BatDrone: A 9.83M-focal-points/s 7.76s-Latency Ultrasound Imaging System with On-Chip Per-Voxel RX Beamfocusing for 7m-Range Drone Applications

Abstract

Imaging technologies are important in the era of drones/mobile vision systems. A CIS provides high-fidelity images with low latency, but it is vulnerable to low/complex light conditions and lacks depth information; LiDAR provides depth information but is bulky and over 50W in power [1]. An Ultrasound Imaging Sensor (UIS), on the other hand, is a promising choice for depth sensing under all-light conditions (day/night/complex); it also consumes orders of magnitude less power than LiDAR with a smaller form-factor. However, conventional UIS systems often rely on off-chip image processing [2]-[4] and use multi-shot TX beamforming, causing multi-Watt power consumption and slow response. Moreover, many are designed for short-range biomedical diagnostics [2]-[5] and are difficult to adapt to mid-range fast-response (sub-50ms) imaging applications such as drone/mobile vision systems. To address these issues, we present a UIS SoC that adopts 1-shot TX with a fully on-chip per-voxel RX beamfocusing (PV-RXBF) (Fig. 32.1.1). The SoC shows 7.76s image reconstruction latency at 9.83M focal-points/s while consuming 142.3mW and covering 7m range.