1) Bias: the diode is reverse-biased above breakdown by an excess voltage. 2) Absorption: a photon generates an electron-hole pair in the depletion region. 3) Avalanche: the strong electric field (>3×10^5 V/cm) drives impact ionization and a self-sustaining avalanche, yielding a fast current pulse. 4) Timestamp: the pulse leading edge, read by a discriminator and a time-to-digital converter (TDC), marks the photon arrival time. 5) Quench: a quenching circuit (passive/active) lowers the bias below breakdown to stop the avalanche. 6) Reset and dead time: the diode returns to its operating bias; during the dead time it cannot register photons. In dToF LiDAR, repeated time-of-flight measurements are accumulated into a histogram (TCSPC) from which distance is derived; SPAD arrays build a 3D depth image.
Conventional photodetectors (PIN photodiodes, and even APDs in linear mode) cannot detect a single photon or time its arrival with picosecond precision at chip scale. SPAD addresses detection at extremely low light levels and precise time-of-flight measurement, which is a prerequisite for digital dToF LiDAR, low-light 3D imaging, and single-photon counting techniques.
The depleted p-n junction where a strong electric field drives impact ionization and a self-sustaining avalanche after photon absorption.
A doped structure around the active area that prevents premature edge breakdown and confines the avalanche to the defined region.
A passive (ballast resistor) or active (discriminator that lowers the bias) circuit that stops the avalanche and resets the diode; it sets the dead time.
Official
Readout electronics that record the avalanche timing; in dToF LiDAR it accumulates repeated measurements into a time-of-flight histogram (TCSPC).
Official
Carriers trapped during the avalanche are released during recovery and trigger secondary avalanches, producing non-Poissonian statistics.
Thermally generated carriers cause false avalanches, limiting sensitivity, especially at room temperature and in InGaAs.
At high light levels the detector saturates because it registers no photons during the dead time, distorting counts.
R.J. McIntyre's work on microplasma instability and avalanche gain statistics, foundational for avalanche detectors.
Cova et al. in Applied Optics systematized single-photon detection with avalanche photodiodes and quenching circuits.
Demonstration of SPADs in standard CMOS technology opened the way to low-cost integrated arrays and on-chip readout.
Introduction of stacked SPAD dToF sensors (e.g. Sony IMX459) for automotive LiDAR; growth of megapixel arrays.
Voltage above breakdown; higher improves PDE but increases DCR and afterpulsing.
Recovery time after an avalanche; sets the maximum count rate (active quenching: ~1–2 ns).
Probability that an incident photon triggers an avalanche; depends on wavelength and excess bias.
False counts from thermally generated carriers; CMOS SPADs can reach <10 Hz.
Uncertainty of the timestamp; modern CMOS SPADs ~50 ps.
Si (visible/NIR), Ge-on-Si and InGaAs/InP (telecom infrared).
Pixel count and active-area fraction; modern arrays >1 Mpx with >70% fill factor.
