Detectors
Choosing the receive side: avalanche photodiodes vs PIN photodiodes
Most of a ranging link's margin is recovered at the receiver, not the transmitter. Here is when an APD earns its 10x price, how gain and noise trade off, and what the specification numbers actually mean.
Updated 2026-10-07 · 4 min read
Why the receiver decides your range
In the range equation, range goes as the square root of received power. That means a receiver that detects 10x less light buys you 3.2x the distance - the same as increasing transmit energy 10x, but without touching the laser, the driver, the optics or the eye-safety classification.
Receiver margin is also cheaper than transmitter margin. Doubling the receive aperture costs one larger lens. Doubling transmit energy costs a bigger diode, a hotter driver and possibly a different safety class.
APD vs PIN in one paragraph each
A PIN photodiode converts photons to carriers with unity gain. Simple, linear, temperature-stable, and cheap. Its limit is that the signal current is tiny - at 1 nW of received power and 0.5 A/W responsivity you have 0.5 nA of signal competing with the thermal noise of the transimpedance amplifier.
An avalanche photodiode multiplies each photogenerated carrier through impact ionisation, giving internal gain of 10 to 200. The signal is amplified before the preamplifier sees it, so the preamplifier noise stops dominating. The price is excess noise: the multiplication process is random, so gain comes with a noise penalty that grows faster than the signal at high M.
The gain-versus-noise trade
Signal grows as M. Avalanche noise grows as Mx, where x is between 0.2 and 0.5 for silicon and around 0.7 to 1.0 for InGaAs. Total noise roughly follows:
At low gain the amplifier term dominates and SNR improves with M. Past an optimum - typically M = 50 to 150 for silicon, M = 10 to 40 for InGaAs - the avalanche noise takes over and more gain makes things worse.
This is why our 905 nm APDs specify an optimum gain of M = 100, and why a bias control loop that tracks M is worth designing. Running an APD at fixed bias across a 40 K temperature swing is the most common receiver mistake: at 0.7 V/°C of breakdown drift, an uncorrected 40 K shift moves the operating point far enough to lose several dB.
Reading the APD specification
Take LCC2-905APD-500/230:
| Parameter | Value | What it means for your budget |
|---|---|---|
| Responsivity | 80 A/W at M = 100 | High gain already applied; compare at equal M |
| Response time | 0.5 ns typ | ~700 MHz bandwidth, enough for 1 - 5 ns pulses |
| Dark current | 0.4 nA at M = 100 | Sets the noise floor before any signal arrives |
| Total capacitance | 1.0 pF | Low capacitance is what makes 0.5 ns possible |
| Breakdown voltage | 100 - 200 V | Your bias supply must reach this |
| Voltage temp. coefficient | 1.4 V/°C | Bias compensation needed: 56 V of drift over 40 K |
That last row is the one people miss. A 1.4 V/°C coefficient is comfortable to compensate. A 1.1 V/°C part with a strong-ambient-light filter - the LCC1 version - is a better fit for outdoor equipment where bias resolution matters less than daytime immunity.
When a PIN is the right answer
Short range. Under about 50 m, there is plenty of return signal and the APD's gain is unnecessary. A PIN with a good transimpedance amplifier is simpler, cheaper and has no bias drift to manage.
Very high bandwidth. Above roughly 1 GHz, APD gain bandwidth product becomes the limit and well-designed PIN receivers compete.
Multi-channel arrays. A 16-element APD array exists, but per-channel bias control for 16 elements is a serious engineering effort. If your architecture can tolerate the sensitivity loss, a PIN array with per-channel TIA gain is easier to build and calibrate.
When you need linearity. APDs compress at high optical input. If you are measuring amplitude rather than time-of-flight, a PIN is the honest choice.
Practical checklist
- Match the active area to the spot, not to the maximum available. A 500 µm APD has ~5x the capacitance of a 230 µm part, which directly costs bandwidth. If your return spot is 100 µm, buy the small one.
- Budget the filter. A 5 - 10 nm bandpass at the laser wavelength is what makes daylight operation possible. Without it, an APD's gain amplifies sunlight just as happily as signal.
- Compensate the bias. Either a temperature-compensated bias network or a closed-loop controller that holds gain at M = 100. Budget 2 - 4 dB of range if you skip this.
- Match the bandwidth to the pulse. BW ≈ 0.4 / tp. Wider is not better - it is louder.
- Protect it. APDs fail from ESD and from forward current. The datasheet's 1 mA forward-current limit is not a suggestion, and every bench test should have a series resistor and a grounded wrist strap.
Put the transmit and receive sides together with the selector - it pairs each recommended transmitter with a matched detector.
Related reading
905 nm vs 1535 nm vs 1550 nm: choosing a wavelength for ranging and LiDAR
The wavelength decision sets your detector, your eye-safety class, your maximum usable energy and your weather performance. Here is how the three bands actually compare for ranging work.
Selection guideHow to choose a 905 nm pulsed laser diode: from range target to part number
A repeatable procedure for turning a range requirement into pulse energy, pulse energy into peak power and pulse width, and peak power into the smallest package that survives your duty cycle.
OpticsFast-axis collimation and fibre coupling: getting the beam from the die to the target
An edge emitter's 25 degree fast axis wastes most of its energy on empty sky. Here is what a fast-axis collimator actually recovers, why it is a precondition for fibre coupling, and when you should not use one.