A technique to solve issue of Richardson-Lucy deconvolution for analyzing RTN effects on SRAM margin variation

Abstract

This paper demonstrates the issues of Richardson-Lucy (R-L) deconvolution technique when applying it to analyze the Random Telegraph Noise (RTN) effects on SRAM margin modulation and proposes the techniques to address the issues. Unlike the R-L based deconvolution, the proposed technique successfully circumvents the issue of ringing error thanks to eliminating any operations of differential and division. This effectiveness has been demonstrated for the first time with applying it to a real analysis for the effects of the RTN on the overall SRAM margin variations. It has been shown that the proposed technique can reduce its relative errors of the RTN deconvolution by 1014 and 104 times compared with the cases of the R-L and without segmentation, respectively. This enables to reduce the cdf error of the convolution of the RTN with the Random Dopant Fluctuation (RDF) (i.e., fail-bit-count error) by 10-orders of magnitude of the conventional algorithm.