29.7 CCSA: A 394TOPS/W Mixed-Signal GPS Accelerator with Charge-Based Correlation Computing for Signal Acquisition
Jieyu Li,Weifeng He,Bo Zhang,Liang Qi,Guanghui He,Mingoo Seok
DOI: https://doi.org/10.1109/isscc42615.2023.10067275
2023-01-01
Abstract:A global positioning system (GPS) is one of the essential features in wearable devices, such as smartwatches. However, the power consumption of the GPS receiver remains a key bottleneck in extending the battery life of those devices. Among various digital signal processing tasks a GPS receiver performs, signal acquisition is the most computationally-complex task. It involves calculating the correlation of two pseudo-noise (PN) sequences, which requires a large number (10-350B) of multiply-and-accumulate (MAC) operations for each satellite [1]. The GPS receiver also performs signal acquisition frequently since it can lose track of satellites due to obstruction, interference, multipath, etc. To reduce power consumption, [1], [2] proposed switched-current-based accelerators. However, current-mode circuits hinder supply voltage scaling for further energy savings. Also, they require frequent calibrations across process, voltage and temperature (PVT) variations. This paper presents CCSA, a mixed-signal GPS accelerator with Charge-based correlation computing for Signal Acquisition. It employs 4096 mixed-signal multipliers, which take digital inputs and produce charge-domain outputs (charges). These charges are dumped and redistributed on a global output wire whose settled voltage level represents the correlation result. We prototyped the CCSA test chip in 28nm CMOS. It achieves 114-to-394TOPSIW energy efficiency with 0.05-to-3.56TOPS throughput across 0.34-0.9V. It exhibits $8.2\times$ higher energy efficiency than state-of-the-art [1] at the same throughput.