Errors in solving inverse problem for reversing RTN effects on VCCmin shift in SRAM reliability screening test designs

This paper discusses how much error of VCCmin estimation does happen and proposes how to reduce the errors in solving an inverse problem for reversing the relationships between effects of Random Telegraph Noise (RTN) and/or Random dopant fluctualtion (RDF) on lifetime overall SRAM margin variations. Several calculation techniques with various MATLAB-built-in deconvolution-functions are compared with regard to error of deconvolution. The proposed technique successfully reduces the error of deconvolution thanks to eliminating the need for any operations of differential, division, and maximum-likelihood gradient sequence. This advantage over the MATLAB-built-in deconvolution-functions has been demonstrated for the first time with applying it to a real reliability screening test design for the effects of the RTN on the lifetime SRAM margin variations. It has been shown that the proposed technique can reduce the estimation errors for fail-bit-count by 1014-fold and VCCmin shift by about 80mV compared with the MATLAB-built-in deconvolution-functions.