Statistical Write Stability Characterization in SRAM Cells at Low Supply Voltage

Abstract

Four write stability metrics for the characterization of six-transistor SRAM cells were experimentally evaluated and compared at low supply voltage (V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DD</sub> ). A silicon-on-thin-BOX technology with reduced body doping was used to achieve low voltage operation. It was confirmed that both bitline and wordline methods are preferable in that they yield metrics that follow normal distributions, which is practically beneficial for the yield estimation. On the contrary, different from at high V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DD</sub> , both write static noise margin from write butterfly curve and write N-curve current (I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W</sub> ) exhibit non-normal probability distributions. The origins of the non-normality are analyzed in detail.