Dual pulse repetition rates with high background noise tolerance for memory-efficient SPAD LiDAR.

Abstract

Single-photon avalanche diode (SPAD) sensors for flash light detection and ranging (LiDAR) typically have high memory overhead. The widely adopted memory-efficient two-step coarse-fine (CF) process suffers from degraded background noise (BGN) tolerance. To alleviate this issue, we propose a dual pulse repetition rate (DPRR) scheme while maintaining a high histogram compression ratio (HCR). The scheme involves emitting narrow laser pulses at two different high rates in two phases, generating histograms and locating their respective peaks; then the actual distance can be derived based on the peak locations and repetition rates. Additionally, in this Letter, we propose spatial filtering within neighboring pixels with different repetition rate sets to cope with multiple reflections, which can potentially confuse the derivation, owing to the existence of several possible peak combinations. Compared with the CF approach, with a same HCR of 7, the simulations and experiments demonstrate that this scheme can tolerate 2 × BGN levels, accompanied with an increase in the frame rate by 4 ×.