Ultra-low power digital integrated circuit design

• Main Author: Le, Van Loi
• Other Authors: Kim Tae Hyoung
• Format: Theses and Dissertations
• Language: English
• Published: 2019
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/85238; http://hdl.handle.net/10220/49202
• Institution: Nanyang Technological University

Relentless advancement of process technology has tremendously brought massive performance boosts and much more integration to modern digital circuits and systems. However, this comes at the cost of the increasingly high power consumption. In addition, there have been increasingly on-going demands for portable hand-held devices and wireless sensor nodes. Nevertheless, their small form factors can accommodate only limited-sized batteries, necessitating low-power consumption. Thus, low power has become one of the most critical design constraints in integrated circuit design. Flip-flops (FFs) are fundamental components in modern digital ICs; they dominate the total area and power of the overall systems. FFs typically consume large dynamic power in systems on chips (SoCs) due to redundant transitions of their large number of internal clocked nodes. Several low-power FFs have been recently proposed for reducing or eliminating power penalty coming from the redundant transitions. However, they suffer from either large area penalty, which is too costly since FFs typically occupy a significant amount of logic area in SoCs, or functional failures at low-voltage operations, which makes these FFs inapplicable for low-voltage applications. In this work, we present a novel FF circuit that eliminates the redundant transitions of local clocked nodes for achieving extremely low-power consumption without extra transistors while further improving performance and minimum operating voltage. Voltage scaling is one of the most widely used techniques for reducing power consumption in low power digital circuits and systems. In fact, digital cores on today’s processors have been increasingly voltage-scaled, and the number of cores has also been significantly increased. Thus, level shifters (LSs) have increasingly become one of the major concerns, especially in aggressively voltage-scaled systems. The recently proposed low-voltage LSs exhibit the trade-off between power, area, and delay. In this work, we propose an ultra-low voltage, area, and energy-efficient LS for fast and wide-range level conversion. The proposed LS exploits a novel reduced-swing buffer design to achieve low static power consumption while a pass transistor enhances the speed of the fall transition so that the conversion delay can be significantly improved. In addition, the proposed LS has the smallest area compared to the previous ultra-low voltage LSs. Gesture recognition has increasingly become one of the most popular human-machine interaction techniques for smart devices. Recently introduced gesture recognition systems suffer from excessive power consumption and large area, limiting their applications for ultra-low power IoT and wearable devices. Thus, a low-power and low-area gesture recognition system is required. In this work, we present two accurate low-power and area-efficient real-time gesture recognition systems for smart wearable devices. The proposed systems utilize efficient memory-reduced gesture recognition engines for achieving area-efficient and low-power accurate gesture recognition. In addition, the feature extraction architectures exploit parallelisms for obtaining real-time gesture recognition operations and remove fixed-pattern noises for further improving accuracy.