Quantifying Energy and Latency Improvements of FPGA-Based Sensors for Low-Cost Spectrum Monitoring

摘要

There is a recent interest in large-scale RF spectrum monitoring using low-cost crowdsourced spectrum sensors. A major challenge here is improving the latency and energy usage of signal processing algorithms on the sensor. This improves operational cost and effectiveness and also makes the sensors more responsive to the monitoring task. We specifically consider the case of signal detection using such sensors. Typical crowdsourced implementation using a low-cost software radio connected to a Raspberry Pi or a smartphone as host is not energy-efficient and incurs significant latencies. We propose use of field-programmable gate array (FPGA) to improve both metrics for the signal detection task. Our benchmarking shows significant improvements with FPGA platforms relative to using a Raspberry Pi or smartphone, upto a factor of 73 in terms of latency and a factor of 29 in terms of energy usage.