Static random-access memory designs based on different FinFET at lower technology node (7nm)
Metadata[+] Show full item record
The Static Random-Access Memory (SRAM) has a significant performance impact on current nanoelectronics systems. To improve SRAM efficiency, it is important to utilize emerging technologies to overcome short-channel effects (SCE) of conventional CMOS. FinFET devices are promising emerging devices that can be utilized to improve the performance of SRAM designs at lower technology nodes. In this thesis, I present detail analysis of SRAM cells using different types of FinFET devices at 7nm technology. From the analysis, it can be concluded that the performance of both 6T and 8T SRAM designs are improved. 6T SRAM achieves a 44.97% improvement in the read energy compared to 8T SRAM. However, 6T SRAM write energy degraded by 3.16% compared to 8T SRAM. Read stability and write ability of SRAM cells are determined using Static Noise Margin and N- curve methods. Moreover, Monte Carlo simulations are performed on the SRAM cells to evaluate process variations. Simulations were done in HSPICE using 7nm Asymmetrical Underlap FinFET technology. The quasiplanar FinFET structure gained considerable attention because of the ease of the fabrication process  – . Scaling of technology have degraded the performance of CMOS designs because of the short channel effects (SCEs) , . Therefore, there has been upsurge in demand for FinFET devices for emerging market segments including artificial intelligence and cloud computing (AI) , , Internet of Things (IoT)  –  and biomedical  – which have their own exclusive style of design. In recent years, many Underlapped FinFET devices were proposed to have better control of the SCEs in the sub-nanometer technologies , ,  – . Underlap on either side of the gate increases effective channel length as seen by the charge carriers. Consequently, the source-to-drain tunneling probability is improved. Moreover, edge direct tunneling leakage components can be reduced by controlling the electric field at the gate-drain junction . There is a limitation on the extent of underlap on drain or source sides because the ION is lower for larger underlap. Additionally, FinFET based designs have major width quantization issue. The width of a FinFET device increases only in quanta of silicon fin height (HFIN) . The width quantization issue becomes critical for ratioed designs like SRAMs, where proper sizing of the transistors is essential for fault-free operation. FinFETs based on Design/Technology Co-Optimization (DTCO_F) approach can overcome these issues . DTCO_F follows special design rules, which provides the specifications for the standard SRAM cells with special spacing rules and low leakages. The performances of 6T SRAM designs implemented by different FinFET devices are compared for different pull-up, pull down and pass gate transistor (PU: PD:PG) ratios to identify the best FinFET device for high speed and low power SRAM applications. Underlapped FinFETs (UF) and Design/Technology Co-Optimized FinFETs (DTCO_F) are used for the design and analysis. It is observed that with the PU: PD:PG ratios of 1:1:1 and 1:5:2 for the UF-SRAMs the read energy has degraded by 3.31% and 48.72% compared to the DTCO_F-SRAMs, respectively. However, the read energy with 2:5:2 ratio has improved by 32.71% in the UF-SRAM compared to the DTCO_F-SRAMs. The write energy with 1:1:1 configuration has improved by 642.27% in the UF-SRAM compared to the DTCO_F-SRAM. On the other hand, the write energy with 1:5:2 and 2:5:2 configurations have degraded by 86.26% and 96% in the UF-SRAMs compared to the DTCO_F-SRAMs. The stability and reliability of different SRAMs are also evaluated for 500mV supply. From the analysis, it can be concluded that Asymmetrical Underlapped FinFET is better for high-speed applications and DTCO FinFET for low power applications.
Table of Contents
Introduction -- Next generation high performance device: FinFET -- FinFET based SRAM bitcell designs -- Benchmarking of UF-SRAMs and DTCO-F-SRAMS -- Collaborative project -- Internship experience at INTEL and Marvell Semiconductor -- Conclusion and future work
M.S. (Master of Science)