Design Techniques for High Performance Wireline Communication and Security Systems

Abstract

As the amount of data traffic grows exponentially on the internet, towards thousands of exabytes by 2020, high performance and high efficiency communication and security solutions are constantly in high demand, calling for innovative solutions. Within server communication dominates todays network data transfer, outweighing between-server and server-to-user data transfer by an order of magnitude. Solutions for within-server communication tend to be very wideband, i.e. on the order of tens of gigahertz, equalizers are widely deployed to provide extended bandwidth at reasonable cost. However, using equalizers typically costs the available signal-to-noise ratio (SNR) at the receiver side. What is worse is that the SNR available at the channel becomes worse as data rate increases, making it harder to meet the tight constraint on error rate, delay, and power consumption. In this thesis, two equalization solutions that address optimal equalizer implementations are discussed. One is a low-power high-speed maximum likelihood sequence detection (MLSD) that achieves record energy efficiency, below 10 pico-Joule per bit. The other one is a phase-shaping equalizer design that suppresses inter-symbol interference at almost zero cost of SNR. The growing amount of communication use also challenges the design of security subsystems, and the emerging need for post-quantum security adds to the difficulties. Most of currently deployed cryptographic primitives rely on the hardness of discrete logarithms that could potentially be solved efficiently with a powerful enough quantum computer. Efficient post-quantum encryption solutions have become of substantial value. In this thesis a fast and efficient lattice encryption application-specific integrated circuit is presented that surpasses the energy efficiency of embedded processors by 4 orders of magnitude.