Margin normalization and propagation in analog VLSI

Abstract

We propose a margin propagation algorithm as an alternate decoding method to usual probability propagation algorithm. The decoding based on margins depends only on basic additions and subtractions and thus can be easily implemented in analog VLSI in a manner that is is independent of the device characteristics. We present an analog VLSI implementation of the margin propagation network using CMOS transistors and provide experimental results obtained from a system fabricated in 0.5μ CMOS process. Preliminary results indicate that margin based decoding is robust to input noise and parameter mismatches and can provide superior performance as compared to usual probability propagation.