Book Description

Next Big Thing Is Surely Small: Nanotechnology Can Bring Revolution. Nanotechnology leads the world towards many new applications in various fields of computing, communication, defense, entertainment, medical, renewable energy and environment. These nanotechnology applications require an energy-efficient memory system to compute and process. Among all the memories, Static Random Access Memories (SRAMs) are high performance memories and occupies more than 50% of any design area. Therefore, it is critical to design high performance and energy-efficient SRAM design. Ultra low power and high speed applications require a new generation memory capable of operating at low power as well as low execution time. In this thesis, a novel 8T SRAM design is proposed that offers significantly faster access time and lowers energy consumption along with better read stability and write ability. The proposed design can be used in the conventional SRAM as well as in computationally intensive applications like neural networks and machine learning classifiers [1]-[4]. Novel 8T SRAM design offers higher energy efficiency, reliability, robustness and performance compared to the standard 6T and other existing 8T and 9T designs. It offers the advantages of a 10T SRAM without the additional area, delay and power overheads of the 10T SRAM. The proposed 8T SRAM would be able to overcome many other limitations of the conventional 6T and other 7T, 8T and 9T designs. The design employs single bitline for the write operation, therefore the number of write drivers are reduced. The defining feature of the proposed 8T SRAM is its hybrid design, which is the combination of two techniques: (i) the utilization of single-ended bitline and (ii) the utilization of virtual ground. The single-ended bitline technique ensures separate read and write operations, which eventually reduces the delay and power consumption during the read and write operations. It's independent read and write paths allow the use of the minimum sized access transistors and aid in a disturb-free read operation. The virtual ground weakens the positive feedback in the SRAM cell and improves its write ability. The virtual ground technique is also used to reduce leakages. The proposed design does not require precharging the bitlines for the read operation, which reduces the area and power overheads of the memory system by eliminating the precharging circuit. The design isolates the storage node from the read path, which improves the read stability. For reliability study, we have investigated the static noise margin (SNM) of the proposed 8T SRAM, for which, we have used two methods – (i) the traditional SNM method with the butterfly curve, (ii) the N-curve method A comparative analysis is performed between the proposed and the existing SRAM designs in terms of area, total power consumption during the read and write operations, and stability and reliability. All these advantages make the proposed 8T SRAM design an ideal candidate for the conventional and computationally intensive applications like machine learning classifier and deep learning neural network. In addition to this, there is need for next generation technologies to design SRAM memory because the conventional CMOS technology is approaching its physical and performance boundaries and as a consequence, becoming incompatible with ultra-low-power applications. Emerging devices such as Tunnel Field Effect Transistor (TFET)) and Graphene Nanoribbon Field Effect Transistor (GNRFET) devices are highly potential candidates to overcome the limitations of MOSFET because of their ability to achieve subthreshold slopes below 60 mV/decade and very low leakage currents [6]-[9]. This research also explores novel TFET and GNRFET based 6T SRAM. The thesis evaluates the standby leakage power in the Tunnel FET (TFET) based 6T SRAM cell for different pull-up, pull-down, and pass-gate transistors ratios (PU: PD: PG) and compared to 10nm FinFET based 6T SRAM designs. It is observed that the 10nm TFET based SRAMs have 107.57%, 163.64%, and 140.44% less standby leakage power compared to the 10nm FinFET based SRAMs when the PU: PD: PG ratios are 1:1:1, 1:5:2 and 2:5:2, respectively. The thesis also presents an analysis of the stability and reliability of sub-10nm TFET based 6T SRAM circuit with a reduced supply voltage of 500mV. The static noise margin (SNM), which is a critical measure of SRAM stability and reliability, is determined for hold, read and write operations of the 6T TFET SRAM cell. The robustness of the optimized TFET based 6T SRAM circuit is also evaluated at different supply voltages. Simulations were done in HSPICE and Cadence tools. From the analysis, it is clear that the main advantage of the TFET based SRAM would be the significant improvement in terms of leakage or standby power consumption. Compared to the FinFET based SRAM the standby leakage power of the T-SRAMs are 107.57%, 163.64%, and 140.44% less for 1:1:1, 1:5:2 and 2:5:2 configurations, respectively. Since leakage/standby power is the primary source of power consumption in the SRAM, and the overall system energy efficiency depends on SRAM power consumption, TFET based SRAM would lead to massive improvement of the energy efficiency of the system. Therefore, T-SRAMs are more suitable for ultra-low power applications. In addition to this, the thesis evaluates the standby leakage power of types of Graphene Nanoribbon FETs based 6T SRAM bitcell and compared to 10nm FinFET based 6T SRAM bitcell. It is observed that the 10nm MOS type GNRFET based SRAMs have 16.43 times less standby leakage power compared to the 10nm FinFET based SRAMs. The double gate SB-GNRFET based SRAM consumes 1.35E+03 times less energy compared to the 10nm FinFET based SRAM during write. However, during read double gate SB-GNRFET based SRAM consume 15 times more energy than FinFET based SRAM. It is also observed that GNRFET based SRAMs are more stable and reliable than FinFET based SRAM.