Analyzing the effect of concurrent variability in the core cells and sense amplifiers on SRAM read access failures
Résumé
Conventionally, the access failures in SRAMs are treated at core cell level by means of differential bit line voltage analysis. In this work it is shown that under the assumption of random process variability, the conventional approach no longer suffices. It still holds that the differential bit line voltage is degraded by the variability in core cell transistors, but the way this voltage difference is interpreted by the sense amplifier to complete the read operation is influenced by random variability affecting its transistors. Case studies show how variability affecting the sense amplifier can degrade or improve its ability to read the data stored by the core cell, which is itself affected by variability. Using principal component analysis and the SB-SI method, we performed a parametric analysis of the sense amplifier/core cell system and we evaluated the joint probability of access failure. A three times increase in the failure probability has been observed when compared to cell's failure probability. Also, the minimum variability value for which a failure is observed is ~2.5X smaller when joint variability is assumed compared to the case when only the core cell is affected by variability.
Mots clés
joint probability
failure probability
parametric analysis
random process variability
read operation
random variability
sense amplifier cell system
sense amplifier core cell system
voltage difference
sense amplifiers
Arrays
SRAM cells
Discharges (electric)
Threshold voltage
Measurement
differential bit line voltage analysis
core cells
core cell transistors
core cell level
concurrent variability
SRAM read access failures
SB-SI method
probability
principal component analysis
failure analysis
SRAM chips
Sense Amplifier
SRAM Memory
Failure Probability Estimation
Transistors