Detector Selection & Comparison — Abridged Guide

Quick-reference guide to optical detector types, performance metrics, and selection. For full derivations and worked examples, see the Comprehensive Guide.

Comprehensive Detector Selection Guide →

1.Introduction to Optical Detection

Optical detectors convert photons into electrical signals. They divide into two classes: photon detectors (wavelength-selective, fast, high sensitivity) and thermal detectors (wavelength-flat, slow, lower sensitivity). Point detectors produce a single signal; array detectors provide spatial or spectral imaging.

Start every detector selection by classifying your measurement: what wavelength, how much power, and how fast? These three questions eliminate most of the candidate pool immediately.

2.Detector Classification & Taxonomy

Five major subcategories exist: photoemissive (PMTs — highest gain, single-photon capable), photovoltaic (photodiodes, APDs, SiPMs — versatile solid-state), photoconductive (PbS, PbSe, HgCdTe — mid-IR coverage), thermal (pyroelectric, bolometer, thermopile — broadband, room temperature), and array (CCD, CMOS, EMCCD — imaging and multichannel spectroscopy).

Type	Gain	Speed	Best For
PIN Photodiode	1	ps–ns	General purpose, high linearity
APD	10–200	ns	Moderate-gain speed applications
PMT	10⁵–10⁷	ns	Single-photon, UV-VIS
SiPM	10⁵–10⁶	ns	Compact photon counting
PbS/PbSe	1	µs	Mid-IR, low cost
HgCdTe	1	ns–µs	MWIR/LWIR, cooled
Pyroelectric	1	ms	Broadband reference
CCD/CMOS	1	Frame-limited	Imaging, spectroscopy

If your wavelength is 190–1100 nm and you do not need single-photon sensitivity, a silicon photodiode is almost always the right starting point. It is the cheapest, simplest, most reliable detector available.

3.Responsivity and Quantum Efficiency

Responsivity

R = \eta \frac{e\lambda}{hc} \approx \eta \frac{\lambda\,(\text{nm})}{1240} \quad \text{(A/W)}

Responsivity (A/W) measures photocurrent per watt of incident light. It increases linearly with wavelength for constant QE because longer-wavelength photons carry less energy, so more photons arrive per watt. QE is the more fundamental metric — it tells how efficiently photons convert to electrons regardless of wavelength.

QE ↔ Responsivity Conversion

\eta = \frac{R \times 1240}{\lambda\,(\text{nm})}

When comparing detectors at different wavelengths, convert to QE rather than comparing responsivity directly. A detector with R = 1.0 A/W at 1550 nm (η = 80%) is not more efficient than one with R = 0.4 A/W at 500 nm (η = 99%).

4.Noise Sources in Photodetectors

Shot Noise

i_{sn} = \sqrt{2eI_{ph}\Delta f}

Johnson Noise

i_J = \sqrt{\frac{4k_BT\Delta f}{R_L}}

The dominant noise source depends on operating conditions. High-light/high-speed measurements are shot-noise-limited. Low-light/low-speed measurements are often Johnson-noise or amplifier-noise limited. Photoconductive detectors (PbS, PbSe) add strong 1/f noise below ~1 kHz — always use chopped illumination with these.

Noise adds in quadrature (root-sum-of-squares). Only the largest noise term matters significantly — if one noise source is 3× larger than all others combined, the total is within 6% of the dominant term alone.

5.Noise Equivalent Power and Detectivity

NEP

\text{NEP} = \frac{i_n}{R} \quad \left(\frac{\text{W}}{\sqrt{\text{Hz}}}\right)

Minimum Detectable Power

P_{min} = \text{NEP} \times \sqrt{\Delta f}

Specific Detectivity

D^* = \frac{\sqrt{A_d \cdot \Delta f}}{\text{NEP}} \quad (\text{Jones})

NEP is the most practical sensitivity metric — it directly predicts the minimum detectable power for any bandwidth. D* normalizes for detector area and bandwidth, making it the correct metric for comparing detector quality across different sizes. Lower NEP = more sensitive system; higher D* = better intrinsic detector material.

Cutting your measurement bandwidth by 100× improves your detection limit by 10×. Before upgrading the detector, consider whether a narrower bandwidth (lock-in amplifier, longer integration) achieves the same result at lower cost.

6.Bandwidth, Speed, and Temporal Response

RC-Limited Bandwidth

f_{3dB} = \frac{1}{2\pi R_L C_j}

Bandwidth–Rise Time

t_r \approx \frac{0.35}{f_{3dB}}

Detector bandwidth is limited by either the RC time constant (junction capacitance × load resistance) or the carrier transit time across the depletion region, whichever is slower. Larger detector areas have higher capacitance and lower bandwidth. Reverse bias reduces capacitance and increases speed.

For a 50 Ω system, estimate bandwidth as f_3dB ≈ 3.2 GHz / C_j(pF). For a 10 kΩ transimpedance, it drops to f_3dB ≈ 16 MHz / C_j(pF). The load resistance choice is the single biggest lever on photodiode bandwidth.

7.Spectral Coverage by Detector Material

Cutoff Wavelength

\lambda_c = \frac{1240}{E_g\,(\text{eV})} \quad \text{(nm)}

Material	Range	Notes
Si	190–1100 nm	Default for UV-VIS-NIR, lowest cost
Ge	800–1800 nm	Broadband NIR, higher dark current
InGaAs	900–1700 nm	Best NIR performance, telecom standard
Ext. InGaAs	900–2600 nm	SWIR, requires TE cooling
PbS	1–3.2 µm	Low cost mid-IR, slow
PbSe	1–5 µm	Low cost mid-IR, slow
InSb	1–5.5 µm	MWIR, requires 77 K cooling
HgCdTe	1–16 µm	Tunable bandgap, requires cooling
Thermal	All	Flat response, low D*, room temp

Detector material is selected by wavelength first. Within a given spectral band, the choice between materials balances dark current, speed, cooling requirements, and cost.

At 1550 nm, InGaAs outperforms Ge by ~100× in dark current and ~10× in D*. Never use Ge where InGaAs covers the wavelength — the only exception is if you need broader spectral coverage spanning the Si-to-InGaAs gap.

8.Internal Gain Mechanisms

PMT Gain

G = \delta^n

APD Excess Noise Factor

F(M) = kM + (1-k)\left(2 - \frac{1}{M}\right)

Detector	Gain	Excess Noise (F)	Voltage	Best For
PMT	10⁵–10⁷	1.1–1.2	1000–2000 V	Photon counting, UV-VIS
Si APD	10–200	2–5	100–500 V	Moderate-gain visible/NIR
InGaAs APD	10–40	5–10	50–80 V	Telecom, lidar (1550 nm)
SiPM	10⁵–10⁶	~1.1 (per cell)	25–75 V	Compact photon counting

Gain helps only when amplifier noise dominates. In the shot-noise-limited regime, gain adds excess noise without improving SNR. PMTs have the highest gain with the lowest excess noise; APDs trade lower gain for solid-state simplicity.

If your signal is above ~10 nW on a well-designed photodiode receiver, you probably do not need a gain detector. Below ~100 pW, you almost certainly do.

9.Practical Selection Criteria

Beyond the core optical specs (wavelength, sensitivity, speed), practical constraints often determine the final choice: cooling (field-portable → no cryogenics), magnetic environment (MRI → no PMTs), dynamic range (power meter → photodiode), cost, and size. Operating mode matters too — photovoltaic for low-noise DC, photoconductive for high-speed.

The cheapest, simplest detector that meets the SNR requirement with ≥ 20 dB margin is almost always the right choice. Over-specifying the detector wastes budget and adds complexity without measurable benefit.

10.Detector Selection Workflow

A six-step decision framework: (1) wavelength → material, (2) signal level → gain requirement, (3) speed → bandwidth class, (4) spatial needs → point vs. array, (5) calculate SNR to verify adequacy, (6) apply practical constraints (cooling, cost, size, environment).

The most common selection mistake is choosing a detector by a single spec (e.g., highest QE) rather than evaluating the full system SNR. A detector with 95% QE paired with a noisy amplifier loses to one with 70% QE and a well-matched low-noise amplifier.

Continue Learning

Comprehensive Guide →Detector SNR & Comparison Calculator →

The Comprehensive Guide includes 7 worked examples, 6 SVG diagrams, 3 data tables, and 10 references.