Time-of-flight sensing measures distance by timing how long light takes to return from a target, and it has moved from lab technique to a commodity depth-sensing layer in phones, cars and robots, with the current research frontier in how single-photon detector arrays are read out rather than in the ranging principle itself.
Summary
Time-of-flight (ToF) ranging recovers distance from the round-trip delay of an emitted optical signal. Two implementation families dominate. Direct ToF (dToF) timestamps individual photon arrivals against the laser fire, which in practice means single-photon avalanche diodes (SPADs) feeding time-to-digital converters and building a per-pixel histogram whose peak is the range. Indirect ToF (iToF) modulates a continuous-wave source and infers delay from the phase of the returned light, using pixels closer to conventional CMOS imagers. dToF buys ambient-light robustness and long range at the cost of timing electronics per pixel; iToF buys pixel simplicity at the cost of ambiguity and multipath sensitivity. The same acronym is used for an unrelated analytical technique, time-of-flight secondary ion mass spectrometry (ToF-SIMS) and ToF mass spectrometry generally, which appears widely in materials and chemistry work such as MOF thin-film characterisation and autonomous SIMS campaigns; this page is about optical ranging.
The parameters that decide a dToF system are photon budget against solar or scene background, detector dead time, laser pulse width and the quantisation resolution of the timing electronics. A Cramer-Rao lower bound analysis for SPAD dToF shows that pile-up caused by dead time does more than lose counts: it statistically couples the distance estimate to the photon flux rate, degrading precision, and it sets an optimal rather than maximal illumination level. The same analysis finds that adding photon-number-resolving capability to the detector yields rapidly diminishing returns. In other words, the physics ceiling is now reasonably well mapped, and design effort has shifted to architecture.
On the supply side, the market scan places STMicroelectronics as the leading SPAD player on the strength of CMOS-integrated SPAD and vertically integrated manufacturing at consumer and automotive volume, with Sony Semiconductor Solutions and Hamamatsu Photonics in a second tier and the patent landscape showing ST ahead of Sony and PixArt. At the application end, ToF modules are now treated as commodity parts: an Arducam ToF depth camera supplies point clouds for a fruit sugar prescreening classifier, and ToF sits alongside camera and ultrasound in a fused depth-estimation stack for sub-100 mW microcontroller platforms.
Evidence base
Open questions
- Does asynchronous per-pixel peak detection hold up against strong solar background, where pile-up couples range and flux estimates, or does its SNR-threshold trigger degrade exactly when it is most needed?
- Can DMD-based spatial modulation that turned a 64x64 SPAD array into an effective 256x256 at 670 m be reduced to a fieldable module, or does it stay a bench technique?
- Does in-pixel digital compute justify its silicon area and power against simply shipping more conventional SPAD pixels from a vertically integrated incumbent?
- Can a fabless SPAD challenger second-source away from its largest competitor’s fab, and on what timeline?
Assessment drafted 2026-08-31 from up to 16 KB sources using the technology-scorecard framework; scores are a draft read pending review.