Design and Test of Embedded SRAMs
A. Pavlov · 2005
Embedded SRAMs can occupy the majority of the chip area in SoCs. The increased process spreads of modern scaled-down technologies and non-catastrophic defect-related sensitivity to environmental parameters can compromise the stability of SRAM cells, which is quantified by a low Static Noise Margin (SNM). A Stability Fault (SF) can present itself in a cell whose SNM is so small that it can accidentally flip in the worst-case operating conditions. In this work, we conduct a comprehensive SRAM SNM sensitivity analysis and identify the major factors causing low SNM. Based on this study, we propose a Weak Cell Fault Model, which can be used in fault simulations to mimic an SRAM cell with a compromised SNM. Furthermore, we have derived an analytical expression for the SNM of the recently proposed loadless 4T SRAM cell. Reading a 6T SRAM cell with bit lines precharged to VDD may not detect several types of defects in the pull-up path of the cell. Such defects can cause the SFs. Regular SRAM March Tests are shown to have extremely limited ability to detect SRAM cells with potential SFs. The traditional Data Retention Test (DRT) is costly in terms of the test time and fails to detect open defects in the cell’s pull-up path of less than 50MΩ, even if provided unlimited pause time at 150C (0.13μm technology). These factors show the disparity between the existing SRAM test practices and the need for an economical and low PPM cell stability tests. We introduce the SRAM Cell Stability Detection Concept explaining the mechanism of the weak cell detection. Based on this concept, we propose three novel programmable Design for Testability (DFT) techniques capable of detecting the SFs and replacing the DRT. For verification of the proposed techniques, we have designed two fully functional SRAM test chips: an asynchronous SRAM (CMOS 0.18μm technology) and a synchronous SRAM (CMOS 0.13μm technology). The simulation and measurement results have proven