromachined shielded circuits, the Thru-Reflect-Line (TRL) calibration technique is used [11,12] where standards and various circuit components having identical upper and lower shielded regions are fabricated using the fabrication process techniques described above. A one tier de-embedding method is implemented for on-wafer probing to calibrate the system reference plane along the transmission line [13]. As a result, all transitions between the ANA and the newly defined circuit reference plane are taken into account and the appropriate coefficients are computed to establish an accurate calibration. B. Micromachined Circuit Component Characterization Simple discontinuities have been implemented to show the realization of conventional circuits in a micromachined configuration. This paper investigates a series open-end stub and a low-pass filter and compares measured results to theoretical ones. The series open-end tuning stub has physical dimensions shown in Figure 7. Comparison between measurements and full wave analysis results is shown in Figure 8. As observed from this figure, the theoretical and experimental results exhibit a shift in the resonant frequency of about 6.9% since the micromachined resonance occurs at 29 GHz compared to the 27 GHz response of the modeled circuit. Although the overall circuit perfonnance is similar, the discrepancy in the resonant frequency can be attributed to the variations between the actual modeled circuit and the one fabricated and measured. The resonant frequency is affected by the fact that the measured line length behaves electrically shorter since there is rounding of the corners and edges of the stub fingers caused during fabrication. This is not accounted for in the model. In addition, metal thickness has been neglected although it has 1. Pitch is defined as the spacing between the center conductor to the ground probe tips. Page 8

been found to contribute considerably to frequency shifts [14]. Lastly, the difference in the magnitude between measurement and theory may be attributed to the fact that the theoretical model assumes a lossless system while in reality the circuit inheritly has both conductor and dielectric loss. A five-section stepped-impedance lowpass filter, designed as shown in Figure 9, has high and low impedances of 10OQ and 20Q respectively and is surrounded by the cavity structure described above. In Figure 10 measurements are shown and compared to theoretical results derived from quasi-static models where conductor and dielectric losses are included. Regarding conductor losses, care was taken to incorporate the specific metallization thickness and the appropriate surface resistivity corresponding to the various sections of microstrip line widths [15]. To realize 100 and 20 ohm impedance steps, 20 gm and 380 4m wide conductor lines are used with slot widths of 210pm and 30pm. Electrically the low impedance sections propagate a coplanar waveguide mode due to the narrow slot width while the high impedance sections propagate a microstrip mode. These sections of lines are connected in series resulting in a mixed mode excitation, thus creating parasitics which cannot be easily accounted for in the quasi-static model. Despite this limitation, however, it can be seen that the measurement response is similar to the model. This is due to the relatively low frequency of operation which reduces the effect of the parasitic mechanisms. The total loss in the system, plotted on Figure 11, shows good agreement between theory and measured results thus confirming the effectiveness of the micromachined integrated shield and the elimination of loss associated with radiation effects. Although the PUFF model incorporates the loss effects into the ideal transmission line theory, fullwave analysis for a lossless system using FDTD is much more reliable (accurate) in predicting the behavior of the micromachined component. Puff is limited in this respect, since it cannot account for parasitics or coupling due to radiation, whereas, the FDTD model has the advantage in that it models the electrical performance based on the exact physical circuit dimensions in both horizontal and vertical dimensions. The results in Figure 12 derived by FDTD show excellent agreement between measurement and theory with the exception of loss. Page 9

IV. CONCLUSION In the above, simple circuit geometries used in conventional planar line designs have been implemented in micromachined form; namely, a series tuning stub and a stepped impedance lowpass filter. The results show that micromachining offers great design capabilities to high frequency application, while preserving electrical performance. The development of micromachined circuits for the miniaturization of conventional microwave circuits has been proven successful. The fabrication of these lines has been presented and measured results have been shown and compared to quasi-static and full-wave theoretical ones. The data presented prove that monolithic integration of the shield using micromachining techniques allows for the development of circuit components that offer comparable performance to conventional circuits. V. ACKNOWLEDGEMENTS This work has been supported by the Office of Naval Research under contract No.N00014-92-J-1070 and the NASA Center for Space Terahertz Technology. The authors would like to thank Mr. Thomas M. Weller and Mr. Chen-Yu Chi for the technical discussions. Page 10

VI. REFERENCES [1] R. F. Drayton and L. P. B. Katehi, "Micromachined Circuits for Mm-Wave Applications", Proceedings of 23rd European Microwave Conference, Madrid, Spain, pp. 587-588, Sept. 1993. [2] N. I. Dib and L. P. B. Katehi, "Impedance Calculation for the Microshield Line", IEEE Microwave and Guided Wave Letters, vol. 2, No. 10, pp. 406-408, October 1992. [3] M. El-Shandwily and N. Dib, "Spectral Domain Analysis of Finlines with Composite FerriteDielectric Substrate," International Journal of Electronics, vol. 68, No. 4, pp. 571-583, April 1990. [4] S. Wedge, R. Compton and D. Rutledge, PUFF Computer Aided Designfor Microwave Integrated Circuits, Version 2.0. [5] K. Kunz and R. Luebbers, The Finite Difference Time Domain Method for Electromagnetics, Florida: CRC press, 1993. [6] G. Mur, "Absorbing boundary conditions for the finite-difference approximation of the timedomain electromagnetic-field equations," IEEE Transactions on Electromagn.agetic Compatibility, vol. 22, No. 11, pp. 377-382, Nov. 1981. [7] K. Mei and J. Fang, "Superabsorbtion-A method to improve absorbing boundary conditions," IEEE Transactions on Antennas and Propagation, vol. 40, No. 9, pp. 1001-1010, Sept. 1992. [8] V. Betz and R. Mittra, "Comparison and Evaluation of Boundary Conditions for the Absorption of Guided Waves in an FDTD Simulation," IEEE Microwave and Guided Wave Letters, vol. 2, No. 12, pp. 499-501, Dec. 1992. [9] X. Zhang and K. Mei, 'Time-domain finite difference approach to the calculation of the frequency-dependent characteristics of microstrip discontinuities," IEEE Transactions on Microwave Theory and Techniques, vol. 36, No. 12, pp. 1775-1781, Dec. 1988. Page 11

[10] D. Sheen, S. Ali, M. Abouzahra and J. Kong, "Finite-Difference Time-Domain Method to the Analysis of Planar Microstrip Circuits," IEEE Transactons on Microwave Theory and Techniques, vol. 38, pp. 849-857, July 1990. [11] E.W. Strid and K.R. Gleason, "Calibration Methods for Microwave Wafer Testing", 1984 IEEE MTT-S International Microwave Symposium Digest, pp. 93-97. [12] G. Engen and C. Hoer, "Thru-Reflect-Line: An improved Technique for Calibrating the Six-Port Automatic Network Analyzer", IEEE Transactions on Microwave Theory and Techniques, vol. 27, No. 12, Dec. 1979, pp. 987-993. [13] M. Maury, S.March, and G. Simpson, "LRL Calibration of VectorAutomatic Network Analyzers", Microwave Journal, pp.387-391, May 1987. [14] W. Heinrich, "Full-Wave Analysis of Conductor Losses on MMIC Transmission Lines," IEEE Transactions on Microwave Theory and Techniques, vol. 38, No. 10, pp. 1468-1472, Oct. 1990. [15] T. E. van Deventer, "Characterization of Two-Dimensional High Frequency Microstrip and Dielectric Interconnects", Ph.D dissertation, The University of Michigan, Dec. 1992. Page 12

Figures Upper Cavity Lower Cavity FIGURE 1. Three - dimensional cross section of micromachined lines where the shield and line are integrated monolithically. Page 13

DESIGN/ ANALYSIS Analyze using FDTI Desired Response Achieved? YES FABRICATIONI/ Fabricate Circuit TESTING Test Circuit Response Theory and Measured - Response compare wit] performance margins? YES STOP FIGURE 2. Design Procedure for Micromachined Circuits. Page 14

dielectric mask oxide layer -/ (a) Print circuits (b) Etch lower grooves Lower substrate-filled cavity (c) Remove Oxide Metallize Lower Grooves FIGURE 3. Lower Wafer Development. (a) Transmission lines are printed on the top surface.(b) Lower cavity is formed by etching v-grooves with an etch angle of 54.75 degrees in <100> silicon using anisotropic etchant. (c) Lower cavity grooves are metallized below the line. Page 15

EDP I bi-layer mask /4 metal layer oxide layer VTT (1) (2) (a) (1) Alignment marks and probe windows (2) Upper cavity (b) Metallize cavity (c) Final cross-sectional view: (1) alignment mark, (2) probe window, (3) upper cavity FIGURE 4. Upper Wafer Development. (a) Etch the probe windows and alignment marks from both sides while etching the cavity from one side only. (b) The cavities are metallized. (c) Final view of a section of the upper wafer. Page 16

FIGURE 5. Photograph of circuit from top view where the probe windows are shown in relation to the transmission line wafer. Page 17

A' mnment Window Probe Window Air-filled Upper Cavity Lower Substrate-filled Cavity FIGURE 6. Completely Shielded Micromachined Circuit after lower and upper wafer alignment. Page 18

I 10801m I FIGURE 7. Series Open.End Tining Stub. wl= 60gm I I I g = 30m Page 19

0 1 -5-. - S1- S1FDTD '0 C. - -.- -- S21-FDTD '.0 -10 " '... -... 10 20 30 40 Frequency, GHz FIGURE 8. Measured vs. FDTD results for the Micromachined Shielded Series Open End Stub. Page 20

TABLE 1. Circuit Dimensions!- li:;|ll ^ l lllllllll I- Line Lengths (microns) - S t11 — 9^ x* -*l= 988 i 20Rm:'~'.: j~:~~. - - -:::-:.: - t.,,,,, - w....... -,,...............A '....... "i" iiii it g - --;iijiii 1 2=703 210Dm -- a. I La3=940 X ___1 -=722 FIGURE 9. Stepped Impedance Lowpass Filter Circuit Dimensions with low impedance sections of 20 ohms and high impedance sections of 100 ohms. Page 21

i i I I i 0 -5.. I I v v v v w 1... 4 s- — 4 -r la I.0 TD * -10 ' -15 -20 - -25 -..!... I I!I N "I ', I /, \ I U \ 'i, I,III: hi ii II I ' I - o- PUFF Mc ed - -- -PUFF Model -30 i I... i.... 3.... 1. 10 15 20 25 30 35 40 Frequency, GHz 0 -5 m oI I" I -10 -15 -20 -25 -30 10 15 20 25 30 35 40 Fequcy, OGH FIGURE 10. Comparison between the PUFF quasi-static model and measured results. Page 22

0 -5 -I....... *4 ----------- S -10 - -15 -2I I 10 15 20 25 30 35 requency, GHz FIGURE 11. Comparison between measured and PUFF modeled Loss response. 40 Page 23

0 -5; / \ - MeaSrured -/ -' -o-a S21-Measured ' -'/ --- sil-FDTD -10o 'a -- - S21-FDTD -15-. ', -20 l l.-30,.,, 10 15 20 25 30 35 40 Frequency, GHz FIGURE 12. Measured and FDTD results for the Completely Shielded 5-section Lowpass Stepped Impedance Filter. Page 24

Figure Captions Figure 1: Three - dimensional cross section of micromachined lines where the shield and line are integrated monolithically. Figure 2: Design Procedure for Micromachined Circuits. Figure 3: Lower Wafer Development. (a) Transmission lines are printed on the top surface.(b) Lower cavity is formed by etching v-grooves with an etch angle of 54.75 degrees in <100> silicon using anisotropic etchant. (c) Lower cavity grooves are metallized below the line. Figure 4: Upper Wafer Development. (a) Etch the probe windows and alignment marks from both sides while etching the cavity from one side only. (b) The cavities are metallized. (c) Final view of a section of the upper wafer. Figure 5: Photograph of circuit from top view where the probe windows are shown in relation to the transmission line wafer. Figure 6: Completely Shielded Micromachined Circuit after lower and upper wafer alignment. Figure 7: Series Open-End Tuning Stub. Figure 8: Measured vs. FDTD results for the Micromachined Shielded Series Open End Stub. Figure 9: Stepped Impedance Lowpass Filter Circuit Dimensions with low impedance sections of 20 ohms and high impedance sections of 100 ohms. Figure 10: Comparison between the PUFF quasi-static model and measured results. Figure 11: Comparison between measured and PUFF modeled Loss response. Figure 12: Measured and FDTD results for the Completely Shielded 5-section Lowpass Stepped Impedance Filter. Page 25

Submitted to the IEEE Transactions on MTT, September 21, 1994 Development of Self-Packaged High Frequency Circuits Using Micromachining Techniques Rhonda Franklin Drayton, Student Member, IEEE, Linda P. B. Katehi, Senior Member, IEEE NASA Centerfor Space Terahertz Technology The University of Michigan Ann Arbor, MI 48109-2122 ABSTRACT The objective of this paper is to present a new concept in packaging: the development of self packaged high frequency monolithic circuits using silicon micromachining technology. These circuits employ miniaturized housings to shield individual passive circuit components (e.g. CPW) or active elements or combinations of them. Primary advantages of such self-packaged configurations at these frequencies include reduction in overall size and weight and increased isolation between neighboring circuits. These characteristics make the presented micropackaged components appropriate for high density, multilevel interconnect circuits. In this paper, the fabrication procedures used to develop self-packaged components are described and performance curves of measured and theoretical results for typical high frequency circuit geometries are shown. I. INTRODUCTION Microwave and millimeter wave circuit design relies heavily on the use of planar transmission lines such as microstrip, stripline, and coplanar waveguide to achieve design flexibility as well as ease in mounting active components. While microstrip and stripline have been utilized the most in the development of passive circuits, limitations in the realization of desired circuit layout have introduced the use of coplanar waveguide or slotline, due to the inherent advantages posed by their geometries, to create what is known as uniplanar circuits. While the combination of the aforementioned lines offers versatility in circuit design, it intensifies parasitic mechanisms such as electromagnetic coupling and parasitic radiation which compromise performance and make 1

success in design conditional upon circuit performance requirements. Therefore, the development of a novel circuit geometry that offers the capability of self-packaging while preserving the structural characteristics of conventional uniplanar technologies can provide a new dimension to high frequency circuit design. Individual shielding to specific circuit components, provided in monolithic form, allows for more compact circuit configurations and successfully addresses the issue of high density. This individual shielding of circuit components is achieved through the use of monolithic cavities which are referred to as "micropackages" since they can exist in miniature form within a larger housing configuration. These self-packaged circuit geometries can be fabricated through the novel use of well established Si or GaAs micromachining technologies. In typical applications, circuits are packaged in metallic housings which tend to introduce unwanted parasitics and multiple resonances and ultimately lead to overall performance degradation. In addition to the above, conventional housing structures can be rather costly and inappropriate for use in lightweight system designs. Micropackaging, on the other hand, can result in increased circuit density while performance requirements are attained through reduction of parasitic electromagnetic coupling in the substrate (substrate modes) and air (space waves). Circuit micropackaging is achieved through cavity regions which are fabricated in materials such as silicon or GaAs and are designed to electrically and/or hermetically encapsulate planar circuit components or subsystems. While preserving monolithic character, micropackaging allows for easy integration with other more conventional circuit arrangements and shielding environments. In applications related to cellular and mobile communications, the resulting self-packaged lightweight circuits offer controllable parasitics and are appropriate for systems with demaiding performance requirements. Although micromachining of silicon is a well established technology for sensor and biomedical applications, for the first time this technology is being applied to develop self-packaged circuit components for high frequency applications. The development of completely shielded micromachined circuits is presented in this paper. The circuit components may be of microstrip or coplanar 2

waveguide (CPW) type and are surrounded by an air-filled cavity in the upper region and a substrate-filled cavity in the lower region. Both cavities are integrated monolithically with the circuits, as shown in Figure 1. Since a comprehensive understanding of fabrication capabilities is critical for the development of these circuits, an extensive study of various fabrication techniques has been performed and a brief description of the findings is presented. To show applicability to high frequency circuit design, simple components such as tuning stubs and filters are developed and measured data are compared to theoretical results. In part (a), a general overview of the design approach is given followed by an extensive discussion of the fabrication procedures developed and implemented. In the part (b), measurement considerations are discussed, fundamental line characteristics are presented, and experimental results for various circuit components such as tuning stub elements and filters are shown. The structures presented in this paper represent only one approach for implementing micromachining techniques to high frequency circuit design. The use of silicon fabrication processes can allow the circuit designer to alter the substrate geometry in a number of creative ways in an effort to optimize circuit performance and reduce its size. By using other approaches, effective forms of hermetic micropackaging mounts or other protective packaging types may be developed for use in hostile environments. Since micromachining has only been recently used in high frequency applications, the development of micromachined circuits is still in its infancy. Therefore, in an effort to outline the potential of this technology, this article presents the first generation of micromachined, self-packaged circuit components. II. DEVELOPMENT OF MICROMACHINED CIRCUITS In order to illustrate the versatility of micromachining, this effort concentrates on developing circuits that reside in partially or completely shielded environments. As it has been extensively discussed in the literature, poor circuit performance in circuits operating in open environments is attributed to free-space radiation and substrate mode excitation, while in shielded environments it 3

is mainly due to package resonances. The open environment described herein refers to circuits which are printed on a dielectric substrate and are free to radiate into free space. On the other hand, the term "shielded geometries" implies circuits which can be partially or completely shielded through cavities in the upper and/or lower regions. Since substrate modes occur in the dielectric substrate, introducing physical alterations to the substrate itself can result in elimination of these parasitic waves and improvement of circuit performance. In conventional packaging mounts unwanted parasitics can cause package resonances which interfere with circuit performance and provide very strict limitations to design. The proposed approach eliminates these resonances by developing micropackages (shielded structures) that are small in size, can follow individual circuit paths, and can be fabricated monolithically with the circuit. The result is a selfpackaged circuit that can be effectively used with other conventional circuits within larger conventional housings either hybridly using flip-chip technology or monolithically in miniaturized packaged configurations. The following sections outline the steps needed to design the circuits mentioned above and extensively describe the silicon (Si) fabrication processes required for their development. ILA Design The first step in the design of high frequency circuits is the specification of the geometrical parameters needed to provide the desired electrical performance. While there exists a wide variety of commercially available computer aided design (CAD) tools for low-frequency applications, circuit design software for high-frequency applications is currently unavailable. Despite the lack of high-frequency CAD software, there is a wealth of high frequency circuit simulation tools that can be used indirectly in circuit design which produces satisfactory results but requires very long design cycles and extensive computation times. The micromachined self-packaged circuits presented here have been designed through such an iterative approach. In this iteration scheme, initial design is based on low-frequency software models that use quasi-static approximations [1-2]. Given this design, the geometrical parameters are changed in an iterative manner and the new cir 4

cuit performance is predicted using the high-frequency analysis software which is based on fullwave models in frequency or time domains. When the predicted electrical performance closely matches the desired one, this iterative cycle is terminated [3-10]. Additional issues that need to be addressed during the design stages are package resonances and substrate mode excitation. As it has been extensively described in the literature [11-12], electrical packages can greatly affect circuit performance either through package resonances or through proximity coupling. The first effect is mostly related to the dimensions of the package while the latter is due to cross-coupling of neighboring circuits. Unfortunately these two mechanisms require contradicting measures which lead to design trade-offs. Ideally, a given circuit has an optimum package size that is small enough to eliminate resonances within the range of operating frequencies and that is physically far enough away from the circuitry so that it does not interfere with the circuit's electrical performance. Although circuits in open environments do not face the previously described problems, they are prone to parasitic radiation which is mostly associated with the excitation of substrate modes. Since the excitation of these modes is mostly dependent on the operating frequency and the physical thickness of the substrate, careful layout configurations in less dense circuit environments can sometimes reduce parasitic radiation, provided there is flexibility in circuit placement. In practical applications, however, circuit requirements greatly limit the flexibility in rearranging the location of the various circuit components such that any layout modifications, at best, can only weakly reduce parasitic loss [13]. During the characterization efforts performed, preliminary findings indicate that substrate modifications alone may have a substantial impact on substrate mode excitation. While researchers have been aware of the issues discussed above since the late seventies when planar monolithic and hybrid circuit approaches became fundamental to high-frequency circuit design, the existing at that time technology could not provide alternative solutions to substrate modes excitation. Recent advances in silicon micromachining techniques, however, allow for unique, yet simple practical solutions to the above problems. With the development of micropackages, dimensions 5

can be chosen simply by using waveguide and cavity models to predict geometrical dimensions that avoid unwanted resonances. In addition, the shape of the cavity can be designed so that it follows the circuit and does not physically affect its performance. For open circuits, substrate modifications using micromachining can be implemented to eliminate these unwanted substrate modes entirely, resulting in improved circuit performance. II.B Fabrication Procedures The micromachined circuits described here are made on a two silicon wafer system having a <100> orientation and primarily utilize silicon micromachining processes. Although fairly standard in sensor applications, these techniques are relatively new to high frequency circuit design. This section presents an extensive discussion on the fabrication steps required for the development of self-packaged circuits and provides a detailed description of the employed processes. All circuits presented herein are of coplanar type, are printed on Si and are shielded by miniature cavities in either or both upper and lower surrounding regions. As seen in Figure 1, the upper region consists of a metallized air-filled cavity while the lower region is a substrate-filled cavity that is metallized on the lower side. Since the circuits are printed on the lower wafer, high resistivity silicon with er = 11.7 is required. In this configuration the upper and lower cavity regions provide ground plane equalization through direct contact with the ground planes of the coplanar waveguide lines. Issues important to consider during design include the fabrication of accurate alignment marks between the various wafers, the design of appropriate feeding lines for individual circuit excitation, and the development of enclosing environments that can be integrated monolithically with the specific circuits without degrading electrical performance. The following sections address these concerns comprehensively while describing the technique used for individual wafer development. The lower wafer shown in Figure 2 is a high resistivity, single-side polished, 350 j~m thick silicon substrate having a dielectric mask of silicon dioxide that has been thermally grown to 1.2 pm 6

thickness. To develop the circuits, the planar lines are printed using standard photolithographic techniques. Since electroplating is used to achieve the 3 micron metal thickness needed for Ka band operation, a seed layer of titanium/gold/titanium (Ti/Al/Ti) is initially evaporated. After the circuits and alignment marks are printed, the lower side of the wafer is patterned photolithographically and regions needed to form the lower cavity are defined using an infrared (IR) alignment procedure. With the substrate-filled cavities being underneath the individual circuits, alignment windows are needed and through this process are defined together with the lower cavities on the backside of the wafer. Prior to anisotropically etching the silicon, the oxide is removed from the areas that define the cavity sidewalls using buffered hydrofluoric acid (BHF). The anisotropic etchant, ethylene diamine pyrocatechol (EDP), uses the <111> crystal plane as an etch stop for <100> silicon wafers resulting in etch angles profiles of 54.750 [14]. The created cavities have pyramidal sidewalls as seen in Figure 2b, where the lower cavity is shown to have its largest width near the upper surface and smallest width at the bottom of the cavity (Figure 2c), simulating an inverted pyramid. The lower cavity shielding is formed when the oxide underneath the CPW ground plane is removed and a seed layer of Ti/Au is evaporated. To provide appropriate ground plane thickness to this lower shielding package, the seed layer is electroplated on the backside of the wafer to 3 microns. The upper wafer shown in Figure 3 contains both upper cavities and alignment marks that are formed by etching from both sides of the wafer. Since these cavities only provide ground plane equalization and shielding without interfering with the signal path, it is not necessary to use high resistivity Si. The upper shielding in this work is developed using 500 gum thick low resistivity Si with 7500A of thermally grown oxide on both sides. After defining the probe window and alignment marks using photolithography, a lift-off procedure is employed to open the areas to be etched while providing an additional masking layer of Ti/Au metal on the backside of the wafer. This layer serves two purposes, it offers protection on the back of the cavity regions and acts as a mask during backside IR alignment. On the lower side, the cavities are defined and the oxide is 7

removed to expose the silicon surface as shown in Figure 3. The patterns are then etched in EDP to a desired dimension which is monitored through "etch rulers" that consist of rectangular widths corresponding to specific etch depths. Since this wafer must be handled frequently after etching the multiple cavities and windows, additional mechanical strength can be provided by including a structural beam in the layout as seen in Figure 4, where the beam is located in the middle of the probe window. After fabrication the cavities in the upper wafer are aligned to the planar circuits and cavities in the lower wafer and the two wafers are securely bonded together to complete the formation of the micropackage (Figure 5). Since the above procedures concentrate on the development of a completely shielded configuration, it should be noted that designs requiring partial shielding can be easily done by following the necessary wafer development scheme. III. EXPERIMENTAL RESULTS AND DISCUSSION Ill.A Measurement Considerations In order to measure circuit performance up to 40 GHz, conventional on-wafer characterization is employed in conjunction with the Thru-Reflect-Line (TRL) calibration technique [15-17]. The measurement set-up uses an HP 8510B Network Analyzer that operates up to 40 GHz, an Alessi probe station, and Cascade Microtech ground-signal-ground probes that have a probe pitch1 of 150 pim. This calibration is achieved using standards that have upper and lower shielded regions identical to those of the circuits of interest and are developed using the fabrication procedures described above. A one tier de-embedding technique is used for on-wafer probing which calibrates the system reference plane to a point within the shielded transmission line. This results in characterization of all transitions located between the input and output ports of the ANA and the newly defined reference plane. 1. Pitch is defined as the separation between the center conductor and the ground plane pin. 8

Measurements have been performed on partially and completely shielded configurations and are completed in two steps. Circuits with upper-half shielding, only, are characterized first. In this arrangement, the cavities are mounted on the coplanar waveguide (CPW) circuits and the whole structure is mounted on duroid with er = 2.2 and thickness of 3.175 mm to prevent the formation of a parallel plate waveguide between the ground planes of the CPW and the wafer chuck. The next step of this investigation deals with the characterization of various circuit geometries which are completely shielded as it has been described in previous section. Since the circuits under test are completely isolated from the outside environment, they are placed directly onto the probe station wafer chuck in the measurement set-up. III.B Characterization of Partially or Completely Shielded Circuits The characterization of partially as well as entirely self-packaged circuits can provide a very comprehensive understanding of the effect of micropackaging on circuit performance. The following sections present a systematic theoretical and experimental investigation which has been performed on a variety of circuits. These circuits can be grouped into three categories: (a) circuits with upper shielding (US) (b) circuits with lower shielding (LS) and (c) circuits with complete shielding (CS). III.B.I Delay Lines and Cavity Dimensions To accurately characterize the presented micromachined geometries, the first issue to address is the development of feeding lines needed to provide the appropriate excitation. Since the measurement system reference impedance is 50 ohms, the feedline dimensions are determined to meet this requirement. For the circuit components presented in this paper, the feedlines were designed using CAD tools available at the University of Michigan [1]. For the on-wafer probe station, coplanar waveguide feedlines were designed to have circuit dimensions of 100 gm center conductor width and 60 gm slot width. These were determined using design equations from Ghione et. al [18]. Matching impedances in the shielded region resulted in planar line dimensions of 180 Am 9

center conductor width and 130 gm slot width. The cavity dimensions for these lines are given in Table 1 of Figure 6. The micromachined circuits in this study can have several transitions which minimize the mismatch between the probe and shielded geometries (Figure 7). Specifically, completely shielded structures and partially shielded embodiments have similar transitions although one of the shielding regions is absent. The first transition occurs between two 50 ohm sections of grounded CPW (GCPW) where the first section (A-B) is a probe feeding pad having center conductor width of 100 jm and slot width of 60 gm that tapers (B-C) outward to a wider line having center conductor width of 180 gm and slot width of 130 jm. The next transition is a discontinuity (D-D) that occurs between the open GCPW line and the completely shielded GCPW. At this transition the conducting line dimensions remain the same but an upper shield has been integrated monolithically and has dimensions that are chosen such that the impedance of the line is not effected by the presence of the shield. As mentioned earlier, the upper region is an air-filled cavity with a width of 1200 jm and height of 280 gm while the lower region is a dielectric-filled cavity that is designed narrower and has width of 950 jm and height equal to 350 gjm. The lower cavity dimensions provides the minimum cavity width in which the sidewalls will not interfere with the field confinement. Since the calibration reference planes are located inside this shielded region, all measured circuits have similar feedline transitions to allow use of the same calibration standard set. The response of a through line for open CPW and upper shielded CPW has been measured and results are shown on Figure 8. As indicated by these results, the open CPW line suffers from parasitic radiation in the form of substrate modes which is indicated by the ripple shown in the data toward the higher frequency end of the band. When the same line is packaged in the upper air region only, coupling into substrate modes is reduced resulting in a flatter response over the entire frequency range. 10

For the lower shielded (LS) and completely shielded (CS) through line, a plot is shown in Figure 9 for the line attenuation in dB/mm which includes the effects of both dielectric and conductor loss. The results shown on the figure indicate that the lower shielded line has performance comparable to data presented by Taub [19] for losses in coplanar waveguide which is printed on silicon wafers with similar resistivity (3000 ohm-cm) and with the same aspect ratios. It is interesting to observe that attenuation in the completely shielded line (CS) is slightly higher than the lower shielded line (LS) due to additional conductor loss that is present in the upper shield. In conclusion, the data shown indicate that both LS and CS cases provide a favorable alternative to the use of grounded CPW which is known to suffer from excitation of substrate modes. In the open-end series stub section discussed later, a more thorough description of micropackaging effects on circuit performance will be presented in order to show the benefits of either partial or complete shielding compared to open environment circuit designs. II.B.2 Short-End Tuning Stub Various simple discontinuities are implemented to show the realization of conventional circuits into micromachined self-packaged design configurations. Figure 10 shows the physical dimensions of a series short-end tuning stub element [20] located within a completely shielded embodiment while Figure 11 shows a comparison between measured and full wave analysis results. The theoretical and experimental data are plotted and show similar electrical performance except for a shift in the resonant frequency which is attributed to the geometrical variations between the model and actual circuit geometry. Since the theoretical results do not account for losses, the difference in the level of the transmission coefficients is expected. Another geometrical variation that may also be responsible the observed differences is the trapezoidal shape of the thecavity walls as oppose to the rectangular shape assumed by the models. Furthermore, the theoretical model assumes identical widths whereas the upper cavity is actually wider than the lower cavity (see Figure 6). Both of these effects can alter the circuit parasitic capacitances, thus, modifying the bandwidth and resonant frequency. Lastly, since the model assumes perfectly rectangular comers 11

in the stub fi; -s, the measured line lengths can appear electrically shorter since the stub fingers suffer from rounding of corners and edges as a result of the fabrication procedures. III.B.3 Filters A five-section stepped-impedance CS lowpass filter has been designed as shown in Figure 12 with high and low impedances of 100 0 and 20 a respectively. Figures 13 and 13 show a comparison between measurements and theoretical results derived from quasi-static models with both conductor and dielectric losses are accounted. In the PUFF [2] model, care was taken to incorporate the specific metallization thickness and the appropriate surface resistivity which correspond to the various sections of microstrip line widths. To realize 100 and 20 ohm impedance steps, 20 pm and 380 pm wide conductor lines are used with slot widths of 210 pm and 30 pm. In the low impedance section the line excites a coplanar waveguide mode due to the narrow slot width while the high impedance section excites a microstrip mode. This mixed mode operation produces parasitic inductances and capacitances that cannot be easily accounted for in the quasi-static model. Despite this limitation however, at relatively low operating frequencies the effects of such parasitics are reduced and, as seen in Figures 13 andl3, the measurements agree very well with the theoretical data. The total loss in the system, as shown in Figure 15, shows good agreement between theory and measured results indicating that the circuit has negligible radiation loss and confirms the effectiveness of the micropackage. The level of loss can be attributed to the aspect ratios of the low and high impedance sections which are known to cause higher loss in both the coplanar waveguide mode [21] and the microstrip mode [22]. Additional validation of the response of the self-packaged filter is provided through a comparison between theory and experiment seen on Figure 13. The plotted data show excellent agreement between measurements and the finite difference time domain model applied for the analysis of a completely shielded (CS) geometry where loss effects have been neglected. 12

III.B.4 Open End Tuning Stub The open-ended tuning stub [20] shown in Figure 17 will be used to illustrate the electromagnetic effects from a variety of micropackaging configurations. Performance curves will be shown for the upper (US), lower (LS) and completely shielded (CS) configurations. These results will be compared to open CPW and a discussion will be presented on the nature of the various effects on the circuit response. In the case of open CPW, loosely bound fields tend to leak power strongly and destructively into substrate modes. When shielding occurs in either or both regions, the propagating modes become tightly bound to the line and radiation is reduced significantly compared to the case of the open CPW. The circuit performance shown in Figure 13 compares the open CPW to the CS geometry. The open CPW shows performance degradation above 25GHz due to the excitation of a strong substrate mode, while completely shielded circuits exhibit a very smooth response even at higher frequencies due to reduction of parasitic radiation and surface waves. When comparing the US to the LS configuration (Figure 19) for the same circuit geometry, the LS has the smoothest response due to the substantial reduction of substrate leakage. However, although the circuit performance for the US show a few ripples at the higher frequencies, overall performance improvements are observed as a nearly symmetric response is obtained. While the LS and CS scattering parameter measurements appear virtually identical in Figures 13 and 19, existing differences will become more apparent when observing the total loss data. A comparison between the total loss of self-packaged components and open environment components, as seen in Figure 20, shows that open configurations exhibit the highest overall loss. Of the self-packaged components, the US configuration has the highest loss which is mainly due to conductor loss and excitation of a strong surface waves. The lowest loss is presented by the LS geometry since the fields are primarily confined in the dielectric-filled cavity causing a reduction of parasitic radiation effects into air. In these circuits the LS package size is chosen to be small enough to suppress unwanted resonances. Lastly, the CS geometry exhibits loss performance that 13

is compromised slightly when compared the to LS case due to the presence of the upper shield. In general, however, micropackages with partially or completely shielded environments can offer significant improvement in electrical performance over open environment circuits. In instances where maximum reduction in loss is needed, the lower shielded configuration is the ideal choice while optimum isolation in either or both regions requires use of the completely shielded configuration. IV. CONCLUSION The development of micromachined cavities for micropackaging of high frequency circuits has been proven successful. The fabrication of these structures has been presented and measured data have been compared to theoretical results. The response of simple circuit geometries used in conventional planar line designs were implemented as through lines, series tuning stubs, and a stepped impedance lowpass filter. Since the derived data show that the monolithic incorporation of a shielding cavity with the circuit results in improved performance, "self-packaged" configurations provide the ability to evaluate specific designs more comprehensively. Lastly, as a result of the good agreement between measured data and theoretical predictions for basic components, the above study strongly indicates the potential of micropackaged circuits in high frequency circuit applications. V. ACKNOWLEDGEMENTS This work has been supported by the Office of Naval Research under contract No.N00014-92 -J-1070 and the Army Research Office. The authors would like to thank Dr. Nihad Dib for theoretical support and Mr. Thomas M. Weller and Mr. Chen-Yu Chi for the many technical discussions. 14

VI. REFERENCES [1] N. I. Dib and L. P. B. Katehi, "Impedance Calculation for the Microshield Line," IEEE Microwave and Guided Wave Letters, vol. 2, No. 10, pp. 406-408, October 1992. [2] S. Wedge, R. Compton and D. Rutledge, PUFF Computer Aided Design for Microwave Integrated Circuits. Version 2.0. [3] N. Dib and L. Katehi, "Modeling of Shielded CPW Discontinuities Using the Space Domain Integral Equation Method (SDIE)," Journal of Electromagnetic Waves and Applications, vol. 5, nos 4/5, pp. 503-523, 1991. [4] M. El-Shandwily and N. Dib, "Spectral Domain Analysis of Finlines with Composite FerriteDielectric Substrate," International Journal of Electronics, vol. 68, No. 4, pp. 571-583, April 1990. [5] K. Kunz and R. Luebbers, The Finite Difference Time Domain Method for Electromagnetics, Florida: CRC press, 1993. [6] G. Mur, "Absorbing boundary conditions for the finite-difference approximation of the timedomain electromagnetic-field equations," IEEE Transactions on Electromagnetic Compatibility, vol. 22, No. 11, pp. 377-382, Nov. 1981. [7] K. Mei and J. Fang, "Superabsorbtion-A method to improve absorbing boundary conditions," IEEE Transactions on Antennas and Propagation, vol. 40, No. 9, pp. 1001-1010, Sept. 1992. [8] V. Betz and R. Mittra, "Comparison and Evaluation of Boundary Conditions for the Absorption of Guided Waves in an FDTD Simulation," IEEE Microwave and Guided Wave Letters, vol. 2, No. 12, pp. 499-501, Dec. 1992. 15

[9] X. Zhang and K. Mci, '"Time-domain finite difference approach to the calculation of the frequency-dependent characteristics of microstrip discontinuities," IEEE Transactions on Microwave Theory and Techniques, vol. 36, No. 12, pp. 1775-1781, Dec. 1988. [10] D. Sheen, S. Ali, M. Abouzahra and J. Kong, "Finite-Difference Time-Domain Method to the Analysis of Planar Microstrip Circuits," IEEE Transactions on Microwave Theory and Techniques, vol. 38, pp. 849-857, July 1990. [11] J. D. Montgomery, "Hybrid MIC North America Markets," Microwave Journal, vol. 32, No. 4, pp. 32-34, April 1989. [12] H.R. Malone, "Antenna/MMIC Packaging Techniques for Commercial Applications," 1992 IEEE Antenna and Propagation Society International Symposium Digest, vol. 3, pp. 1264-1268. [13] W. P. Harokopus, Jr., "High Frequency Characterization of Open Microstrip Discontinuities," Ph.D dissertation, The University of Michigan, Dec. 1991. [14] K. E. Petersen, "Silicon as a Mechanical Material," Proceedings of IEEE, vol. 70, No. 5, pp. 420-457, May 1982. [15] E.W. Strid and K.R. Gleason, "Calibration Methods for Microwave Wafer Testing," 1984 IEEE M7T-S International Microwave Symposium Digest, pp. 93-97. [16] G. Engen and C. Hoer, "Thru-Reflect-Line: An improved Technique for Calibrating the SixPort Automatic Network Analyzer," IEEE Transactions on Microwave Theory and Techniques, vol. 27, No. 12, Dec. 1979, pp. 987-993. [17] M. Maury, S.March, and G. Simpson, "LRL Calibration of Vector Automatic Network Analyzers," Microwave Journal, pp.387-391, May 1987. 16

[18] G. Ghione and C. Naldi, "Coplanar Waveguides for MMIC Applications: Effect of Upper Shielding, Conductor Backing, Finite-Extent Ground Planes and Line-to-Line Coupling," IEEE Transactions on Microwave Theory and Techniques, vol. 35, No. 3, pp. 260-267, March 1987,. [ 19] T. S. R. Taub arnd P. G. Young, "Attenuation and ~eff of Coplanar Waveguide Transmission Lines on Silicon Substrates," Eleventh Annual Benjamin Franklin Symposium on Antenna and Microwave Technology in the 1990's, May 1993, pp. 8-11. [20] N. Dib, L. P. B. Katehi, G. E. Ponchak, R. N. Simons, "Theoretical and Experimental Characterization of Coplanar Waveguide Discontinuities for Filter Applications," IEEE Transactions on Microwave Theory and Techniques, vol. 39, No. 5, pp.873-882, May 1991. [21] W. Heinrich, "Full-Wave Analysis of Conductor Losses on MMIC Transmission Lines," IEEE Transactions on Microwave Theory and Techniques, vol. 38, No. 10, pp. 1468-1472, Oct. 1990. [22] T. E. van Deventer, "Characterization of Two-Dimensional High Frequency Microstrip and Dielectric Interconnects," Ph.D dissertation, The University of Michigan, Dec. 1992. 17

LIST OF FIGURES FIGURE 1. Three-dimensional cross-section of a micropackaged circuit where the shield and line are integrated monolithically. FIGURE 2. Lower Wafer Development. (a) Transmission lines are printed on the top surface.(b) Lower cavity is formed by etching v-grooves. (c) Lower cavity grooves are metallized below the line forming direct contact to the upper ground planes. FIGURE 3. Upper Wafer Development. (a) Probe windows and alignment marks are etched from both sides while the cavity etched from one side only. (b) The upper cavity is metallized. (c) The upper wafer sectional view after processing. FIGURE 4. Photograph of circuit from the top view where the probe windows are shown in relation to the circuit wafer. FIGURE 5. Completely shielded micropackaged circuit after lower and upper wafer alignment. FIGURE 6. Dimensions for the completely shielded micropackaged circuit. FIGURE 7. Various transitions of a completely shielded micropackaged circuit used for onwafer probing. FIGURE 8. Effective dielectric constant for an upper shielded and open CPW through line. FIGURE 9. Attenuation constant for a completely shielded and lower shielded through line. FIGURE 10. Series short-end tuning stub dimensions. FIGURE 11. Comparison between FDTD model and measured results for reflection (S 11) and transmission (S21) coefficients of a completely shielded short-end tuning stub. FIGURE 12. Dimensions of a 5-section stepped impedance lowpass filter having low impedance sections of 20 ohms and high impedance sections of 100 ohms. FIGURE 13. Comparison of reflection coefficient between the PUFF model and measured results for a 5-section stepped impedance lowpass filter. FIGURE 14. Comparison of transmission coefficient between the PUFF model and measured results for a 5-section stepped impedance lowpass filter. FIGURE 15. Loss comparison between the PUFF model and measured response for a 5 -section stepped impedance lowpass filter. FIGURE 16. Comparison between the FDTD model and measured results for the reflection (S11) and transmission (S21) coefficients of a 5-section stepped impedance lowpass filter. FIGURE 17. Series open-end tuning s tub circuit dimensions. FIGURE 18. Comparison of reflection (S 11) and transmission (S21) coefficients between open and completely shielded CPW environments of an open-end series stub

FIGURE 19. Comparison of reflection (S ll1) and transmission (S21) coefficients between upper and lower shielded CPW environments for an open-end series stub. FIGURE 20. Loss comparison of an open-end series stub in open (CPW), upper (US), lower (LS) and completely shielded (CS) environments.

LIST OF FIGURES FIGURE 1. Three-dimensional cross-section of a micropackaged circuit where the shield and line are integrated monolithically. FIGURE 2. Lower Wafer Development. (a) Transmission lines are printed on the top surface.(b) Lower cavity is formed by etching v-grooves. (c) Lower cavity grooves are metallized below the line forming direct contact to the upper ground planes. FIGURE 3. Upper Wafer Development. (a) Probe windows and alignment marks are etched from both sides while the cavity etched from one side only. (b) The upper cavity is metallized. (c) The upper wafer sectional view after processing. FIGURE 4. Photograph of circuit from the top view where the probe windows are shown in relation to the circuit wafer. FIGURE 5. Completely shielded micropackaged circuit after lower and upper wafer alignment. FIGURE 6. Dimensions for the completely shielded micropackaged circuit. FIGURE 7. Various transitions of a completely shielded micropackaged circuit used for onwafer probing. FIGURE 8. Effective dielectric constant for an upper shielded and open CPW through line. FIGURE 9. Attenuation constant for a completely shielded and lower shielded through line. FIGURE 10. Series short-end tuning stub dimensions. FIGURE 11. Comparison between FDTD model and measured results for reflection (S 11) and transmission (S21) coefficients of a completely shielded short-end tuning stub. FIGURE 12. Dimensions of a 5-section stepped impedance lowpass filter having low impedance sections of 20 ohms and high impedance sections of 100 ohms. FIGURE 13. Comparison of reflection coefficient between the PUFF model and measured results for a 5-section stepped impedance lowpass filter. FIGURE 14. Comparison of transmission coefficient between the PUFF model and measured results for a 5-section stepped impedance lowpass filter. FIGURE 15. Loss comparison between the PUFF model and measured response for a 5 -section stepped impedance lowpass filter. FIGURE 16. Comparison between the FDTD model and measured results for the reflection (S 11) and transmission (S21) coefficients of a 5-section stepped impedance lowpass filter. FIGURE 17. Series open-end tuning s tub circuit dimensions. FIGURE 18. Comparison of reflection (S11) and transmission (S21) coefficients between open and completely shielded CPW environments of an open-end series stub

FIGURE 19. Comparison of reflection (Sll) and transmission (S21) coefficients between upper and lower shielded CPW environments for an open-end series stub. FIGURE 20. Loss comparison of an open-end series stub in open (CPW), upper (US), lower (LS) and completely shielded (CS) environments.

FIGURES Upper Cavity (air-filled) Lower Cavity (substrate-filled) FIGURE 1. Three-dimensional cross-section of a micropackaged circuit where the shield and line are integrated monolithically.

dielectric mask...oxide layer ~~ ~~/ (a) Print circuit (b) Etch lower grooves ^ Lower substrate-filled cavity (c) Remove oxide and metallize lower grooves FIGURE 2. Lower Wafer Development. (a) Transmission lines are printed on the top surface.(b) Lower cavity is formed by etching v-grooves. (c) Lower cavity grooves are metallized below the line forming direct contact to the upper ground planes.

bi-L -- --- - — 2 - EDP FI. I I 41 4 ayer mast metal layer / <:.-~ - oxide layer / UTT (1) (2) (a) (1) Alignment marks and probe windows (2) Upper cavity (b) Metallize cavity (c) Final cross-sectional view: (1) alignment mark, (2) probe window, (3) upper cavity FIGURE 3. Upper Wafer Development. (a) Probe windows and alignment marks are etched from both sides while the cavity etched from one side only. (b) The upper cavity is metallized. (c) The upper wafer sectional view after processing.

FIGURE 4. Photograph of circuit from the top view where the probe windows are shown in relation to the circuit wafer.

Probe Window Alignment Window Air-filled Upper Cavity / Lower Substrate-filled Cavity FIGURE 5. Completely shielded micropackaged circuit after lower and upper wafer alignment.

Upper Cavity 500 Lm I 350 gm Lower Cavity TABLE 1. Cross Sectional Dimensions of Shielded Circuits Cavity height (glm) width-max (gLm) width-min (g(m) Upper 280 1200 800 Lower 350 950 500 FIGURE 6. Dimensions for the completely shielded micropackaged circuit.

A BC * I: i D;-'-p-i I1.... -.. T....-..:t....:. -![..i...:....:-:::...........::.-.-::::::::.::::::.:::::::-:::::::::::::;::::,L; r~ rr-rrr I ~ I i ~iL:. ~.~~~~. -~I.~-~.~.I~. ~:~ ~~:~:-.,:~.. ~~I~~~. ~~~~~I~~ ~:~:~:~:~:):~:~:-:I C --- —----------------------------- -- -- -- -- ---- -- -- -- -- -- -- -- -- -- -- -- - ---- - - - - 1 — A -- ----------------- ---------------- A — -- -- -- -- -- -- -- -\ — -- -- -- -- -- -- - - - -- -- -- -- -- -- -- — \- -- -- -- -- -- -- -- -. I -irriiiii~~r -.... -............... k ~~I. fIf,....-_. A BC A BC D ( Upper Cavity Lower Cavity Grooves FIGURE 7. Various transitions of a completely shielded micropackaged circuit used for onwafer probing.

4 z 7- 1 h~ -- Upper Shielded g - X 10 15 20 25 30 35 40 FREQUENCY (GHz) FIGURE 8. Effective dielectric constant for an upper shielded and open CPW through line.

0.3... I... I -....... Lower Shielded -- Completely Shielded X 0.2 z 0 z 0.1 — <..... -o. 0 -,, '.' I ' '.. I 10 20 30 FREQUENCY (GHz) 40 FIGURE 9. Attenuation constant for a completely shielded and lower shielded through line.

1 1020 gml s=176 pm w=133 pgm.................................................................................................................................................................................... ---------- -- ---....................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................... wl= 57pm Ig =48 pIm FIGURE 10. Series short-end tuning stub dimensions.

0 -10 | -20 -30 i I i l l I10 20 30 40 FREQUENCY (GHz) FIGURE 11. Comparison between FDTD model and measured results for reflection (S11) and transmission (S21) coefficients of a completely shielded short-end tuning stub.

TABLE 2. Actual Circuit Dimensions - vk ^ -: -:..:-...I Line Lengths (microns) ~~~~~~~~~........................................................ -........-3.........,,: __ _- _ -..:........................-....= 9 4 210 gm =;^g^^ 2 i '"^'"^-I *'!4=722................. - -....-.g......5=988 FIGURE 12. Dimensions of a 5-section stepped impedance lowpass filter having low impedance sections of 20 ohms and high impedance sections of 100 ohms.

0 -5 co Z I-4 -10 -15 -20 -25 -30 10 15 20 25 30 35 40 FREQUENCY (GHz) FIGURE 13. Comparison of reflection coefficient between the PUFF model and measured results for a 5-section stepped impedance lowpass filter.

0 z orw m z FW -5 -10 -15 -20 -25 -30 10 15 20 25 30 35 40 FREQUENCY (GHz) FIGURE 14. Comparison of transmission coefficient between the PUFF model and measured results for a 5-section stepped impedance lowpass filter.

0 mz --- rurfr Moaei. -15 - -20 10 15 20 25 30 35 40 FREQUENCY (GHz) FIGURE 15. Loss comparison between the PUFF model and measured response for a 5 -section stepped impedance lowpass filter.

0 ~5 '. - - f - -S21, measured n \./ E - Sl.....,, FDTD model ~ 2 --- S2, FDTD model -15 -25 10 15 20 25 30 35 40 FREQUENCY (GHz) FIGURE 16. Comparison between the FDTD model and measured results for reflection (S 11) and transmission (S21) coefficients of a 5-section stepped impedance lowpass filter.

1086gm......................................................................................................................................................................................................................... 1........................................... -.......................... I............................................................................................................................................................................. -............................. -............ I.............................................. -......... I............................................. I............................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................ I.................. s= I 804m.................................................................................................................................. w l = 6 0 4 m............................................................................................................................................... - Il l -, "........... I......................................................................................................................................................................................... w=130gm I - - - 0 - 11 g=48pm FIGURE 17. Series open-end tuning stub circuit dimensions.

0 I 1 - I 1 v I - - - - - i.. o.. _.............. ^. I... "-.......s CPW -. -o —S: CPW ~.S, CS,, CS -.-s Cs T -10 - -15 - -20 - CPW = Open Coplanar Waveguide CS = Completely Shielded.... I... I... I I,i i! I 10 15 20 2 25 30 3 35 40 FREQUENCY (GHz) FIGURE 18. Comparison of reflection (S11) and transmission (S21) coefficients between open and completely shielded CPW environments of an open-end series stub.

0....... uSl, s U -S2US1 o 219S, L No -- ". " S -15 -- - US = Upper Shielded LS = Lower Shielded 10 15 20 25 30 35 40 FREQUENCY (GHz) FIGURE 19. Comparison of reflection (S11) and transmission (S21) coefficients between upper and lower shielded CPW environments for an open-end series stub.

-5 s -10 co F -15 0 -20 10 15 20 25 30 35 40 FREQUENCY (GHz) FIGURE 20. Loss comparison of an open-end series stub in open (CPW), upper (US), lower (LS) and completely shielded (CS) environments.

Micromachined Circuits for lillimeter- and Sub-millimeter-Wave Applications Linda P. B. Katelti, Gabriel M. Rebeiz, Thomas M. Weller, Rhonda F. Drayton, Heng-Ju Cheng and John F. Wliitaker The University of Michigan Ann Arbor, Ml 48409-2122 Abstract This article reviews recent developments in Si micromachining, as it applies to the design of millimeter- and sub-millimeterwave circuits. Advances in semiconductor-processing techniques have made it possible to create monolithic geometries which are not bound to the surface of the wafer. These use a third dimension, to provide shapes and exhibit electrical functions, which are not possible through standard microstrip or coplanar-waveguide technologies. A variety of recently developed circuits are presented herein, which demonstrate the numerous capabilities micromachining offers to the circuit and antenna designer. 1. Introduction Microwave and millimeter-wave planar-integrated circuits and antennas are the central nervous system in communication and radar systems. In the past decade, the microwave field has been experiencing a technological revolution due to advances in solidstate devices, insulating and semiconducting materials, and circuitanalysis techniques. Furthermore, the planarization of transmission lines to microstrip, stripline, or coplanar-waveguide form provides great flexibility in design, reduced weight and volume, and compatibility with active devices and radiating elements. Ther are drawback to these planr gometris-ower. These include frequency-dependent mechanisms, such as parastic coupling and radiation, as well as inceased ohmic loss, dielectric loss, and dispersion. These effects can seriously deteriorate electrical performance, and can lead to costly and time-intensive design cycles [1-2]. Suppression of so these ectromagnetic mechanisms, such as parasitic radiaionind coupling, has led to improved performance, but requires vey sophisticated pa es, which add considerably to the volume, weight, and cost of the circuit [3-5]. Elimination of other deleterious effects, such as dispersion and high ohmic loss, which are inherent in conventional planar lines, will require fundamentally new approaches to planarization and circuit integration This article addresses the development of relevant transmission-line technology, aimed at providing superior performance in the afeetioned respects. Recent advances in semiconductor-processing techniques offer an historic opportunity to disinguish the monolithic from the planar character, and to provide integration along all the directions of three-dimensional space [9-13]. The capability to add one more dimension, and a few more parameters, in the circuit design, can lead to revolutionary geometries. These reduce ohmic loss and eliminate electromagnetic-parasitic effects, without affecting the monolithic character. Operating frequencies are thereby extended, and circuit and array performance are improved. A variety of monolithic, quasi-planar transmission-line geometries, which will support transverse-electromagnetic (TEM) wave propagation, can be developed using semiconductor micromachining. These lines are characterized by: * Superior electrical performance * Low weight and small volume * Easy fabrication * Great potential for low cost The characteristics shown above may be prioritized according to pre-eisting needs, resulting in a number of approaches which use micromachining for millimeter- and sub-millimeter-wave circuit design. The evolution of micromachined circuits and antennas for these frequency ranges is still in its infancy. However, presented here is a description of recent accomplishments in this area, with emphasis on the effort performed at the University of Michigan There are two techniques which have shown promise, and which have extensively used micromachining to realize novel circuits. The first utilizes a membrane-supported transmission-line configuration [9- 11], and has superior preformance as the highest priority. The second technique introduces new concepts in packaging for minituried circuits, using integratedshiding cavities, and empsizes size/volume/cost reduction [12-14]. The merits of each approah, in relation to electrical performnce, fabrication, and compatibility, will be presentd, and the impact of the newborn technologies on the state of the art will be discussed. 2. Membrane-supported circuit and antenna technology Mmbrane-supported radiatin elements, such as the integrated-horn antenna were developed for the first time, in 1987, by G. M Rebeiz and D. B. Rutledge. They have been researched, since t988, at the University of Michigan, by G. M Rebeiz and L. P. B. Katehi, under the support of the NASA Center for Space Terahertz Technology. During these past five years, antennas have been successfully used in a wide variety of applications, which cover the millimeter- and sub-millimeter-wave spectum [6-8]. The fabrication of these radiating elements ld to the establishment of a membrane technology which is reliable, repeatable, and easy to employ. In view of these capabilities, a membrane-supported tanission line, called microshield, was proposed at the 1991 MTT-S International Microwave Symposium [9]. The microshield, which is only one of the possible membrane-supported geometries, may be considered as an evolution of the conventional microstrip or coplanar waveguide. It is characterized by zero dielectric loss, very;low radiation loss, reduced electromagnetic interference, and compatibility with conventional technologies, such as microstrip and coplanar waveguide. This transmission medium creates the basis of a new monolithic technology, which can provide generic IEEE Antennas and Propagation Magazine, Vol. 35, No. 5, October 1993 1045-9243/93/$03. 0001993 EEEE 9

designs appropriate for circuit and antenna applications in the millimeter- and sub-millimeter-wave region. Membrane supported transmission lines are quasi-planar configurations, in which a pure, non-dispersive TEM wave is propagated through a two-conductor system, embedded in a homogeneous environment. Homogeneity of the environment can be accomplished by using a 1 5-mm-thick dielectric membrane, or a few-micron-thick diaphragm, to support the signal lines, while ground is provided by a metallized-micromachined cavity. This cavity is fabricated in Si wafers, using etchants such as KOH (potassium hydroxide) or EDP (ethylene diamine pyrocatechol), and is then metallized, using evaporation or plating. The signal lines are created by metal deposition on the membranes or diaphragms, or can be grafted onto the supporting structure by lift-off techniques. Similar processing methods are available for GaAs wafers. Due to the cavity shielding, and the pure TEM character of the propagating mode, these lines possess a large single-mode frequency band (DC to >1 THz), and have very low losses and zero signal dispersion. Consequently, circuit components made of these line geometries can provide electrical performance which is superior to those of conventional planar circuits. The success of membrane-supported circuits relies on the development of thin-film dielectric membranes or diaphragms, which possess good electrical and mechanical properties. These thin-film layers are grown on Si or GaAs wafers, and are used to support the planar-conducting-strip lines. In view of the previously mentioned performance objectives, the thin films must have low losses at microwave and millimeter-wave frequencies, as well as compatibility with semiconducting and conducting materials. The mechanical considerations include reduced sensitivity to applied pressure and temperature variations, along with increased membrane or diaphragm sizes. In the example circuits presented in this article, the membranes are in a tri-layer SiO2/Si3N4/SiO2 configuration, with constituent thicknesses of 7000, 3000, and 4000 A, respectively. This size makes the membrane transparent to propagating signals at frequencies as high as 3 THz, and provides a near-homogeneous, air-filled environment to the propagating electromagnetic wave. This leads to very fast wave velocities, zero dispersion, and zero dielectric loss. The base layer (7000 A) is grown using thermal oxidation, while the final two layers are formed using a chemicalvapor-deposition process. The effective dielectric constant of the 1.4-mm-thick membrane is approximately 4.9, using a simple averaging of the oxide (E, = 3.9) and nitride (e, = 75) layers. The material has very low loss from DC (p> 1014 f-cm) to well into the THz frequency range. To date, membrane dimensions as large as 2.0 mm x 1.9 cm have been fabricated for use in microshield lines, with the average size being around 1.8 mm x 1.0 cm. The total fabrication yield on these membranes is greater than 95 percent. As shown in Figure 1, various configurations of membrane lines may require the use of two or three wafers, to provide a cavity shield on one or both sides. This cavity shield is necessary for some uniplanar geometries, where multiple ground planes or other conductors are in close proximity to the signal line. Figure la shows a coplanar-stripline geometry, made of two high-resistivity <100> Si wafers, in which the electric field is confined to the surface area between the two strips. In this case, shielding by the cavity is not required and, depending on the application, the line may operate in a variety of environments [15, 16]. The geometry shown in Figure lb, the microshield line, is a similar two-wafer Micromachined Micromachined Coaxal Line Shielded I Microshie ld Line Line -\"\\\\\\\~N \s\ s\\'s~.,W~ \ 350 gm i-3 350m Two high-resistivity <100> Si wafers b)/1. p c_ 1.5 pn thick metallization c) I. —~i '1 I <100> Si wafers Thin-film membrane 800 am d) e) -- - i2 200 >m Figure 1. Illustrated here are five possible configurations of the micromachined quasi-planar transmission lines. The dimensions given are characteristic of lines designed to operate up to 40 GHz. The metallization, shown in black, is typically 1.5 microns thick. The line features will vary from 10 to 1000 microns. MEMBRANE. MEMBRANE 2) ETCHWAFER 1) PATTERN MEMBRANE = 3) METALLIZE FRONT 4) METALLEZE BACK 5) ENCLOSE CAVITY CENTER CONDUCTOR UPPER GROUND PLANES MEMBRANE Fixure 2. Fabrication process for a microshield transmission line. 10 IEEE Antennas and Propagation Magazine, Vol. 35, No. 5, October 1993

arrangement, where the signal line, printed on a thin-film membrane or diaphragm, is shielded on one side by a metallized cavity. Here, the field is confined between the signal line and the metallized cavity on the lower side, while it is free to radiate in free space on the top side. Geometries of this form are very appropriate for monolithic antenna and receiver applications, where the feeding lines and other circuit components are printed on the same wafer. Figure Ic shows a three-wafer configuration made of <110> highresistivity silicon wafers. In this structure, the signal line is shielded by a metallized cavity on both sides. Such a line suffers no radiation loss, and also retains the dispersionless characteristics mentioned above. A line of this type is a monolithic version of the coaxial line, and is expected to have very low propagation losses. 2. 1 Fabrication The fabrication of the microshield line, shown in Figure lb, begins with a Si wafer which has membrane layers on both sides The membrane is a 1.4-mm-thick composite SiO2Si3NA/SiO2 material which, as stated before, lis grown using a combination.of thermal oxidation and chemical-vapordeposition processes. The transmission-line structure is formed by first removing photolithographically-defined regions fom the membrane layer on the back side of the wafer. This pattern serves as an etch mask, allowing pyramidal cavities to be opened in the wafer using anisotropic Si etchants, such as ethylene diamine pyrocatechol (EDP) or potassium hydroxide (KOH). When the wafer has been etched through to the front side, the upper-membrane layerwi behave as an etch stop, thrb forming the exposed mem surface with surrounding silicon walls below (see Figurei2). The final step in the fabrication process is the metallization of the line pattern on the upper-membrane urface, and of the cavity sidewalls underneath. A photoresist lift-off technique, with metal evaporation, is used to deposit te circuit pattern on the top side of the membrane layer. Evaporation is also used for metal deposition from the back side onto the cavity sidewalls, but her some shadowing is required, to protect selected areas of the nmbrane from metallization. A diced or micromachined silicon mask is used to shield the region between the inner edges of the upper ground planes and the center conductor, thus preventing metallization of the membrane below the slots. The finished circuit is then minted on a ground plane, to fully enclose the lower shielding cavity. Typical cavity dimensions for circuits designed for the 10-40 GHz range are 355 pm high by 1000-2000 pm wide. These dimensions are reduced as the operating flequency increases The slot- and signal-line widths will vary, depending on the desired characteristic impedance level of the transmission line, but generally range from 10 rm to I mm. 2.2 Measurements The procedure described above has been used to design and fabricate very-low-loss, ultra-fast lines, and high-performance circuits. Recent measurements have shown astonishing performance. wax AJGaAS IIa~ ISI-GaAS~ LT-GaAS I SI-aAS I-f A AI wax LT-JZ rjl embrane I + - wax Bake LT-GaAs im Ai/Au Figure 3. Major steps in the preparation of coupled strip lines for electro-optic sampling experiments. A low-temperature GaAs (LT-GaAs) layer is removed from its substrate support, and attached to the membrane using van der Walls forces. The metal lines are then printed over the LT-GaAs slices. 1.2 1 E E 0. z tc.o CD Q) 0.8 0.6 0.4 0.2 2.2.1 Loss measurements C - membrane substrate 50 150 250 350 450 I 50 150 250 350 450 Loss measurements on membrane-coupled strips (MCS) and membrane-coplanar waveguide (MCW) were performed, using electro-optic sampling from 50 GHz to 500 GHz [15-16]. In order to generate and measure electrical pulses having picosecond duration, the optoelectronic technique of photoconductive switching, along with a sub-picosecond-temporal-resolution electro-optic sanpling charazterization system have been employed Frequency (GHz) Figure 4. A photograph of multiple membrane-coupled strip (MCS) lines, which cross two LT-GaAs slices and are supported by the 1.5-micron-thick membrane (dark area). Also shown are loss measurements using electro-optic sampling on lines with GaAs and membrane substrates. IEEE Antennas and Propagation Magazine, V61. 35, No. 5, October 1993 11

'Measurement error: 0.07ps [11, 15]. For the measurement of transients in the time domain, reflections from any source near the probing point will cause a loss of accuracy in the acquisition of clean, easily analyzed signals. Therefore, it was necessary to integrate a 200-rm-wide photoconductive-generator patch with the signal lines, which are fabricated on the membrane. The sub-picosecond-lifetime photoconductor employed was a 1-rim-thick layer of GaAs, grown by molecular-beam epitaxy at the low substrate temperature of -200~C (known as LT-GaAs). This photoconductor was lifted off its substrate, through etching techniques, to form a free-standing film supported on a wax layer, and was then grafted to the membrane substrate using van der Walls bonding techniques [20]. The MCS was then defined on the membrane using standard photolithographic techniques, as summarized in Figure 3.., 5 I.L UoM- MCS IAT AT&T-CAT o. 0. a. 1.0 0.8 0 6 *:.,. *ir.r *C X r r r~ rrerrrr * *........~........, *, C<XIIIIIIIII.I..../.. /... * * * ~...... M E...... r... ~~ /~~ I l — - I..............,...... 0.4 ~/ * * * 0.4: 1.0 2.0 3.0 imn -4.0 5.0 Figure 4 shows the total loss, in Np/mm, from 50 GHz to 500 GHz, for a coupled stripline (w = s = 20 pm) printed on a 100 -micron-thick GaAs substrate (), and supported by a dielectric membane ( —). The picture of multiple MCS lines, included in the figure, clearly shows the well-defined boundaries of th dielectic membrne, and the two verticallyorientd LT-aAs patches, which are under the nes. These experimentl results confim the lowloss behavior of MCS, and the superiority of this tnission line over conventional substrate-printed lines. Figure 5. Pulse rise-time vs. distance for the MCS line and the CAT line. BACK-TO-BACK MICROSHIELD TO CPW TRANSITON IHMICROSHIELDELD 1.5 mm CPW 0. V 0 z 7: -10. -20. REFLECIION COEFFICIENT [Sll].... I I.. I I...A <0 Ity h -M LOSS [l- ISll -S2t12] Im [0 z, -30. 10. 20. 30. 40. 10. 20. 30. 40. FREQUENCY [GHz] FREQUENCY [GHz] C Figure 6. A diagram of a back-to-back microshield-to-coplanar-waveguide (CPW) transition. The photograph shows the back side of the circuit, with the ground plane removed. The middle CPW section is 1.5 long. Also shown are the measured S11 values, and the total loss performance. 12 IEEE Antennas and Propagation Magazine, Vol. 35. No. 5, October 1993

In Figure 5. the measured pulse risetime vs. distance, for the same membrane lines, shows negligible dispersion, compared to recently published results for the coplanar air transmission line (CAT) The CAT, which was developed at AT&T, was considered to be the fastest transmission-line geometry available to date The measured results shown in Figures 4 and 5, however, indicate that the membrane-supported line has the lowest loss and fastest phase velocity of any monolithic transmission line ever built 2.2.2 CPW-to-microshield transition The primary motivation for designing a CPW-to-microshield transition lies in the interest of using membrane geometries as part of a complex monolithic circuit, which combines a variety of uniplanar technologies. Figure 6 shows the measured performance of a back-to-back transition, from 10 GHz to 40 GHz. This transition has been designed in such a way as to preserve the form of the field, and to equalize the characteristic impedance in the different lines. As a result, from 10 to 40 GHz, the reflection (Sl1) is below -22 dB and the total loss, (I-S|li2 -S212f), is less than -10 dB [10, 17]. Note that these data include the effects of two 5-SECTION b- z c2 0. -10. -20. -30. -40. 1 -20. 25. 30. 35. 40. 7-SECTION 0. TOP VIEW OF -IVrSECTION FILTER 1356 680 iI! 1320 265 Q F 1 40 m -= m ul a - c: 0 < -10. -20. -30. -75 Q -40. L 20..2. W.. 25. 30. 35. 40. (DIMENSIONS IN MICRONS - NOT TO SCALE) FREQUENCY [GHz] Figure 8. The measured and theoretical response of five- and seven-section microshield low-pass filters. transitions and a connecting section of line. This performance clearly demonstrates the ability to achieve low-loss CPW/microshield integration in millimeter-wave applications. 2.2.3 Microshield stepped-impedance low-pass filters Stepped-impedance filters are commonly used for applications which do not require a sharp cutoff, such as in the rejection of outof-band mixer harmonics. Illustrated in Figure 7 is a schematic of a five-section 0.5 dB-ripple Chebyshev filter, which has been fabricated and tested. Also shown is a photograph of a microshield thru-line (top), and both a five- and a seven-section low-pass filter. The measured performance of these filters is shown in Figure 8 [10, 17], along with the theoretical response which was obtained using Figure 7. A schematic diagram of a microshield five-section stepped-impedance filter. The photograph shows a thru-line (top) iand a 5- and a 7-section filter. IEEE Antennas and Propagation Magazine, Vol. 35, No. 5, October 1993 13

ideal-transmission-line theory'. The agreement between the experimental and theoretical data is quite good, and the performance illustrates the potential for low-loss microshield circuit components. 3. Nlicromachined quasi-planar circuits Micromachining may also be used effectively to develop miniaturized circuits, in which the package has been fabricated and integrated with the rest of the circuitry. Circuit design requires the use of two silicon wafers, one of which has upper-half shielding cavities that have been etched anisotropically, while the other wafer has the planar circuits that have been printed using photolithographic techniques, with or without the lower cavities (see Figures Id, le) [12, 14]. 3.1 Fabrication Micromachined circuits, consisting of conducting lines and metallized cavities, are developed in two stages, using a two-wafer system. In this section, a comprehensive discussion of the fabrication of upper-half-shielded (Figure Id) as well as completelyshielded (Figure Ie) micromachined circuits is presented. In order to provide a shielded environment in the upper-half region of the planar lines, cavities and alignment marks are created on a low-resistivity Si wafer (385 gtm) having a 1.4-um-thick, trilayer dielectric mask, consisting of Si02/Si3N4/SiO2. The alignment marks are etched completely through the wafer. However, the cavity heights are less than the total wafer thickness, thus requiring a separate etch depth. To prepare the wafer for the anisotropic etching in these areas, the etch pattern is first defined in the tri-layer mask, using standard photolithographic techniques. Removal of the mask requires the use of buffered HF, and a plasma etcher for the upper oxide and nitride layers, respectively. The wafer is then repatriated to expose and remove the lower oxide only in the alignment-mark pattern, while preserving the oxide mask in the cavity areas. Using the tri-layer as a mask for the anisotropic etchants, KOH or EDP, an initial depth is obtained for the alignment marks. Next, the remaining oxide in the cavity areas is removed, using buffered HF, and the wafer is then etched to produce both the alignment windows, which extend vertically through the wafer, and the desired cavity depth. Finally, the cavity regions are metallized using evaporation techniques. For the upper-half-shielded circuits, planar lines are defined on a high-resistivity Si wafer using standard photolithographic and UPPER CAVITY FABRICATION W ETCH ALIGNMENT METALLIZE BACKSIDE WINDOW (1) & CAVITY (2) UPPER-HALF - - SHIELDED CIRCUITS PRINT LINES AND ALIGN AND SECURE ALIGNMENT MARKS WAFERS COMPLETELY -- SHIELDED U i CIRCUITS ETCH LOWER GROOVES PRINT LINESAND ANGN AND SECURE METALLIZE BACKSIDE WAFERS Figure 9. The development of micromachined shielded circuits. IThe theoretical data include quasi-static corrections for the linelength extension at impedance steps [18]. Figure 10. Photographs of micromachined shielded circuits. (a) A top view of five upper-half-shielded circuits, with alignment marks on the far left and right sides. (b) A close-up view of three circuits with the shielding in place. metal-evaporation techniques. The upper cavity is then aligned to the planar circuits via microscope, and the two wafers are attached using regular adhesion methods or Si-to-Si electrobonding. A similar procedure is followed for completely-shielded circuits which incorporate cavities beneath the planar lines. Prior to the planar-line metallization, however, cavity sidewalls are established by anisotropically etching grooves on the backside of the highresistivity wafer. This wafer is subsequently metallized from the back, to complete the lower-side shielding. The fabrication process for both upper-half and completely-shielded circuits is illustrated in Figure 9. 3.2 Measurements As with the membrane-supported circuits, the micromachined shielded circuits are measured using an HP 8510B Network Analyzer system, which operates at frequencies up to 40 GHz. Measurements are obtained from Cascade Microtech highfrequency GSG probes, via an Alessi probe station. In order to suppress parallel-plate and microstrip modes between the circuits and the wafer chuck of the probe station, the circuit wafer is placed on 1/8" thick 5880 RT/Duroid, with E, = 2.2. The thru-reflect-line (TRL) calibration is performed, to eliminate the effects of connectors, cables, and probes from the measured data, and to accurately establish a known reference plane. After calibrating, scattering-parameter measurements of the circuits are obtained, to compare theoretical and experimental data. The circuits described here are shielded by miniature micromachined cavities in the upper-half region, which are 800 rtm wide and 200 Kim high, as shown in Figure Id. Each circuit is fed by open coplanar waveguide, which serves as a transition from the IEEE Antennas and Propagation Magazine, Vol. 35, No. 5, October 1993

GSG probes to the actual micromachined circuit (see Figure 10). Initial characterization of these lines was obtained by measuring the scattering parameters of a delay line. Figure 11 shows that the total loss of the line is about 008 dB per guided wavelength,,A. Additional circuits have also been measured as part of this effort, including an open-end series stub, designed to resonate at 30 GHz, and an open-end gap (see Figures 12a, 12b). Figure 13 shows a 0 -10 ': -20 O -30 ~-40 -50 -60 comparison between experimental data for these circuits, and theoretical results which were derived using a full-wave spacedomain integral-equation method [19]. The discrepancy between theory and experiment is due to the fact that the theoretical data assume completely shielded and lossless circuits, while the experiments have been performed on upper-half-shielded circuits, with ohmic losses and radiation losses present. The measured total loss for the open-end series stub and open-end gap, implemented in both CPW and the upper-halfshielded micromachined line, is given in Figure 14. The data show that the total loss is 3 dB lower for the micromachined circuits, from 10 GHz to 40 GHz. This reduction is due to lower parasitic radiation, and it emphasizes the effectiveness of the integrated cavity. Even further reduction of the loss is expected for the completely shielded micromachined circuits, which are presently in development at the Radiation Laboratory of the University of Michigan.. 4. Conclusions Two approaches for the development of micromachined circuits for millimeter- and sub-millimeter-wave applications have been discussed. These new technologies provide a means to overcome the drawbacks typically associated with conventional planar-transmission lines, such as microstrip and coplanar waveguide. The examples shown demonstrate the use of micromachining to effectively incorporate a third dimension into the geometry of monolithic circuits, thus introducing additional parameters in the design, and resulting in structures with superior electrical performance (membrane-supported circuits), and reduced volume and weight (micromachined-shielded circuits). The membrane-coupled strip (MCS) line, for example, has been experimentally verified as the fastest monolithic transmission line to date, and has demonstrated extremely low propagation losses up to 500 GHz. High-performance circuit components, such as low-pass filters, have also been fabricated using the microshield configuration, and low-loss integration of this line with conventional coplanar waveguide has been achieved. Finally, micromachined circuits, which incorporate miniaturized shielding packages in a monolithic design, have been described. Preformance of these circuits shows that the low-volume, low-weight packaging is effective in reducing signal loss, without degrading the desired frequency response. The capabilities of micromachining certainly extend beyond the limited number of applications presented in this article. The recent advances in Si and GaAs processes allow us to extend circuit b) g=30ugm 0 5 10 15 20 Frequency, I... 25 30 35 40 GHz Ax = 2,176mm w=130gm s=180gm Figure 11. The delay-line characteristics for upper-halfshielded micromachined circuits. a) g = 30Lm w=130gm s=180gm - wl= 60gm w=130pm s=180um - - 1086,gm Figure 12. Circuit descriptions for (a) an open-end series stub and (b) an open-end gap. IEEE Antennas and Propagation Magazine, Vol. 35, No. 5, October 1993 15

Total Loss, dB ryrZ3 SI P. CA) z 0 0n 0 0 ft 0 (A. (A C *1 CL i -4r ii I.0 (0 I............................................. I................................. -.9 I.......:::e.............. 4.......... - -— I ( - ----- - - -. -- - -,% v (2 LA I~ I I 4 I E0* 0 LA Di -C 0 % t. 00 11 4 S-parameter, degrees S-parameter, dB ft0 "I 0 Total Loss, dB I I I I 9, pis S-parametqrs, degree$ I I I B IB '." S-parameter, dBi tA r / I,. I 4: '. - 4.A. I. P" Is I. a a t9 4..9. 6 ' 1 W g K 0-a '-V tI wA LA

geometries beyond single planes in novel ways, and to investigate techniques to improve current millimeter- and sub-millimeter-wave systems by reducing size and cost, and by providing superior performance. Similar opportunities exist to experiment with configurations that can provide new functions, and introduce new concepts in circuit design. 5. Acknowledgments This work has been supported by the Office of Naval Research, the NASA Center for Space Terahertz Technology, the Jet Propulsion Laboratory, and the National Science Foundation. 6. References 1. L. P. Dunleavy and L. P. B. Katehi, "Shielding Effects in Microstrip Discontinuities," IEEE Transactions on Microwave Theory and Techniques, MTT-36, No. 12, pp. 1767-1774, December 1988. 2. T. E. van Denventer, L. P. B. Katehi, and A. Cangellaris, "An Integral Equation Method for the Evaluation of Conductor and Dielectric Losses in High Frequency Interconnects," IEEE Transactions on Microwave Theory and Techniques, MTT-37, No. 11, pp.. 1964-1972, December 1989. 3. W. P. Harokopus and L. P. B. Katehi, "Characterization of Microstrip Discontinuities on Multi-layer Dielectric Substrates Including Radiation Losses," IEEE Transactions on Microwave Theory and Techniques, MTT-37, No. 11, pp. 2058-2066, December 1989. 4. R. S. Pengelly and P. Schumacher, "High-Performance 20 GHz Package for GaAs MMICs," Microwave Systems News & Circuit Techniques, pp. 10-19, January 1988. 5. N. L. VandenBerg and L. P. B. Katehi, "Broadband Vertical Interconnects Using Slot-Coupled Shielded Microstrip Lines," IEEE Transactions on Microwave Theory and Techniques, MTT40, No. 1, pp. 81-88, January 1992. 6. D. B. Rutledge, D. P. Neikirk, and D. P. Kasilingam, "Integrated Circuit Antennas," in K J. Button (ed.), Infrared and MillimeterWaves, Vol. 10, New York, Academic Press, pp. 1-90. 7. G. M. Rebeiz, D. P. Kasilingam, P. A. Stimson, Y. Guo, and D. B. Rutledge, "Monolithic Millimeter-Wave Two-Dimensional Horn Imaging Arrays," IEEE Transactions on Antennas and Propagation, AP-28, pp. 1473-1482, September 1990. 8. G. M. Rebeiz and D. B. Rutledge, "Millimeter-Wave and Submillimeter-Wave Antenna Structure," US Patent 4888, 597. 9. N. I. Dib, W. P. Harokopus, L. P. B. Katehi C. C. Ling. and G. M. Rebeiz, "Study of a Novel Planar Transmission Line," presented at the 1991 IEEE MTT-S International Symposium, Boston, MA, June 1991. 10 T. M Weller, G M. Rebeiz, and L. P. B. Katehi, "Experimental Results on Microshield Transmission Line Circuits," presented at the 1993 International Symposium of the IEEE Microwave Theory and Techniques Society, Atlanta, Georgia, June 1993. 11. T. M. Weller, L. P. B. Katehi, G. M. Rebeiz, H. J. Cheng, and J. F. Whitaker, "Fabrication and Characterization of Microshield Circuits," presented at the 1993 International Symposium on Space Terahertz Technology, Los Angeles, California, March 1993. 12. R. F. Drayton and L. P. B. Katehi, "Experimental Study of Micromachined Circuits," presented at the 1993 International Symposium on Space Terahertz Technology, Los Angeles, California, March 1993. 13. R. F.- Drayton -and L. P.. B. Katehi,,"Microwave Characterization of Microshield Lines," 40th ARFTG Conference, Orlando, Florida, December 1992.:14. R- F. Drayton and L. P. B. Katehi "Micromachined Circuits for Mm-Wave Applications," accepted for presentation at the 1993 European Microwave Conference, Madrid, Spain, September 1993. 15. H. J. Cheng, J. F. Whitaker, T. M. Weller, and L. P. B. Katehi, "Transmission of Ultra-High-Bandwidth Pulses on a LowDistortion Coplanar Stripline," presented at the 1993 TEEE LEOS Topical Meeting on Optical-Microwave Interactions. 16. H. J. Cheng, J. F. Whitaker, T. M. Weller, and L. P. Katehi, "Transmission of Ultra-High-Bandwidth Pulses on a LowDistortion Coplanar Stripline," submitted to IEEE Guided Wave Letters. 17. T. M. Weller, L. P. B. Katehi, and G. M. Rebeiz, "High-Performance Microshield-Line Components," submitted to the IEEE Transactions on Microwave Theory and Techniques. 18. E. O. Hammerstad and F. Bekkadal, A Microstrip Handbook, ELAB Report, STF 44 A74169, N7034, University of Trondheim, Norway, 1975. 19. N. I. Dib, L. P. B. Katehi, G. E Ponchak, and R. N. Simons, "Theoretical and Experimental Characterization of Coplanar Waveguide Discontinuities for Filter Applications," IEEE Transactions on Microwave Theory and Techniques, MTT-39, No. 5, May 1991. 20. E. Yablonovitch et. al., Applied Physics Letters, 56, pp. 2419 -2421, 1990. ------------------------ ------— ~YI#U I II)~)()(Q ~ mU IEEE Antennas and Propagation Magazine, Vol. 35, No. 5, October 1993

Submitted to the IEEE Transactions on MTT, February 11, 1994. High Performance Microshield Line Components Thomas M. Weller, Student Member, IEEE, Linda P. Katehi, Senior Member, IEEE, and Gabriel M. Rebeiz, Senior Member, IEEE NASA/Center for Space Terahertz Technology University of Michigan Ann Arbor, MI 48109-2122 Abstract Several millimeter-wave passive components have been fabricated using the microshield transmission line geometry, and their performance is presented herein. Microshield is a quasi-planar, half-shielded design which uses a thin dielectric membrane (1.5 tm) to support the conducting lines. This approach provides a nearly homogeneous, air-filled environment and thus allows extremely broad-band TEM operation. The paper examines the conductor loss and effective dielectric constant of microshield lines and presents results on transitions to conventional coplanar waveguide, right-angle bends, different stub configurations, and lowpass and bandpass filters. Experimental data is provided along with numerical results derived from an integral equation method. The microshield line is shown to be very suitable for high performance millimeter- and submillimeter-wave applications. I. INTRODUCTION The microshield line was introduced in 1991 [1] as an alternative transmission medium to coplanar waveguide for millimeter- and submillimeter-wave applications. It is a partially shielded, quasi-planar transmission line in which the center conductor and upper ground planes are surrounded by air and supported by a 1.5 jm-thick dielectric membrane, as shown in Figures 1 and 2. This configuration allows single-mode, TEM wave propagation over a very broad bandwidth with minimal dispersion and zero dielectric loss. Furthermore, 1

Membrane Lower.... I wg^l' ' '^t Lower Shiclding Cavity j '' '| Figure 1: The microshield transmission line geometry (not to scale). The dark lines indicate metallization, which is typically 1-2 microns thick. the metallized lower shielding cavity minimizes signal cross-talk between adjacent lines and eliminates radiation into parasitic substrate modes. A description of these and other advantages of the microshield geometry can be found in previous communications [2, 3, 4], along with a summary of the fabrication process. The objective of this paper is to present results from a detailed study which emphasizes microshield circuit design and measurement in the 10-40 GHz frequency range. An important part of the project has been the development of a good transition between the microshield line and the measurement probe contact pads, which must be printed on the surrounding high-resistivity silicon substrate for mechanical support. This transition is covered in detail in section II. The conductor loss of microshield, discussed in section III, has been measured on lines of different aspect ratios and values as low as 0.03 dB/mm at 35 GHz have been obtained. These results demonstrate the potential for low-loss, millimeter-wave interconnect lines and passive circuit elements. The data are also consistent with the findings from recent opio-electronic sampling experiments in which low-loss signal propagation on membrane lines was measured up to 1 THz, well beyond the upper frequency limits of similar GaAs and quartz supported lines [5]. In Section IV, some of the advantages offered by the very low effective dielectric constant of the microshield line are covered. The performance of several microshield circuits such as transitions to cpw lines, right-angle bends, series stubs, lowpass filters, and bandpass filters is presented in Section V. 9

Figure 2: Photograph of the backside of a wafer, shown with the ground plane removed. The circuit in the center is a 5-section stepped-impedance lowpass filter (refer to Figure 15).:3

A theoretical analysis is provided as a validation of the measurements and also as a means to predict the performance beyond the 40 GHz limit of the measurement setup. The theory involves a full-wave analysis which is based on a space domain integral equation (SDIE), similar to that described in [6] but capable of analyzing non-shielded geometries. The SDIE is derived by replacing the slots in the metallization by fictitious metal surfaces, and then imposing magnetic currents on the surfaces which must satisfy the field continuity conditions. The method of moments is used to discretize the integral equation and solve for the magnetic currents, or electric fields, in the slots, from which the scattering parameters can be determined. It is assumed that all conductors are infinitely thin and perfectly conducting, and that the shape of the lower shielding cavity is rectangular. In addition, the effect of the thin membrane layer is approximated by assuming that the lower cavity is homogeneously filled with a material which has a dielectric constant slightly greater than 1 (see section IV). In this paper it is shown that the characteristics of the microshield line are well-suited for specialized millimeter-wave and submillimeter-wave applications. Obviously, integrating the thin dielectric membrane on silicon or GaAs substrates requires additional steps in the fabrication process, but in many cases this is offset by eliminating the procedures for airbridges and substrate thinning. Furthermore, the circuit yield is very high, exceeding 95% in a research laboratory, and the substrate etch can be one of the last steps in the process, thereby increasing yield. These factors, combined with the very broad-band, low-loss performance, make microshield a viable transmission line option for systems operating at mm-wave frequencies. II. MEASUREMENT AND SYSTEM CALIBRATION In this section the measurement setup is described along with some details regarding the line standards needed for calibration. The measurements are made using a microwave probe station and an HP 8510B vector network analyzer. Since the membranes cannot withstand repeated probe contact, the contact pads are made with grounded coplanar waveguide (gcpw) on the silicon substrate, and a transition is made onto the microshield line. The conductor 4

backing of the cpw pads is a continuation of the ground plane which completes the lower shielding of the microshield. A one-tier TRL technique [7] is utilized to calibrate to reference planes which are on the microshield line. It was found that the geometry of the probe contact pads and the transition section have an important effect on the quality of the calibration. The initial design, shown in Figure 3, consisted of 55Q gcpw contact pads followed by a 735 /im long gcpw impedance matching section. This configuration was intended to act as a low-loss transition to the 75Q microshield line, the effects of which could be removed with the TRL calibration. Even though the non-calibrated return loss from the thru and delay lines was consistently better than 12 dB, the performance of this design was rather poor and the repeatability suffered due to unwanted interaction between the calibration lines and nearby circuits. Typical results for the calibrated insertion loss of a microshield delay line are also shown in Figure 3. The cause of the measurement errors is the leakage of power into undesired substrate modes along the length of grounded cpw line, which cannot be accounted for correctly using the standard HP 8510B TRL calibration software. The discontinuities at the probe contact pads, impedance steps, and the end of the microshield cavity are all possible sources of excitation for undesired parallel plate modes and TMn and TEn surface waves [8]-[16]. Also, as reported in [12], the finite lateral extent of the ground planes can result in distinctly predictable resonances in gcpw delay lines. The leakage mechanism associated with the transition in Figure 3 exhibits a consistent structure across the 20-40 GHz band, apart from the sharp resonance seen around 35 GHz. The circuit dimensions suggest that the primary problem is leakage into the TEM parallel plate mode [10]. Different authors have shown that the ground plane spacing (S+2W) must be kept small compared to the ground plane width, G, and the substrate height, H, in order to maintain propagation in the cpw-like mode [13, 15, 16]. Furthermore, since the power leakage increases with distance, a shorter length of gcpw between the probe contact pads and the microshield line will also reduce this problem. One way to eliminate the leaky modes is 5 K a

I CCp V) -1. 40. FREQUENCY [GHzl 330t H=350 Figure 3: A transition from the cpw microwave probes to the microshield line, and the measured S21 of a microshield delay line using this design. All dimensions are in microns. 6(

to integrate metallized cavities in the back of the substrate near the probe contact points. These modifications were implemented into the transition design shown in Figure 4, and the significant improvement in the calibration is demonstrated in the same figure. The ripple near 13 GHz appears to be a resonance which occurs when the total length of the 'thru' calibration line becomes a half-wavelength. Also, in a later design, the ripple near 36 GHz was reduced by completely eliminating the 300 /m length of gcpw line. III. ATTENUATION The attenuation characteristics of the nicroshield line have been investigated by comparing measured data on different microshield geometries against experimental and theoretical results for coplanar waveguide which are found in the literature [16, 18, 191. At low frequencies, the two types of lines are expected to perform equally well, since dielectric and radiative losses are relatively small. Thus the attenuation is dominated by conductor loss, which should be about the same in microshield and cpw due to the similar geometry of the metallization. At higher frequencies, however, the microshield line gains an increasing advantage since it has essentialy zero dielectric loss and does not radiate energy into the substrate, unlike a substrate-supported cpw line. This characteristic has been verified by recent electro-optic sampling experiments, which provided attenuation data on membrane lines up to 1000 GHz [5]. The comparison between the microshield and cpw attenuation is shown in Figure 5. The microshield data in curves A and B was generated by averaging the measured insertion loss for several lines with lengths between 2.3 mm and 8.5 mm, while the gcpw data of point H was taken from measurements of a microshield-to-gcpw transition on a high-resistivity silicon wafer (p > 2000 $-cm). The cpw curves which are shown for comparison were chosen either because the dimensions are similar to the microshield lines (curves C,F.G) or to illustrate the effects of reducing the line geometry (curves D,E). All pertinent geometrical parameters for the cpw and microshield examples are given in the table beneath the figure, and it is 7

.25 ----------------- -------.00 1 z 0 CL -.25 -.50 -.75.~o,~~.~. —~ ---........ooo-o ~. oo.o., —..,.. I,,aI -1.00 -10. A A.......... --- ------—.. 15. 20. 25. 30. 35. FREQUENCY [GHz] 40. I - A.-o Melized ' Backside | Cavitier T-I 2(X x 200 H=350 Pmbe Contacts GCPW t MICROSHIELD ---— I 680 x 680 I V 8 80 3 i2m =300 V2 Ai I QC Figure 4: Measured S21 of two microshield delay lines of different length, using the modified transition from the cpw probes. The cavities are etched from the backside of the wafer and are pyramidal in shape; th dotted line is the incavity size at the bottom of the wafer. All dimensions are in microns. 8

noted that the conductor thickness is only about 2 skin depths at 25 GHz for all but two cases. Several aspects of the information shown in Figure 5 deserve comment. First, the curves indicate that the loss for microshield and cpw lines of similar size is nearly the same at low frequencies, as expected. Also, the previous assertion that the microshield line performs better as the frequency increases is supported by the trends in the data. The second point is that the results for the microshield lines are consistent with the cpw loss behavior outlined in [17] by Jackson; namely, for a constant ground plane separation, an aspect ratio (S/S+2W) of 0.83 (curve A) gives much higher loss than an aspect ratio of 0.63 (curve B). A smaller shielding structure, however, may cause the behavior to deviate from that of the cpw line, particularly as the cavity size is reduced to the point that the lower ground plane and sidewalls affect the configuration of the fields. A third point relates to a comparison between the cpw and microshield data in terms of loss per electrical length, as the wavelength in the air dielectric is about 2.5 times longer than in GaAs and about 1.5 times longer than in quartz. The calculated loss for cpw in points F and G is about 0.1 dB/Ag lower than the measured loss in curve A for the microshield line, which has identical center conductor and slot widths. This is a relatively small difference, however, some of which can be explained by errors in the measured and/or theoretical data. The results presented here confirm that the microshield line is free from unexpected or excessive conductor-loss mechanisms, and has performance which is comparable to conventional substrate-supported coplanar waveguide at lower frequencies. Furthermore, the absence of dielectric-related loss and the ability to maintain non-dispersive, single-mode propagation over a very broad bandwidth lead to low attenuation well into the millimeter-wave frequencies. IV. ADVANTAGES OF AN AIR SUBSTRATE Due to the air substrate of the microshield line, the effective dielectric constant, Er,eff, is 1The technique used in [18] assumes zero conductor thickness. 9

.25 E E co la z P: <:D z LL <.20.15.10.05.00 5. 15. 25. 35. 45. FREQUENCY [GHz] Curve Line E- Substr S Wi H t Zo Q Data Ref. A pshield 1.0 Air 250 25 355 1.2 75 Meas - B usshield 1.0 Air 190 55 355 1.2 100 Meas - C cpw 12.8 GaAs 232 84 100 - 50 Calc [18] D cpw 12.8 GaAs 69 28 100 - 50 Calc 18]' E cpw 12.9 GaAs 88 16 500 1.0 30 Meas [19] F cpw 12.9 GaAs 250 25 500 1.0 30 Calc [16] G cpw 4.0 Quartz 250 25 250 1.0 50 Calc [16] H gcpw 11.7 Si 50 125 355 1.2 73 Meas - Figure 5: Attenuation for microshield and coplanar waveguide lines. S is the center conductor width, W the slot width, H the substrate height, and t the metal thickness (in um). The width of the lower shielding cavity for the microshield lines is 1800 microns. 10

1.25 I.... I..I.W=25 I I.. I G. O: S=250 ain.W=25 pm, G —300) im 1.20 - --- S=190 r,W=55 Am. G=300 n - 1.15,1.10. ~ - - 1.05 - 1.00.... 10. 15. 20. 25. 30. 35. 40. FREQUENCY [GHz] Figure 6: Measured effective dielectric constant, Er,ef f, on two microshield lines with different aspect ratios. The cavity heigth (H) is 350 /m. very close to 1. Because of the membrane, however, Ere ff is slightly increased since a fraction of the fields will be contained within the thin dielectric layers. The membrane is a 1.5 pmthick tri-layer conposite of SiO2/Si3N4/SiO2 with constituent thicknesses of approximately 7000A/3000A/4000A. The dielectric constant of the oxide is 3.9, and that of the nitride is 7.5. As shown in Figure 6, the measured fr,eqf changes from around 1.09 to 1.15 as the slot width is reduced from 55 to 25 pm. The increase is a result of greater field confinement in the slot and thus in the membrane, and represents a decrease of nearly 3% in the guided wavelength. This dependence on the slot width is similar to the characteristics of cpw lines on a substrate such as GaAs or quartz [19]. There are several potential advantages of the nearly homogeneous air-dielectric. These include the absence of substrate moding problems, which in turn reduces dispersion and losses due to parasitic radiation. Another advantage is the improved electrical performance of discontinuities such as shorts, opens, steps and bends. As an example, the effective length extension2 of a grounded coplanar short-end on substrates with.,=1.25 and 12.9 2The effective length extension, ALt,c is the position at which an ideal short circuit would result in the same impedance seen at the physical end of the line. 11

(GaAs) is shown in Figures 7 and 8. The value of 1.25 was chosen to yield an effective dielectric constant around 1.12, in order to simulate the imicroshield line. The plots are theoretical results from the full-wave analysis described in the introduction, and from an expression by Getsinger which is applicable for quasi-TEM lines printed on an infinitelythick dielectric [20]. In Figure 7, all lines have S=W=100ljm with substrate heights of 350 and 700pm, while in Figure 8, all lines have a characteristic impedance around 65 f. Since the end-effect is completely inductive for pure-TEM operation, the physical size of the length extension is nearly independent of the dielectric constant. The shorter wavelength in the GaAs substrate, however, makes the extension become longer electrically. A similar argument can be made regarding the performance of coplanar open-ends. Under the quasiTEM conditions, the physical length of the effective length extension, ALoc, is roughly independent of Ce, for small values of ALoc. Again, this results in an increased electrical length when the substrate dielectric constant is higher, and these end effects can become significant since ALc is typically at least twice as large as ALc [21]. Thus, the air-filled environment of the microshield geometry offers the optimum conditions for making circuit performance less susceptible to these frequency dependent phenomena. V. MICROSHIELD CIRCUIT COMPONENTS V.1 Transitions to Coplanar Waveguide An electrically short, low-loss transition between microshield and gcpw is important for measurement calibration purposes and for applications which could integrate circuit components configured in both types of lines. Essentially any means of coupling power into or out of a microshield circuit will require transmission lines printed on a solid substrate to provide mechanical strength at the connection. It has been found that the similarities of the field configurations in cpw and microshield make broad band, low-loss transitions quite easy to achieve [3], despite the large difference in the dielectric constants of the transmission media. Although a rigorous optimization has not been conducted, designs based on a simple matching of quasi-static characteristic impedance 12

15. - 10.~- \ii =- H=700 mL, =1 2.9 --— r --- Getsinger. E,=12.9 5. GaA Membrane ] 0. 5. 10. 15. 20. 25. 30. 35. 40. 45. FREQUENCY [GHz] Figure 7: Effective length extension (in degrees) for a short-end discontinuity on lines with equal aspect ratios of (S/S + 2W) = 0.33. The characteristic impedance for the lines are 135 f (e7=1.25) and 55 ~? (ec=12.9). The line parameters are S=100 pm, W=100 pm and the substrate height H is given in the legend. 15. 10. I... I I. I I I. I. I -..- I - - - -. -. -. - -.. - I - -,=1.25 ---- Getsinger, E=1.25 - H=350 am, =12.9 - H=700 mun, e= 1 2.9 ------ Getsinger, e,=12.9 <3 d 5. - GaAs. Membrane.= = i~~...I I... I.... I.... 0.. - -.-..- -.... - -.. - 5. 10. 15. 20. 25. 30. 35. 40. 45. FREQUENCY [GHz] Figure 8: Effective length extension (in degrees) for a short-end discontinuity on lines with equal characteristic impedance (65 Q). The parameters are S=165 /Lm, W=10 /m for the line with e,=1.25, and S=40 itm, W=67 jm for the line with e,=12.9. The substrate height H is given in the legend. 13

values have demonstrated very good performance up to 40 GHz. Three designs are given in Figure 9; the 'wide' design has slot widths which are nearly twice the substrate height, and thus will propagate a mode which is similar to that of microstrip, whereas the 'match' and 'taper' designs have narrower slots, leading to propagation of a cpw-like mode. The measured data on these transitions is shown in Figure 10 and the lower plot shows that a return loss below 20 dB across the entire band has been obtained. For the cpw curves, the ground plane beneath the transition section was removed. Also, the noise in the data is due to the calibration, since the original probe-to-microshield transition was used for these measurements (Figure 3). The effect of a field nlismatch between the microshield and gcpw lines is evident by comparing the 'wide' and 'match' gcpw designs, which have a similar characteristic impedance but different configurations of the electromagnetic fields. As mentioned, the 'match' design, like the microshield, will have a primarily cpw-like mode since the fields will tend to concentrate in the narrow slots. The slot widths in the 'wide' design, however, lead to propagation in more of a microstrip-like mode since there is greater influence from the lower ground plane. This influence can be examined by computing the change in the characteristic impedance as the substrate height (H) is varied. If H goes from 350 sm, as used in the current lines, to 1 cm, the changes in Zo for the microshield and 'match' gcpw lines are only 5.8% and 6.2%, respectively, but for the 'wide' gcpw the change is 31%. The field mismatch is clearly reflected in the data, which shows that Sl,match is 4-6 dB lower than Sl,uidc across the entire band. Additional transitions were tested which have smaller geometries for both the microshield and gcpw lines than those in Figure 9, and are therefore more appropriate for compact circuit configurations. The performance for two of the designs is shown in Figure 11. These circuits were measured using the modified probe-to-microshield transition (Figure 4) and thus the data has much less structure than in the previous figure. The strong resonances occur at frequencies for which the length of the gcpw sections are multiples of 2_. Each transition 14

Air - A,' W Metal Silicon IT L gp shield (g)cpw p. shield Design H L S W D IZJ WIDE 350 1460 120 650 0 71 MATCH 350 1460 60 180 0 75 TAPER 350 1460 70 175 140 71 Figure 9: Three transitions from 75Q microshield line (S=350 /m, W=35 /m) to cpw and grounded cpw. The dimensions for the (g)cpw lines are given in the table in microns, and the characteristic impedance values are for the grounded cpw configuration (i.e. with a lower ground plane). 15

I f% -I. r -I - CWM80 a WIDE CPWc w75 MATCH -20. -20. - CPW 102 Q -30. D -30. 10. 20. 30. 40. 10. 20. 30. 40. FREQUENCY [GHz] FREQUENCY [GHz] -10., M7CPW76 f C- CCPW 71 TAPER -20. -30. 10. 20. 30. 40. FREQUENCY [GHz] Figure 10: Measured S1 for three transitions from 75 f microshield line to cpw and grounded cpw (gcpw). The plots are labeled in the upper right hand corner to correspond to the designs in Figure 9. 16

0. ---- A: S2 -10. - *- BS, -~ --- B: S2 -15 L t21.. z < -20. -25.-30. w 10. 20. 30. 40. FREQUENCY [GHz] Figure 11: Measured scattering parameters for two transitions from 75 Q microshield line to cpw and grounded cpw (gcpw). The center conductor width (S) and slot width (W) for the microshield are 250 and 25, respectively. In curve A, the gcpw dimensions are S=50, W=125 and the length of the line is L=1460. In curve B, the gcpw dimensions are S=30, W=80 and L=3500. All dimensions are in microns. has a reasonably low reflection across the entire band, and it is expected that a rigorous full-wave analysis could lead to further design improvements. V.2 Right-Angle Bend A very common circuit 'element' in millimeter-wave systems is the right-angle bend. This structure gains significance with increasing frequency due to the parasitic capacitance and inductance which are associated with the abrupt change in the field orientation. These problems, combined with the mode-conversion which can also occur, result in a high reflection of the incident power (large S11). In microstrip form, the typical means of improving the return loss is to miter the outside corners of the signal line [22]. This approach is less effective for coplanar waveguide [23], and alternative approaches such as air-bridges and dielectric overlays are adopted [24]. Botlh of these techniques are meant to offset the effects of different electrical path lengths along the two slots. Without this type of compensation, the asymmetry of the cpw right-angle bend will lead to excitation of the unwanted slot-line 17

mode. Many of the problems inherent to bends which are printed on conventional cpw-line can be minimized with the microshield geometry. The absence of the high dielectric constant material has two important effects: it leads to a reduction in the parasitic capacitance and, for a fixed physical size, it reduces the difference in electrical path lengths through the two slots. The shielding cavity, furthermore, provides continuous ground plane equalization without introducing additional discontinuities, thus improving upon the conventional airbridge. A comparison between the measured performance of a microshield-line bend and a typical cpw-line bend on GaAs [23] is shown in Figure 12. The cpw bend has a much smaller geometry and incorporates air-bridges which are placed very close to the discontinuity. The microshield bend has an Si, which is at least 8 dB lower over the 10-40 GHz band. Even when the lower ground plane was removed and the microshield bend was mounted on a styrofoam block, the S11 stayed below -20 dB, indicating that conversion into the slot-line mode is minimal. The noise in the data below 18 GHz is due to calibration error. V.3 Open-End and Short-End Series Stub Another common coplanar waveguide component is the open-end series stub, which was previously analyzed by Dib et al. [6]. The passband resonance of the stub occurs when the mean length is 4-, and the associated radiation loss (1 - |IS 12 - IS21&12) is very low due to the location of the stub within the center conductor. The stop-band resonance occurs when the stub length is 2 and is considerably stronger than the passband resonance, resulting in higher radiation losses. To demonstrate these characteristics, a comparison between measured and theoretical results for a typical open-end stub is given in Figure 13. The theoretical data, which are in good agreement with the measured results, extend out to 70 GHz in order to show the stop-band response. Since the attenuation in the measured circuit is dominated by conductor loss, the radiation loss is calculated using the theoretical scattering parameters. Small errors in the numerical results caused erratic behavior of the radiation loss above 60 GHz, and these results are not shown. Also, the shift in the predicted frequency 18

0. I. I t 0 S+2W 0. --- CPW... -15. -20... -30.... -35.-40.. ----~ ~ 10. 15. 20. 25. 30. 35. 40. FREQUENCY [GHzJ Figure 12: Measured performance of right-angle bends fabricated using a microshield line and a cpw line on GaAs. The cpw bend has S=39, W=24, H=90 and uses 2 gm-high airbridges on each side of the bend [231. The microshield line has S=250, W=25 and H11=350. All dimensions are in microns. response is believed to result partly from using an inaccurate value for the effective dielectric constant (refer to Section I), as the slot widths in the actual circuit were 3-5 gum smaller than expected due to fabrication problems. As presented in section IV, this causes Er,eff to increase and therefore reduces the resonant frequency. The performance seen in the figure is typical of designs with gap widths L2 as small as 60 tlm, as was verified experimentally and theoretically. Furthermore, the response is more narrow band than a stub with similar transverse dimensions printed on silicon [25]. The short-end series stub, like the open-end stub, is useful for a variety of circuits including bandstop filters, pin diode switches and attenuators, and has been reported by several authors [6, 26, 27]. The geometry offers the same advantages of compactness and low radiation loss. Unlike the open-end version, however, it has a much wider 3-dB bandwidth than a comparable stub printed on alumina or GaAs [25] (roughly 70% compared to 35%). The performance for a typical design which is a at r 30 GHz, along with results obtained from resone i blieedto eslt arly ro usnganinacuatevaue ortheefectvedieecr4 19

-5. L **- - "' ' Cu 1 1 i -20.-,- ' C:,.-'-., ': -25. ' -- [MEAS] -30' A t -' [ CALCI ------ Rad. Loss [CALC] -35.. - { I 10. 20. 30. 40. 50. 60. 70. FREQUENCY [GHz] Figure 13: Measured and calculated S-parameters for a microshield open-end series stub. The stub dimensions are L=2500, L2=200, S=250, S2=80, W=25, and W2=25. All dimensions are in microns. 20

S L S11 -5. - Rad. Loss [CALC -10., 0 S.. 30. 40. 50. FREQUENCY [CALC The stub dimensis are L=2500, 2=200, =250, 80, W 25, and W225. All dimensions are in microns. the full-wave analysis, are shown in Figure 14. The shift in the frequency response is similar to the one in Figure 13 and carries the same explanation. As with the open-end stub, small variations in the dimensions have no significant effects on the shape or bandwidth of the response, but may change the resonant frequency if the mean length is altered. V.4 Lowpass Filters -35. - FREQUENCY fGHzI The stepped-impedance approach to lowpass filter design is a roshied short-ly easy technique to use, and it is well suited for applications which do not require a sharp roll-off in the insertion loss. Often, however, the filter specifications call for high rejection over multiple-octave bandwidths, a requirement which may be difficult to meet using conventional substrate-supported lines due to the propagation of higher order modes. Thus the very broad, single-mode band 21

width of the microshield line can provide superior filter performance in this respect. In addition, the absence of the dielectric-related loss mechanisms, both attenuation and parasitic radiation, results in very low passband insertion loss. Stepped-impedance filters using 5-, 7-, and 9-sections have been designed and tested. Some of the results were previously presented in [3], where it was shown that the measured performance compared very well with ideal transmission line theory, as a result of the pure TEM nature of the microshield propagating mode. In Figure 15, the measured data for a 5-section filter is compared with results from the full-wave analysis described in section I. The rejection is greater than 20 dB up to 75 GHz, which is about 1.5 octaves above the 3-dB point at 26 GHz. The measured passband insertion loss of this filter is between 0.2 and 0.5 dB from 20-23 GHz. The noise in the measured data is due to the transition design shown in Figure 3. As mentioned previously, a short-end series stub can be utilized to make circuits such as bandstop filters by cascading multiple stubs in series. It is also possible to take advantage of the broad bandwidth of this stub to design high performance lowpass filters. One example is the addition of a short-end stub onto the end of the stepped-impedance lowpass filter described above. The stub has a length of 940/im, with other dimensions identical to those listed in Figure 14, and resonates around 90 GHz. By comparing the theoretical performance of this new filter in Figure 16 to the performance without the stub in Figure 15, it is clear that there is a significant improvement in the stopband of the filter. Another example is a very compact, easy to design lowpass/bandstop filter with a corner frequency of 52 GHz, which is simply the series combination of two 90 GHz stubs. This design has a total length of 1.98 mm, using a 100 min spacing between the stubs, and thus would have very low conductor loss along with the low radiation loss. The predicted performance is shown in Figure 17. By reversing the second stub to make the circuit longitudinally symmetric, the roll-off is more gradual and thus the low-band ripple in S21 is reduced. The predicted peformance for this configuration is shown in Figure 18. 22

75 n (EDGE OF LOWER SHIELDING CAVITY) L, L2 L3 L2 L, -0 C: z < s 0.. -5. -10. -15. -20. -25. -30. -35. - 2(0. 30. 40. 50. 60. 70. 80. FREQUENCY [GHz] Figure 15: Measured and calculated lengths of the different sections are LI S-parameters for a microshield lowpass filter. The = 745, L2 = 1200, and L3 = 1189 (microns). 23

-0. -155 - -20. '-1 - ~ -2S.- - -30. - IL -35.- - 44)... I. I...... 10. 20. 30. 40.. 60. 70. 80. FREQUENCY [GHz] Figure 16: Theoretical performance of the filter shown in Figure 15 with a 90 GHz short-end series stub attached. 24

I.,,-x l lI | -0 cU z Figure 17: Theoretical figuration). 0. -10. - 1 -30. o tS s -35. S -40. -.. i 10. 30. 50. 70. 90. 110. 130. FREQUENCY [GHz] performance of two 90 GHz short-end series stubs (asymmetric con 25

Is --- —----------- - -Ea~ &~ lQ 3 -z Figure 18: Theoretical figuration). -5. -25. -30. -35. - -. ' I 21 -25. ~.: -30. 10. 30. 50. 70. 90. 110. 130. FREQUENCY [GHz] performance of two 90 GHz short-end series stubs (symmetric con 26

V.5 Low-Loss Bandpass Filters By cascading multiple open-end stubs in series, it is quite simple to realize a bandpass response with high out-of-band rejection and low loss. A three-section design was fabricat.ed using the stub dimensions given in Figure 13, with each section separated by 150 tm. The measured response, shown in Figure 19, has an insertion loss of only 1.0 dB from 22-32 GHz, which is competitive with the best shielded bandpass filters using suspended stripline [28]. The performance of this microshield filter could be further improved using thicker metallization3 or larger slot widths to minimize the conductor loss. The calculated data is generated by using the scattering l)arallleters found from the full-wave analysis of a single stub and treating the filter as three non-coupled elements in series. The agreement between the measured and calculated performance is quite good and indicates that there is very little electromagnetic coupling between the stubs, even though the stub separation is only 150 im. The response of the filter can also be modeled almost exactly by cascading the nieasured results for a single stub, and this approach accurately predicts the 1.0 dB insertion loss. VI. CONCLUSIONS This paper presents a comprehensive introduction to the microshield transmission line and several passive circuit components. The measured performance from 10-40 GHz is very good and in certain cases superior to the performance which is characteristic of conventional coplanar waveguide circuits. Clearly, there are differences between the microshield line and a substrate-supported line, including larger circuit dimensions due to the low dielectric constant and the use of the thin dielectric membrane. At the same time, however, there are the advantages of very broad-band TEM operation, a wide range of characteristic impedance, no requirement for air-bridges, and the possibility of using much thicker substrates at higher frequencies without the need for mechanical thinning. At submillimeter 3AII circuits fabricated in this studily had a metallization thickness of 1 pm. 27

0 z 0

wavelengths, furthermore, the larger electrical size could simplify fabrication. Thus, the microshield line shows significant potential for high-performance millimeter wave applications, and is a strong candidate for submillimeter wave designs. ACKNOWLEDGEMENTS This work was supported by the NASA Center for Space Terahertz Technology and the Office of Naval Research. The authors thank Ms. Rhonda Franklin Drayton (UoM) and Mr. George E. Ponchak (NASA-Lewis/UoM) for their valuable discussions and considerable help in making the measurements, and Dr. Nihad Dib (UoM) for his expert help in the theoretical analysis. REFERENCES [1] N. I. Dib, W. P. Harokopus, L. P. Katehi, C. C. Ling, and G. M. Rebeiz, "Study of a Novel Planar Transmission Line," 1991 IEEE MTT-S Digest, Boston, pp. 623-626. [2] N. I. Dib and L. P. Katehi, "Impedance Calculation for the Microshield Line," Microwave and Guided Wave Letters, Vol. 2, No. 10, Oct., 1992, pp. 406-408. [3] T. M. Weller, G. M. Rebeiz, and L. P. Katehi, "Experimental Results on Microshield Line Circuits," 1993 IEEE MTT-S Digest, pp. 827-830. [4] T. M. Weller, L. P. Katehi, G. M. R.ebeiz, H. J. Cheng, and J. F. Whitaker, "Fabrication and Characterization of Microshield Circuits," Proceedings of the 4th International Symposium on Terahertz Technology, University of California at Los Angeles, Los Angeles, CA, 1993, pp. 223-237. [5] H. J. Cheng, J. F. Whitaker, T. M. Weller, and L. P. Katehi, "Terahertz-Bandwidth Characterization of Coplanar Waveguide on Dielectric Membrane via Time-Domain Electro-Optic Sampling," To be presented at the 1994 MTT-S Internation Microwave Symposium. [6] N. I. Dib, L. P. Katehi, G. E. Ponchak, and R. N. Simons, "Theoretical and Experimental Characterization of Coplanar Waveguide Discontinuities for Filter Applications," IEEE Trans. MTT, vol. 39, No. 5, May, 1991, pp. 873-882. [7] G. F. Engen and C. A. Hoer, "Thru-Reflect-Line: An Improved Technique for Calibrating the Dual Six-Port Automatic Network Analyzer," IEEE Trans. MTT, Vol. MTT-27, No. 12, Decenmber, 1979, pp. 987-993. 29

[8] M. Tsuji, H. Shigesawa, and A. A. Oliner, "New Interesting Leakage Behavior on Coplanar Waveguides of Finite and Infinite Widths," 1991 IEEE MTT-S Digest, pp. 563-566. [9] H. Shigesawa, M. Tsuji, and A. A. Oliner, "Conductor-Backed Slot Line and Coplanar Waveguide: Dangers and Full-Wave Analyses," 1988 IEEE MTT-S Digest, pp. 199-202. [10] M. A. Magerko, L. Fan, and K. Chang, "Multiple Dielectric Structures to Eliminate Moding Problems in Conductor-Backed Coplanar Waveguide MICs," IEEE Microwave and Guided Letters, Vol. 2, No. 6, June, 1992. pp. 257-259. [11] R. W. Jackson, "Mode Conversion Due to Discontinuities in Modified Grounded Coplanar Waveguide," 1988 IEEE MTT-S Digest, pp. 203-206. (12] L. Wen-Teng, C-K. C. Tzuang, S-T. Peng, C-C. Chang, J-W. Huang, and C-C Tien, "Resonant Phenomena in Condlctor-Backed Coplanar Waveguide (CBCPW)," 1993 IEEE MTT-S Digest, pp). 1199-1202. [13] M. Riaziat, I. J. Feng, R. Majidi-Ahy, and B. A. Auld, "Single-Mode Operation of Coplanar Waveguides." Electronics Letters, 19th November, 1987, Vol. 23, No. 24. pp. 1281-1283. [14] M. Riaziat, I. J. Feng, and R.. Majidi-Ahy, "Propagation Modes and Dispersion Characteristics of Coplanar Waveguides," IEEE Trans. MTT, Vol. 38, No. 3, March, 1990, pp. 245-251. [15] E. M. Godshalk, "Generation and Observation of Surface Waves on Dielectric Slabs and Coplanar Structures," 1993 IEEE MTT-S Digest, pp. 923-926. [16] W. H. Haydl, W. Heinrich, R. Bosch, M. Schlechtweg, P. Tasker, and J. Braunstein, "Design data for millimeter wave coplanar circuits," 1993 European Microwave Conference Proceedings, pp. 223-228. [17] R. W. Jackson, "Considerations in the Use of Coplanar Waveguide For Millimeter-Wave Integrated Circuits,' IEEE Trans. MTT, Vol. 34, No. 12, December, 1986, pp. 1450-1456. [18] M. Zhang, C. Wu, K. Wu, and J. Litva, "Losses in GaAs Microstrip and Coplanar Waveguide," 1992 IEEE MTT-S Digest, pp. 971-974. [19] W. H. Haydl, J. Braunstein, T. Kitazawa, M. Schlechtweg, P. Tasker, and L. F. Eastman, "Attenuation of Millimeterwave Coplanar Lines on Gallium Arsenide and Indium Phosphide Over the Range 1-60 GHz," 1992 IEEE MTT-S Digest, pp. 349-352. [20] W. J. Getsinger, "End-Effects in Quasi-TEM Transmission Lines," IEEE Trans. MTT, Vol. 41, No. 4, April. 1993, pp. 666-672. [21] K. Beilenhoff, H. Ilingbeil, W. Heinrich, and H. Hartnagel, "Open and Short Circuits in Coplanar MMIC's," IEEE Trans. MTT, Vol. 41, No. 9, September, 1993, pp. 1534-1537. [22] R. Chadlha. and K. C. Gupta., "Compensation of Discontinuities in Planar Transmission Lines." IEEE Trans. Alicrowalve Theory Tech., Vol. 30, No. 12, December, 1982, pp. 2151-2156. 30

[23] Amjad A. Q. Omar, "An Accurate Solution of 3-D Coplanar Waveguide Circuits," Ph.D. Thesis, University of Waterloo, Waterloo, Ontario, Canada, 1993. [24] R. N. Simons and G. E. Ponchak, "Modeling of Some Coplanar Waveguide Discontinuities," IEEE Trans. Microwave Theory Tech., Vol. 36, December, 1988, pp. 1796-1803. [25] N. I. Dib, "Theoretical Characterization of Coplanar Waveguide Transmission Lines and Discontinuities," Ph. D. Thesis, University of Michigan, Ann Arbor, MI, 1992. [26] A. K. Sharma and H. Wang, "Experimental Models of Series and Shunt Elements in Coplanar MMICs," 1992 IEEE MTT-S Digest, pp. 1349-1352. [27] P. A. R. Holder, "X-band microwave integrated circuits using slotline and coplanar waveguide," The Radio and Electronic Engineer, Vol. 48, No. 1/2, January/February, 1978, pp. 38-42. [28] W. Menzel, "Broadband Filter Circuits Using an Extended Suspended Substrate Transmission Line Configuration," Proceedings of the 22nd European Microwave Conference, Aug. 24-27, 1992, pp. 459-463. 31

Stubmit.d to the IEEE Microwave and Guided Wave Letters on November 1, 1994. A 250 GHz Microshield Bandpass Filter Thomas M. Weller, Student Member, IEEE, Linda P. Katehi, Senior Member, IEEE, and Gabriel M. Rebeiz, Senior Member, IEEE The Radiation Laboratory University of Michigan Ann Arbor, MI 48109-2122 Abstract - A four-section, planar bandpass filter has been designed, fabricated and tested at 130-360 GHz. The filter is based on the microshield line, a half-shielded transmission line in which the conducting lines are supported on a 1.4 gm-thick dielectric membrane. The insertion loss of the filter is less than 1.5 dB at 250 GHz, demonstrating the excellent performance of the microshield geometry. Also, a Monte Carlo routine was developed in conjuction with a semiempirical/semi-analytical model to allow the S-parameters of the filter to be derived from scalar power measurements. I. INTRODUCTION The potential for high performance planar circuits which are implemented in conventional substrate-supported microstrip or coplanar waveguide is severely limited at sub-millimeter wave frequencies. This shortcoming is primarily due to dielectric and radiation losses, and the most common solution is to utilize waveguide-based designs [1]. One quasi-planar transmission line which can overcome these difficulties is the microshield line. Introduced in 1991 [2], microshield is a half shielded geometry, similar to CPW, which uses a thin (around 1.4 tm) dielectric membrane to support the conducting lines and upper ground planes above a metallized shielding cavity. It is characterized by low loss, broad band TEM propagation and has previously been demonstrated in circuits operating at Ka-band [3, 4] and W-band [5]. In the work presented here, microshield is utilized to fabricate the first planar bandpass filter at 250 GHz. This paper also describes a quasi optical measurement technique which was developed in order to determine the filter S-parameters 1

in the presence of reflections at the filter input and output ports, due to the receiving antenna and the microbolometer used for power detection, respectively. II. BANDPASS FILTER The bandpass filter consists of a compact configuration of open-end series stubs which provides both low loss and longitudinal symmetry. Application of the open-end stub to filter design has been examined in previous publications, in which it was noted that the location of the stub within the center conductor leads to greater field confinement and thus lower radiation loss than is characteristic of straight gap-coupled resonators or stubs with lateral extension [6, 7]. The design applied here utilizes four quarter-wavelength stubs which are resonant at 250 GHz, with the second and fourth stubs reversed relative to the first and third (Figure 1). III. MEASUREMENT TECHNIQUE At frequencies above 115 GHz, there are currently very few options available for microwave circuit characterization. Electro-optic sampling and waveguide-based approaches are possible techniques, but were not available for this project. Therefore, an alternative technique based on quasioptical methods was developed which uses a semi-empirical/semi-analytical model to extract the circuit S-parameters. The objective of the quasi-optical measurement is to determine the amount of power transmitted through the filter as a function of frequency, using the circuit shown in Figure 2. In this figure, RF power is supplied from a Gunn diode/tripler pair and coupled into the circuit through a CPW-fed, double folded-slot antenna [8]. A thin-film bismuth microbolometer detector [9] is placed at the output of the filter, in parallel with an RF short which is positioned approximately a quarterwavelength away (L4). The RF short is a silicon oxide (SiO) thin-film capacitor which allows DC biasing of the bolometer detector and presents an open circuit at the output of the filter. The power measurement is made by modulating the RF signal with a 1 KHz square wave and using a lock-in amplifier to detect the 1 KHz voltage across the bolometer. As a means of calibration, a similar circuit is utilized in which the filter is replaced by a straight length of microshield transmission line. Using the measured absolute responsivity (in Volts/Watt) of the bolometers in each circilt. 2

one can determine the ratio of the power received through the filter (Pf/it) to that received through the transmission line (PTL)IV. S-PARAMETER EXTRACTION The remaining step involved in characterizing the filter is to extract its S-parameters using the experimental values for P itt/PTL. An equation which relates the scalar power measurement data and all the pertinent vector quantities in the system is derived from the circuit model in Figure 2: Pfilt/PTL I= S2la i1- rdrbe-2y(L+2+L3)2 ) X 1 + (Sa- s1,) rd.rae-2(LI+3) - Sla (rd e-21L, + r e-2L) 12 (1) where 7 is the propagation constant of the microshield line. In order to resolve the unknowns in Eqn. 1, a Monte Carlo routine (MCr) has been implemented which uses a combination of empirical and analytical data for initial values. Estimates for ra and rb are computed using measured values for the bolometer DC resistance and values for 7 extrapolated from 10-40 GHz measurements. Initial values for the antenna reflection coefficient (rd.), on the other hand, have been determined using a full-wave analysis [8]. After specifying upper and lower limits for all the parameters in Eqn. 1, the MCr is then used to fit them to the model. An important point is that the initial range for the magnitude of the filter S-parameters is between 0 and 1, i.e. they are unconstrained. However, the quantity IS11,12 + IS21,o12 is restricted to a range which encompasses the expected overall filter loss. Also, the range for the phase is 80~ about the values predicted by a full-wave analysis of the filter. In order to investigate this method of solution, the filter S-parameters from the full-wave analysis were used along with randomly generated values for the remaining unknowns to compute exact values for Pfilt/PTr The MCr was then applied to this hypothetical data set and the accuracy proved to be very good. As shown in Figure 3, the MCr-extracted 1S21,al follows the true distribution almost exactly and the maximum passband error is about 0.45 dB. As a consequence of the weaker dependence of Eqn. 1 on ISll,a it is more difficult to resolve this parameter. Likewise, it is oni1l possible to determine the phase of the S-parameters when the magnitude of the respective unk nr s 3

is quite large. Even then, due to the rapid phase variation in the 4-element filter. the results can only approximate the overall distribution. V. RESULTS AND CONCLUSION Using the measurement technique described in section III and the S-parameter extraction technique described in section IV, the performance of the filter was measured at eight frequencies from 130-360 GHz. Due to the limited RF power and the low responsivity of the bolometers (typically around 2 Volts/Watt), the dynamic range is limited to approximately 20 dB. As shown in Figure 4, the S-parameters extracted from the raw measured data are very close to the predicted response from the full-wave analysis. The measured pass-band insertion loss for this four-section filter is around 1.5 dB, and the low frequency stop-band provides excellent rejection. These results confirm that high-performance, planar circuits are realizable at millimeter- and submillimeter-wave frequencies using the microshield line configuration. ACKNOWLEDGEMENTS This work has been supported by the NASA Center for Space Terahertz Technology and the Office of Naval Research. The authors would also like to thank Sanjay Raman and Steve Gearhart for their help in making the measurements, and Dan Ross for discussions regarding the Monte Carlo routine. 4

REFERENCES [1] L. Q. Bui, D. Ball, and T. Itoh, "Broad-Band Millimeter-Wave E-Plane Bandpass Filters, IEEE Trans. MTT, pp. 1655-1658, December 1984. [2] N. Dib, W. Harokopus, P. Katehi, C. Ling, and G. Rebeiz, "Study of a Novel Planar Transmission Line," 1991 IEEE MTT-S Digest, pp. 623-626. [3] T. M. Weller, G. M. Rebeiz and L. P. Katehi, "Experimental Results on Microshield Transmission Line Circuits," 1993 IEEE MTT-S Digest, vol. 2, pp. 827-830. [4] T. M. Weller, L. P. Katehi and G. M. Rebeiz, "High Performance Microshield Line Components," To appear in IEEE Trans. MTT, March 1995. [5] S. V. Robertson, L. P. Katehi and G. M. Rebeiz, "W-Band Microshield Low-Pass Filters," 1994 IEEE MTT-S Digest, vol. 2, pp. 625-628. [6] D. F. Williams and S. E. Schwarz, "Design and Performance of Coplanar Waveguide Band-Pass Filters," IEEE Trans. MTT, vol. MTT-31, pp. 558-566, July 1983. [7] N. Dib, P. Katehi, G. Ponchak and R. Simons, "Theoretical and Experimental Characterization of Coplanar Waveguide Discontinuities for Filter Applications," IEEE Trans. MTT, pp. 873-882, May 1991. [8] T. M. Weller, L. P. Katehi and G. M. Rebeiz, "Single and Double Folded-Slot Antennas on Semi-Infinite Substrates," Submitted to the IEEE Ant. Propagat, Sept. 1994. [9] D. P. Neikirk, W. W. Lam and D. B. Rutledge, "Far-Infrared Microbolometer Detectors," International Journal of Infrared and Millimeter Waves, vol. 5, no. 3, pp. 245-278, 1984. 5

Figure 1: The four-section microshield bandpass filter (not to scale). The metallization pattern is shown in (a), where L=250, S=50 and W=10. A cross-sectional view is shown in (b), where H=200, Wi =320 and W2=40. The geometry is comprised of three silicon wafers which are used to support the membrane and filter metallization, form the cavity sidewalls, and complete the lower ground plane, respectively. In (b), metallization is indicated by the dark lines. All dimensions are in microns. Figure 2: Diagram of the circuit used in the quasi-optical measurement, where DUT stands for device under test. The upper illustration shows a circuit which has a bandpass filter as the DUT. The reflection coefficients ra and rb may be different, in general, to allow for fabrication tolerances between the bolometers of the filter circuit and the transmission line circuit. Figure 3: Comparison between the MCr-extracted values (markers) and the "exact" predicted filter response from a full-wave analysis (lines), using a hypothetical set of system parameters to compute P/ilt/PTL. Figure 4: Comparison between the results extracted from raw measured data (markers) and the predicted filter response from the full-wave analysis (lines). 6

w 0 s \.\,NN N z IX ~ XX X X XXsis XXXWXXXX L L L L (a) (b) 7

Antenna Filter - RF Short -- '. m-a - or ~ m I I Bolometer Bolometer rds..J ra,b L.. Folded-Slot Antenna [Sa]: DUT = Filter [Sb]: DUT = Trans. Line To Lock-in Amplifier 8

MAGNITUDE [dBl iS) -:1 0, S "'' I.a, a I

-0. a I 21 a40. 100-2 0 0.@ 30.40. A10 FRQUNC[Gz 10

Submitted to IEEE Trans. Antennas Propagat., September 1994. Single and Double Folded-Slot Antennas on Semi-Infinite Substrates Thomas M. Weller, Student Member, IEEE, Linda P. Katehi, Senior Member, IEEE, and Gabriel M. Rebeiz, Senior Member, IEEE The Radiation Laboratory University of Michigan Ann Arbor, MI 48109-2122 Abstract - In this paper, design guidelines are presented for single and double folded-slot (DFS) antennas on a dielectric half-space. The DFS antenna consists of two folded-slot elements which are separated by a half-wavelength, and are fed from one end using a coplanar waveguide transmission line. A space-domain integral equation technique, based on the free-space Green's function, has been used to compute the resonant circumference and self and mutual impedances of single foldedslots on various substrates. In addition, the input impedance and radiation patterns of several DFS designs are included. Specific examples are presented for air, quartz and silicon substrates, and measured data for DFS antennas on air and silicon substrates is provided to validate the calculated results. Finally, details pertaining to the derivation and evaluation of the moment method admittance terms are described. I. INTRODUCTION As is well known, planar antennas consisting of patches, dipoles or slots, fed by a microstrip transmission line, are extremely useful due to their low cost, light weight, and flexibility of design. They have been extensively analyzed (e.g. [1]) and are used in applications which range from complex conformal arrays for radars to quasi-optical spatial power combiners [2]. As the popularity of the coplanar waveguide (CPW) transmission line has increased significantly in recent years, antenna elements which are suitable for a CPW-feed configuration have also become important [3, 4, 5]. In this paper, design guidelines for CPW-fed, single and double folded-slot (DFS) antennas on a dielectric half-space are presented. The DFS antenna consists of two planar, folded-slot elements which are displaced along a CPW line by a half-wavelength, resulting in a compact geometry which provides a modest amount of beam-shaping. In order to characterize the antennas, a space domain integral equation technique has been employed which determines the magnetic current, or electric field, in the apertures of the metallization. It is a standard formulation based on the expansion of the unknown current in terms of 1

subsectional basis functions, the coefficients of which are solved for using the method of moments with Galerkin testing. A unique feature of this implementation, however, is the approach used to calculate the admittance terms for overlapping piecewise sinusoidal basis functions, in particular the handling of the 1/r singularity. A general outline of the formulation is presented in section II, while the details pertaining to the evaluation of the admittance terms are the subject of the appendix. The calculated and experimental results presented here are intended to provide a thorough introduction to the characteristics of single and double folded-slot antennas on semi-infinite substrates. The data primarily pertain to air (e,=1), quartz (Er=4) and silicon (E,=11.7) substrates. In section III, the resonant circumference, self impedance, and mutual impedance of the folded-slot are examined. This is followed by an analysis of several double folded-slot designs in section IV, which includes the input impedance (Zi,), directivity and far-field patterns. II. THEORY The formulation used here is a space domain integral equation technique based on the free-space dyadic Green's function. In this approach, continuity of the tangential electric (E) and magnetic (H) fields in the apertures of the CPW feed line and antenna is enforced, resulting in an integral equation involving the aperture electric field (or magnetic current, M) as the unknown parameter. Since M = E x ni, where n' is the outward unit normal in each half-space, continuity of Eta is satisfied by setting Msup = -Msub = M. In this expression, Msup and Msb refer to the equivalent magnetic currents on the surfaces of the upper and lower dielectric half-spaces, respectively. The continuity of Htan is expressed in terms of the following equation: h x [G,,(r, t) + Gub(r )] * M()ds' = J(1) where G is the free-space dyadic Green's function and J is the current excitation which is assumed with the ideal current source model. To solve this equation, the aperture is first subdivided and piecewise sinusoidal basis functions are utilized to approximate the magnetic current distribution. These functions are of the form: T(C,() = f(C)p(() sin(kin (C-n-1)) if n- < C < (n sin(kkm (n-i) f(C) = sin(km (C._,+1) if (n_ < ~ < (n+ [ '^ ^ if~n< Cn -H p(~) = 1 if (n-1 <. < n nt = (n0 (else 1nl = |Cn-Cn~l (2) 2

where km is a scaling parameter which is typically equal to the free-space wavenumber. By applying the method of moments with Galerkin testing to Eqn. 1, the following system of equations is obtained: ( (Y~Z> (Y> ) ( {MA} ) - {Jz} ) (3) (YzX) (Yzz) I AIz} f jz} ( where (Y(((, = x, z)) represents a block of the admittance matrix, {May} is a coefficient vector for the current expansion functions, and JC is the weighted integral of JC over a subsection. The mechanics involved in deriving expressions for the admittance terms have been presented in previous publications, and the particular method adopted here was utilized by Barkeshli and Volakis [6]. Although the details of that analysis will not be repeated here, the approach taken to accurately compute the self-cell term when using the PWS basis functions is outlined in the appendix. After solving Eqn. 3 for {M}, the normalized self impedance is determined by analyzing the magnetic current standing-waveform in the antenna feedline [7]. A cubic-spline interpolation can resolve the values and positions of the current minima and maxima, from which the guided wavelength (Ag), standing-wave ratio (swr), and distance from the reference plane to the first current maxima (dmax) can be calculated. The impedance is then found using ideal transmission line equations. The mutual impedance between two folded-slots is determined by inverting the admittance matrix and extracting the (ij) element of the resultant impedance matrix, where i refers to the subsection at the input port of the first antenna and j refers to the subsection at the input of the second antenna. This element, Zj, directly relates the voltage at port i due to a source current at port j. Finally, the directivity and far-field patterns are determined with standard expressions from the literature (e.g. [8]) since the complete magnetic current distribution in the antenna elements is known. III. SINGLE FOLDED-SLOT ANTENNA This section provides calculated results for the resonant frequency and self and mutual impedances of the single folded-slot antenna on different semi-infinite substrates (Figure 1). As with the simple slot antenna, the radiation pattern is simply that of a magnetic dipole, and thus is not included in this analysis. The frequency behavior of the self impedance for the three designs described in Table 1 is given in Figure 2. It is seen that the first useful resonance coincides with the peak in the real part of the self impedance, since the first zero crossing of the imaginary part is associated with a near short-circuit impedance. The dependence of this first resonance on the dielectric constant of the substrate material, c, is examined in Figure 3. This data shows the circumference of the antenna normalized to the free-space wavelength at the resonant frequency, where the circumference is given by C = 2(La + Sa + Wa2 - Wal). The curves can be approximated by using a quasi-static 3

expression for the effective dielectric constant: C/A +1)/2 (4) where Co is the value of C/Ao for E, = 1.0. For the designs considered in Figure 3, this constant is equal to 0.97 and 0.99 for SalLa = 0.01 and 0.04, respectively. Equation 4 underestimates the full-wave result, however, indicating that the actual effective dielectric constant is a few percent lower than the quasi-static approximation. It is also worth noting that the results in Figures 2 and 3 take into account the effects of the CPW feedline, as discussed in section II. The feedline will slightly reduce the effective slot circumference, thereby increasing the resonant frequency. Finally, the plots in Figure 4 contrast Z,,lf for a folded-slot (design C) and a comparable slot antenna. The peak in Re(Z,,f ) at the first resonance for the folded-slot is four times smaller than for the slot, which is the dual of the relationship between folded-dipoles and standard dipole antennas [8]. The self impedance of each antenna at its respective second resonance, however, is 16 Ql. The mutual coupling between two or more folded-slot antennas is useful information for designing arrays. Shown in Figures 5-7 is the mutual impedance, Z12, between two folded-slots of designs A-C, respectively, at the first resonance. The antennas are in the broadside configuration with each input port positioned as shown in Figure 1, and the separation distance S is measured between the center of the folded slots. As opposed to the self impedance data, these calculations do not include the CPW feedlines. Ideally, the CPW lines do not radiate, however, such that negligible differences are incurred by their omission. By comparing the plots in Figure 7 to those in Figure 8 for the comparable slot antenna pair, it is clear that the impedance coupling between the slots is four times as strong as the coupling between folded-slot antennas, as found above for ZeIf. The impedance coupling between two folded-slots of design C at the second resonance (Circumference = 0.47A0) is shown in Figure 9. In this case, the coupling is nearly the same as for the slot antenna at its second resonance (La = 0.33Ao). IV. DOUBLE FOLDED-SLOT ANTENNA The design of the double folded-slot antenna is primarily based on two parameters: the foldedslot elements and the connecting line between them. General guidelines for each of these are given below, and typical performance characteristics of the designs given in Table 2 are discussed in the following sections. In what follows, the frequency at which the E- and H-plane patterns are peaked at broadside and the two sidelobes in the E-plane are symmetric is defined as the center frequency, f,. IV.1 Design Guidelines - Folded-Slot Element - The total circumference, measured through the center of the slots, should be approximately equal to 1.0-1.04 Ag, where Ag is the guided wavelength of the prop 4

agating CPW mode at fc. In computing this wavelength, the same quasi-static approximation used in Eqn. 4 has been applied. Widening the slot width Wa2 decreases the center frequency and also tends to increase the reactive component of Zrn around fi. If the slot width is too narrow, however, there tends to be greater variation in Zi,, versus frequency. Using a width for Wa2 around Ag/60 is a good compromise between the two effects. - Connecting Line - The separation between the two elements must provide a 180~ phase shift for the guided CPW mode and the radiating space wave, and this may be accomplished using a meander line such as that shown in Figure 10. In this design the path length through the connecting line (2Lc1 + Lc2 + 2Lc3) must be approximately equal to 0.5 Ag and the physical distance (2Lc1 + L,2) must be approximately equal to 0.5 Aant, where Aant is the wavelength in the dielectric half-space. The factor for each length, as derived from the full-wave analysis, ranges from 0.503-0.51 (versus 0.5), and increases as the slot width Wa2 narrows. A similar design with a symmetric meander on each side of the center line could also be used, however the single-meander version is adequate and does not cause significant slotline mode excitation, since each bend is complimentary to a bend in the opposite direction. Also, note that a comparable end-fed double slot antenna requires a 360~ element separation in order to feed each antenna in phase. The DFS geometry, therefore, is much more compact. IV.2 Radiation Patterns Since each of the slots is equivalent to a magnetic dipole, the E-plane for the double folded-slot antenna is parallel to the feedline and normal to the plane of the antenna. With the correct phasing between the elements, the main beam points at 0 = 90~, i. e. broadside. The H-plane is parallel to the slots and has one lobe that is symmetric about 4 = 90~, which is simply the pattern for a single magnetic dipole. In order to demonstrate typical characterisitcs, the E- and H-plane patterns for design B are shown in Figures 11 and 12. Due to limitations in the experimental setup, the measured patterns are only strictly valid from about 30~ to 150~. It can be observed that a deeper null exists in the E-plane toward 0 = 180~, and also that the sidelobe facing the feedline rises quickly as the frequency increases. Without the additional beam shaping which is provided by a dielectric lens, the half-power beamwidths of the DFS antenna are approximately 50~ in the E-plane and 76~ in the H-plane. For the air substrate geometry, standard approximations (e.g. [8], pg. 32) predict a directivity of about 7 dB from these beam sizes, and the moment method results fall into the range of 7+.25 dB. On the quartz and silicon substrates, the directivity is greater since there is preferential radiation into the higher dielectric constant material [1]. Typically, patterns with the main lobe near broadside and sidelobes below -12 dB are possible over 8-10% bandwidths. 5

IV.3 Input Impedance The input impedances for the five geometries given in Table 2 are presented in Figures 13-16. By comparing these results to the impedance of a single slot or a single folded-slot, it is seen that the general characteristics around the first resonance are similar. For the DFS, however, the resonance is shifted down in frequency relative to the other two types of antennas, occurring when the slot circumference is only around 0.8 AV. Consequently, the center frequency, which is given in the last column of Table 2, occurs in a band where the impedance is slowly varying. This leads to the broad-band nature of the double folded-slot antenna. The results for the air-substrate designs A and B are shown in Figures 13 and 14, respectively. The impedance is between 55-75 Ql, with an imaginary part less than ~jlO Q2, over a 20-25% bandwidth. The discrepancy between the experimental and calculated values is primarily due to interference from the (large) flange mount coaxial connector which is 2 cm from the antenna. Low sidelobe levels exist at the same frequencies at which the agreement is very good, but the sidelobes illuminate the connector outside the center frequency band (see Figure 11). In Figure 15, calculated values for the low frequency designs on quartz and silicon are shown. As expected, the impedance is scaled down by approximately (Er,eff) relative to the levels for the air substrate, where Er,eff is the effective dielectric constant for the propagating CPW mode. Finally, the characteristics of a higher frequency design on silicon are given in Figure 16, in order to validate the full-wave analysis when the meander line is included. The parameters for this geometry differ from the design rules stated in Section IV.1, however this was the only measured data available to the authors. In general, the agreement between the calculated and experimental values is good, with the exception of the ripple which is present in the measurements. In this case, the additional structure can be explained by reflections at the interface between free-space and the Stycast lens (Er = 12) that was attached to the substrate in order to simulate the semi-infinite dielectric half-space. A matching layer could be used to minimize this interference. V. CONCLUSION This paper has presented a broad investigation into the characteristics of single and double folded-slot antennas on semi-infinite substrates. These antennas are naturally suitable for coplanar waveguide feedline configurations and are therefore easily integrated with active devices. The double folded-slot antenna is a compact planar geometry which measures approximately one half of a guided wavelength on each side. It provides a broad-band input impedance and a directivity of at least 7 dB, depending on the substrate. Also included in the paper was the outline of a rigorous full-wave analysis that was used to characterize the antennas, along with design guidelines for DFS antennas on air, quartz and silicon substrates. 6

ACKNOWLEDGEMENTS This work was supported by the Office of Naval Research and the NASA Center for Space Terahertz Technology. The authors would also like to thank Steve Mollenkopf for his help in making the measurements. APPENDIX This appendix provides an outline of the method which was used to compute the self-cell term of the admittance elements in Eqn. 3. It was shown in [6] that the co-polarized term, Yth, takes the following form in each dielectric half-space above and below the plane of the antenna metallization: C = 2jY [k2ACC - B(] G(flf') (5) where Yi is the intrinsic admittance of the dielectric medium, k is the wave number and A = Jds' Jds T,(C', ')T,(/) Be = Jds'JdsW(~', )To(() e-jkr G(f/f') = 4irr r = /(- c)2+ ( )2 (6) where T((, ) denotes the derivative of T with respect to (. The variables To and T, represent the testing and expansion functions, respectively, and the corresponding unprimed and primed coordinates refer to the domain of the 'observation' and 'source' cells. Due to the form of the PWS functions shown in Eqn. 2, YC( is a summation of four sub-contributions which result from the interactions of the two sides of the expansion and test functions. Using the standard technique of adding and subtracting the singularity, each of the sub-contributions can be rewritten using this general form: I = fds' ds co(km + a) [cos(kr) -1 jsin(kr)] +/3 ds' ds cos(kC + (7) where a and / are independent of the observation coordinates C and f The first integral on the right-hand side of Eqn. 7, II, is well-behaved in the limit as r -+ 0 and is therefore easy to handle numerically. It is the second integral, 12, for which an analytical solution is required, however this is not possible without some type of approximation to the integrand. In this case, a second order expansion of the cosine function has been used which puts 12 in the following form: 12 I IJd(2Jd - ((l,0 - (Ck )) (8) r` 7

For typical cell sizes, this approximation introduces less than 0.3% error. Furthermore, using relations from [9], a closed form solution for the interior integral may now be obtained: I2 |/ds' [[( - ( + km)) {n( + r) + ln( + r)} 2 km (Q + 1km)) {r + C2 ln( + r)} (k2 )J r + ln( +r) _ 3n( + r)] (9) -'2 6 3 6 f ~ where ( = c-~/ r = 2 +F2 (10) The evaluation of the cross-polarization admittance elements, Yet and Yte, can be handled in a similar manner. 8

REFERENCES [1] M. Kominami, D. M. Pozar and D. H. Schaubert, "Dipole and Slot Elements and Arrays on Semi-Infinite Substrate," IEEE Trans. Antennas Propagat., vol. AP-33, no. 6, June 1985, pp. 600-607. [2] S. Kawasaki and T. Itoh, "40 GHz Quasi-Optical Second Harmonic Spatial Power Combiner Using FETs and Slots," 1992 IEEE MTT-S Digest, pp. 1543-1546. [3] S. V. Robertson, N. I. Dib, G. Yang, and L. P. Katehi, "A Folded-Slot Antenna for Planar Quasi-Optical Mixer Applications," 1993 IEEE AP-S Digest. [4] H. S. Tsai, M. J. W. Rodwell and R. A. York, "Planar Amplifier Array with Improved Bandwidth Using Folded-Slots," IEEE Microwave and Guided Wave Letters, vol. 4, no. 4, April 1994, pp. 112-114. [5] X. H. Yang and W. X. Zhang, "Coplanar Waveguide Antenna Arrays for MIC/MMIC at Millimetre Wave Frequencies," Electronics Letters, vol. 26, no. 18, Aug. 1990, pp. 1464-1465. [6] K. Barkleshli and J. L. Volakis, "Electromagnetic Scattering from an Aperture Formed by a Rectangular Cavity Recessed in a Ground Plane, " Journal of Electromagnetic Waves and Applications, vol. 5, no. 7, pp. 715-734, 1991. [7] N. I. Dib, "Theoretical Characterization of Coplanar Waveguide Transmission Lines and Discontinuities," Ph.D. Thesis, Radiation Laboratory, University of Michigan, 1992. [8] C. A. Balanis, Antenna Theory - Analysis and Design, New York: John Wiley & Sons, Inc., ch. 2-3 and 8, 1982. [9] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products, San Diego, CA: Academic Press, Inc., 1965. [10] T. M. Weller, L. P. Katehi, and G. M. Rebeiz, "High Performance Microshield Line Components," To be published in IEEE Trans. on MTT. [11] S. M. Mollenkopf, personal communication, May 1994. 9

List of Figures 1 Single folded-slot antenna on a dielectric half-space (not to scale). The cross-hatched areas indicate metallization................................ 12 2 Calculated self impedances for folded-slot designs A, B, and C from Table 1........ 13 3 Resonant circumference of a folded-slot antenna versus the dielectric constant of the semi-infinite substrate. The antenna dimensions are Wal/La = 0.01, Wa2/La = 0.02, La = 10 mm, and Sa is given in the figure. The results include the effects of the CPW feedline with dimensions Wf = 0.1 mm and Sf = 0.4 mm............ 13 4 Calculated self impedances for folded-slot design C from Table 1 and a comparable slot antenna. The dimensions of the slot are Wa/La = 0.02 azd La = 10 mm..... 14 5 Calculated mutual impedance for two folded-slot antennas of design A from Table 1 versus separation distance. The antennas are aligned in a broadside configuration, and the computations are performed at the first resonance of a single folded-slot (Circumference = 0.97Ao)................................. 14 6 Calculated mutual impedance for two folded-slot antennas of design B from Table 1 versus separation distance (broadside configuration). The computations are performed at the first resonance of a single folded-slot (Circumference = 0.62Ao).. 15 7 Calculated mutual impedance for two folded-slot antennas of design C from Table 1 versus separation distance (broadside configuration). The computations are performed at the first resonance of a single folded-slot (Circumference = 0.40Ao).. 15 8 Calculated mutual impedance for two slot antennas on a silicon substrate (Er = 11.7) versus separation distance, where the antenna dimensions are Wa/La = 0.02 and La = 10 mm. The antennas are aligned in a broadside configuration, and the computations are performed at the first resonance of a single slot (La = 0.19Ao)... 16 9 Calculated mutual impedance for two folded-slot antennas of design C from Table 1 versus separation distance (broadside configuration). The computations are performed at the second resonance of a single folded-slot (Circumference = 0.47Ao). 16 10 Double folded-slot antenna on a dielectric half-space (not to scale). The E-plane pattern is in a plane parallel with the feedline and normal to the page, at 4 = 90~ and 0 = 0-180~; the feedline is at 0 = 180~....................... 17 11 E-plane patterns for the double folded-slot antenna design B, at three different frequencies............................................ 18 12 E- and H-plane patterns for the double folded-slot antenna design B at 5.7 GHz... 18 13 Calculated and measured input impedance for design A from Table 2......... 19 14 Calculated and measured input impedance for design B from Table 2......... 19 10

15 Calculated input impedance for designs C and D from Table 2............... 20 16 Calculated and measured input impedance for design E from Table 2. The measured data is from [11]...................................... 20 11,

TOP VIEW Wf Sf -^ CPW FEEDLINE END VIEW END VIEW 1 'r^^ 1 DIELECTRIC Y a I X INFINITY Figure 1: Single folded-slot antenna on a dielectric half-space (not to scale). The cross-hatched areas indicate metallization. Table 1: Parameters for different folded-slot antenna designs (refer to Figure 1). tr refers to the dielectric constant of the lower half-space and all dimensions are in millimeters. 12

125. 100. -- A-( =1.0) i; B-(E=4.0) i a \! z uJ U C-T cn 75. 50. t 25. ~-~ C-(Er=11*7) ~ *1 *.. I I' I~ I\ - - I I = I I I *1 I 1 II I I. I 0 0 I I /f 0. -25. -50. - - - I - - - - - - - - 2.5 7.5 12.5 17.5 22.5 FREQUENCY [GHz] Figure 2: Calculated self impedances for folded-slot designs A, B, and C from Table 1. 1.0, -.8.6 *A —0.04. SA/LA=0.04 -- -- SA/L,-O.O1, 2. I - I. I. I - 1. 3. 5. 7. 9. 11. 13. DIELECTRIC CONSTANT (et) Figure 3: Resonant circumference of a folded-slot antenna versus the dielectric constant of the semi-infinite substrate. The antenna dimensions are Wai/La = 0.01, Wa2/La = 0.02, L. = 10 mm, and Sa is given in the figure. The results include the effects of the CPW feedline with dimensions Wf = 0.1 mm and Sf = 0.4 mm. 13

a U IL 180. 150. 120. 90. 60. 30. 0. -30.5 -60. - 2.5 4.5 6.5 8.5 10.5 12.5 FREQUENCY [GHz] Figure 4: Calculated self impedances for folded-slot design C comparable slot antenna. The dimensions of the slot are Wa/La (Er = 11.7) from Table 1 and a = 0.02 and La = 10 mm. C: z LU z 60. 40. 20. 0. -20. -40. -60..0.5 1.0 1.5 2.0 25 3.0 3.5 4.0 SEPARATION [S/XA Figure 5: Calculated mutual impedance for two folded-slot antennas of design A (c, = 1.0) from Table 1 versus separation distance. The antennas are aligned in a broadside configuration, and the computations are performed at the first resonance of a single folded-slot (Circumference = 0.97A,). 14

- CL: u z In w I^3 30. 20. 10. 0. -10. -20. -30..0.5 1.0 1.5 2.0 2.5 SEPARATION [S/XA] Figure 6: Calculated mutual impedance for two folded-slot antennas of design B (Er = 4.0) from Table 1 versus separation distance (broadside configuration). The computations are performed at the first resonance of a single folded-slot (Circumference = 0.62Ao). 15. 15..-.. ----........,....,. - RE[Z12] -10. - -15. --—,..-.00.25.50.75 1.00 1.25 1.50 SEPARATION [S/XJ Figure 7: Calculated mutual impedance for two folded-slot antennas of design C (Er = 11.7) from Table 1 versus separation distance (broadside configuration). The computations are performed at the first resonance of a single folded-slot (Circumference = 0.40Ao). 15

60. a U z 0 40. 20. 0. -20. -40. -60..00.25.50.75 1.00 1.25 1.50 SEPARATION [S/AJ Figure 8: Calculated mutual impedance for two slot antennas on a silicon substrate (cE = 11.7) versus separation distance, where the antenna dimensions are Wa/La = 0.02 and La = 10 mm. The antennas are aligned in a broadside configuration, and the computations are performed at the first resonance of a single slot (La = 0.19Ao). 7.50....5.00 RE[Z12].00! -5.00 -7.50 -7.50,......00.25.50.75 1.00 1.25 1.50 SEPARATION [S/IX Figure 9: Calculated mutual impedance for two folded-slot antennas of design C (Er = 11.7) from Table 1 versus separation distance (broadside configuration). The computations are performed at the second resonance of a single folded-slot (Circumference = 0.47Ao). 16

TOP VIEW WI f A-.- z Ir f 1 XL/i CPW FEEDI END VIEW LINE FOLDED SLOT DIELECTRIO C C: * x INFINITY Figure 10: Double folded-slot antenna on a dielectric half-space (not to scale). The E-plane pattern is in a plane parallel with the feedline and normal to the page, at 4 = 90~ and 0 = 0-180~; the feedline is at 0 = 180~. Design E, Wf /Sf Wal /W2 /Sa/La. W / S / Lc / L2 / LC3 fc A 1.0.100/3.30.415/.415/.415/25.0.415/.415/12.6/0.00/0.00 6.06 B 1.0.100/3.30.415/1.66/.415/25.0.415/.415/12.6/0.00/0.00 5.68 C 4.0.264/1.58.264/.528/.264/15.8.264/.264/4.17/4.17/1.65 6.05 D 11.7.165/.495.165/.330/.165/9.92.165/.165/2.44/2.44/1.30 6.05 E 11.7.060/.030.040/.040/.150/2.55.040/.070/.455/.700/.545 23.8 Table 2: Parameters for several double folded-slot antenna designs (refer to Figure 10). ~r refers to the dielectric constant of the lower half-space. All dimensions are in millimeters and the center frequency, fc, is in GHz. 17

5. z S-. 0 w 0. -5. -10. -15. -20. _1s; 0. 30. 60. 90. 120. 150. 180. THETA [DEG] - E PLANE Figure 11: E-plane patterns for the double folded-slot antenna design B, at three different frequencies. 5. rl" 0 E — m.) 0. -5. -10. -15. -20. -25. Figure 12: E- and H L-._ 0. 30. 60. 90. 120. 150. 180. THETA [DEG] PHI [DEG] plane patterns for the double folded-slot antenna design B at 5.7 GHz. 18

- a - z2 200. 150. 100. 50. 0. -50. -100. 4.0 4.5 5.0 5.5 6.0 6.5 7.0 FREQUENCY [GHz] Figure 13: Calculated and measured input impedance for design A from Table 2. PI-" U w H & Eh Po z 0-4 200. 150. 100. 50. 0. -50. -100. 4.0 4.5 5.0 5.5 6.0 6.5 7.0 FREQUENCY [GHz] Figure 14: Calculated and measured input impedance for design B from Table 2. 19

150. Eu z CQ M0. 100. 50. 0. -50. 4.0 4.5 5.0 5.5 6.0 6.5 7.0 FREQUENCY [GHz] Figure 15: Calculated input impedance for designs C and D from Table 2. 150. C: w21 u II wl A4 F-* 100. 50. 0. -50. L-m 10.0 15.0 20.0 25.0 30.0 FREQUENCY [GHz] Figure 16: Calculated and measured input impedance for design E from Table 2. The measured data is from [11]. 20

APPENDIX C APPENDIX C I Development of Novel Planar Transmission Lines for High-Frequency Passive Circuit

INTERNATIONAL SYMPOSIUN N SIGNALS, SYSTEMS AND ELECTR i:N!hi S PROCEEDINGS PARIS September 1-4, 19 'i CNAM 292, rue St-Martin 75741 PA'RIS Cedex 03

Theoretical and Experimental Characterization of Microshield Circuits Linda P.B. Katehi, Nihad I. Dib, and Rhonda F. Drayton University of Michigan, 3228 EECS Building, Ann Arbor, MI 48109-2122, USA ABS "RACT A theoretical and experimental study of the microshield line, a recently proposed new type of monolithic planar transmission line, is presented. The Space Domain Integral Equation method is used to analyze a specific discontinuity of the proposed line. It is shown that microshield line discontinuities tend to radiate less than the conventional coplanar waveguide (CPW) ones. In order to validate the theory, measurements are performed on scaled models in the range 10-18 GHz. The experimental and theoretical results are in good agreement. Moreover, it is shown that the microshield line can provide a wide range of characteristic impedances due to many available parameters of design. 1 INTRODUCTION The microshield line, a new monolithic structure which can provide generic shapes appropriate for circuit and array applications, was proposed in [1]. This new geometry is characterized by low radiation loss [1], reduced electromagnetic interference [2] and compatibility to antennas of microstrip or aperture type [3]. In this configuration, the ground plane is warped to totally or partially surround the inner strip conductor as a shielding microcavity. Such a structure can be made monolithically using dielectric etching and metal deposition techniques as will be described below (see Fig. 1). In these lines the in-:c conducLor can be printea on a Lmin substrate layer (Fig. 3(b)) or can be suspended in air by using membrane technology (Fig. 3(a)). As it has been demonstrated [4, 5], membranes can be effectively employed to provide high-efficiency monolithic antennas and arrays operating at very short wavelengths. One of the advantages of the microshield line is that there is no need for via-holes or air-bridges for ground equalization because of the existence of the shielding microcavity. In addition, a wide range of characteristic impedances can be obtained due to the many available Parameters in design. Moreover, preliminary theoretical an:.sis has shown that the microshield line discontinuities tend to radiate less than the CPW ones [1]. On the other hand, the presence of conductor losses is expected to limit the use of microshield lines and circuits to the lower end of the sub-mm-wave spectrum. 2 FABRICATION The fabrication of microshield circuits is dependent on the thin dielectric membrane technology and the anisotropic etching of the supporting wafers. A Si membrane is a 3-layer SiO2/SisN4/SiO2 structure which must be slightly in tension to yield flat and rigid self-supporting characteristics. The SiO2 layers are made by high temperature thermal oxidation (1100~), while the Si3N4 layer is made by chemical vapor deposition at 800~ [4]. After the development of the three-layer structure, the membranes are fabricated in two steps. First, an opening in the silicon-nitride layers is defined on the back of the wafer, and then the silicon is etched until a transparent membrane appears. The etching solution is anisotropic and yields pyramidal cavities which can be exploited for microshield transmission-lines. Figure 1 gives the most important steps of the procedure in a simplified way. Once the membranes are fabricated, it is easy to lithographically define several different microshield geometries for impedance and propagation constant measurements. The microshield geometry can create the basis of a new technology and can lead to novel ideas in antenna and array design. Figure 2 shows a picture of a low-pass microshield filter exciting an aperture type trapezoidal antenna. This circuit has been made as a proof of fabrication feasibility and it does not present an optimized design. In fact, in order to successfully design this antenna configuration many important issues associated with the microshield geometry have to be studied and understood. Some of these issues are (i) dispersion (ii) parasitic radiation, (iii) transitions from conventional planar lines to microshield structures and (iv) electromagnetic coupling to neighboring circuit elements.

Si or GaAs Waffer Step 1 Step 2 I p. Membrane __......... i a.... Etch Metallize l — Step 3 7 I Metallize 3 THEORY The theoretical nmethod is based on a space domain integral equation (SDIE) method [1, 6]. The original boundary problem is split into two simpler ones by introducing an equivalent magnetic current M, on the slot aperture. This surface magnetic current radiates an electromagnetic field in the two regions, above and below the slots, so that the continuity of the tangential electric field on the surface of the slots is satisfied. The remaining boundary condition to be applied is the continuity of the tangential magnetic field on the surface of the slot apertures which leads to the desired integral equation. This integral equation is solved numerically using the method of moments where the unknown magnetic current is expanded in terms of appropriate basis functions. Then, Galerkin's method is applied to reduce this equation to a linear system of equations which is solved to provide the field distribution on the membrane slots. From the field distribution, the scattering parameters and radiation loss from the discontinuity under consideration are computed as functions of frequency using standard techniques [6]. The details of the SDIE method may be found in [1, 6]. 4 SCALED MODEL MEASUREMENTS Although the microshield line is intended for sub-mm wave applications, its experimental characterization can be best performed on scaled models. This is due to the fact that 2-port scattering parameters measurements are limited to 100 GHz from the state of the art vector systems. In addition, since the microshield line is a passive structure, most of the information obtained from the scattering parameter measurements, excluding conductor losses, can be scaled with frequency to the sub-mm wave region. In these models, the membrane of the microshield line has been replaced by a 30 mils duroid layer (e, = 2.2) as shown in Figure 3(b). In this configuration, the strip conductor is printed on this duroid layer and is shielded by a cavity milled on a metal block. The RF measurements were performed in the 10-18 GHz frequency range on the HP 8510B Automatic Network Analyzer. To determine the circuit performance, a one-tier Thru-Reflect-Line (TRL) calibration has been performed to eliminate the effects of the connector discontinuity from the measured data. 5 RESULTS The theoretical technique presented previously has been applied to a number of membrane and dielectric microshield and conventional coplanar discontinuities in order to compare circuit performance. The numerical results Step 4 Shield the back of the Cavity Figure 1: Fabrication process for the microshield line. Figure 2: A membrane microshield line low-pass

Membrane Conductor 1 / h a -— > I (a) -5 -8 Dielectric I Conductor / I In V-d cA c -11 -14 -17 - Dieecric Microhield Membrane Miroshield CPW '...... --- — (b) I h I _ 11'....... 10 11 12 13 14 15 16 17 11 <- a - Figure 3: (a) A simplified cross section of the membrane microshield line. (b) The dielectric microshield line suitable for scaled measurements. of this study have been presented in [1] and provide very strong indications about the exciting low-loss performance of microshield structures. In this paper, the same methodology is applied to similar microshield discontinuities and the theoretical results in the form of scattering parameters are compared to experimental ones performed on scaled models. The specific discontinuity discussed here is the open-end coupled microshield line configuration with an inner conductor width of 60 mils, slot-width of 20 mils and microcavity width and height given by 100 mils and 50 mils, respectively. Figure 4 shows the radiation loss factor of such a discontinuity in the microshield line and is compared to that corresponding to conventional CPW. It can be seen that radiation from the coupled microshield openends is approximately one half the radiation from coupled open-ends made of conventional CPW. In addition, for this specific discontinuity with the above dimensions, the dielectric microshield line and the membrane microshield line have almost the same radiation loss factor. For the same configuration, Figure 5 shows the scattering parameters and the loss factor as were evaluated theoretically and measured experimentally. The comparison between theory and experiment is very good. The ripple seen in the measured data is mainly due to the very small coupling between the two lines which intensifies the effect Frequency (GHz) Figure 4: Radiation Loss factor of coupled open-ends di continuity (S1=120 mils, g=20 mils). of connector repeatability. Other results on scaled mode will be shown in the Symposium. In addition to the scale models, a variety of membrane microshield discontinuiti, have been fabricated and measurements of their scatterir parameters using a Cascade probe station are underwa Results from these measurements will be presented in t] Symposium. Figure 6 gives the characteristic impedance of a rr croshield line as a function of S/(S+2*W) for a given ca ity size. This characteristic impedance has been obtain( using a quasi-TEM analysis based on the Point MatchiI Technique [7]. Recently, a closed form expression of tl characteristic impedance of the microshield line has bet derived using Conformal Mapping techniques [2], whir will be presented in the symposium too. 6 ACKNOWLEDGMENTS This work has been supported by the Office of Naval R search under contract No. N00014-92-J-1070 and by tl NASA Center for Space Terahertz Technology. The ai thors would like to thank Prof. G. Rebeiz for his scientii support. References [1] N. Dib, W. Ilarokopus. L. Katehi, C. Ling, G. R

beiz, "Study of a Novel Planar Transmission Line", 1991 IEEE MTT-S International Microwave Symposium Digest. pp. 623-626, June 1991. [2] N. Dib and L. Katehi, "Impedance Calculation for the Microshield Line, " submitted to the IEEE Microwave and Guided Wave Letters, April 1992. [3] L. Rexberg, N. Dib and L. Katehi, "A Microshield Line Loop Antenna for Sub-mm Wavelength Applications, " accepted for presentation in the 1992 AP-S Symposium, Chicago. [4] G. Rebeiz, D. Kasilingam, P. Stimson, Y. Guo and D. Rutledge, "Monolithic Millimeter-Wave TwoDimensional Horn Imaging Arrays," IEEE Trans. on Antennas and Propagation, pp. 1473-1482, Sept. 1990. [5] S. Gearhart, C. Ling, G. Rebeiz, H. Davee and G. Chin, "Integrated 119 pm Linear Corner-Cube Array, " IEEE Microwave and Guided Wave Letters, pp. 155-157, July 1991. [6] N. Dib, P. Katehi, G, Ponchak, and R. Simons, "Theoretical and Experimental Characterization of Coplanar Waveguide Discontinuities for Filter Applications," IEEE Trans. on Microwave Theory and Techniques, pp.873-882, May 1991. [7] D. Rowe and B. Lao, "Numerical Analysis of Shielded Coplanar Waveguides, " IEEE Trans. on Microwave Theory and Techniques, pp. 911-915, Nov. 1983. --- S.,,.th --- SL teory ------- L factor. thoor * S. expim * S, expriment Los factor, experiment.... I... I I I I I. * * * I. * I * I * * * * UI P~ 0~o ~ ~~000 0 -20 -40 —.....'....i... ~n r *.*....I.-;....-_... —... *.-. 0^ 0 0 0 '~a a a* *a. -- --...................i 10 11 12 13 14 15 16 17 18 Frequency (GHz) Figure 5: Scattering parameters and loss factor of coupled open-end dielectric microshield lines discontinuity (S1=120 mils, g=20 mils). ' 400 ' -- -- ----- — I I I 350 - S=20p 83 300 soo 250 \...... S300 pm 200 M soo '- "S=700? W.~ 150 S 100 "L~~~ --- —-- 50 0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 S/(S+2*W) Figure 6: Characteristic impedance of a membrane microshield line (a=1200 znm. h —290;m).

40th ARFTG CONFERENCE-DIGEST Fall 1992 * * I. -4 '. - il.-.1... - I -. - - * J- *'-I -.. - - - - ' 'i:.';'. ).; '.4 Ramada Magate.....-Orlando, Florida roe ~ ~ - A d d O F SMOROI.1 -7-Ily

MICROWAVE CHARACTERIZATION OF MICROSHIELD LINES Rhonda F. Drayton and Linda P.B. Katehi NASA Center for Space Terahertz Technology University of Michigan Ann Arbor, MI 48109-2122 ABSTRACT An experimental characterization of a new type of monolithic planar transmission line, the microshield, is presented. Theoretically, it is shown to have the tendency to radiate less than the conventional coplanar waveguide (CPW). To characterize the line, experimental measurement of a scattering parameters of the microwave model is essential. This paper presents characterization through the design, fabrication and measurement of the line and a discontinuity such as the two coupled open-end line. The frequency dependence of the effective dielectric constant is measured and compared to the computed values. Although proposed for applications at lower sub-mm-wave frequencies, fabrication in a hybrid fashion allows for use at microwave frequencies. Present applications of microshield technology are seen in Microtech Cascade probes and feed throughs for package connections in CPW interconnects. In addition, it is possible to use the microshield to connect MMIC's by designing the shield width to accommodate an MMIC. Therefore characterization of the microshield line and discontinuities is necessary. 1 INTRODUCTION The microshield is a novel monolithic planar transmission line that can provide generic shapes useful for circuit and array applications [1]. The geometry of the line exhibits lower radiation loss than the conventional coplanar waveguide (CPW) line, reduced electromagnetic interference [2] and compatibility to antennas of microstrip and aperture type [3]. Geometrically, the microshield can be made such that the ground plane totally or partially surrounds the inner conductor strip as a shielding microcavity (Figl). This structure is made monolithically by incorporating dielectric etching and metal deposition techniques as well as thin film membrane technology. Since the dielectric substrate can be reduced to a micron, which is significantly smaller than a wavelength, a pure TEM mode is possible. Consequently, these lines are designed to operate in the lower sub-mm-wave region of the spectrum (frequencies up to 500 GHz). IMembranel Center Conductorl i 1Microcavit Figure 1: Microshield Transmission Line 171

Experimentally, characterization is obtained through a scaled model of the microshield line due to limitations in state of the art vector measurement systems to 100 GHz. Since the line is passive and linear, most characteristics, excluding conductor losses, can be obtained from the scattering parameter measurements which can be scaled with frequency to the lower sub-mmwave region. These conductor losses which can be obtained from other studies on the CPW lines are expected to offer limitations in the use of the microshield lines for circuit applications. As an advantage, the necessity for airbriges or via holes is eliminated since the center conducting line is in such close proximity to the microcavity for which the ground planes are tied together. In addition, a wide variety of characteristic impedances are obtained through variations in the design parameters such as microcavity height and width as well as the conductor and slot width. Finally, radiation losses are significantly lower than those of the conventional CPW or microstrip due to the microcavity enclosure. Although the primary focus is on the characterization of the microshield for lower submm-wave applications, at microwave frequencies a hybrid form of the microshield can be observed in Microtch Cascade probes and feed through packaging using CPW interconnects. In addition, it is possible to use the microshield for MMIC excitation by designing the shield width to accomodate the device. Therefore, characterization of the microshield line and discontinuities is necessary. 2 MICROWAVE MODEL The fabrication of the microshield circuits is dependent on thin dielectric membrane technology [4] as well as anisotropic etching techniques. The structure consists of an inner strip conductor printed On thin substrate layer. This conductor is then shielded by the microcavity to form the microshield line. The scaled nicrowave model is designed for operation in the frequency range of 10-15 GHz (Fig 2). This microwave model requires the design of a test fixture and circuits that effectively simulate geometrical characteristics of the microshield transmission line. Therefore, the microwave model looks very similar to the inverted microstrip line. Figure 2: Microwave Model of the Microshield Transmission Line 172

The design criteria for the fixture requires matching of the fields at the discontinuity between coaxial excitation and transmission line. Thus, the resulting characterization is a tri-fold process. First, various transitions are evaluated to provide maximum excitation to the line. Second, the circuits are evaluated to determine the minimum thickness and dielectric constant of the substrate which are feasible for accurate characterization of the line. Thirdly, the results of the above processes are combined to evaluate the overall performance of the line. The test fixture is designed so that a cavity encloses the conducting line. A cavity is machine milled into an aluminum block and then a conducting line, which is printed on a substrate, is center aligned over the cavity. Center positioning of the conducting line ensures excitation of the CPW mode and suppression of the slot mode in the line. Once alignment is obtained, the circuit substrate backing is supported by a PVC block on its edges to offer mechanical strength to the substrate surface and support for the coaxial connectors. After mounting the circuit, the connectors are placed on the circuit and secured for measurement 3 CIRCUIT DESIGN AND FABRICATION CONSIDERATIONS The circuits are designed for 10-15 GHz on duroid, er = 2.2, with substrate thickness of 30 mils. While thinner substrates have been tested, flexibility in the dielectric backing resulted in poor ground plane contact This leads to leakage and poor measurement repeatability. For improvement, ground planes are included on the circuit surface to ensure contact of the cavity ground to the CPW ground. While this enhanced the performance, it is observed that a substrate thickness of 30 mils provides the most repeatable measurements for characterization. To ensure excitation of the CPW mode, the ground planes of the CPW line are extended inside the cavity. Connection between the coaxial line and the CPW line are obtained by field matching rather than impedance matching. It is observed that the CPW line of 72 ohms matches optimally to coaxial line of 50 ohms [5]. Implementing the above considerations results in a fixture that is easily characterized during calibration. The circuits are fabricated using standard photolithography and wet etching techniques. Due to the frequency range of interest, conventional wet etching techniques are sufficient given circuit dimensions. 4 MEASUREMENT CONSIDERATIONS Characterization of the fixture and discontinuities are obtained through the Thru-ReflectLine (TRL) calibration technique which eliminates the effects of the connector discontinuities from the measured data. Since precision calibration standards of the microshield microwave model are not available for the ANA, a two-tier de-embedding technique is utilized to characterize the line[5]. The two-tier de-embedding technique consists of calibrating the ANA using precision coaxial standards and then performing a calibration of the test fixture using the calibration standards designed for the line. For field matching, the transmission line is terminated with linear tapers. Since the cavity structure is the basis for mechanical support in the fixture design, coaxial excitation was obtained using Omni Spectra flange mount tabs. These connectors have a Teflon diameter of 187 mils and a tab width of 60 mils. The connectors are modified to include ground connections for the CPW line and are mounted to the circuit via the test fixture. Proper contact is maintained by using silver conducting paint. Experimentally, the propagation constant of the line is obtained from the measurement of a long delay line (Fig 3). For the characterization of discontinuity effects, the two-coupled open 173

end line is measured. The scattering parameters are shown in Fig 4 and Fig 5 while the radiation loss factor is shown in Fig 6. The comparison between theory and experiment is very good. The ripple seen in the measured data is mainly due to the very small coupling between the two slot lines which intensifies the effect of connector repeatability. 1.6 - 1.5 1A 4 w 13 " Theorctical.. c, f.4 - Exp mt -- IExperimeet I *. I.. i.. I.. 1.2 L1 1...II I I I I I 10 11 12 13 Frequency, GHz 14 IS Figure 3: Effective Dielectric Constant of the Microshield Microwave Model -I 0 'U —, -10 a ID 2 S.0 Sll - Tbeory -- Experiment ~. ~. I. -30 -40 I I I I 10 11 12 13 Frequency, GHz 14 15 Figure 4: Two Coupled Open End Line Scattering Parameter Magnitude (S1=20 mils and g = 10 mils) 174

180 90 in to V. w 4 0 -v Theory -Experiwmet I i i I -180 I I I I 10 11 12 13 Frequency, GHz 14 1s Figure 5: Two Coupled Open End Line Scattering Parameter Phase (SI=2Onis and g= l0 mils) ID' I. 0 "N ci CZ..010 -20 -30 I 0 Theory Experiment... I.. I.. I.. I I I 0 -40 10 11 12 13 Frequency, GHz 14 15 (Si=2Omis and g= l1niiS) 5 CONCLUSION 17 5

A preliminary characterization is shown for the microwave model of the microshield transmission line. Circuit design and fabrication procedures have been presented. In addition, the experimental setup to measure the scattering parameters has been described. The results of this prelimenary study indeed show less radiation than the conventional coplanar waveguide (CPW) lines in range of 10-15 GHz. 6 ACKNOWLEDGEMENTS The authors thank Prof. Gabriel Rebeiz and Mr. George Ponchak for their scientific support. This work has been supported by the Office of Navel Research (Contract No. 029226) and the NASA Center for Space Terahertz Technology. REFERENCES [1] N. Dib, W. Harokopus, L. Katehi, C. Ling, G. Rebeiz, "Study of a Novel PlanarTransmission Line", 1991 IEEE MTT-S International Microwave Symposium Digest,, pp. 623-626, June 1991. [2] N. Dib and L. Katehi, "Impedance Calculation for the Microshield Line," submitted to the IEEE Microwave and Guided Wave Letters, April 1992. [3] L. Rexberg, N. Dib and L. Katehi, "A Microwhield Line Loop antenna for Sub-mm Wavelength Applicatons, " accepted for presentation in the 1992 AP-S Symposium, Chicago, n. [4] G. Rebeiz, D. Iasilingam, Y. Guo, P. Stimson and D. Rutedge, "Monolithic MillimeterWave Two-Di ensional Horn Imaging Arrays, " IEEE Trans. on Antennas and Propagation, pp. 1473-1482, Sept. 1990. [5] R. Simons and G. Ponchak, " Modeling of Some Coplanar Waveguide Discontinuities," IEEE Trans. on Microwave Theory and Techniques, Vol. 36, pp. 1796-1803, Dec. 1989. 176

ASI\ FOURTH INTERNATIONAL SYMPOSIUM ON | uC _ SPACE TERAHERTZ TECHNOLOGY IEEE The University of Michigan SYMPOSIUM PROCEEDINGS Royce Hall, UCLA March 30-April 1, 1993 UCLA Los Angeles, California Sponsored by: NASA Office of Advanced Concepts and Technology (OACT), University Space Engineering Research Centers Program, with cooperative sponsorship by the Microwave Theory and Techniques Society of IEEE. Organized Jointly by: The University of Michigan's NASA Center for Space Terahertz Technology and UCLA's Center for High Frequency Electronics.

Page 238 Fourth international Symposium on Space Terahertr Technology EXPERIMENTAL STUDY OF MICROMACHINED CIRCUITS Rhonda Franklin Drayton and Linda P.B. Katehi NASA Center for Space Terahertz Technology University of Michigan Ann Arbor, MI 48109-2122 1.0 Abstract Planarization of transmission lines such as microstip, stripline, and coplanar provide enhanced flexibility in array design, compatibility to active devices and radiating elements, as well as reduced volume and weight. However, many unwanted frequency dependent mechanisms are introduced from these planar lines which cause degradation in the electrical performance of the circuit. As a result, the development of new geometries which reduce the effect of such mechanisms as ohmic loss and electromagnetic interference without affecting the monolithic character will result in extended operating frequencies and improved electrical performance of circuits and arrays. This paper presents the development of micromachined circuits which provide shielding of conventional lines using micromachining techniques. Presented is a description of the fabrication and characterization of these circuits. In addition, the study will show information about the propagation characteristics of the line and scattering parameters for a variety of micromachined circuits with comparison to theoretical and conventional circuit geometries. 2.0 Introduction The thrust for the development of cost effective, lightweight electrical systems which exhibit improved electrical performance can be attributed to communications, commercial and military applications. One application in the area of communications is for the development of miniature satellites, microsats, which contribute to reduction in weight and improved fuel economy. Commercial applications will benefit from this development of high intelligence highway systems which require parallel processing of vast amounts of information transfer to maintain efficient operation of such systems. On the other hand, military applications will require development of PIN cards used for identifying military personel.Thus, the impetus for investigating micromachined circuits is due to the strong desire to develop inexpensive circuits that provide improved electrical performance with reduced volume and weight. Currently, monolithic quasi-planar geometries which exhibit electrical and geometrical resemblance to waveguide shielded transmission lines and circuits are possible to implement due to recent advances in semiconductor processing techniques. Silicon based monolithic waveguides have been successfully developed for use in the sub-millimeter wave frequencies using silicon

Fourth International Symposium on Space Terahertz Technology Page 239 micromachining techniques by JPL II and University of Nlichiganl21]. The waveguide is split along the broad wall and each half is tormed by etching a channel through an appropriately onented silicon wafer. Finally, a polyamide bonding technique is used to glue the etched channels to a smooth silicon wafer. In addition their performance has been measured and has been found very satisfactory[ 1]. Similarly, a shielded transmission line structure, Figure 1, can be developed by replacing the waveguide with a cavity which is etched into Si using KOH and then metallized using evaporation techniques. Then the planar transmission line is created by metal deposition on the Si semiconductor substrate and attached to the original wafer using electrobonding techniques. The results are lines and circuits which evolve from the conventional metallic waveguide and waveguide-shielded components with the advantages mentioned above. -- -. -.... FIGURE 1. Micromachined CPWV As a result, this study will present the development of micromachined circuit, which use an etched cavity as a shield to planar geometries. After describing the fabrication and experimental procedures needed to characterize these novel circuits, scattering parameter measurements will be shown. Finally, results regarding the propagating characteristics of the line and a variety of circuit geometries will be presented and compared to theoretical models as well as conventional geometries. 3.0 Fabrication Procedure In order to implement the development of a micromachined circuits, technology resulting from recent advances in semiconductor processes is used. These circuits are created by attaching the shielding structure to the planar line structure using electrobonding techniques. Although the shield and lines require the use of standard photolithography, the planar lines use a bilayer process technique for metal definition while the shield uses anisotropic wet etching methods for the development of the cavity structure. Initially, a Si wafer having a membrane (SiO2/Si3N4/Si02) dielectric surface is used to develop the cavity structure since depths of varying heights are required. Using photolithoeraphv to define the areas to be etched, the first oxide layer is etched using BHF while the nitride layer is etched using plasma etching. After repatterning the wafer to protect the areas to be etched later, the second oxide layer is removed while the remaining membrane layers serve as a mask for any unpatterened sections of wafer. Once the initial depth is obtained for the alignment marks, the remaining cavity areas are then exposed and etched to the required depth. The result is the development of a structure which has sections etched entirely through the wafer, the alignment marks,

Page 240 Fourth International Symposium on Space Terahertz Technology and sections etched partially through, the cavities. Once the etch is complete, the shielding portion of the micromachined circuit is metallized using evaporation techniques. Secondly, transmission lines are designed which consist of calibration standards and various circuit geometries necessary to characterize the electrical performance of the micromachined circuits. The lines, developed on a separate wafer, use photolithography to pattern circuit areas and the bilayer process to develop the appropriate line depth and width with the use of evaporation techniques. Lastly, the micromachined circuits are developed by aligning the cavity structure to the transmission line structure. Once aligned, they are secured using electrobonding methods. 4.0 Measurement Procedure Characterization of the circuits require the use of the following experimental arrangement: an HP 85 10 OB Network Analyzer modified to operate up to 40 GHz, an Alessi Probe Station, and Cascade Probe Tips. To reduce the effects of possible substrate modes in the high resistivity silicon substrate, the circuits are isolated from the wafer chuck by 1/8" thick 5880 duroid, er=2.2. To accurately measure the circuit performance of the micromachined circuits, the Thru-ReflectLine (TRL) calibration technique is used. In order to perform this calibration, standards were fabricated using the above process. A one tier de-embedding technique was used for on-wafer calibration which calibrates the system reference plane along the transmission line. As a result, the calibration takes into account all transitions between the ANA and the reference plane. The micromachined circuit, Figure 2, requires several transitions along the feed line to excite the micromachined circuit. The first transition is a taper between two 50 ohm sections of cpw-line. This transition is necessary to provide excitation to the desired line via the Cascade 150 gim pitch probes with minimal mismatch to the probes. The input port, having center conductor width of 100im and slot width of 60 pm, is transitioned into a line with center conductor width of 180 pm and slot width of 130 Agm. The second transition occurs between cpw and micromachined cpw circuit The micromachined circuit is of coplanar type line which includes a micromachined shielding cavity having height of 200 im and width of 800 rin. Subsequently, the required input line of all circuits measured will be of this type. ~.. "W1,",..-. —~,,.... 7t:.. ~ Il.. -." FIGURE 2. Micromachined circuit with various transitions

Fourth International Symposium on Space Terahertz Technology Page 241 5.0 Measurements and Discussion 5.1 Philosopy of thle Experimental Study In order to facilitate a complete preliminary study of micromachined circuits, it is necessary to compare the measured results to both theoretical models as well as conventional geometries. The theoretical model used for the micromachined circuits is based on the space domain integral equation (SDIE). With the appropriate Green's function for the lower half space and upper cavity region of the two dimensional circuit, unknown magnetic surface currents used to replace the slots, Ms, can be determined for a given electric source, Js, using the method of moments. Once Ms is known, Galerkin's method is applied to reduce the above integral equation to a linear system of equations.For a given field distribution, the scattering parameters of the equivalent network is computed using standard techniques. [3,4] The coplanar waveguide circuits, Figure 3b, are design to be identical in geometry to the micromachined planar circuits. Figure 3a. However, the design includes separate calibration standards for the cpw and micromachined circuits. The resulting measurement will therefore show the response of micromachined circuits as compared to the theoretical model and conventional coplanar waveguide. FIGURE 3. (a) Micromachined Circuits and (b) Open CPW Circuits 5.2 Measurements A through line is used to provide information about the propagating characteristics of the micromachined transmission line. Figure 4 below shows the effective dielectric of a line having length of 2088(m and dimensions as stated above. The results show good agreement between the measured micromachined circuit and the theoretical model, which is based on Full Wave Analysis[3]. The two circuit designs presented are an open end series stub in Figure 5a, designed to have gap spacing of 30 g~m, and open end gap in Figure 5b, designed to resonate at 30 GHz with resonate stub length of 1086pm

r-..AL -.- -n-A:.. c......- C~.,nrp Trnhrt T~echno)lou Page 242 tourtn inrernatwunui aymfpnaum unv ok' n.-..-... — 0 r 201 2* x 35I 3: Frequewy,GHz FIGURE 4. Effective Dielectric Constant and slot width of 30 inm. The scattering parameters for the above circuits have been measured and are compared to the theoretical model for the micromachined cpw circuit. In addition they are compared to conventional cpw circuits of the same type. (a) Open End Series Stub (b) Open End Gap FIGURE 5. (a) Open End Series Stub and (b) Open End Gap For the open end series stub in Figure Al.a, theory and experiment are in excellent agreement both in magnitude and phase as seen in Figures Al.b and A I.c. For the same circuit, comparison with the cpw line is shown in FiguresA2.a and A2.b. Thle line geometry is identical for both circuits, however, the micromachined circuit provides a shielding environment. In Figure A2.c, the total loss of this micromachined circuit varies between 6 dB and 2 dB at the high end below that of conventional cpw lines. The open end gap in Figure Bl.a, shows very close agreement in magnitude and phase with the theoretical results as seen in Figures Bl.b and Bl.c. As stated initially, it is very important to avoid distortion of the circuit performance when the shield is included. As a result of Figures B2.a and B2.b, comparison between the micromachined circuit and cpw show similarities in the overall circuit performance. The distinction between the lines occur when observing the total loss of the circuits. In Figure B2.c, the total loss, which is again reduced, indicates improvement in the overall performance of the circuit due to the micromachined cavity of the line.

Fourth International Symposiurn on Space Traertz Technology Page 243 To indicate the potentili tur this cn.:;luc\. i-LUIc () shows a geomeItrcai dejr:pi(onl ot a -milled c..:uit as compared to the new milcr);nlhincd circuits. 'lhe illed reprc.c:, nc cilrct:lt 'hilc the micromachined circuit represents 5 dit'erent circuits housed on thie silico1r,aik'cr thiat is about 1/6 of a 3" wafer..r;,..'.. ~ o. FIGURE 6. Milled Circuit and Micromachined Circuit In addition to the physical reduction in the size and volume, circuit geometries ot similar type indicate promise for applications at terahertz frequencies. However, of the two, the micromachined circuits offer the most flexibility in testing due to state of the art measurement systemsavailable namely the probe station. These circuits and system offer more repeatable measurements along with ease in the fabrication methods used. Figure 7 shows the loss performance of similar geometries between the micromachined and milled circuit. Since the circuits are passive in nature, losses, except ohmic, can be scaled appropriately to the terahertz frequencies to give reasonable indicators of expected circuit performance. w=1301m s=180~m g=30-m I l __ _ _ __ g=s1 =508.mn g sl| w=5081im - s=1,564sim... a 4 i -0::,,,,! U, 'S —,-. —. 10-10...20-.... 215 I.-.-I: 3 o0 -0 - -20 - -30 - - —,- IIIpenwmI r ~~- ThowrpJh - ti -, I I I -: I i t I, 10 15 20 Z5 30 35 40 Preqmtacy, on 10 11 12 13 14 Frequotcy, GHz 15 (a) (b) FIGURE 7. Losses of (a) Micromacnined circuit vs. (b) Milled circuit

Page 244 Fourth International Symposium on Space Terahertz Technology 6.0 Summary Micromachined circuits have been built and tested. The results show that the scattering parameter measurements are in very good agreement with the theoretical model used. The study also shows comparison between the cpw and micromachined cpw. The results indicate consistently that the shielded geometry results in reduction in the total loss of the various circuits measured. When compared to machine milled circuits of similar geometry, the losses are comparable. As a result, micromachined circuits show encouraging preliminary results which warrant further development of these lines for more complicated geometries. Acknowledgments This work was partially supported by the Office of Naval Research and the NASA Center of Space Terahertz Technology. The authors would also like to thank Nihad Dib for theoretical model contributions. References 1. M. Yap, Y-C Tai, W.R. McGrath and C. Walker, "Silicon Micromachined Waveguides for Millimeter and Submillimeter Wavelengths," presented in the 3rd International Symposium on Space Terahertz Technology, Ann Arbor, MI, March 1992. 2. Linda P.B. Katehi," Low-Loss Transmission Lines for Terahertz Frequency Applications," IEEE Proceedings, November 1992, pp.1771-1787. 3. N.I. Dib, P.B. Katehi, "Modeling of Shielded CPW Discontinuities Using the Space Domain Integral Equation Method (SDIE)," Journal of Electromagnetic Waves and Applications (JEWA), Vol. 5, No 4/ 5, pp. 503-523. 4. N.I. Dib, L.P.B. Katehi, G.E. Ponchak and R.N. Simons, "Theoretical and Experimental Characterization of Coplanar Waveguide Discontinuities for Filter Applications, IEEE Trans. on MTT, Vol. 39, No. 5., May 1991.

Fourth International Symposium on Space Terahertz Technology Page 245 Figures g = 30rm II w=13Lm - II s=180/m w = 60pn 1086gm (a) Physical Geometry 0 -4 AQ 12 2 -12 -16 -20 10 15 20 25 30 Frequo cy, GHz 35 40 S a! 6: (a) Magnitude of Scattering Parameter ISO Expnowww...- I -! I150 - so o ----- AnC 10 S1 20 25 30 35 Frnoecv. OH: (b) Phase of Scattering Parameter 40 FIGURE Al. Micromachined CPW, Theoretical vs. Experimental: Open End Series Stub

Page 246 Fourtn International Symposiu onSpace Terahertz Technology 122 RO 1: if Q b 0* 0" -4 a -12 -16 * 20 I. Mag (S 1 2) jia C 0 miem(311) } 4 I Cp 10 Is 20 25 30 Frvquancy. 0Hz 35 40 (a) Magnitude of Scattering Parameters a, 10 is 20 25 30 35 40 Fneusc~v. 0Hz (b) Phase of Scattering Parameters W w, 'a 0 -15 - -I -20 10 I5 20 25 30 Frequency, GHz 35 40 (c) Total loss FIGURE A2. Micromachined CPW vs. CPW LIne: Open End Series Stub

Fourth International Symposium on Space Terahertz Technology g=30nLm I I w=130gm - s= 180rn om (a) Circuit geometry Page 247 I I I - a w -I -1 -1 LO- My(SS) -s - - --- - - -- -— E —xpe nen --— l - -- I 4 —Torei |:,PC,Zo I I. - i i 10 15 20 25 30 35 Frquecy, GHz (b) Magnitude of scattering parameters 40 ~ 00 ns 50 S.-loo a -150 -200 10 15 20 25 30 35 40 FIrquency, OHt (c) Phase of scattering parameters FIGURE B1. Micromachined CPW, Theoretical vs. Experimental: Open End Gap

Page 248 Fourth International Symposium on Space Terahertz Technology c1 b I a 00 (a) Magnitude of Scattering Parameters 10!5 20 25 30 35 40 Frequency, GHz (b) Phase of Scattering Parameters O- - -.5. o t,. t 5 - I| — CV|!; o -1-.-m -:t --- —----- I --- - -- - *30 10 15 20 25 3 35 40 FPquncy. GHz (c) Total Loss FIGURE B2. Micromachined CPW vs. CPW Line: Open End Gap

Fourth International Symposiumrn on Space Terahertz Technology Page 223 FABRICATION AND CHARACTERIZATION OF MICROSHIELD CIRCUITS T. M. Weller, L. P. Katehi, and G. M. Rebeiz U of M NASA Center for Space Terahertz Technology H. J. Cheng and J. F. Whitaker Center for Ultrafast Optical Science Abstract Microshield is a partially shielded, planar transmission line in which the center conductor and upper ground planes are surrounded by air and supported by a 1.5 pm thick dielectric membrane. This configuration allows single-mode, TEM wave propagation over a broad bandwidth with very low dispersion and zero dielectric loss. A lower shielding cavity is used to minimize radiation loss, vary the characteristic impedance, and prevent power loss into substrate modes. These properties highlight the potential of microshield as a high performance medium for submillimeter wave transmission. Here we present an overview of the fabrication of microshield lines, and repcrt on the recent progress in the characterization of various circuit structures. Theoretical and experimental results are shown for lowpass filters, transitions to grounded coplanar waveguide, and different resonant stub designs. A description of an ultrafast electro-optic sampling exoeriment using membrane-based coplanar strip transmission lines is also included, and preliminary data is presented for line losses up to 500 GHz. I Introduction In 1988 a subrillimeter wave integrated-horn antenna was developed which contained a dipole suspended on a 1.5 Am thick dielectric membrane inside a silicon cavity [Ij. The success of this technique has led to the application of the membrane technology to a variety of antenna geometries, including wideband millimeter wave power meters and corner cube arrays at 1.2 THz [2. 3]. The potential for low-loss, non-dispersive electromagnetic wave transmission eventually led to the introduction of the microshield transmission line in 1991 [4]. Microshield is 4 partially shielded, planar transmission line in which the center conductor and upper ground planes are surrounded by air and supported by the dielectric membrane. This configuration. which is shown in Figure 1, allows single-mode. TEM wave propagation over a broad

r-..-sL 4qtlC,,,rciuf on SPace Terahert: Technoloc Page 224 tourrn InJcrnrruswu^ Jyu, -- -r -r Upper Ground Planes Center I qIqb '. I I Conductor - Silicon Membrane Wafer ---------------------- Lower Shielding Cavity Figure 1: Illustration of the microshield transmission line geometry. bandwidth with very low dispersion and zero dielectric loss. The lower shielding cavity is used to minimize radiation loss, vary the characteristic impedance, and prevent power loss into substrate modes. It also serves to equalize the upper ground planes, thereby eliminating the need for airbridges or via-holes. Based on the potential high-performance from this geometry, the microshield line is currently being investigated as an alternative for submillimeter wave-guiding applications. The characteristics of the microshield geometry were first examined in the theoretical study by Dib, et al [4, 5]. It was shown, for example, that discontinuities.in microshield radiate less than comparable coplanar waveguide circuits, primarily due to the field confinement provided by the lower shielding cavity. Even on uniform transmission lines, frequency-dependent radiation into the substrate, which increases with the relative permittivity, can be significant at submillimeter wave frequencies. This loss mechanism is clearly minimized with microshield, since the conductors are completely surrounded by air. Finally, characteristic impedances ranging from around 30 n to over 280 SI can easily be obtained by varying the center conductor width, upper ground plane spacing, and the height and width of the shielding cavity. Accurate, closed-form expressions for the characteristic impedance have been derived using a conformal mapping approach [5]. Here we present an overview of the fabrication of microshield lines and report on the recent progress in the characterization of various circuit structures. Theoretical and experimental results

Fourth International Symposium on Space Terahertz Technology Page 225 from 20-40 GHz are shown for lowpass filters, transitions to grounded coplanar waveguide (gcpw). and two resonant stub designs. The insertion loss of a 5-section filter was measured at <.3 dB at 25 GHz, and the response compared nearly identically with ideal transmission line theory predictions. A low-loss transition to gcpw was also fabricated which had an S11 below -15 dB across the frequency range. The last section describes an ultra-fast optoelectronic sampling experiment using coplanar strip transmission lines printed on the dielectric membranes. Preliminarv data on the loss characteristics up to 500 GHz demonstrates a definite advantage of membrane lines versus lines printed on high dielectric constant materials such as silicon or GaAs. II Microshield Fabrication The fabrication of microshield begins with a silicon wafer which has membrane layers on both sides. The membrane is a composite SiO2/Si3N4/SiO2 material which is grown using a combina-_ tion thermal oxidation/chemical vapor deposition process. The wafers used here are 340 /lm thick, high-resistivity silicon wafers with a < 100 > orientation. The microshield structure is formed by first removing photolithographically defined regions from the membrane layer on the back side of the wafer. This pattern serves as an etch mask, allowing pyramidal i cavities to be opened in the wafer using anisotropic Si etchants, such as ethylene diamine pyrocatechol (EDP) or potassium hydroxide (KOH). When the wafer has been etched through to the front side, the upper membrane layer will behave as an etch stop, thereby forming the exposed membrane surface with surrounding silicon walls below. The final step in the fabrication process is the metallization of the circuit pattern and the lower cavity sidewalls. A tri-layer photoresist/aluminum/photoresist technique with metal evaporation is used to deposit the circuit pattern on the top side of the membrane laver. Evaporation is also used for metal deposition from the backside onto the cavity sidewalls, but here some shadowing is required to protect selected areas of the membrane from metallization. A diced or.-cro-machined silicon mask is used to shield the region between the inner-edges of the upper ground planes and 'The pyramidal shape of the cavity is due to the varying rate at which the etchant attacks different surface orientations inside the silicon.

Page 226 Fourth International Symposium on Space Terahert- Technology the center conductor, thus preventing metallization of the membrane below the slots. The finished circuits are then mounted on a g:ound plane to fully enclose the lower shielding cavity. III Theoretical Characterization A rigorous full-wave analysis is being developed for use in the theoretical characterization of microshield circuits. The formulation involves a space-domain integral equation (SDIE) that is generated through the application of equivalent currents. This approach has previously been applied in a variety of electromagnetic problems, such as the characterization of coplanar waveguide [61, rectangular waveguide junctions [7], and scattering from microstrip patches [8]. This method is also well-suited to the study of microshield transmission lines, as it provides a relatively simple means of studying the different types circuit geometries which are possible. These geometries include step changes in the height and width of the shielding cavity and abrupt changes in the dielectric material, in addition to discontinuities in the metallization patterns of the signal line and upper ground planes. The first step in the analysis is to simplify the circuit geometry by applying the equivalence principle. Here fictitious metal surfaces are introduced in the slots of the metallization which transform each region of a circuit into either a closed metallic cavity (inside) or a ground plane in free space (outside). Upon each side of these surfaces magnetic currents are imposed which act as equivalent sources. To determine the strength of these unknown currents the condition of continuous tangential electric and magnetic fields is enforced over each of the fictitious planes. Since M = E x n, where n is the outward unit normal in each region, continuity of Eton is strictly satisfied by setting Mower =-Supp = M. The continuity of Htan is expressed in terms of the following integral equation: nix f (Gext(F, ir) + Gint(r r')l. M,(7r)ds' = J, (1) where Gextint are the magnetic field dyadic Green's functions external and internal to the metallic cavities, respectively, and Js is the current assumed with the ideal current source model, which

Fourth International Symposium on Space Terahertz Technology Page 227 is used as the excitation. The solution to (1) is obtained by first dividing each surface into subsections, and then expanding the unknown magnetic currents using localized piecewise-sinusoidal basis functions. Equation (1) is then enforced in an average sense over each of the subsections, and is solved by using the method of moments (Galerkin's method) to generate a system of linear equations based on the mutual coupling, or admittance, between each of the currents. The solution of this system yields the complex coefficients of these basis functions, which is equivalent to finding the tangential electric field. This field is in the form of pure standing waves, such that standard transmission line theory may be then used to extract the scattering parameters of the circuit under analysis [6]. The formulation summarized above is used for the characterization of circuits which have discontinuities in just the upper metal layers. By extending this approach, it will be possible to analyze circuits such as T-junctions, shunt stubs with conformal cavities, and transitions between microshield and other planar lines such as coplanar waveguide. This extension primarily involves the use of equivalent current sheets in multiple planes, which serves to separate cavity discontinuities into homogeneous regions. The work in this area is currently in progress. IV Theoretical and Experimental Results In this section we present results on the performance of grounded coplanar waveguide (gcpw) to microshield transitions, stepped-impedance lowpass filters and different resonant stub designs. All measurements were made from 20-40 GHz, using an Alessi microwave probe station connected to an HP 8510B network analyzer. The system was calibrated using the TRL (thru-reflect-line) technique. For the theoretical data, ideal transmission line theory was used to characterize the lowpass filters and proved to be very accurate. The resonant stub designs were characterized using the full-wave analysis, and this section includes a study of the effect of shield-geometry variation on the stub perforiancep The formulation for characteriiat.ion of the gcpw-microshield transiio.ua is currently being developed, and thus only experimental data is presented.

Page 228 Fourth International Symposium on Space Terahert: Technology IV.1 (Grounded) CPW to Microshield Transitions The primary motivation for investigating the gcpw-to-microshield transition lies in optimizing the design for circuit measurements. As mentioned. these measurements are made using a microwavp probe station. Since the membranes cannot reliably withstand the force of repeated probe contact, the contact pads for must be formed in gcpw, which necessitates the use of transitions from the gcpw to the microshield line under test. In order to calibrate the system, these transitions need to be designed for low reflection. There were four micreshield-gcpw-microshield transitions which were measured. The structures consist of a section of grounded cpw of length at 40 GHz, which is positioned in the center of a microshield line. The design is illustrated in Figure 2. In all cases the microshield geometry has a center conductor width, s, of 350 ym and a slot width, w, of 35 pm. The characteristic impedance for this line is i75. For the four different designs, s + 2w in the gcpw was also held constant at 420 pm, and center conductor widths of 35, 70, 140, and 280 pm were tested. These geometries result in characteristic impedances of 88, 71, 54, and 36 Q, respectively. The measured SIn for the transitions is given in Figure 3. The best performance, less than -15 dB from 20 to 40 GHz, is obtained with the 70 pm transition, which is the design with the closest impedance match. This transition also had a measured loss, (1- I| S1 12 - I S21 |2), below -15 dB across the entire band. IV.2 Stepped-Impedance Lowpass Filters Stepped-impedance lowpass filters were investigated since they are relatively easy to design and could demonstrate the low loss, TEM characteristics of microshield. In order to obtain the impedance ratios required to implement the chosen designs, a center conductor width ratio of 33: 1 was necessary. The conductor widths and the impedances are shown in the schematic diagram of the 5-section filter given in Fiagire 1 The measured performance for 5- and 7-section 0.5 dB ripple Chebyshev designs is shown in Figure 5. Also shown is the theoretical response obtained using ideal transmission line theory,

Fourth International Symposium on Space Terahert: Technology Page 229 (UPPER GROUND PLANE) MICROSHIELD - --.0 T. MICROSHIELD (UPPER GROUND PLANE) Figure 2: Illustration of the microshield to grounded cpw (coplanar waveguide) transition. The center conductor width, S, the slot width, W, and the length of the gcpw, L, are all indicated. L is A at 40 GHz. The slots in the gcpw are shaded to indicate that the substrate dielectric changes from air in the microshield to silicon in the gcpw section. 0. C) -0 I U. o Z -5. -10. -15. -20. -25. 35 MICRONS.-. — 70 MICRONS ----- 140 MICRONS -..... 280 MICRONS -30. ' 20. 25. 30. 35. 40. FREQUENCY [GHzl Figure 3: Measured Sll response of four different microshield to grounded cpw transitions. The microshield center conductor width is 350 microns. Each plot corresponds to a different center conductor width for the gcpw. In all cases, 2s+w is 420 microns for both microshield and gcpw lines.

r-..-L Cnm...C.. t TePraknhet~ Technology Page 230 tourn inrernanuaurl Jyruv& "bs v- - 0 -ce(EDGE OF LOWER SHIELDINIG CAVrm| are given in /im and the different impedance sections are indicated. The shaded regions are the (UPPEM GROUND PLANE) Figure 4: Diagram of the 5-section stepped-impedance lowpass filter (not to scale). Dimensions are given in Hm and the different impedance sections are indicated. The shaded regions are the upper ground planes, which have been marked to indicate the positions of the lower shielding cavity sidewalls. The distance between the cavity sidewalls is 1800 /im. with empirically-derived corrections for the effective line extension at impedance steps [9]. The agreement is quite good, and the differences near cutoff (shown in the insets) result from a peak in the measured loss which is not accounted for by the theory. IV.3 Resonant Stubs In this section the performance of a microshield shorted resonant stub is presented. The design which was implemented consists of a stub which is formed in the center conductor, and.is of length A- at the resonant frequency. This will be referred to as the in-line series shorted stub (ILS). A comparison of the measured and calculated results is first shown, followed by an illustration of the effects of varying the shielding cavity dimensions. In addition, the characteristics of the ILS stub are compared to those of a stub which is formed in the upper ground planes, which is referred to as the ground plane series stub (GPS). It is noted here that the measured ILS circuit had cavity dimensions of height, hc = 340 pm, and width, wc = 1800 am, as measured at the top of the cavity. The measured and predicted performance of the ILS stub is shown in Figure 6, where Ckt. A refers to the theoretical data. A schematic of the upper metallization pattern is also shown in the figure. The deep resonance at 28 GHz is shifted down in frequency relative to the ideal

Fourth International Sympo.siuz on Space Terahert: Tec ii:ology Page 23' 5-SECTION 0. -10. - -20. -30. -40.20. 25. 30. 35. 7-SECTION 40. 0..-E ad;^:> -10. -20. -30. -40. L 20. 25. 30. 35. FREQUENCY fGHzl 40. Figure 5: Measured and theoretical response of a 5- and a 7-section stepped-impedance microshield filter.

'......,,..:t,,,ni umr),>.,., on r,.. -eranertZ Tecnnoios' -5.; 10 35 -U 20. 25. 30. 35. 40. FREQUENCY [GHz] Figure 6: Measured and calculated response of the in-line series shorted stub, where Ckt. A refers to the calculated data. In the diagram on the right (not to scale), L, = 2500 and L = 120. All.- I Ck A L dimensions are in pMm. response, since the physical stub length is A1 at: 29.3 GHz, when account is taken of the slTort transverse slot lengths, Lt. In Figure 7, the effects of a reduced cavity size and upper shielding are demonstrated. The Ckt. B results represent the predicted performance with a lower cavity of dimensions he = 100 pm and uWc = 420 pm, which is the practical minimum cavity size for this circuit. Clearly, the cavity has only second order effects, which can be explained by the small slot width of only 35 pum. The Ckt. C results represent the predicted performance for the geometry of Ckt. A when shielded symmetrically above and below. In this case near-ideal stub performance is obtained, as the resonance has shifted to % 29.5 GHz. As mentioned, a comparison has been made between the in-line series shorted stub and an alternative design called the ground plane series stub. The theoretical performance of each stub, using the symmetric shielding configuration, is shown in Figure 8. The upper metallization pattern for the GPS is also given in this figure, where the total stub length is 2370 ATIm, corresponding to a resonant frequency of around 32 GHz. The GPS has a much narrower 3-dB bandwidth, roughly 14% versus 52% for the ILS, but will radiate much more strongly in an open environment.

Fourth Internatipnal Symposium on Space Terahert: Techn:igy Page 233 0. -5. 10 z 0 7. -10. -15. -20. -25. -30. 40. FREQUENCY [GHz] Figure 7: Theoretical response of different in-line series shorted stub geometries. Ckt. A has lower shielding cavity dimensions of 1800 x 340 mm (width by height), the Ckt. B cavity is 420 x 100 1um, and Ckt. C has the same cavity dimensions as A but is symmetrically shielded on the top and bottom. 0. -5. 10 0:2 -10. -15. -20. 35 350 -.t - L. 4 Eta -25. -30. 40. -KikQJQUL.NY [iUHzJ Figure 8: Theoretical response of the in-line series stub (ILS) versus the ground plane series stub configuration (GPS). The circuits are symmetrically shielded and have the same cavity dimensions as Ckt. A in Figure 7. In the diagram on the right (not to scale). Lt = 420 rim and Ls = 1950 Mm.

Page 234 Fourth International Symposium on Space Terahertz Technology V Electro-optic Sampling Experiment Another interesting structure that has been fabricated on the micron-thick membranes is a simple coplanar strip transmission line (CPS). This structure, devised in order to push the limits of high-frequency, guided-wave transmission, should be capable of supporting millimeter- and submillimeter-wave electronic signals with very little induced distortion. This is due to the fact that the degree of phase velocity dispersion experienced in hybrid-mode propagation on planar transmission lines, as well as the amount of frequency-dependent radiation of energy into the substrate of such structures, increases with the permittivity of the substrate. Therefore, the electric field of a signal travelling on a uniform CPS on a membrane substrate experiences essentially a homogeneous dielectric above and below the plane of the membrane, i.e. air. This leads to propagation that has no radiation loss, no substrate dielectric loss, and no dispersion due to a frequency-dependent effective permittivity. In this preliminary investigation, we have demonstrated that pulses having subpicosecond rise times and single-picosecond durations propagate for lengths as great as 5 mm without loss of fidelity. The usable frequency content of these pulses extends from dc to nearly 1 THz, and by taking Fourier transforms, we have computed the attenuation and phase velocity of the time-domain signals on the line. Between 50 and 500 GHz, it has been observed that dispersion is negligible and attenuation is very small, arising presumably only from ohmic losses in the metal conductors. In order to generate and measure electrical pulses of such short duration and high frequency content, the optoelectronic technique of photoconductive switching, along with an ultra-high-speed electro-optic sampling characterization system, have been employed. These techniques are both based on the use of laser pulses of extremely short duration (< 100 fs). The photoconductive source requires the ultra-short laser pulse to convert a dc bias to a subpicosecond electrical transient, while electro-optic sampling uses the optical pulse with an electro-optic crystal to produce a sampling gate having a duration of only several tenths of a picosecond. A probe tip fashioned from the electro-optic crystal is dipped into the electric field above the transmission line under test, and without any electrical contact the propagating waveform is measured. For the measurement of transients in the time domain, reflections from any source near the probing

Fourth International Svyposium On S.nr.: Terahertz Technoltp U'..'. point cause a loss of accuracy in the acquisition of clean, easily analyzed signals. Therefore. wa: necessary to integrate a 200-pum-wide photoconductive-generator patch with the CPS fabricated on the membrane. This was a convenient alternative compared with transporting a short,!,kci-i'ai transient onto the CPS via inductive wirebonds or through the use of an unwieldy flip-chip geoimetry. The photoconductor of choice, having subpicosecond carrier lifetime, high dark resistivity, high responsivity, and high dielectric breakdown, was a l-pm layer of GaAs grown by molecular beam epitaxy at the low substrate temperature of - 200 deg C (LT-GaAs). This photoconductor was lifted off its substrate to form a free-standing film through etching techniques, and then grafted to the membrane substrate using a van der Waals bonding technique. The CPS was then defined on the LT-GaAs and the membrane using standard photolithographic techniques. The pulse output after optical excitation of the photoswitch was measured at distances of 1, 3, and 5 mm from the generator, with no discernable distortion except for a small decrease in peak amplitude at the 5 mm propagation distance. The measured pulse rise time was < 0.75 ps. and the FWHM (full width at half maximum) was < 1.2 ps. When the Fourier transforms of the time-domain signals were compared, the phase velocity between 50 and 500 GHz was found to be flat, indicating the absence of dispersion. The attenuation, shown in Figure 9, was found to increase only slightly, to about 0.S dB/mm at 500 GHz. This result can be directly compared to the propagation of a subpicosecond pulse along a CPS of the same dimensions on a GaAs substrate. While the attenuation at 50 GHz is only about 0.7 dB/mm for this structure, at 500 GHz it is in excess of 9.5 dB/mm. This is mainly due to the extreme radiation loss present even in uniform lines that are fabricated on high-permittivity substrates. A much larger degree of phase-velocity dispersion also results from propagation on the GaAs-substrate line, leading to pulses with rise times and durations that rapidly expand with increasing propagation distance. VI Conclusion This paper has presented a review of the recent progress made in the experimental and theoretical characterization of microshield transmission line circuits. Low loss transitions from microshield to grounded coplanar waveguide have been successfully designed and measured, and

'r T. - rT —L- AI- -. Page 236 Fourth International Symposium on.pace I rahertz ecnnoiogy 10.00 10.00 i-l-I I I 1 -- GaAs substrate E 1.00 E 1. ~- - membrane substrate w =s =20 pm 0.10 I I I I I I I I 50 100 150 200 250 300 350 400 450 500 frequency (GHz) Figure 9: Measured attenuation of a coplanar stripline (CPS) printed on a GaAs substrate compared to a line printed on a dielectric membrane. The measurements were made using an electro-optic sampling technique. The conductor and slot widths of the CPS are both 20 um. stepped-impedance lowpass filters demonstrated very good performance and compared quite well to ideal transmission line theory. A brief study of different microshield resonant stub designs has also been presented. Finally, a preliminary study using an ultra-fast electro-optic sampling technique has shown that coplanar strip lines, printed on the dielectric membranes, show strong potential as a low loss transmission medium for THz-frequency applications. VII Acknowledgements This work has been supported by the Office of Naval Research under contract No. N00014 -92-J-1070, the NASA Center for Space Terahertz Technology, and the NSF under the Center for Ultrafast Optical Science. The authors would like to thank Prof. G. Rebeiz's group for their assistance, and particularly Mr. Walid Ali-Ahmad.

Fourth International Symposium on Space Terahertz Technology Po 23 References [1] G. M. Rebeiz, D. P. Kasilingam, Y. Guo, P. A Stimson and D B. Rutledge. 'Monolithic MillimeterWave Two-Dimensional Horn Imaging Arrays" IEEE Trans. on Antennas and Propagation, Vol. 38, pp. 1473-1482, Sept. 1990. [2) Curtis C. Ling and Gabriel M. Rebeiz, "A Wide-Band Monolithic Quasi-Optical Power Meter for Millimeter- and Submillimeter-Wave Applications ", IEEE Trans. on Antennas and Propagation, Vol. 39, pp. 1257-1261, Aug. 1991. [3] Steven S. Gearhart, Curtis C. Ling, Gabriel M. Rebeiz, Hemant Davee, and Gordon Chin, "Integrated 119-/nm Linear Corner-Cube Array ", IEEE Microwave and Guided Wave Letters, Vol. 1, pp. 155-157, July 1991. (4] N. Dib, W. Harokopus, P. Katehi, C. Ling, and G. Rebeiz, "Study of a Novel Planar Transmission Line," 1991 IEEE MTT-S International Microwave Symposium Digest, Boston, pp. 623-626. [5] N. I. Dib, P. B. Katehi, "Impedance Calculation for the Microshield Line," Microwave and Guided Letters, Vol. 2, No. 10, Oct. 1992, pp. 406-408. [6] N. I. Dib and P. B. Katehi, "Modeling of Shielded CPW Discontinuities Using the Space Domain Integral Equation Method (SDIE)," Journal of Electromagnetic Waves and Applications,April 1991. [7] Z. Shen and R. Hua, "Moment method analysis of rectangular waveguide T-junctions having arbitrary cross-sections", Imt. J. Electronics, 1991, Vol. 71, No. 3, 463-469 [8] K. Barkeshli and J. Volakis, "Electromagnetic Scattering from an Aperture Formed by a Rectangular Cavity Recessed in a Ground Plane," Journal of Electromagnetic Waves and Applications, Vol. 5, No. 7, pp. 715-734, 1991. [9] E. 0. Hammerstad and F. Bekkadal, A Microstrip Handbook, ELAB Report. STF 44 A74169, N7034. University of Trondheim, Norway, 1975.

Mon. a.m. STUDY OF MICROMACHINED QUASI-PLANAR LINES Rhonda Franklin Drayton* and Linda P.B. Katehi NASA Center for Space Terahertz Technology University of Michigan Ann Arbor, MI 48109-2122 Recent industrial trends in space and communication technology show the development of systems that have high rate information transfer as well as improved fuel economy. High rate transfer results from the improved electrical performance of circuit components while better fuel economy results from the development of smaller lightweight systems. Currently, planarized transmission lines such as microstrip, stripline, and coplanar waveguide provide enhanced flexibility in antenna and array design, compatibility with active devices and radiating elements, and reduction in size and weight. However, such lines produce unwanted frequency dependent mechanisms such as parasitic radiation and coupling which interfere with wave propagation. Consequently, a need exists for the development of transmission lines without the above limitations which exhibit improved electrical performance while offering additional reduction in the size and Weight of circuits. With recent advances in semiconductor processes, namely micromachining, Si or GaAs semiconductor substrate materials may be utilized to create planar transmission lines and their shielding cavities in monolithic form. Serving a dual purpose, the cavity can reduce electromagnetic interference and parasitic radiation while introducing additional design parameters. Geometrically, the planar transmission lines are created by metal deposition on a Si substrate whereas the shielding is created by etching a cavity into Si using KOH. After metallizing both surfaces using evaporation techniques, the two wafers arc attached by electrobonding. The result is the development of lines that offer improved electrical performance and the potential to incorporate an integrated package. In this paper a novel transmission line, which exhibits the advantages discussed above, is presented. A discussion of the fabrication procedure required to develop this line as well as the measurement technique used for characterization are presented. In addition, various line properties such as propagation constant and characteristic impedance are shown and compared to conventional lines, such as microstrip and coplanar waveguide. Quasi-Planar Transmission Line 30

BB-3 EXPERIMENTAL RESULTS ON MICROSHIELD TRANSMISSION LINE CIRCUITS T. MN. W'eller, G. M. Rebeiz, and L. P. Katehi NASA Center for Space Terahertz Technology The f niterqity of.ichignn:.-1z.A-lbor.IIl 4109-2122 Abstract - Presented here are the first experimental data on microshield circuits which use the membrane technology. Microshield shows potential for improved performance relative to current planar transmission line geometries. This paper will describe results on stepped-impedance filters and coplanar waveguide-to-microshield transitions. I. INTRODUCTION This paper presents for the first time an extensive experimental study of microshield transmission line circuits which are fabricated using membrane technology [1]. The microshield line was introduced in 1991 and has been shown, using proven theoretical techniques, to exhibit desirable waveguiding properties [2]. These properties include pure TEM propagation and zero dispersion in a wide, single mode frequency band, low radiation loss. zero dielectric loss, and a broad range of possible line impedance. The lower shielding cavity, furthermore, provides ground plane equalization and thus eliminates the need for air bridges or via holes. Several types of microshield circuits have now been fabricated and measured, including filters, stubs, and transitions from grounded coplanar iwaveguide (gcpw) to microshield. The transitions are of immediate interest because the preferred measurement technique requires initial contact to gcpw, as discussed below. Additionally, transmission lines with drastic step changes in the dielectric material may provide novel capabilities and are worthy of investigation. Stepped-impedance filters are examined since they are simple to design and are Shown to effectively demonstrate microshield performance. The fabrication and characterization of the transitions and filters are reviewed here. II. FABRICATION AND DESIGN The foremost characteristic of microshield circuit fabrication is the utilization of the membrane technology described by Rebeiz, et al [1]. For the work described here, 340 pim thick < 100 > silicon wafers, with 1.5 pm SiO^! Si-.V4/SiO2 membrane layers, have been used. The mli croshield structure is created by first removing photolitlographically defined regions from the membrane layer on the back side of the wafer. This pattern serves as an etch mask, allowing pyramidal cavities to be opened in the wafer using anisotropic Si etchants, such as ethylene diamine Pyrocatechol (EDP). The upper membrane layer will behave as an etch stop, thereby forming the exposed membrane surface with a surrounding cavity below [3]. The final steps in the fabrication process are the metallization of the circuit pattern and the cavity sidewalls. A tri-layer photoresist/aluminum/photoresist technique with metal evaporation is used to deposit the circuit pattern on the top side of the membrane layer. Evaporation is also used for metal deposition on the cavity sidewalls, but here some shadowing is required to protect selected areas of the membrane from backside metallization. A diced or micromachined silicon mask is used to shield the region between the inner-edges of the upper ground planes, thus preventing metallization in the slots. The finished circuits are then mounted on a ground plane to fully enclose the lower shielding cavity. A photograph of a 5- and a 7-section lowpass filter is shown in Figure 1. An important design objective in this project is to develop a low loss transition from grounded coplanar waveguide to microshield. This is necessary because the measurement device is a microwave probe station, and the probe contacts must be formed in gcpw since the membranes cannot reliably withstand multiple probing attempts While the design of a transition should include consideration of the characteristic impedance, it must also be considered that each of these lines is a mixed coplanar-microstrip guiding structure, for which the relative slot widths, center conductor width, and ground plane separation all determine the mode of propagation. Guidelines for the design of gcpw can be found in [4, 5], while those for the microslrheld line are found in:[3. Alt llLoiii:._;iansition design was chosen a priori for measurement purposes, three other variations were tested and their performance is reviewed in the following section. 827 CH3277-1/93/0000-0827$01.00 C 1993 IEEE 1993 IEEE MTT-S Digest

III. EXPERIMENTAL RESULTS A. (Grounded) CPW to Mficroshield Transitions The primary motivation for investigating the gcpw-tomicroshield transitions lies in optimizing the design for measurement purposes, as discussed above. One can envision, however, alternative reasons for utilizing short segments of cpw along a microshield line (or vice versa): a simple means of fixed phase delay or advance, as the phase ve-. l y will be lower in cpw tllai. rn microshield; mechanical support to facilitate mounting hybrid active devices along a microshield line; and obtaining very large impedance ratios for filter purposes are all possibilities. To our knowledge, such planar transmission line circuits with similar metallization, yet drastic dielectric changes, have not previously been explored. There were four microshield-gcpw-microshield transitions which were measured. The structures consist of a section of grounded cpw of length x '2 — at 40 GHz, which is positioned in the center of a microshield line. The design is illustrated in Figure 2. In all cases the microshield geometry has a center conductor width, s, of 350 pm and a slot width, w, of 35 pm. The characteristic impedance for this line is 75 Qf. For the four different designs, s + 2w in the gcpw was also held constant at 420 pm, and center conductor widths of 35, 70, 140, and 280 pm were tested. These geometries result in characteristic impedances of 88, 71, 54, and 36 fl, respectively. The measured Si, for the transitions is given in Figure 3. The best performance, less than -15 dB from 20 to 40 GHz, is obtained with the 70 pm transition, which is also the design with the closest impedance match. This transition also operates with very low radiation loss, as shown in Figure 4. It should be noted, however, that the SI, response is very dissimilar to that expected from ideal transmission line theory, which would predict a gradual rolloff to a perfect match at 40 GHz. These discrepancies may be partially explained by the field-configuration considerations mentioned above, as the microshield line is operating in a strongly cpw-like mode, whereas the grounded cpw line will exhibit more of a hybrid cpw/microstrip characteristic due to the smaller height-to-slot width ratio. B. Stepped-Impedance Low Pass Filters "Cpped-imped?-^ filth — -... '" used for applications which do not require a sharp cutoff, such as in the rejection of out of band mixer harmonics. They are a desirable test circuit for the new microshield technology primarily because they are quite easy to design and their performance may be used to qualify impedance calculations, the degree of TEM mode propagation, and the level of conductor and radiation loss. A schematic diagram of the 5-section filter desi given in Figure 5. The measured performance for 5 -section 0.5 dB ripple Chebyshev designs is shown in F 6. Also shown is the theoretical filter response obtainx ing ideal transmission line theory, with empirically-de corrections for the effective line extension at inmpe, steps [61. The agreement is quite good. and the differ, near cutoff can be explained by the peak in the radi. loss. shown in Figure 7 Also. no correction for cond loss is included in the theoretical data. IV. CONCLUSION Experimental results on microshield stepped-impedanc ters and transitions to grounded coplanar waveguide been presented. The filter response compared very to ideal transmission line theory when effective line le! corrections for impedance steps were employed. The transition performance demonstrates an S I below -1.' from 20 to 40 GHz. V. ACKNOWLEDGEMENT This work has been supported by the Office of Naval search under contract No. N00014-92-J-1070 and the N/ Center for Space Terahertz Technology. The authors w, like to thank Prof. G. Rebeiz's group for their assista and particularly Mr. Walid Ali-Ahmad. VI. REFERENCES [1] G. M. Rebeiz, D. P. Kasilingam, Y. Guo, P. A. Stir and D. B. Rutledge, "Monolithic Millimeter-Wave 1 Dimensional Horn Imaging Arrays", IEEE Trans. on tennas and Propagation, Vol. 38, pp. 1473-1482, Sept. 1 [2] N. Dib, W. Harokopus, P. Katehi, C. Ling, and G. beiz, 'Study of a Novel Planar Transmission Line," IEEE MTT-S International Microwave Symposium Di Boston, pp. 623-626. [3] N. I. Dib, P. B. Katehi, "Impedance Calculation fox Microshield Line," Microwave and Guided Letters, Vc No. 10, Oct. 1992, pp. 406-408. [4] G. Ghione and C. Naldi, "Coplanar Waveguides for M Applications: Effect of Upper Shielding, Conductor E ing, Finite-Extent Ground PLanes, and Line-to-Line pling," IEEE Trans. Microwave Theory Tech.,, Vol. N 35, No. 3, March 1987. [5] Y. C. Shih and T. Itoh, "Analysis of Conductor-Ba Coplanar Waveguide," Electronics Letters,. Vol. 18, Nc Ju:.'. 19S2. [6] E. 0. Hammerstad and F. Bekkadal, A Microstrip I book, ELAB Report, STF 44 A74169, N7034, Universi Trondheim, Norway, 1975. 828

Figure 1: Photograph of a 5- and a 7-section microshield stepped-impedance low pass filter. The upper metallization is 340 /m from the lower ground plane, producing the shadows which are seen around the center conductor and upper ground plane edges. The thin center conductor sections are 40 /m wide and the width of all ground planes is 0.42 cm. The grounded coplanar waveguide-tomicroshield transitions occur where the line narrows on either end of the circuits. (UPPER GROUND PLANE) 0. MICROSHIELD go ---- m -5. - c -10. Z -15. 0 I -20. -25. MICROSHIELD NW (UPPER GROUND PLANE) -30. - 20. Figure 2: Illustration of the microshield to grounded cpw (coplanar waveguide) transition. The center conductor width, S, the slot width, W, and the length of the gcpw, L, are all indicated. L is - at 40 GHz. The slots in the gcpw are shaded to indicate that the substrate dielectric changes ir.. - r n...L - r - l: -. In *,A:-... c to. --...... -,,I LA, - A I...............,.. COr,.-..-, '' I.. 25. 30. 35. 40. FREQUENCY [GHz] -- 35 MICRONS -—. — 70 MICRONS ---- 140 MICKUNS -.... 280 MICRONS Figure 3: Measured S1l response of four different microshield to grounded cpw transitions. The microshield center conductor width is 350 microns. Each plot corresponds to a different center conductor width for the gcpw. In all cases, 2s+w is 420 microns for both microshield and gcpw lines. 829

.5 4 g-. v:., Co,=... I I - I.. - J.\j. I. A 5-SECTION.2 - It) Z) 2 0 1 0 1 2U ). 25. 30. 35. 40. FREQUENCY [GHz] Figure 4: Measured loss factor, (1- 1 Si 12 - S21 12), for the 70 micron microshield to grounded cpw transition. 7-SECTION w z z 0 25. 30. FREQUENCY [GHz] 40. Figure 5: Diagram of the 5-section stepped-impedance lowpass filter (not to scale). Dimensions are given in pm and the different impedance sections are also indicated. The shaded regions are the upper ground planes, which have been marked to indicate the positions of the lower shielding cavity sidewalls. Figure 6: Measured and theoretical response of a 5- and a 7-section stepped-impedance microshield filter..5.4.3 0 t L. 40. FREQUENCY [GHz] Figure 7: Measured loss factor, (1- I Sl 12 - I S21 12), f< the 5- and 7-section low pass filters. 830

Page Si n 1 02 Fifth In:ernartonai SJtmnposzum on Space Terahert: Technology Millimeter and Submillimeter Wave Microshield Line Components Thomas M. Weller, Stephen V. Robertson, Linda P. B. Katehi, and Gabriel M. Rebeiz NASA Center for Space Teraherz Technology The University of Michigan AnnArbor, MI 48109 Abstract Recent efforts at the University of Michigan have resulted in the successful development of high-perfoimance planar components which operate in Ka- and W-Bands. These circuits are fabricated using silicon micromachining techniques, and they are placed on a thin dielectric membrane which is shielded by a metallized cavity on one side (see Fig. 1). This type of line, called microshield line, has exhibited very low radiation and ohmic losses for frequencies as high as 1 THz. A variety of microshield components have been fabricated and their performance has been measured at the facilities of the Radiation Laboratory at the University of Michigan and NASA Lewis Research Center. These circuit components exhibit superior performance compared to planar components made with conventional microstip or coplanar technology. In this paper, basic transmission line properties such as dispersion and attenuation are first examined. This is followed by a sample of the measured results obtained at Ka-Band for right angle bends, low-pass filters, and band-pass filters. Furthermore, extension of this technology to higher frequencies is explored. Top Side Metallization Membrane Circuit Wafer s w Metallized Cavity Side Walls Metallized Ground Plane Wafer Figure 1. Microshield line uses micromachining techniques to suspend components in air on a thin dielectric membrane.

Fifth International Symposium on Space Terahertz Technology Page S0 1. Introduction The need for high performance microwave circuit components which could operate in the THz freqeuncy range led to the development of the microshield line in 1991 [1]. Microshield line uses silicon micromachining and membrane technology to locally remove the substrate underneath a conventional CPW type transmission line. The resulting configuration propagates a pure TEM wave in a homogeneous air dielectric with very low loss and virtually no dispersion over an extremely wide bandwidth [2]. In this work we present results showing the high performance characteristics of microshield line. Measurements of loss and effective dielectric constant show the low loss and low dispersion as expected. Microshield line components are compared to conventional planar components. In particular, right angle bends show better return loss performance when fabricated with microshield line instead of CPW. Also, planar filters for high frequency applications which normally suffer due to the presence of a dielectric substrate are shown to operate effectively when implemented using microshield line. 2. Fabrication Microshield circuits are fabricated on high-resistivity (> 2,000 -cm) silicon wafers with a 1.5 iLm-thick SiO2/Si3N4/SiO2 composite deposited on both sides [3]. After the circuit metallization is patterned on the top side of the wafer using evaporation and lift-off, infrared alignment is used to pattern the backside membrane. The membrane layers are then etched to expose the silicon, at whiconh point thylenediamene pyrocatechol (EDP) anisotropic silicon etchant is used to define cavities underneath the circuit metallization. Once the silicon has been etched away, the top side membrane remains in tension to support the circuits. Next, the cavity side walls are metallized using a shadow mask evaporation which prevents an RF short from occurring between the signal line and the upper ground planes. Finally, the wafers are placed on a metallized ground plane wafer to completely shield the cavity (2-D geometry shown in Fig. 1). 3. Microshield Line Characteristics A. Effective Dielectric Constant Due to the air substrate of the microshield line, the effective dielectric constant, r,eff, is very close to 1. The presence of the membrane, however, causes a slight increase in eeff since a fraction of the fields will be contained within the thin dielectric layers which have

;e S04 Fifth International Symposium on Space Tere;er: Technology relative dielectric constants of 3.9 (SiO2) and 7.5 (Si3N4). As shown in Figure 2, the measured reff changes from around 1.09 to 1.15 as the slot width is reduced from 55 to 25 gm. This increase is a result of greater field confinement in the slot and thus in the membrane, and represents a decrease of nearly 3% in the guided wavelength. This dependence on the slot width is similar to the characteristics of CPW lines on a substrate such as GaAs or quartz [4]. As the line geometry approaches single micrometer dimensions which are comparable to the membrane thickness, reff could increase to around 2.0. 2 l ' ' i. I. i i | i i r i X i ~ ll |. I " l l l... 1.8 s 190.rn, w 55 Lm ----- s = 250 Jnm, w = 25 m 1.6 1.4 1.2 10 15 20 25 30 35 40 Frequency (GHz) Figure 2. Effective dielectric constant at Ka-Band for two geometries of microshield line. B. Attenuation The attenuation characteristics of the microshield line have been investigated by comparing measured data on different microshield geometries against experimental and theoretical results for coplanar waveguide which are found in the literature. At low frequencies, the two types of lines are expected to perform equally well, since dielectric and radiative losses are relatively small and attenuation is dominated by conductor loss, which should be about the same in microshield and CPW due to the similar geometry of the metallization. At higher frequencies, however, the microshield line gains an increasing advantage since it has essentially zero dielectric loss and does not radiate energy into the substrate, as with a substrate-supported CPW line. This characteristic has been verified by recent electro-optic sampling experiments, which have provided attenuation data on membrane lines up to 1000 GHz [2].

Fifth International Symposium on Space Terahertz Technology Page S05 The comparison between the microshield and CPW attenuation is shown in Figure 3. The curves shown for comparison were chosen either because the CPW dimensions are similar to the microshield dimensions (curves CwFG) or to illustrate the effects of reducing the line geometry (curves D,E). All pertinent geometrical parameters for the CPW and microshield examples are given n Table 1 beneath the figure, and it is notedat athe conductor thickness is only about 2 skin depths at 25 GHz for all but two cases (note: the technique used in [5] assumes zero conductor thickness). 0.25 0.2 Io.J 0 0.15 -0.1,- D D...-"' H C -------- d FO -ro A 153 ' a 0.05 0 1( 3 15 20 25 30 35 40 Frequency (GHz) Figure 3. Comparison of microshield and CPW attenuation. Table I. Attenuation for microshield and coplanar waveguide lines. S is the center conductor width, W the slot width, H the substrate height, and t the metal thickness (in urn). The width of the lower shielding cavity for the microshield lines is 1800 grn. Curve Line Er Substr. S W H t Zo (|) Data Ref A shield 1.0 Air 250 25 355 1.2 75 Meas - B. tshield 1.0 Air 190 55 355 1.2 100 |Meas - C CPW 12.8 GaAs 232 84 100 - 50 alc [5]1 D CPW 12.8 GaAs 69 28 100 - 50 Calc [5] E CPW 12.9 GaAs 88 16 500 t.O 30 Meas [61 F CPW 12.9 GaAs 250 25 500 1.0 30 Calc [41 G CPW 4.0 Quanz 250 25 250 1.0 50 Calc [(4] H GCPW 11.7 Si 50 125 355 1.2 73 Meas I. J.: -, -:.-: -'..

Page 8 6 Fifth International Symposium on Space Tcrar:.: T.:hnoIogS The resul:s present:ed here confirm lha: he microshield line is free from unexpectec or excessive conductor-loss mechanisms, and has performance which is comparable:o conventional substrate-supported coplanar waveguide at lower frequencies. Furthermore,:he absence of dielec ric-related loss and he abilitry:o maintain non-dispersive, sing e-moce propagation over a very broad bandwidth lead to low attenuation well into:he milimeterwa.ve frequerces. 4. Microshield Line Components A. RightAng,'e Bends A very common circuit 'element' in millimeter-wave systems is the right-angle bend. Th.s strucntur gains significance with increasing frequency due to the parasitic capaci* tace urLd irductance which are associated with the abrupt change in the field orientation. These problems, combined with mode-conversion, result in a high reflection of the incident power 'large S 11'. In coplanar waveguide designs, the typical means of improving the returi loss is to utilize air-bridges or dielectric overlays [7]. Both of these techniques are meant to offset the effects of different electrical path lengths along the two slots. Without th3s type of cornpensatlolrn, the asyrrnietry of the CPW ri ght-angle bend will lead to excitation of the turwartted slot-lEi.e rrode. Many of the problerrs:x.erent to berds which are printed orL conventional CPW can be rriimized with the microshield geometry. The absence of the high dielectric constant material )has two important effects: it leads to a. nduction in the parasitic capacitance and, for a, fixed physi.cJ sxze, it redi.ces the dLffererncc ir, electriczl pathi lergths throu gh the two slots. The shielding cavity, fiDtenrmore, provides continuous ground plame equalization without introducing additionall dscontinuities, thus improving u,pon conventional airbridges. A comparison betweer the measured peformnance of a raicroshi,eld-line bend and a typical CPW bend on 3aAs [8] iLs shown in Figuxe 4; the microshield bend has an SI which is a2 least 8 dB lower over the 10-40 GHz band. The noise in the data below 18 GHz,s dte to call.bratio error.

Fifth International Symposium on Space Terahertz Technology?:^e S07 0 - F I I I.... I.... *... * -: - Microshield Une, -5 250 UIM, w-25 m, h350 L m -10 - --- CPW on GaAs (w/Air Bridges), - s-39 pm, w=24 rm, h-90 pmm -... --- -25 -30 -35 -40 I, 10 15 20 25 30 35 40 Frequency (GHz) Figure 4. Performance of microshield and CPW right-angle bends. B. Low-Pass Filters The stepped-impedance approach to low-pass filter design is a relatively easy technique to use, and it is well suited for applications which do not require a sharp roll-off in the insertion loss. Often, however, the filter specifications call for high rejection over multiple-octave bandwidths, a requirement which may be difficult to meet using conventional substrate-supported lines due to the propagation of higher order modes. Thus, the very broad, single-mode bandwidth of the microshield line can provide superior filter performance in this respect In addition, the absence of dielectric-related loss reduces both attenuation and parasitic radiation, resulting in very low passband insertion loss. Stepped-impedance filters using 5-, 7-, and 9-sectons have been designed and tested. Some of the results were previously presented in [9], where it was shown that the measured performance compared very well with ideal transmission line theory, as a result of the pure TEM nature of the microshield propagating mrode. In Figure 5, measured data for a 5-section filter is compared with results from a full-wave moment-method analysis. The rejection is greater than 20 dB up to 75 GHz, which is about 1.5 octaves above the 3-dB point at 26 GHz. The measured passband insertion loss of this filter is between 0.2 and 0.5 dB from 20-23 GHz.

Page 8C8 Fifth International Svymosium on Space Terahert: Technology -5 f " ---- S - 'MoM' - -- 1: -:MoM C -— ~ --- S.., - measured -15 't15 - --- S21 - measured 9 -2: | | i;. * 23 33 43 5C ^ Frqu (GHz) -25! Figure. Performac of a 5-secon stepped-impae low-pass filter. -3C II -35 -43 23 33 43 5C S: 7: 8: Frequis ns( (GHz) Figure 5. Performance of a 5-section stepped-impedance low-pass filter. C. Band-Pass Fi'ters By cascading multiple open-end tunilxg stubs in ser.es, it is quite sarnple to realize a bandpass response wth high out-of-baid rejection ixbd low irisetioa loss. Thle geometry of a three-stage design is shown ii Fgire 6, in which each section is separated by 150 rX,.. The rmeasimsd response, shown ii Figre 7, h.as ar. irsertior loss of only 1.0 dE from 22-32 GHz, wfich is competitive with the best shielded bandpass filters using suspended stripl;ne [10]. The perforrCance of this microshield flter could be fiurter irrlproved usirng thicker rcletlLizt.oln (since the orle rricror thickr.ess is eqUivwent to just over two skirrdepths) or hjrger slot widths to minirriize the conductor loss. The cJciliated data included Ln the plot is genemabted by utsing the scattering parameters found from the full-wave analysis of a single stub section lxid treating the filter as three non-coupled elements in series. The agreercment between the rleasured and calculated performance is quite good and indicates that there is very little electromagnetic coupling between the stubs, even though the stub separation is only 150 Wi.TL The response of the filter can also be modeled almost exactly by csscadcnLg the measured results for a. single stub, and this approach accurately priedicts the 1.0 dB insertion loss.

Fifth International Symposium on Space Terahert: Technology?;se S09 Figure 6. Geometry of a microshield line bandpass filter. -5 - - ~ ' CD 35' - S, -measured ' - - S1 - measured -40 * I I I a 10 20 30 40 50 60 Frequency (GHz) Figure 7. Performance of a microshield line bandpass filter. 5. Conclusions Microshield line has been shown to be a very low loss, low dispersion transmission line well into the sub-millimeter wave frequency range. This work has demonstrated microshield circuit components with excellent performance at frequencies as high as 40 GHz, and continuing work at the University of Michigan has realized a 90 GHz lowpass filter and a 250 GHz band-pass filter. This recent work will be presented in detail in forthcoming publications.

3< ' oJ Fztn *n;rrL2:;r.. irrpus:: 2;;:;. rZ Jr 6. Acknowledgments This work was supported by the NASA Center for Space Terahertz Technology and:he C ffice of Naval Research. References [1] N.I. Dib, W.R Harokopus Jr., L.P.B. Katehi, C.1. Ling, and ',.M. Rebeiz, "S:udy of a Novel Plazxr Tirursrr:.sslor. Line," 1991 r!EEE M77r-S I,gest, pp. 523-62 5. [2] H. J. Cheng, J. F W.itacer, T. M. Weller, azid L. P. Katehi, 'Terahertz-Bandwidth Chzractaizatiori of Coplanhur Wavegaide on Dielectric Membrane via Time-Domain Electo-Optic Sampling," Presented a thie 15954' M;Tr-S international Microwave SymposimLn [3] G.M. Rebeiz, D.P. Kasilingam, Y. Guo, P.A. Stirnson, D.P. Rutledge, "Monolithic Millimeter-Wz.ve Two-Dirnersi.onaJ Horn Irriag.ng Arrays," rEEE Trans. Antennas and Propagation, Vol. 38, pp. 1473-1482, September 1 9 ). [4] W. H. Iaydl, W. Heinrich, R. Bosch, M. Schlechtweg, P. Tasker, mad J. Braunstein, "Eesigr, data fo:- iilliir.eter wave coplanar circuits," 1992 Eruropean Microwave Conference Proceedings, pp. 223-228. [5] M. Zhang, C. Wu, K. VWu., arld J. Litv2., "Losses ln GaAs Microstrip and Coplzman Waveg4ide," 1952 IEEE AMT-S j5 igest, pp. 971-974. [6] W. HL Haydl, J. Bramunsteii, T. Kitazawa, M. Schlechtweg, P. Tasker, ard L. F. Eastmran, "Attenuation of Milirrieterwave Coplmar Ljies on Gall.urr Arsenide and Indiumr Phosphide Over the Range 1-60 GHz," 1992 IEEE MATT-S D'gest, pp. 349 -352. [7] R. N. Sirrlons ar.d G. E. Ponclak, "Modeling of Sorrme Coplanaur Waveguide DiscontinLities," I'EEE Trans. Microwave Tieor Tec., Vol. 36, DEecember, 1988, pp. 1796 -1803. [8] ArnjaLd A. Q. OrrWu, "An Accurate Solution of 3-D Coplsjhar Wz,vegiide Circuits," Ph.E. Thesis, University of Waterloo, Waterloo, C ntario, Canada, 1-993. [S] T. M. Weller, G. M. Rebeiz, ig.d L. P. Krtehi, "Experirental Results on Microshield Line Circuit,"' 1993 rEEE M7r-S Digest, pp. 827-830. [10] W. Menzel, "Broadband Filter Circuits Using an Extended Suspended Substrate Transmission Line Configuraon," Proceedings of tie 22nd Europeva Microwave Conference, Aug. 24-27, 1992, pp. 459-453.

t 4Otc-C Conference Proceedings RD EUROPEAN WAMICVRwAE 11 [l i 2CONFERENCE Monday 6th to Thursday 9th September 1993 Palacio de Congresos, Madrid, Spain In co-operation with: LUjau UPM - UniverSidad Politecnica de Madrid CICYT - Comision Interministerial de Cienca y Tecnologia CAM - Comunidad Autonoma de Madrid IEE - The Institution of Electrical Engineers IEEE - The Institute of Electrical and Electronics Engineers - Region 8 - MTr Society URSI - The International Union for Radio Science In association with: Reed iLB Exhibition I Companies (UK) 90 Calverley Road, Tunbridge Wells, Kent TN1 2UN, UK Telephone: +4~4 (0)892 F514027 Telex: 95604 MEPNCL G Fax: +44 (0)892 541022

-,,^..!'u C b,,'c A,,: - re,;n r croncve Corfe:'er: - - Micromachined Circuits for Mm-Xave Applications Rhonda F. Drayton and Linda P.B. Katehi The University of Michigan Ann Arbor, MI 48109 USA ABSTRACT A novel transmission line which utilizes micromachining techniques in the development of circuits for millimeter-wave applications is presented. Micromachined circuits incorporate the fabrication of an upper-half shielding environment with the circuit geometry. The circuit development requires the use two silicon wafers where one has cavities that have been etched using anisotropic etching while the other one has the planar circuits which have been printed using standard photolithographic techniques. Afterwards, the two wafers are aligned and secured using bonding techniques to form a circuit that has an upper-half shielded cavity incorporated monolithically. This paper presents a detailed description of the fabrication procedure necessary for upper-half shielded (UHS) circuits as well as for completely shielded (CS) circuits. Experimental results of the upper-half shielded circuits will be presented and compared to theoretical data as well as data for conventional coplanar waveguide circuits of the same type. INTRODUCTION Uniplanar transmission lines, such as microstrip, coplanar waveguide and slotline have provided enhanced flexibility in high-frequency circuit design, reduced weight and volume, and compatibility to high-speed active devices. Planarization, however, gives rise to unwanted frequency dependent mechanisms such as parasitic coupling and radiation which deteriorate electrical performance and lead to costly and time intensive design cycles. Suppression of these electromagnetic mechanisms has led to improved performances but has required very sophisticated packages which considerably add to the volume, weight and most important cost of the circuits. Consequently, the development of new packaging techniques which minimize cost and reduce volume and weight can have a major impact in today's technology and can lead to high volume markets. This paper presents the development of miniaturized circuits of coplanar waveguide type where the package has been fabricated and integrated with the rest of the circuit using Si-based Micromachining [1],[2]. The characterization of these circuits has been completed both theoretically and experimentally with very close agreement between theory and experiment. In the following sections the fabrication process of micromachined circuits is described and theoretical as well as experimental data are presented and discussed for a variety of three-dimensional structures. FABRICATION Micromachined circuits, consisting of conducting lines and metallized cavities, are developed in two stages using a two-wafer system. In this section a comprehensive discussion is presented on PAGE I

the development of completely shielded as well as upper-half shielded micromachined circuits. First. the cavity structure, Figure 1, is etched into a low resistivity Si wafer (353|imn) having a 1.4 pmn dielectric tri-layer of oxide/nitride/oxide. This wafer requires the production of alignment marks which are etched through the entire wafer along with cavities which are etched through partially. In this process. the areas to be etched are defined using standard photolithographic techniques. Since the tri-layer must be removed in the areas of silicon tc be etched, the upper oxide layer is etched using buffered HF while the middle nitride layer is etched by a plasma etcher. The wafer is then repatriated to expose and remove the lower oxide layer in the areas that will be etched entirely while preserving the lower oxide layer in the areas that will be etched partially. Alignment Mark Upper Cavity Metallization _ V /... -- -. => -=> 1st Etch State 2nd Etch State Final State FIGURE 1. Development of cavity regions on the upper wafer for Micromachined circuits. Since the ti-layer dielectric is used as a mask when anisotropically etching, with KOH or EDP, the initial desired depth is etched into the silicon surface. Next, the remaining lower oxide layer in the cavity areas is removed using buffered HF and the wafer is etched for the remaining depth to produce both the alignment windows and appropriate cavity depth. Circuits..'. -.:-.:::?:.........::.::..::::....... ^T Lower Cavity Lower Wafer FIGURE 2. Development of cavity regions and circuits on the lower water for the Micromachined circuits. For the development of completely shielded circuits, cavities must be incorporated below the surface of the circuit layer. This is done by anisotropically etching grooves in the backside of a high resistivity Si wafer (353gm) having the tri-layer dielectric mask as described above. Next, planar CPW-like geometries are printed, using photolithographic and metal evaporation techniques for a lift-off procedure. Once complete, the cavity structure is then metallized using evaporation techniques. Completely Shielded Circuit Upper-Half Shielded Circuit FIGURE 3. Micromachined Completely Shielded and Upper-Half Shielded Circuit in two-dimensions. Finally, alignment of the upper cavity to the planar circuits is obtained via microscope and the two wafers are attached using regular adhesion methods or Si-to-Si electrobonding. While the procedure PAGE 2

described above concen ra es on the development of completely shielded circuits. the upper-half shielded circuil follows he same procedure with he exclusion of ihe lower cavity etching step. RESLLTS AND DISCUSSION MEASUREMENT SYSTEM: The micromachined circuits are measured using an HP 851 SB Network Analyzer sys em, which opera.es a. frequencies up to 40 GHz. Measurements are obtained from Cascade Micro.ech high frequency GSGS probes via an Alessi probe station. In order to suppress parallel plate and microstrip modes between the circuits and the wafer chuck of the probe station, the circuit wafer is placed on 1/8" thick 5880 RT/EDuroid having er=2.2. The Thru-Reflect-Line TRL' calibration is performed to eliminate the effects of the connectors, cables and probes from he measured data and to accurately establish a known reference plane for the various circuits. After calibrating, scattering parameter measuremen s of the circuits were obtained to compare theoretical and experimental data. CIRCUrIT DESiCRIPTION: (a) Front View (tb) TopF View FIGURE 4. Micromachined Shielded CPW, Cavity Dimensions: height = 200 Pm and width = 800 pm. The fabricated and tested circuits are both open CPW and microrrLachined upper-half shielded CPW, as shown on Figure 4. The c.v.ty region for these circui ts are 800rrli wide and 200cmr high. Each circuit is fed by ungrounded open coplanar waveguide which has feed lines that transition from center conductor width of 100 iun to 180pgm and slot width of 50im to 130 gm. Since the TRL calibration is used, calibration standards are also fabricated to include identical cpw feed line to shielded region transitions for accurate calibration of the rnicromachined circuits. MEASUREMENT RESULTS: The experimental results shown are for the micromachirted upper-half shielded 1UHS) circuit. Initial characterization,s obtained by measuring the scattering parameters of a through line where the performance characteristics are extracted. Below in Figure 5 is a graph that shows that the total loss of the through line, is about 0.38 dB/Xg. PAGE 3

' -20 -a.. F -30..o Ax = 2,176mm w=130im - s=18I0m i 0 5 10 15 20 25 30 35 40 Frequency, GHz FIGURE 5. Through line characteristics for the Micromachined Upper-Half Shielded CPW. (a) Open End Series Stub,-1 OaMwI s=30Wm s=SOm ---- w=1086pA — 0":= iSof (b) Open End Gap g-3Olm w-130pm - I s- I - -~n - FIGURE 6. Circuit description for the (a) Open End Series Stub and the (b) Open End Gap. 0 s-U I -12 -20 in qu 6a 11L 10 15 20 25 30 Frequency, GHz 35 40 VI at SD ito SD la C. S. a I. 0.; 200 --- — I - - 150............-. 100....... I0-.-...:-:... —r.,.p ^ - m 0 --CA If SZ)7- i -.-MicomadhrnedCP w 150 -A. o 100 -S 0. E -50-.100 -C. I I I.........:...........:.....................:..:...........:,,,..- -- Microanchined CPW:. — Theoretical............. i......................... a a a a I i I I I -.................. -u ---............. --- —-.. ' - - * -,.J ft......;...........-..........-.... ---. ----.. ----.......... i I -r - I - * t....................... *'Ari(Rji.: -.............7................. - I -150 -"tN —,% - - - - - - - - - - - - - - - - - - - -.................. I.......... 0 r ($2 1 1 1 1 d -l~^- - 15 0- - -200- I I I I I -AJw - '. - I-1 1 10 15 20 25 30 35 40 Frequency, GHz 10 15 I 20 25 30 Frequency, GHz (b) 35 40 (a) FIGURE 7. Micromachined UHS CPW, Theoretical vs. Experimental.: (a) Open End Series Stub and (b) OpenEnd Gap. PAGE 4

Vaous circuits were measured as pan of this characterization effort which include an open-end senes stub, designed to resonate at 30 GHz, and an open-end gap as shown above in Figure 6. Figure 7 shows a comparison between theoretical and experimental data for the micromachined UHS open-end series stub and open-end gap respectively. The theoretical results were denved using a full-wave spacedomain integral equation technique [31 and agree very well with the experimental results in both the I I,,,, 0 -4, I t+ v r Ic ICa Q. Ci) -4 - -8 - -12 - gsMag( SI).....................; - - Micromachined CPW CPW I-I I - L. e0 Cm_ QC -5 - M —ag(S1 Mag(S 1 )...........-................ T -10- - - Mag(S 12) -15 - -I ~ M T rr- h r~ < P -20.-...... M.. crImachilnpd rCPWI:-CPW I I I I 1 T ju ~. -_ - I ' ' ' I' t. - In- I - -ZU 10 10 I I I I 15 20 25 30 Frequencv, GHz 35 40 10 15 20 25 30 Frequency,;GHz 35 40 v, 0 L. ai o E - 10 15 20 25 30 35 10 15 20 25 30 35 40 Frequncy, GHz Frequency, GHz (a) (b) FIGURE 8. Micromachined UHS CPW vs. CPW: (a) Open End Series Stub and (b) Open End Gap. magnitude and phase. Figure 8 shows a comparison between experimental data for the micromachined UHS CPW and open CPW in the case of the open end series stub and open end gap circuits. The results indeed verify that the micromachined cavity does not affect the electrical performance of the planar circuit. Similarly, Figure 9 below shows experimentally measured total loss for the open end series stub and open end gap. The total loss is reduced by an average of 3 dBs from 10 to 40 GHz for the micromachined UHS CPW as compared to the open CPW. This reduction is due to elimination of radiation in the upper-half region of the circuit and emphasizes the effectiveness of the integrated cavity. Although the primary focus has been on the upper-half shielded circuits, some experimental results for the completely shielded geometries will be presented at the conference. PAGE 5

.10.... [-15 ---/ -. - Micromachined CPW S i Micromachined CPW -. to.. | 0,PW........... -2I I I I i I 30 -10 15 20 25 30 35 40 10 15 20 25 30 35 40 Frequency, GHz Frequency, GHz (a) (b) FIGURE 9. Micromachined UHS CPW vs. CPW Total Loss: (a) Open End Series Stub and (b) Open End Gap. CONCLUSION In conclusion, we have presented the development of micromachined circuits of simple geometries. The fabrication has been shown for both the completely shielded and upper-half shielded circuits. Comparisons have been shown between the upper-half shielded circuits and theory as well as open CPW measurements. Thus, with the improvement of the above loss performance combined with smaller volume and lower weight, micromachined circuits indeed make a very good candidate for development of integrated microwave and millimeter-wave packages. ACKNOWLEDGMENTS This work was partially supported by the Office of Naval Research and the NASA Center of Space Terahertz Technology. REFERENCES 1. M. Yap, Y-C Tai, W.R. McGrath and C. Walker, "Silicon Micromachined Waveguides for Millimeter and Submillimeter Wavelengths," presented in the 3rd International Symposium on Space Terahertz Technology, Ann Arbor, MI, March 1992. 2. Linda P.B. Katehi," Low-Loss Transmission Lines for Terahertz Frequency Applications," IEEE Proceedings, November 1992, pp.1771-1787. 3. N.I. Dib, L.P.B. Katehi, G.E. Ponchak andR.N. Simons, "Theoretical and Experimental Characterization of Coplanar Waveguide Discontinuities for Filter Applications, IEEE Trans. on MT, Vol. 39, No. 5, May 1991. PAGE 6

Presented at the 5th International Symposium on Space Terahertz Technology - May 1994 MICROMACHINED DETECTOR MOUNTS FOR MILLIMETER WAVE APPLICATIONS R. F. Drayton, C. Kidner, J. East and L. P. B. Katehi NASA Center for Space Terahertz Technology University of Michigan Ann Arbor, MI 48109-2122 Abstract High frequency circuit development requires that circuits are miniature in size with good electrical performance. These types of circuits can be achieved by utilizing micromaChining techniques that allow individual planar geometries and shielding environment to be fabricated monolithically. This paper will present the development of such circuits including a detector mount where the advantage of this configuration is a lowcost circuit system with small weight and volume. Since the shield is integrated with the planar geometries, characterization is simplified and the resulting circuit performance is comparable to conventional planar lines. At the higher frequencies this is particularly useful since various geometries can be easily fabricated using silicon micromachining techniques. 1.0 Introduction Micromachined circuits have been studied at the University of Michigan for development of passive circuit components [1] where the individual planar geometries and shielding environment are fabricated monolithically resulting in miniaturization of conventional circuit components. The advantage of this configuration is a lightweight circuit system with small weight and volume. In addition, characterization of such systems is simplified since the shield is integrated with the planar components resulting in circuit performance that is comparable to conventional planar lines. For higher frequency applications, this is particularly useful since various geometries can be easily fabricated to desired dimensional requirements using silicon micromachining techniques. This paper presents the development of a variety of micromachined circuits including a detector mount that utilizes the advantages of micromachining to create the circuit components and a

Presented at the 5th International Symposium on Space Terahertz Technology - May 1994 mounting structure for planar diodes. A description of the development of the system and the subcomponents required along with a discussion of circuit performance is presented. 2.0 Fabrication Issues The mounting structure is a two silicon <100> wafer system which relies primarily on standard fabrication processes and etching techniques. The semiconductor processing steps include standard photolithography techniques for wafer patterning along with lift-off and evaporation techniques for metallization of the patterned areas. Ethylene diamine pyrocatechol (EDP), an anisotropic etchant, is used to etch silicon where it selectively stops on the <111> crystal plane resulting in sidewall angles of 54.7~ as shown in Figure 1. Upper Cavity 54.750 (a) Lower Cavity (b) FIGURE 1. Two wafer silicon system of the transmission line with integrated shielding environment In Figure 1, the top wafer shows the air-filled upper cavity, having a width of 1200 pm and height of 280 pim, that is metallized after etching low resistivity silicon. On the bottom wafer coplanar waveguide transmission lines are printed on the polished side of high resistivity silicon using photolithography with lift-off and evaporation techniques for metallization. Since the etched grooves on the bottotm side of this wafer define the lower substrate-filled cavity, they are metallized last to provide the lower shielded region having a height of 350 pm and width of 950 gm. Once the indi

Presented at the 5th International Symposium on Space Terahertz Technology - May 1994 vidual wafers are fabricated, they are secured using adhesion techniques while alignment is done via alignment windows. 3.0 Circuit Design and Results Since micromachining techniques allow the integration of the shield with the transmission line components, circuit parameters are needed to establish a feedline that matches to 50 ohm systems. For the configuration shown in Figure 1, the transmission lines are coplanar waveguides having a conductor width of 180 gum and a slot width of 130 pgm with above shielding dimensions. For very high frequency circuits, initial circuit performance can be determined using on-wafer measurement techniques in the Ka-band. The results provide good indicators of the expected circuit response when scaled to terahertz frequencies excluding loss performance. In order to accurately measure the circuits, however, the measurement test wafer contains various circuit components having identical upper and lower shielded regions in addition to individual calibration lines, needed for the performance of a Thru-Reflect-Line (TRL) de-embedding [2,3]. Resulting experimental measurements of the micromachined circuits are thus obtained using on-wafer probing techniques [4] via a measurement system which consists of an HP 8510B Network Analyzer, Alessi Probe Station, and Cascade Microtech ground-signal-ground (GSG) probes with 1501xm pitch. FIGURE 2. Detector Mounting Structure Design Scheme

Presented at the 5th International Symposium on Space Terahertz Technology - May 1994 Specific design issues are addressed in order to develop a micromachined detector which requires the use of the circuit shown in Figure 1 [5]. The design procedure begins by determining [6] and evaluating the range of realizable impedances available to implement a lowpass filter with a requirement of -20 dB insertion loss and the open circuit input impedance at the design frequency. Using the cross-sectional dimensions of the shielded regions given above, the high and low impedances are determined where the low impedance sections have a conductor width of 380 jim and slot width of 30 im and high impedance section has conductor width of 20 im and slot width of 210 im. The filter response shown below in Figure 2 for a 5 section stepped impedance filter is compared to results obtained from a finite difference time-domain model [7]. 0~~~ -S- S - Model PI"I /A - -S1 - Model X -10 — si1- Measured -u-S21 - Measured E - -20 30 -. 10 20 30 40 Frequency [GHz] FIGURE 3.5-section Stepped Impedance Lowpass Filter The above response is needed in order to match the output of the filter design. Since the transmission lines are coplanar waveguides, this detector scheme uses the diodes in an anti-parallel configuration across the slots. In order to determine an appropriate input matching network the dc parameters of the diodes must be determined to minimize the power reflected from the diode input.

Presented at the 5th International Symposium on Space Terahertz Technologv - 1May 1994 4.0 Conclusion Subsystem components have been developed for a detector mounting scheme. In order to complete the detector system, in-house or commercially available diodes may be used which determine the type of appropriate input matching network. From a circuit perspective, micromachining can be used to develop detector mounting structures quite easily, where the problem is addressed from a component level prior to cascading these elements into the overall system. 5.0 Acknowledgments This work has been supported by the Office of Naval Research under contract No.N00014-92-J1070 and the NASA Center for Space Terahertz Technology. The authors would like to thank Dr. Nihad Dib for his contribution of theoretical models and Mr. Chen-Yu Chi and Mr. Steve Mollenkopf for technical discussions. 6.0 References [1] R. F. Drayton and L. P. B. Katehi, "Micromachined Circuits for Mm-Wave Application", 23rd European Microwave Conference, Madrid, Spain, pp. 587-588, September 1993. [2] G. Engen and C. Hoer, "Thru-Reflect-Line: An improved Technique for Calibrating the SixPort Automatic Network Analyzer", IEEE Trans. Microwave Theory Tech., Vol. MTT-27, No 12, Dec. 1979, pp. 987-993. [3] M. Maury, S.March, and G. Simpson, "LRL Calibration of Vector Automatic Network Analyzers", Microwave Journal, May 1987, pp.387-391. [4] E.W. Strid and K.R. Gleason, "Calibration Methods for Microwave Wafer Testing", 1984 IEEE M7T-S Digest, pp. 93-97.

Presented at the 5th International Symposium on Space Terahertz lechnology - l\! 1994 [5] Bhartia-Bahl, Microwave Solid State Circuit Design, John Wiley & Sons, Inc., New York, 1988, Chap. 11. [6] N.I. Dib and L.P.B. Katehi, "Impedance Calculation for the Microshield Line", IEEE Microwave and Guided Wave Letters, Vol.2, No. 10, pp. 406-408, October 1992. [7] N.I. Dib, personal communication.

Development of Miniature Microwave Circuit Components Using Micromachining Techniques R. F. Drayton and L. P. B. Katehi NASA Center for Space Terahertz Technology The University of Michigan Ann Arbor, MI 48109 USA ABSTRACT Miniature microwave circuit components are presented that were developed using silicon micromachining techniques. Although micromachining is very well known in sensor applications, it is used, herein, for the first time in microwave applications. This paper presents the development of micromachined lines and describes the theoretical and experimental characterization of circuit components such as filters and tuning stubs. 1.0 INTRODUCTION In the past, planar transmission lines such as microstrip, stripline, and coplanar waveguide have provided flexibility in the design of microwave circuits and ease in mounting active components. In almost all design cycles, these circuits are typically developed and tested in an open environment. Once the performance is characterized, however, these circuits are mounted into a metal housing which may introduce unwanted parasitics and multiple resonances leading to the degradation of the overall circuit performance. In addition, this approach is costly and rather impractical for developing lightweight systems since the main contribution to the overall weight is the circuit housing. Miniaturized high frequency circuits with an integrated housing offer lightweight and controllable parasitics, which makes them appropriate for cellular and mobile communications in view of the system requirements. In an effort to develop miniaturized circuits which perform comparably to conventional transmission lines, micromachining techniques can be used. Ultimately, complex micromachined systems such as detectors, mixers, and integrated receivers can be developed, thus leading to a new advanced technology. However, in order to make the development of such systems possible, realization of primary high frequency circuit components like filters and tuning stubs is required. This paper presents the development of miniature high frequency circuit components which incorporate a monolithic shielding environment. The study includes an discussion of the fabrication procedures as well as an extensive theoretical and experimental characterization of tuning stubs and stepped impedance filters. PAGE 1

2.0 FABRICATION Micromachining processes are fairly standard in sensor applications, however, they are relatively new to high frequency circuit applications. The miniaturized circuits mentioned above are a two silicon <100> wafer system and rely primarily on the micromachining process of silicon etching. This etching is done using ethylene diamine pyrocatechol (EDP) which is an anisotropic etchant to silicon. Other semiconductor processing steps include standard photolithography techniques used for wafer patterning along with lift-off and evaporation techniques used for metallization of the patterned areas. In Figure 1, the top wafer shows the upper cavity, having a width of 1200 rm and depth of 280 em, that is metallized after etching low resistivity silicon. On the bottom wafer, transmission lines of coplanar waveguide type are printed on the polished side of high resistivity silicon using photolithography, lift-off techniques and evaporation techniques for metallization of lpm. Since the etched grooves on bottom side of the wafer define the lower substrate-filled cavity, they are metallized last to provide the lower shielded region having a depth of 350 am and width of 950 nm. Once the individual wafers are fabricated, they are secured using adhesion techniques while alignment is done via alignment windows. This process results in a monolithic shielded transmission line, Figure 2, that is reduced considerably in size compared to conventional structures placed in a metal housing. 3.0 RESULTS AND DISCUSSION In this paper results are presented for a seven section stepped impedance filter and a series shorted stub. To determine the desired characteristic impedance of the miniature circuits, quasi-static models by Dib based on point matching techniques[l] are applied to obtain the appropriate line and cavity dimensions. Once the circuits are fabricated and actual dimensions are d min the measured results are then compared to theoretical ones. The stepped impedance filter was designed using PUFF [2], a quasistatic model that allows for synthesis and analysis of simple circuits of microstrip form. Intricate geometries like the series short stub, however, require the use of more sophisticated models. Among the various full-wave analysis techniques, the space domain integral equation is appropriate to accurately compute the electric field distribution in planar shielded geometries as the ones considerd here. This approach uses Galerkins method to detemine the electromag ic fields and simple network analysis to derive the scattering parameters[3]. The measurement test wafer contains various circuit components having identical upper and lower shielded regions in addition to individual calibration lines, needed for the performance of a ThruReflect-Line (TRL) de-embedding [5,6]. Resulting experimental measurements of the micromachined circuits are thus obtained using on-wafer probing techniques [4] via a easrement system which consists of an HP 8510B Network Analyzer, Alessi Probe Statio, and Cascade Microtech ground-signalground (GSG) probes with 1501m pitch. A seven-section stepped-impedance lowpass filter, designed as show in Figure 3, has high and low impedance of 10O and 200 respectively and is surrouded by the cavity stucure described above. In Figures 4 and 5 measurements are shown and compred to theorecal results derived from quasi-static models where conductor and dielectric losses are included. Regarding conductor es, care was taken to incorporate the specific metallization thickness and the appropriate surface resistivity corresponding to the various sections of microstrip line widths [7]. To realize 100 and '20 ohm impedance steps, 15 pmn and 380 grm wide conductor lines are used with slot widths of 215pm and 30 pm. In the low impedance section, however, the line excites a coplanar waveguide mode due to the narrow slot width PAGE 2

while the high impedance section excites a icrosrip mode. This results in sections of line that operate in a mixed mode and create parasitics that cannot be easily accounted for in the current model. Despite this limi,tid however, it can be seen in Figure 4 that the magnitude and phase measurements agree very well with the model. This is due to the relatively low operating frequency which reduces the effect of the parasitics. The total loss in the system, as shown in Figure 5, shows good agreement between theory and measured results indicating that the circuits have negligible radiation loss thus confirming the effectiveness of the micromachined integrated shield. The level of loss is due to the fact that low anid high impedance sections have aspect ratios that have been shown to cause higher loss for both the coplanar waveguide mode [8] and the microstrip mode[7]. Minimizing the effect of the conductor loss can be done by increasing the metallization thickness to about 3 pm, as opposed to lpm used in this effort. Figure 6 shows the physical dimensions of a series short tuning stub while Figure 7 shows comparison between measurmnts and the full wave analysis results. Since the theoretical results do not account for losses, the difference in the magnitude between measurements and theory is expected. While the overall circuit performance is similar, the discrepancy in the resonant frequency can be attributed to the variations in actual cavity dimensions and those allowable in the model. The model assumes that the cavity is rectangulIr, whereas, in reality the cavity is trapezoidal due to anisotropic etching. Furthermore, the upper cavity is wider than the lower cavity (see Figure 1) while the model assumes identical widths. In addition, the measured line length behaves electrically shorter due to rounding the comers and edges of the stub fingers during fabrication. } Overall, both the series short stub and the low pass filter show good agreement with the thorcal predictions. As a result, the above study indicates the capability of micromachining to provide high frequency circuit components with a performance comparable to conventional circuits. 4.0 CONCLUSION The development of miniature microwave circuits has been proven successful The fabrication of these lines has been psented and measured lts have been shown and compared to quasi-static and fullwave teoretical ones. The data presented show that monolithic integration of the shield using micromachining techniques allows for the development of circuit components that offer comparable performance to convnioal shielded acircuits. To further impoe the circuit performance and reduce conductor loses, ap pate metliztion thickness for the ground plane as well as the conducting line should Mm 'iapid. hL e miniature form of these circuits is verified when comparing the geometrical co waveguide B-plane circuits which are commercially available for a wide range of fi 5.0 ACKNOWLEDGEMENTS This work has been supported by the Office of Naval Research under contract NoN00014-92-J-1070 and the NASA Center for Space Terahertz Technology. The authors would like to thank Dr. Nihad Dib for his contribution of theoretical models. PAGE 3

REFERENCES [1] N.I. Dib and LP.B. Katehi, "Impedance Calculation for the Microshield Line", IEEE Microwave and Guided Wave Letters, Vol.2,No. 10, pp. 406-408, October 1992. [2] S. Wedge, R. Compton and D. Rutledge, PUFF Computer Aided Design for Microwave Integrated Circuits Version 2.0. [3] N.I. Dib, L.P.B. Katehi, G.E. Ponchak andR.N. Simons, "Theoretical and Experimental Characterization of Coplanar Waveguide Discontinuities for Filter Applications, IEEE Trans. on MTT, Vol. 39, No. 5, May 1991. [4] E.W. Strid and K.R. Gleason, "Calibration Methods for Microwave Wafer Testing", 1984 IEEE M7T-S Digest, pp. 93-97. [5] G. Engen and C. Hoer, "Thru-Reflect-Line: An improved Technique for Calibrating the Six-Port Automatic Network Analyzer", IEEE Trans. Microwave Theory Tech., Vol. MTT-27, No 12, Dec. 1979, pp. 987-993. [6] M. Maury, S.March, and G. Simpson, '~LRL Calibration of Vector Automatic Network Analyzrs", Microwave Journal, May 1987, pp.387-391. [7] T.E. van Deventer, "Characterization of Two-Dimensional High Frequency Microstrip and Dielectric Interconnects", PhD dissertation, The University of Michigan, Dec. 1992. [8] S.R. Taub and P.G. Young, "Attenuation and e of Coplanar Waveguide Transmission Lines on Silicon Substrates", Eleventh Annual Benjamin Franklin Symposium on Antenna and Microwave Technology in the 1990's, May 1993, pp. 8-11. PAGE 4

LIST OF FIGURES Upper Cgavity 280 Wri A 350 i 95 500 pr Low~x CaviI FIGURE 1. Miniature Microwave Circuit in Completely Shielded Form - m"mi FIGURE 2. Phm tograph c~ Mlina Mewnwave Circuit in COimalk y Shhrlded Form. ii m 2091p 12 =484W1 14=865pn Ilj=1083pmm l6=484pm 1~r-209pmI FIGURE 3. Stepped Impedance LowPass Filter Circuit Dimensions with low impedance sections of 20 LI and high impedance sections of 100 Q PAGE S

0 co QQ *O r. - ~a cw I En -10 -20 -30 10 20 30 Frequency, GHz 40 rO Q -4. ~-.O 00 O O 6! A: V= w Aq w V I CA 180 90 0 -90 -180 Frequency, GHz FIGURE 4. Theoretical and Experimental Results for the Completely Shielded Stepped Impedance Low Pass Filter. The top view shows the magnitude of the S-parameters whike the lower vkw shows the phase. PAGE 6

0 I - l l r l rT I T r I I i C, O 0 E -10 -20 10 20 3C Frequency [GHz] 40 FIGURE 5. Measured and Theoretical Totl Low Measurement for the Completely Shidded Stepped Impedance Low Pass Fliter w2=20 rai g = 481Lm II w=1331im s=1761n I - wl= 571m a a I I I I 1002 uLm fl=20 Won FIGURE 6. Circuit description for the Series Short Stub. PAGE 7

0 lo e ~ Te ory o g ^\ |S12| -20 0 ~ Theory Measured -30 10 20 30 40 Frequency, GHz FIGURE 7. Measured and Theoretical Results for the Completely Shielded Series Short Toning Stub PAGE 8

Presented at the 1994 International Societyfor Hybrid Microelectronics Symposium, November 1994, Boston, MA.. Development and Characterization of Miniaturized Circuits for HighFrequency Applications using Micromachining Techniques Rhonda F. Drayton, Thomas M. Weller, and Linda P.B. Katehi Radiation Laboratory The University of Michigan 1301 Beal Ave. Ann Arbor, MI 48109-2122 P: (313) 764-0500 F: (313) 747-2106 Abstract Wireless and personal communications systems require lightweight, high frequency circuits that provide compactesas well as opnum electrical performance. In order to advance current capabilities in these areas, two design approaches have rcently been developed at the University of Michigan. One approach that is directed toward circuit minia ation uti izes micromachining techniques to realize integrated packages for individual circuit geometries. Another appoach hat is used to realize high-performance circuits involves the use of dielectric membranes to support the conducting lies. This paper pesents the development of these concepts and discusses the fabrication techniques and the perform nce characteristics of some representative circuits. Key words: micromachining, dielectric membrane, integrated package, coplanar waveguide, high frequency circuits INTRODUCTION High frequency circuit design plays an increasingly important role in the development of complex wireless and personal co unication systems. Thes systems require a wide variety of (monolithic) microwave integrated circuits which utilize planar transmission lines, such as coplanar waveguide, micostrip and slotline. While each of these lines offers greater design flexibility, reduced weit and volume, and simpler integaion with active deyices when compared to nonplanar guiding sct, planarization has resulted in some nndesirable side effcts. Inhret in these structures are unwanted freq cy dependent mechanisms such as parasitic coupling nd radiation, as well as conductor and dielectric loss hat often result in electrical perfomance deadation. In order to reduce the coupling and radiation effects, sophisticated pckages are usually developed to mpove the circuit performance. Unfortmately, they add considerably to the volume and weight of the circuit, and the overall manufacturing cost Transmission loss c be reduced by using a low loss matrial such as alumina, but this requires laser dicing techniques and cannot be utilized in an MMIC implementation. A solution that can address the above issues and provide an inexpensive alternative is the use of micromachining. Unwanted radiation can be suppressed by developing packages that are integrated monolithically with the circuit, while dielectric loss can be eliminated by developing circuits on free-standing dielectric membranes. Presented here are two approaches that use micromachining extensively to realize novel circuits. The first approach introduces packaging of miniarized circuits with integrated cavities, and emphasizes sizevolume/cost reduction. The second approach involves a membrane-supported nsmission line configuration and emphasizes superior performance as its main objective. While both techniques are compatible with silicon and GaAs processes, the work described here is exclusively silicon-based. This paper describes the design of each new geometry, along with an outline of the fabrication procedures. Performance is shown that indicates the effectiveness of the dielectric membrane-supported line as a low loss transmission line. Also included are curves that show the response of a completely shielded tuning stub and a membrane supported right-angle bend at Ka-band.

MICROMACHINED CIRCLIT METHODOLOGY The process steps applied in his work come primarily from a mature silicon micromachining technology. They are standard practice, for example, in developing microstructures for sensor applications. For high frequency circui. design, however, ihe use of micromachining is relaiively new. Therefore, the following sec ions describe the specific fabrication sleps required Do implement the new circuit geometries: monolithically integrated shielded circuis and dielecric membrane-supported circuits. Monolithically Integralt4d Shi(4lded Circuits Circuit miniaturization can be achieved by implementing structures which minimize the size and weigh, of the housing element used to package circuit components. By developing cavities tha, follow the pah of individual circuits and provide good isolation, the effective circuit area is reduced and the overall packing density can be increased. This concept has been explored using micromachining techniques and the resulting procedure will be described in the following section. The micromachined circuits presented here have a shielding environment tat has been integrated using a two-wafer system. Individual circuits are isolated by including a substue-filled cavity beneath each line in the lower wafer and an air-filled cavity in the upper wafer (Figure 1). The integration of windows that are etched entirely through the upper wafer allows repeatable alignment of the wafers. Finally, for measurement prposes the circuits are fed through grounded coplanar,waveguide (GSCF) lines tha transition from microwave probe pads into the shielded lines. The lower wer shown in Figure 2 is a high resistivity, silicon wafer with e of 11.7 and a thickness of 350 pm. To develop the circuis, the planar lines and alignment marks ire printed on the silicon dioxide masking layer using standard photolithographic techniques. Next, the lower side of the wafer is patterned using an infrared alignment to expose the regions that form the lower cavity. Prior to anisotropically etching the silicon, the oxide mask is removed from the these areas using buffered hydrofluoric acid (BHF:. The anisotropic etchant, ethylene diamine pyrocatechol EDP', uses the <111> crystal plane of <10C silicon wafers as an etch stop. This causes an etch angle profile of 54.75~ on the sidewalls of the cavities as seen in Figure 2b, wih the lower cavity having is widest dimension near the upper surface. Prior t;o he final me, allization of the lower cavities. the oxide in the groove underneath the ground plane is removed. This allows direct metal-to-metal contact between the lower cavity and the CPW ground planes. The upper wafer requires pauerning of both sides prior to etching the cavity, probe windows and alignment marks. In this study, a low resistivity silicon wa'er is used having a thickness of 5)X Urn with 7,5,) A of thermal oxide on both sides. In Figure 2d, probe windows and alignment marks to be etched are defined on the lower side of the wafer using a lift-oT procedure where a thin layer of Ti/Au is deposited. Next, cavities and alignmen. windows are defined on lhe opposilt side using an infrared alignment, followed by oxide removal in he open areas using BHF o expose ihe silicon surface for etching. To monitor lhe etch dep i, rectangular marks called eiah rulers are also paluerned that have widhs which correspond.o specific depJhs. The double-sided etching process allows sliuclural beams wilth constant width to be created in the probe windows as shown in Figure 3, to provide addiional mechanical s ength ito wafers with multiple cavities. After etching, lhe cavi ies seen in Figure 2e and 2f are lhen metallized to the appropriate thickness. Completely shielded circuits are developed after alignment and attachment of the upper wafer to the lower shielded planar circuits using a microscope. The resulting configuration is shown in Figure 1 where the heights of the upper and lower cavities are 28C and 35C microns, respectively, and the widths are 1180 and 983 microns, respectively. Dielectric-Membrane Supported Circuit The objective of the membrane-supported circuit technology is to obtain high performance, broadband operation with minimal dielectric loss or radiation into substrate modes. The first application of this technology was in developing the microshield line [1], which is a partially shielded configuration that is very similar to conventional coplanar waveguide, except that the substrate has been removed. This line has demonstrated excellent perfomance in a number of circuit applications from Ka-band to W-band 12,3] and the attenuation characteristics have proven o be excellent up to 1 THz [4]. The membrane approach has also been used to implement an air-substrat microsrip line, which was used to design an extremely low-loss Willnson power divider at 33 13Hz [5]. The basic processing involved in dielectric membrane-supported circuits is similar to that just described for the monolithically integrated shielded circuits. The primary distinction is the use of a thin '1.5 micronlhick' dielectric tri-layer which is composed of SiO2/

Si3N4Si1O2, grown using a combination of thermal oxidation and chemical vapor deposition. This membrane is deposited on boh sides of a silicon wafer and is used both as an ech mask for Ithe micromachining steps and also to suppo. he conducting lines of he transmission line. As with Ihe iniegraied shielded circuits, the main process steps are he metallization of the planar transmission line pa uerns and he etching of ihe silicon caviles and alignmenil marks. The metallizalion is done using standard photolithographic, evaporation and liftoff lechniques, wilh a meial hickness typically around 1-1.2 microns. Meal systems of Ti/I/Au or Ti/Al/Ti/ Au are generally used. The etch seps are performed using photolithographically defined patterns in he membrane, with anisotropic silicon etchants such as EDP or KOH (potassiwn hydroxide]. These sltps are meant to remove silicon from beneath the conduc ing lines in order to form a free-standing membrane. The fabrication sequence for the microshield line is outlined in Figure 4. In this figure, the micromachining precedes the meallization steps but it is also possible to pattern the planar lines on the upper wafer surface before the ezch. Also shown in the illus ration is the metallization of the lower shielding cavity tha results from the silicon nrmoval. This step is comple.ed wilh the aid of a "shadow mask" wafer, which is processed along with the membrane wafer, and has rec angular slots that are etched entirely through. The "shadow mask" is attacted temprarily to the back of the membrane wafer to prevent metallization under the slot between the center conducting line and the upper ground plane. Etched alignment marks are used io align the "shadow mask" to the membrane wafer, and the two wafers are separated after the backside evaporation. The final step in the microshield process is to mount a lower ground plane wafr to the membrane wafer in order to enclose the lower shielding cavity. Silver epoxy or pholoresist may be used as an adhesive. Although the transmission line geometry of the membrane-supported microstrip is different from microshield, the fabrication steps are very similar. The reader is referred to [5] for details. CIRCUIT PERFORMIA NCE The measurement setup used for this work consisted of an HP 8510B Network Analyzer, an Alessi probe station and Cascade Microtech ground-signalground probes with a probe piich of 150 jmn. To accuraiely measure the circuit performance of the micromachined circuits, the Thru-Reflect-Line (IRL, calibration technique is used [6,7] with the standards and circuit components having identical feedlines. A one tier de-embedding method is implemen.ed.o estab lish the system reference plane along the planar transmission line [8]. As a result, all transiions between the ANA and the newly defined circuit reference plane are taken into account and the appropriate error coefficients are computed to establish an accurate calibration. A comparison of the attenuation characteristics of the lines with integrated shielding and the membranesupported microshield line is shown in Figure 5. Due Lo the absence of dielectric loss and the low radiation loss, the microshield line provides the optimum loss performance for a coplanar waveguide-type geometry. This has been proven in eleciro-optic sampling experiments which compared identical coplanar waveguide geomeries on quartz, GaAs and a membrane. These tess demonstraed that the membrane-suppored lines had significantly lower loss as the frequency increased in o the submillimeter-wave region [4]. It is also evident from Figure 5 lha., Lhe lines wilh he in.egrated shielding have very good loss performance. The configuralion which has shielding on only the lower side performs nearly as well as the microshield line and is ypical of convenlional coplanar waveguide. The excess loss, in this case, results from the silicon in the lower half-space of the circuit. WVhen the shielding is added to the upper half-space to eliminate radiation effects and provide circuit-to-circuit isolation, the attenuation is slightly increased (upper curve'. This is believed to result from increased conductor loss. Another example of the perfrmance of circuits with integrated shielding is shown in Figure 6. This fige compares experimental and calculated results for an open-end series stub, a component which is often used for circuits such as bandpass filters and switches. The calculated results were obtained using a finite difference time domain (FDTD) analysis and show reasonably good agreement with the measured data. More importantly, the smooth charactistics of the measurements indicate that the circuit does not suffer from unwanted radiation or resnance effects. Finally, a comparison betwe right-angle bends fabricated using microsield and convtioal coplanar waveguide on GaAs [9] is given in Figure 7. The bend, shown in Figure 8, is a very common circuit 'element' in millimer-wave sstems, and gains significance with inceasing frequenrc due to the parasitics which are associated with the abrupt change in the field orientation. The difference in the path lengths through the inner and outer slots, furthmore, leads to popagation of an unwanted even mode, which radiates more than the desired odd mode. The absence of a substate in the microshield line, however, lea to a reduction in the prasitic capacitanc and reduces the diffeence in electrical path lengths through the two slots for a fixed physical size. 1. has also been demonsira:ed that he

even and odd mode propagation velocities are much closer on a membrane-supported line than on GaAs. Therefore, in the event of even mode excitation, less signal degradation will occur with the microshield line. As shown in the figure, these advantages lead to a reflection which is at least 8 dB lower over the 10-40 GHz band. CONCLUSION This paper has presented two new approaches for high frequency circuit design which are based on micromachining techniques. The first approach centers on miniaturized shielding for individual circuits with the goal of high packing densities without degradation due to crosstalk. The second approach is direced at developing very high performance circuits using membrane-supped transmission lines. Realization of these concepts has been achieved and examples pertinent to high frequency applications for each approach have been shown. The measured data demonstrate the high level of performance which can be achieved. ACKNOWLEDGEMENTS This work has been supported by the Office of Naval Research under contract No.N00014-92-J-1070 and the NASA Center for Space Terahertz Technology. The authors would like to thank Dr. N'had Dib for the finite difference time domain models. REFERENCES N. Dib, W. Harokopus, P. Katehi, C. Ling, and G. Rebeiz, "Study of a Novel Planar Transmission Line," in 1991 IEEE M7T-S, pp. 623-626. 2 T. M. Weller, G. M. Rebeiz and L. P. Katehi, "Experimental Results on Microshield Transmission Line Circuits," in 1993 IEEE MTT-S Digest, 2, pp. 827-830. 3 S. V. Robertson, L. P. Katehi and G. M. Rebeiz, "WBand Microshield Low-Pass Filtes," in 1994 IEEE MTT-S Digest, 2, pp. 625-628. 4 H. J. Cheng, J. F. Whitaker, T. M. Weller, and L. P. Katehi, Terahertz-Bandwidth Characterization of Coplanar Waveguide on Dielectric Membrane via Time-Domain Electro-Optic Sampling," in 1994 IEEE MTT-S Digest, 1, pp. 477-480. s T. M. Weller, L. P. Katehi, M. I. Herman, and P. D. Wamhof, "Membrane Technology (MIST-T) Applied to Microstrip: A 33 GHz Wilkinson Power Divider," in 1994 IEEE MIT-S Digest, 2, pp. 911-914. 6 E.W. Strid and K.R. Gleason, "Calibration Methods for Microwave Wafer Testing", in 1984 IEEE MTT-S International Microwave Symposium Digest, pp. 93-97. 7 G. Engen and C. Hoer, "Thru-Reflect-Line: An improved Technique for Calibrating the Six-Pon Automatic Network Analyzer", IEEE Transactions on Microwave Theory and Techniques, 27, No. 12, pp. 987-993, Dec. 1979. ' M. Maury, S.March, and G. Simpson, "LRL Calibration of Vector Automatic Network Analyzers", Microwave Journal, pp.387-391, May 1987. 9 Amjad A. Q. Omar, "An Accurate Solution of 3-D Coplanar Waveguide Circuits," Ph.D. Thesis, University of Waterloo, Waterloo, Ontario, Canada, 1993.

FIGURES FIGURE 1. A three - dimensional cross-section of micromachined lines where the shield and line are integrated monolithically. FIGURE 3. Photograph of micromachined completely shielded circuits. In this top view, the dark areas represent the upper cavity areas while open areas represent the probe windows needed for on-wafer probing. LOWVER WAFER LPPER WAFER,aP EDP Membrane 1Memne ^ Membrane I Ig F -I[ (5):6: r5i -.a FIGLRE 4. Fabrication steps of microshield line: '1) membrane wafer, '2, pattern membrane, 3' etch wafer, A4; metallize front of wafer, '5) metallize back of wafer, '6' enclose cavity. FIGLRE 2. The lower wafer is developed by TI printing transmission lines on the top surface,.(2 forming lower cavities by etching v-grooves in <1ZX> silicon using an anisotropic etchant hat result in an e ch angle of 54.75, and '3 melallizing the lower cavity beneath the line. The upper wafer is developed by (4} etching probe windows and alignment marks a', from both sides while the cavity (b, is etched from one side only; '5, libe cavities are Lhen metallized as shown in the 3-D section of the upper wafer which include the probe windows 'c).

0.3 i i r- - -- * Microshield (s= 190 w55) 'i ensions a.oe inmplres-. $h~icd., - I -O L: iO. 0.2 -- z - *. 0 g 10 20 30 40 FREQU~CY [GHl *1 FIGURE 5. Loss performance of the microshield line compared Lo both shielded transmission lines. All dimensions are in microns. 1l0 2m 60 gR 1 80 pm............C... l.. -- —.-. 0. -T nmisso-Measured < ' FIG.aEnsmiio 6. FIGURE 7. Photograph of two microshsield right angle e \om.bends. Te top photo shows tde front side of the circuits.erecaand the lower photo shows the backside with the i. nbottom ground panes removed. The shielding cavities shown in e lower photo are 1.6 mm wide. 130 * — [!_____0*?I tI - G-A —. -s=9 w=24 10 20 30 40 Mioeld (s=250, =25) FREQU ENCY IGHzl -1 0 o. FIGURE 6. Performance of an open-end series stub for - -20 the completely shielded transmission line compared to FDTD theoretical results. All dimensions are in microns. T -30 -40 I 10 20 30 40 FREQUENCY [GHzI FIGURE 8. Performance of a right-angle bend on microshield and GaAs. All dimensions are in microns.

ouurlf t c u * UnI 1 Yyj iCiL. I v -II I- nsernaiunio.tinisi^ i c *jniLu3, UC-iter rA4 Micromachined Self-Packaged W-Band Bandpass Filters Stephen V. Robertson, Linda P. B. Katehi and Gabriel M. Rebeiz The Radiation Laboratory Department of Electrical Engineering and Computer Science The University of Michigan Ann Arbor, MI 48109 Abstract Experimental and theoretical results are presented for membrane supported Wband bandpass filters which utilize silicon micromachining technology to create self-packaged, shielded circuits. A coupled line shielded microstrip implementation of a 5-element 0.5 dB equal ripple Chebyshev filter achieves a minimum insertion loss of 3.1 dB with a 6% bandwidth centered at 94.7 GHz. The measured filter performance shows very sharp cutoff with out of band attenuation better than 25 dB and input return loss better than 8 dB. Results are also presented for a 5-element filter that achieves a minimum insertion loss of 1.4 dB with an 11.3% bandwidth centered at 94.7 GHz, and a 3-element filter with 0.6 dB insertion loss and 16.4% bandwidth at 94.9 GHz. Efforts to model filter performance using commercially available software and FDTD techniques are discussed. 1. Introduction Planar microwave and millimeter-wave circuits take advantage of microelectronics processing technology to realize compact, low-cost, high-performance applications. As the operating frequency increases, however, planar circuits fabricated on typical silicon and gallium arsenide substrates begin to suffer from power loss through substrate modes and dispersion caused by the discontinuity in permittivity at the dielectric/air interface. At even higher frequencies, these problems become prohibitive and realization of microwave circuits such as filters and couplers becomes very difficult. Recently, efforts have been made to reduce or eliminate these problems by utilizing silicon micromachining techniques. One method employs selective silicon etching to remove I

the substrate underneath planar circuits which are then supported by thin dielectric membranes in free space [1]. This technique has resulted in a new type of transmission line called mnicroshield line which has very low loss and nearly zero dispersion up to frequencies as high as a few terahertz [2]. Circuit components realized in microshield line have shown excellent performance compared to conventional ones [3],[4], and microshield line has shown excellent performance in planar W-Band low-pass filters [5]. In addition, silicon micromachining allows for the development of miniature self-packaged components appropriate for densely packaged microwave and millimeter-wave circuit applications [6]. In this work, we present a series of self-packaged bandpass filters which have been realized in W-Band using micromachining techniques. The filters are implemented in shielded microstrip coupled line geometries and employ dielectric membranes and micromachined shielding cavities. Measured results presented here show very high performance, low loss bandpass filters that are centered at approximately 94 GHz and have bandwidths ranging from 6% to 16.4%. The filters were designed using commercially available software and low frequency scale modeling. Finite difference time domain analysis [7] was used to verify the design and confirm the measurements. 2. Filter Design The filters were designed for coupled line shielded microstrip implementations using the commercially available software package PARFIL [8] running on an IBM compatible PC. Filter parameters were calculated for a ground plane spacing of 100 pm and an upper shielding height of 500 jxm. The designs used an internal impedance of 90 Q and assumed an air dielectric everywhere. Three 94 GHz filters were designed using PARFIL, with two 5-element filters designed for bandwidths of 4.3% and 8.5%, and a 3-element filter designed for 12.8% bandwidth. The filter designs were optimized by constructing a 47:1 scale model of the 4.3% bandwidth 5-element filter designed for 2.00 GHz. A 76 gm thick polyethylene film was suspended over an aluminum sheet metal ground plane, with circuit metallization patterns 2

defined on the polyethylene using copper tape. Measurements of the model filter on an HP 8720B 2-port network analyzer revealed a filter center frequency of 1.958 GHz, and measured bandwidth of approximately 3.5%, which agreed fairly well with the calculated performance. Through further experimental iteration, it was found that the resonator lengths required shortening by 5.3% to shift the center frequency up to the desired value. FDTD analysis at 94 GHz verified the findings of the low frequency modeling experiment. To facilitate on-wafer probing measurements, a matching network was designed to transition from a CPW-style probe pad to microstrip. Since the filter design impedance was 90 n and the probing system was matched to 50 Q the CPW-to-microstrip transition had a dual purpose: (1) to transform the circuit impedance from 50 Q to 90 Q and (2) to transfer the fields from a horizontally opposed CPW type distribution to a vertically oriented microstrip distribution. To accomplish these two tasks, a Klopfenstein taper and a microshield line geometry were chosen. The Klopfenstein taper was geometrically desirable since it promoted a gradual transfer from the CPW mode to the microstrip mode, and microshield line has been shown to support both of these modes [9]. The taper was designed using a point matching method (PMM) algorithm [10] to calculate the characteristic impedance of the microshield line for various 2-dimensional geometries. For a 220 pgm constant width center conductor, it was found that varying the slot width from 10 prm to 290 pn obtained the desired range of impedances from 55 Q to 81 LI These results were used to design the taper shown in figure 1 based on the method found in [ 11]. At the beginning of the transition, the narrow slots of the microshield line promote a high horizontal field concentration which is compatible with the wafer probe geometry. At the end of the transition, where the impedances approach the higher range, the slots of the microshield line are much wider (as wide as twice the height above the ground plane), forcing the fields to concentrate in the area between the center conductor and the lower ground plane, like a microstrip mode. At the end of the taper, the CPW upper ground planes are terminated, and the pure microstrip mode is launched. 3

3. Measurements The filters were fabricated on 500 gm thick high resistivity wafers with dielectric membranes grown on both sides using anisotropic silicon etching [5] (see figure 2). Three wafers were used to create the self-packaged shielded microstrip line, and figure 3 illustrates a 2-dimensional cross-section of the assembly. Micromachining of the ground plane wafer was implemented using a two step etch process [6] which allowed the creation of windows for on-wafer probing measurements (figure 4). The measurements were performed using an HP 85 10C Vector Network Analyzer with Model 120 Picoprobe 150 pm pitch coaxial probes. A through-reflect-line (TRL) calibration was performed using the NIST de-embedding software and the following circuits: a 4 mm thru line, a 6 mm delay line, and a 2 mm offset open. The measured response of the 5-element, 4.3% bandwidth filter is shown in figure 5. The filter achieves a minimum insertion loss of 3.1 dB and a wider than designed bandwidth of 6% at a center frequency of 94.7 GHz. The plot in figure 6 compares the measured insertion loss with that predicted by the FDTD analysis. The discrepancy in bandwidth occurs since the FDTD analysis does not consider the effect of the thin dielectric membrane. The presence of the membrane slightly increases the capacitances of the even and odd mode impedances in the coupled line sections, thereby increasing the bandwidth and the center frequency of the filter. Previous measurements of microshield line components have shown that the membrane does influence the response of membrane supported circuits [3]. PARFIL is used to include conductor loss in the filter analysis, since the FDTD code does not treat the finite conductivity of the metal. A value of 2.44x10-8 Q-cm for the resistivity of gold is used, and the surface roughness is assumed negligible since the gold was evaporated. The loss tangent is taken to be zero in the air dielectric. The loss in the filter is observed to be due entirely to conductor loss, since the measured insertion loss of 3.1 dB agrees very well with the calculated value of 3.3 dB. 4

The measured S-parameters of two other filters are depicted in figures 7 and 8. The response plotted in figure 7 of a 5-element filter designed for 8.5% bandwidth indicates a minimum insertion loss of 1.4 dB with an 11.3% bandwidth centered at 94.7 GHz. Measurements of a 3-element, 12.8% designed bandwidth filter are shown in figure 8. This filter has a minimum insertion loss of 0.6 dB, a center frequency of 94.9 GHz, and a bandwidth of 16.4%. 4. Conclusions We have presented high performance planar W-Band bandpass filters which are fabricated using silicon micromachining and membrane technologies and are self-packaged. The filters display excellent performance as measured on a W-Band probe station, including low passband insertion loss and high out of band signal rejection. The theoretical results match the measurements very well. 5. Acknowledgments This work was supported by NASA. The authors wish to thank Dr. N. Dib for providing the FDTD code and Mr. G. Ponchak at NASA Lewis Research Center for his time and for the use of the W-Band network analyzer and the wafer probing apparatus. 5

References [1] N.I. Dib, W. Harokopus, L.P.B. Katehi, C.C. Ling, and G.M. Rebeiz, "Study of a Novel Planar Transmission Line," 1991 IEEE MTT-S Digest, pp. 623-6. [2] H. Cheng, J.F. Whitaker, "Terahertz-Bandwidth Characterization of Coplanar Waveguide on Dielectric Membrane via Time-Domain Electro-Optic Sampling," 1994 IEEE M7T-S Digest, Vol. 1, pp. 477-80. [3] T.M. Weller, L.P.B. Katehi, and G.M. Rebeiz, "High Performance Microshield Transmission Line Components," to be published in the IEEE Trans. MiT, March 1995. [4] C.Y. Chi, G.M. Rebeiz, "A Low-loss 20 GHz Bandpass Filter," Submitted to the 1995 IEEE MTT-S International Microwave Symposium, October 1994. [5] S.V. Robertson, L.P.B. Katehi, and G.M. Rebeiz, "W-Band Microshield Low-Pass Filters," 1994 IEEE MTT-S Digest, Vol. 2, pp. 625-8. [6] R.F. Drayton, L.P.B. Katehi, "Development of Miniature Microwave Circuit Components Using Micromachining Techniques," 1994 IEEE M7T-S Digest, Vol. 1, pp. 225-8. [7] N.I. Dib, L.P. Katehi, "Analysis of the Transition from Rectangular Waveguide to Shielded Dielectric Image Guide Using the Finite-Difference Time-Domain Method," IEEE Microwave and Guided Wave Letters, Vol. 3, pp. 327-329, September 1993. [8] WaveCon EM Wave Control Software, P.O. Box 2697, Escondido, CA 92033. [9] T.M. Weller, G.M. Rebeiz, L.P. Katehi, "Experimental Results on Microshield Transmission Line Circuits," 1993 IEEE MTT-S Digest, Vol. 2, pp. 827-830. [10] N.I. Dib, L.P.B. Katehi, "Theoretical Characterization of Coplanar Waveguide Transmission Lines and Discontinuities," PhD Thesis, University of Michigan, November 1992. [ll]D.M. Pozar, Microwave Engineering, Reading, MA: Addison-Wesley Publishing Company, 1990, pp. 321-4. 6

CPW Proibe Pail,,WC tapear micrcostrip line Figure 1. The Klopi-enstein taper used as a CPW-to-microstrip transition. III........... A.4 It Figure 2. Photogzrapi o, tne +~3% designed oandwidth coupled mne i ter a~ter selective removal &71 iconiC uostrate. Thie memorane area appears dar~er than the surrouncing silicon in the picture. '7

metallized shielding wafer membrane I microstrip line 11 I' - - - TT micromachined ground plane wafer micromachined ground plane wafer ground plane metallization Figure 3. Two-dimensional geometry of the micromachined shielded microstrip. micromachined ground plane wafer CPW probe pad / circuit wafer " metallized shielding wafer Figure 4. A 3-dimensional view of the self-packaged membrane microstrip geometry, showing the probe window and exposed CPW probe pad. 8

0 -5 cn 0) ct3 -10 -15 -20 -25 I..Li 2 5.................................................'...............................30 \, A. 75 80 85 90 95 100 105 110 Frequency (GHz) Figure 5. Measured S-parameters of a 5-element bandpass filter with 3.1 dB minimum insertion loss, 6.1% bandwidth, and 94.7 GHz center frequency. 0 -5 A mn ClJ 0) co:3 o.C: ~3 -10 -15 -20 -25 -30 -35 -40 75 80 85 90 95 100 105 110 Frequency (GHz) Figure 6. FDTD results compared with the measured S-parameters of a 5-element bandpass filter. 9

0 -5 m -o CD -0 0) CO (M -10 -15 -20 -25 0 At -!..........;...... t....: A i I 1 -30 75 80 85 90 95 100 105 110 Frequency (GHz) Figure 7. Measured S-parameters of a 5-element bandpass filter with 1.4 dB minimum insertion loss, 11.3% bandwidth, and 94.7 GHz center frequency. 0 -5 Cu3 CD CD co -10 -15 -20 -25 -- -j 75 80 85 90 95 100 105 110 Frequency (GHz) Figure 8. Measured S-parameters of a 3-element bandpass filter with 0.6 dB minimum insertion loss, 16.4% bandwidth, and 94.9 GHz center frequency. 10

Submitted to the IEEE Antennas and Propagation Society International Symposium 1/5/95 Surface Wave Mode Reduction For Rectangular Microstrip Antennas loannis Papapolymerou*, Rhonda F. Drayton and Linda P.B. Katehi The Radiation Laboratory The University of Michigan, Ann Arbor MI 48109 USA I. Introduction Rectangular microstrip antennas are being used for several years in communication systems and other applications (radars, telemetry, navigation, biomedicine), due to their advantages over conventional antennas [1]. With the recent development of micromachining techonology, there is an increased interest in incorporating patch antennas with other components fabricated with this technology. However, micromachined circuits are usually fabricated on high dielectric constant substrates, such as Silicon, and are required to operate in high frequencies. Under these conditions, the excitation of surface waves is pronounced, thus leading to lower efficiency, degradation of the antenna radiation pattern and polarization characteristics and undesired coupling between array elements. Several approaches have been proposed in order to reduce surface wave power excitation [2-3]. In [3], Jackson et al. achieved a complete elimination of the dominant TMo surface wave mode for circular patches, by use of a specific design of the antenna element. In this paper, a similar approach will be followed for the rectangular patch, in an effort to reduce significantly the TMo surface wave mode. II. Analysis Our purpose is to design a rectangular patch antenna operating at f=23 GHz (K-band), with a dielectric constant of Er = 10.8 and thickness of h=25 mils, with a suppressed TMo surface wave mode. Since for the given substrate, higher order modes exist at frequencies above 40 GHz, only the dominant mode needs to be eliminated. 1

Submitted to the IEEE Antennas and Propagation Society International Symposium 1/5/95 According to the equivalence principle and the cavity model, in terms of radiation a rectangular microstrip patch can be modeled as a rectangular loop of magnetic current. From the cavity model, the electric field of the dominant TMO mode for the geometry shown in Fig. 1 is given by: Ez = Acos(n7y/b) (1.1) where a,b are the width and the resonant length of the patch, respectively. Here the assumption is made that the patch is fed in such a way that only the TM,o mode is excited. Thus, the radiating edges will be at y=O,b. The equivalent magnetic current at these two edges is: M = 2E x n = 2Ai (y=0,b). As it is known, a single Hertzian magnetic dipole, oriented in the x direction at a height z' above the ground, will give rise to a TMo surface wave field given by: YP = K(z,z' )H/2)(fMop)sinp (1.2) where ftmo is the propagation constant of the TMo mode and K(z,z') is an amplitude factor that depends on the height of the source and the observation point. By integrating over the two radiating edges of the rectangular patch for the magnetic current distribution, the total surface wave field radiated by the magnetic currents takes the form: P = -4A B(Z ) 2 roP)tan cos(w b sin o)sin(fio M cosb) (11.3) '1 2 2 h where B(z) = K(z, )dz'. 0 In order to derive the above expression the far field approximation for the phase of the radiated surface wave is used. The pattern of (1.3) is shown in Fig. 2 (dashed line). By setting b = nulls are placed at the location of the peaks of ITMo the lobes at 4 = -, so that four minor lobes replace the two major lobes (Fig. 2 2 2- solid line). As a result, the surface wave pattern is reduced. Choosing b according to the above formula, results in greater resonant length than the one chosen for the design frequency. In order to overcome this decrease in operating frequency, the material in a rectangular region under the patch is removed (Fig. 3a,b), thus creating a lower effective dielectric constant which will permit the 2

Submitted to the IEEE Antennas and Propagation Society International Symposium 1/5/95 desired increase in operating frequency. The length of this "cored" region is found from the transcedental equation, which is derived from the field expressions of the dominant mode of the cavity, after applying the boundary conditions at the various interfaces. Il. Results In order to test the accuracy of this analysis, an antenna was designed according to the above procedure and was compared with a conventional patch. Simulations were run using the FDTD method [4]. Measurement of the field inside the dielectric region showed a suppressed electric field for the reduced-surfacewave (RSW) antenna, which was more than 10 db lower than the field of the conventional patch (Fig. 4). Measurement of the reflection coefficient also with the help of FDTD, showed good agreement between the resonant frequencies of the cored patch and the conventional patch. Acknowledgements The authors would like to thank the Office of Naval Research for the financial support of this project and Dr. Nihad I. Dib for providing the FDTD programs. References [1] D.M. Pozar, " Microstrip Antennas", IEEE Proceedings, vol. 80, pp. 79-91, Jan. 1992. [2] N.G. Alexopoulos and D.R. Jackson, "Fundamental Superstrate (cover) Effects on Printed Circuit Antennas", IEEE Trans. Antennas Propagat., vol. 32, pp. 807 -816, Aug. 1984. [3] D.R. Jackson, J.T. Williams, A.K. Bhattacharyya, R.L. Smith, S.J. Buchheit, S.A. Long, "Microstrip Patch Designs That Do Not Excite Surface Waves", IEEE Trans. Antennas Propagat., vol. 41, pp. 1026-1037, Aug. 1993. [4] A.G. Engel, N.I. Dib and L.P.B. Katehi, "Characterization Of A Shielded Transition To A Dielectric Waveguide", IEEE Trans. Microwave Theory Tech., vol. 42, pp. 847-854, May 1994. 3

9G0o Is Igo y I A *....... b h I x 0 a Fig. I 270 Fig.2 A A' 1% I- - - - I I I I I I I I i I I I - - - - - N Fig. 3a A'1...............V... Fig. 3b NbCrO~trip pakhs with I*s-23 GKz -SI, & I -Is c~nverdmonal pa"~ D 0. ~ed plMCh -2 0 -25 K 0 -30 K 0 K -~ K 0 0 -40 0 -45 -So Fi~g. 4

Miicrostrip Patch Antennas on AMicromachined Low-Index MAaterials Rh.orda F. Draytor.*, Iomnnis Papapolyrrerou and Lirda P. B. Katehi The Radiation Laboratory The Department of Electrical Engineering and Computer Science The University of Michigan, Ann Arbor, MI 4 109O USA Introduction Microstrip patch antennas have been used extensively in vehicle and corrmunication systems in the past several decades. These anteruras provide low profile, lowweight, and low-cost along wl.te eas!y Jintegratior, into plar.ar circuit desigrns arLd array confgurations [1]. The conventional planar microstrip georretry, however, has limited bandwidtoh 2-5%, and poor efficiency as well as l.rted power capability when probe fed or rricrostrip fed. There are several factors which influence these parameters, the patch geomrretry and rrmedurn directly arid feedn.rg stz.ctiu.e i-Ad.ectly. The patch anrtelnna is prirmtarily inflLtenrced by the dielectn c constarLt ard thickness of the substrate rriatenal. With this 1L inilrid, the u ghlest bandwidth will be observed in des. gls irrplerrelited orl thick, low.rdex rcaterials [2]. The efficierncy of te antennlLa is affected princriaily by the excitatiorn of substralte r.odes. Whtile these rr.odes are irihelrer.tly elxcited ir. the substrate arid for that reasorl cannot be elinminated e it rely, redicti on will r]esult ir. improved performace sirnce thle eneirgy lost to these modes will now be available as useful power [3]. SJrface waves in low index rraterials are lower thaln those excited in high inde)x orles thereby nralk ag therri the preferred rrledium. ClI.rre.tly, designrs are often implemented on low inde)x materials for high paitch efficiency, arid wi.de banldwidtfh which canr be increased fi.rthler usinrg ww.debaind ijnpedarlce mratchjing networks anrLd or aperture coL.pling irrlethods [2]. In colantrast, circuit applicati on requirerrlents for hi.gh denls ty aL:id comlpactness are best acuhieved onl thirI high. index rr.aten.als. As rrielitiorned albove hloweve-r, these carL have adverse effects on the patch pedrfonnma.ce. SUlice patch. ar.terjnas are relaltively si rrLple to ujrriplerr.e -t,, rniariy appli catiorls, it is irnportar.t to address the requi.rerrie its of bothi the aritenrLa ar.d feedir.g stnJctLures. Ir. thle past ceitaurn desi gn tradeoffs were acceptable due to tecdu~iology lirrdtation]rs. However, withi the aid of new emerging technologies and the vast kaowledge of the patch anrtenna behavior, these issues can now be addressed effectively. A vaiety of approaches have been proposed [4] for improved anternna efficiency where rrLaterial undenleath the patch is modified to alter its electrical properties.

Herein, an alternative approach is proposed which allows for easy implementation using micromachining. The goal of this approach is to develop an "effective" low index material underneath the antenna while the feeding circuit is kept on the high index one. The design of a conventional patch antenna fed by a microstrip is developed using micromachining techniques to decrease the height of the high index material underneath the patch while the material under the feedline remains unaltered. This approach reduces the value of the effective dielectric constant only in the area under the antenna, since it is comprised of a mixed dielectric medium of high index material and air. Design/Fabrication Approach The patch geometry was initially designed using PCAAD CAD tool [5] for resonances at 20 GHz. This CAD tool is based on the uses a cavity model which determines the antenna resonant length (1/2 guided wavelength) and width for a given substrate material and thickness. Since a center probe fed excitation is assumed in PCAAD, for a planar configuration, an inset microstrip is used to match the antenna to the given feedline impedance. This design is implemented to provide a symmetric antenna design. The objective of this work is to explore the implementation of planar designs on localized low index materials. Initial investigation is implemented at 1.07 GHz on stycast, having ~r=12, with thickness of 6.6 mm. These are chosen to simulate the response of a patch on silicon substrate of thickness 350 microns at desired operating frequency of 20 GHz. Since the actual circuit will be used in a planar configuration with horizontal feeding structure, the microwave model is excited by using a coaxial probe to feed a inset microstrip feed line. The geometry is shown in Fig. 1 where the feed line is located on 100% material while the patch is located on 50-50 percent stycast - air dielectric material. When two pieces are brought into contact, the air region runs directly underneath the patch antenna and extends horizontally into the substrate material as shown in the lower geometry of Fig. 1 dashed lines. At the desired operating frequency, the design is implemented using micromachining etching processes to chemically remove the material from underneath the patch. A detailed description of this procedure is described in [6]. An additional wafer is mounted underneath to enclose the cavity below the patch and provide the microstrip ground plane. The circuit wafer consists of high resistivity silicon with Or = 11.7 and thickness of 350 microns while the ground plane wafer can be a low resistivity material. The microstrip feed in this arrangement is excited through a coplanar waveguide transition. Results and Conclusions In Fig.2, a comparison is shown between the microwave model and the theoretical results for the return loss of the patch. The theoretical results are based on finite difference time-domain method for the microstrip inset fed antenna. The FDTD [7]

results presented correspond to a patch design with a silicon substrate and are compared to the modeled results on the stycast material. The measured and theoretical results have similar shape although there is a 100 MHz difference in the resonant response. This resonance shift can be improved by implementing a finer mesh in the direction normal to the cavity where the 50-50 material-air interface lies. The above design will be implemented next on silicon using micromachining techniques. The presentation will describe the fabrication and measurement procedures utilized and will present measured and theoretical results on the performance of the micromachined antenna. This design offers the advantage of reducing the effects of surface wave excitation which can result in improved bandwidth and antenna efficiency. Acknowledgments This work was supported by the Army Research Office and the Office of Naval Research. The authors would like to thank Dr. Nihad Dib for the use of the FDTD programs. References [ 1 ] K. Carver, and J. Mink, "Microstrip Antenna Technology," IEEE Transactions on Antenna and Propagation, Vol. AP-29, No. 1, January 1981, pp. 2-24. [2] D. M. Pozar, "Microstrip Antennas," Proceedings of the IEEE, Vol. 80, No. 1, January 1992, pp.79-91. [3] R. D. Jackson, J. T. Williams, et. al, "Microstrip Patch Designs That Do Not Excite Surface Waves," IEEE Transactions on Antenna and Propagation, Vol. 41, No. 8, August 1993, pop. 1026-1037. [4] M. J. Vaugh, K. Hur, and R. C. Compton, "Improvement of Microstrip Patch Antenna Radiation Patterns, "IEEE Transactions on Antenna and Propagation, Vol 42, No. 6, June 1994, pp. 882-885. [5] D. M. Pozar, PCAAD: Prsonal Co ter Aided An Design. Version 2.1. [6] R. F. Drayton and L. P. B. Katehi, "Micromachined Circuits for Mm-Wave Applications," 23rd European Microwave Conference, Madrid Spain, September 1983, pp. 587-588. [7] A. G. Engel, N. I. Dib, L. P. B. Katehi, "Characterization of a Shielded Transition to a Dielectric Waveguide," IEEE Transactions of Microwave Theory and Technique, Vol. 42, No. 5, May 1994, pp. 847-854.

Figures A A' x wpzch hi -- h TABLE 1. Dimensions A A \ Parameter A Ifeed A' lpatch x Wsubstre Wpach ht hi Value(mm) 50.88 68.26 4.76 153.03 65.09 6.6 3.3 Figure 1. Micromachined patch antenna with effective low index material. 0 -1 a et c3 E -2 -3 -4 -5 -6 -7 0.5 Figure 2. Measured material. 0.75 1 1.25 Frqueamy [GHzl 1.5 and theoretical results of a micibwave model on stycast