Show simple item record

Unreliable Silicon: Circuit through System-Level Techniques for Mitigating the Adverse Effects of Process Variation, Device Degradation and Environmental Conditions.

dc.contributor.authorKarl, Eric Alexanderen_US
dc.date.accessioned2008-08-25T20:52:16Z
dc.date.availableNO_RESTRICTIONen_US
dc.date.available2008-08-25T20:52:16Z
dc.date.issued2008en_US
dc.date.submitteden_US
dc.identifier.urihttps://hdl.handle.net/2027.42/60701
dc.description.abstractDesigning and manufacturing integrated circuits in advanced, highly-scaled processing technologies that meet stringent specification sets is an increasingly unreliable proposition. Dimensional processing variations, time and stress dependent device degradation and potentially varying environmental conditions exacerbate deviations in performance, power and even functionality of integrated circuits. This work explores a system-level adaptive design philosophy intended to mitigate the power and performance impact of unreliable silicon devices and presents enabling circuits for SRAM variation mitigation and in-situ measurement of device degradation in 130nm and 45nm processing technologies. An adaptation of RAZOR-based DVS designed for on-chip memory power reduction and reliability lifetime improvement enables the elimination of 250 mV of voltage margin in a 1.8V design, with up to 500 mV of reduction when allowing 5% of memory operations to use multiple cycles. A novel PID-controlled dynamic reliability management (DRM) system is presented, allowing user-specified circuit lifetime to be dynamically managed via dynamic voltage and frequency scaling. Peak performance improvement of 20-35% is achievable in typical processing systems by allowing brief periods of elevated voltage operation through the real-time DRM system, while minimizing voltage during non-critical periods of operation to maximize circuit lifetime. A probabilistic analysis of oxide breakdown using the percolation model indicates the need for 1000-2000 integrated in-situ sensors to achieve oxide lifetime prediction error at or under 10%. The conclusions from the oxide analysis are used to guide the design of a series of novel on-chip reliability monitoring circuits for use in a real-time DRM system. A 130nm in-situ oxide breakdown measurement sensor presented is the first published design of an oxide-breakdown oriented circuit and is compatible with standard-cell style automatic “place and route” design styles used in the majority of application specific integrated circuit designs. Measured results show increases in gate oxide leakage of 14-35% after accelerated stress testing. A second generation design of the on-chip oxide degradation sensor is presented that reduces stress mode power consumption by 111,785X over the initial design while providing an ideal 1:1 mapping of gate leakage to output frequency in extracted simulations.en_US
dc.format.extent3278148 bytes
dc.format.extent1373 bytes
dc.format.mimetypeapplication/pdf
dc.format.mimetypetext/plain
dc.language.isoen_USen_US
dc.subjectSemiconductor Reliabilityen_US
dc.titleUnreliable Silicon: Circuit through System-Level Techniques for Mitigating the Adverse Effects of Process Variation, Device Degradation and Environmental Conditions.en_US
dc.typeThesisen_US
dc.description.thesisdegreenamePhDen_US
dc.description.thesisdegreedisciplineElectrical Engineeringen_US
dc.description.thesisdegreegrantorUniversity of Michigan, Horace H. Rackham School of Graduate Studiesen_US
dc.contributor.committeememberBlaauw, Daviden_US
dc.contributor.committeememberSylvester, Dennis M.en_US
dc.contributor.committeememberFlynn, Michaelen_US
dc.contributor.committeememberMudge, Trevor N.en_US
dc.subject.hlbsecondlevelElectrical Engineeringen_US
dc.subject.hlbtoplevelEngineeringen_US
dc.description.bitstreamurlhttp://deepblue.lib.umich.edu/bitstream/2027.42/60701/1/ekarl_1.pdf
dc.owningcollnameDissertations and Theses (Ph.D. and Master's)


Files in this item

Show simple item record

Remediation of Harmful Language

The University of Michigan Library aims to describe library materials in a way that respects the people and communities who create, use, and are represented in our collections. Report harmful or offensive language in catalog records, finding aids, or elsewhere in our collections anonymously through our metadata feedback form. More information at Remediation of Harmful Language.

Accessibility

If you are unable to use this file in its current format, please select the Contact Us link and we can modify it to make it more accessible to you.