We explore the 6T and 8T SRAM design spaces through read static noise margin (RSNM), word-line write margin, and leakage for future 10-nm FinFETs. Process variations are based on the ITRS and modeled at device (TCAD) level. We propose a method to incorporate them into a BSIM-CMG model card for time-efficient simulation. We analyze cells with different fin numbers, supply voltages, and temperatures. Results show a 1.8× improvement of RSNM for 8T SRAM cells, the need for stronger pull-downs to secure read stability in 6Ts, and high leakage sensitivity to temperature (10× between 40°C and 100°C). As a specific example, we show how the RSNM of a 6T SRAM cell can be improved by using back-gate biasing techniques for independent-gate FinFETs. We show how WLMN is increased by reducing the strength of pull-up transistors when reverse back-gate biasing is applied on it and how the RSNM can be increased by reducing the strength of access transistor by reverse back-gate biasing of pass-gate transistors. When combining these two techniques, RSNM can be improved up to 25% without compromising cell write ability for any sample. In general, when compared to previous technologies, read stability is untouched, writeability is reduced, and leakage keeps stable.

Peer Reviewed