A 240 × 160 3D-Stacked SPAD dToF Image Sensor With Rolling Shutter and In-Pixel Histogram for Mobile Devices

Abstract

A 240 <inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> 160 single-photon avalanche diode (SPAD) sensor integrated with a 3D-stacked 65nm/65nm CMOS technology is reported for direct time-of-flight (dToF) 3D imaging in mobile devices. The top tier is occupied by backside illuminated SPADs with 16 <inline-formula> <tex-math notation="LaTeX">$\mu {\mathrm{ m}}$ </tex-math></inline-formula> pitch and 49.7&#x0025; fill-factor. The SPADS consists of multiple 16<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula>16 SPADs top groups, in which each of 8 <inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> 8 SPADs sub-group shares a 10-bit, 97.65ps and 100ns range time-to-digital converter (TDC) in a quad-partition rolling shutter mode. During the exposure of each rolling stage, partial histogramming readout (PHR) approach is implemented to compress photon events to in-pixel histograms. Since the fine histograms is incomplete, for the first time we propose histogram distortion correction (HDC) algorithm to solve the linearity discontinuity at the coarse bin edges. With this algorithm, depth measurement up to 9.5m achieves an accuracy of 1cm and precision of 9mm in office lighting condition. Outdoor measurement with 10 klux sunlight achieves a maximum distance detection of 4m at 20 fps, using a VCSEL laser with the average power of 90 mW and peak power of 15 W.