When your signal is a handful of photons — think scintillation counters, LIDAR return pulses, fluorescence spectroscopy, or PET scanners — a regular photodiode won't cut it. You need internal gain to lift a single photoelectron above the amplifier noise floor. Two technologies dominate: the venerable photomultiplier tube (PMT) and the solid-state upstart, the silicon photomultiplier (SiPM).

A PMT is a vacuum tube. A photon strikes a photocathode, ejecting one electron. That electron accelerates through 800-2000 V across a chain of 8-12 dynodes, each releasing 3-5 secondary electrons per impact. Net gain: 10⁶ to 10⁷. Rise times of 1-2 ns are routine. But PMTs are fragile, sensitive to magnetic fields, need HV supplies, and get permanently damaged by exposure to room light while biased.

A SiPM is an array of hundreds to thousands of tiny avalanche photodiodes (microcells, ~20-50 μm each) reverse-biased above breakdown in Geiger mode. Each cell fires a fixed-charge avalanche pulse when a photon triggers it, then quenches via a series resistor (~100 kΩ-1 MΩ integrated on-chip). The outputs sum on a common bus. Bias voltage is only 25-60 V. Gain matches a PMT (~10⁶), and quantum efficiency (PDE, "photon detection efficiency") of 40-60% beats most PMTs (20-30%).

Real-world example: In a handheld gamma spectrometer with a CsI(Tl) scintillator, a 1 MeV gamma ray produces roughly 50,000 optical photons over ~1 μs. Coupled to a 6×6 mm SiPM with 40% PDE and 10,000 microcells, you'll fire around 20,000 cells — well below saturation, giving a linear energy response. Swap in a PMT and you'd need a stable 1000 V supply and mu-metal shielding; the SiPM runs off a boost converter from a 3.7 V lithium cell.

Rule of thumb — SiPM dynamic range: keep your peak photon count below 25% of total microcells to stay linear. Above that, cells recharge (dead time ~50-100 ns) faster than they can re-fire, and response compresses logarithmically. For N microcells and expected photons P with PDE η, linearity holds while ηP < 0.25N.

Watch out for SiPM dark count rate (DCR): thermal carriers trigger avalanches too. At room temperature, expect 50-100 kHz/mm² of dark pulses indistinguishable from single photons. Cool to -20 °C and DCR drops ~50×. PMTs, by contrast, have dark rates of only 100-1000 Hz total — a real advantage for photon-counting at long integration times.