Optical detectors: photodiodes and avalanche photodiodes
PIN and avalanche photodiodes: cut-off wavelength, quantum efficiency, responsivity, gain and excess noise, shot and thermal noise, receiver SNR and speed.
Drafted with Aria, reviewed by the AiCanCode.org team. Spotted an error? Use Give Feedback at the bottom of the page.
Why it matters
The photodetector turns light back into current at the end of every optical link and inside every optical instrument — power meters, spectrophotometers, encoders, smoke detectors, lidar and fibre sensors. Its responsivity, speed and noise decide the weakest signal you can measure and the bit rate you can receive. Choosing between a PIN photodiode and an avalanche photodiode (APD) is a classic sensitivity-versus-cost-and-complexity decision.
Key ideas
Absorption and photocurrent. A photon with energy above the bandgap (hν ≥ E_g) can create an electron–hole pair. In a reverse-biased junction the electric field in the depletion region sweeps the pair apart, producing a current proportional to the optical power. Photons with λ longer than the cut-off λ_c = hc/E_g (λ_c in µm ≈ 1.24/E_g in eV) are not absorbed: silicon (E_g = 1.12 eV) works up to about 1.1 µm, so 1.3 µm and 1.55 µm links use InGaAs or germanium.
Quantum efficiency and responsivity. Quantum efficiency η is electrons collected per incident photon (0 to 1). Responsivity is current per watt: R = I_p/P = ηqλ/(hc), i.e. R ≈ η·λ(µm)/1.24 A/W. For a fixed η, R rises linearly with wavelength up to λ_c and then falls sharply.
PN and PIN photodiodes. In a simple PN diode much light is absorbed outside the thin depletion region, and carriers there move slowly by diffusion. A PIN diode inserts a wide, lightly doped intrinsic layer so nearly all absorption happens in the high-field region: higher η, faster drift-limited response and lower junction capacitance. Speed is limited by the transit time across the depletion layer and by the RC time constant of junction capacitance with the load: f_RC = 1/(2πR_L·C_j). A thicker intrinsic layer improves η but increases transit time — the basic trade-off.
Operating modes. Photoconductive mode (reverse bias) gives a wide depletion region, low capacitance and fast linear response, at the cost of dark current. Photovoltaic mode (zero bias) gives no dark current and the lowest noise for low-frequency, precision light measurement (power meters, solar cells), but is slower. A transimpedance amplifier holds the diode near zero volts and converts current to voltage.
Avalanche photodiodes. An APD adds a high-field multiplication region, biased just below breakdown (tens to hundreds of volts). Photo-generated carriers gain enough energy to create further pairs by impact ionisation, giving an internal gain M (typically 10–100): I = M·R·P. The multiplication is random, adding an excess noise factor F(M) ≈ M^x, where x is about 0.3 for Si, 0.7 for InGaAs and 1.0 for Ge. APDs help when the receiver is limited by amplifier thermal noise; there is an optimum M beyond which multiplied shot noise dominates. The gain is strongly temperature dependent: breakdown voltage rises with temperature, so the bias must be temperature-compensated.
Noise. Shot noise of the photocurrent and dark current, i_s² = 2q(I_p + I_d)B (times M²F for an APD), and thermal noise of the load, i_t² = 4kTB/R_L. Their sum against the signal power I_p² gives the receiver SNR, which sets the bit-error rate in a digital link. Other figures of merit: noise-equivalent power (NEP, the optical power giving SNR = 1 in 1 Hz) and detectivity D* for infrared detectors.
Other detectors. Phototransistors (gain but slow), photomultiplier tubes (very high gain, single-photon, used in spectroscopy), photoconductive cells (LDRs), CCD and CMOS arrays, and thermal detectors (bolometers, thermopiles, pyroelectric) whose response does not depend on photon energy.
Formulas
λ_c = hc/E_g, λ_c(µm) ≈ 1.24/E_g(eV)
λ_c cut-off wavelength, E_g bandgap.
η = (I_p/q)/(P/hν), R = I_p/P = ηqλ/(hc) ≈ η·λ(µm)/1.24
η quantum efficiency, I_p photocurrent (A), P incident optical power (W), R responsivity (A/W), q = 1.602 × 10⁻¹⁹ C.
I_p = q·η·Φ
Φ incident photon flux (photons/s).
I_APD = M·R·P, F(M) ≈ M^x
M avalanche gain, F excess noise factor, x material exponent.
i_s² = 2q(I_p + I_d)B·M²F (M = F = 1 for PIN), i_t² = 4kTB/R_L
I_d dark current (A), B bandwidth (Hz), k = 1.381 × 10⁻²³ J/K, T temperature (K), R_L load (Ω); noise currents in A².
SNR = (M·I_p)²/[2q(I_p + I_d)B·M²F + 4kTB/R_L]
f_RC = 1/(2πR_L·C_j)
C_j junction capacitance (F).
Worked examples
Example 1 (standard). An InGaAs PIN photodiode has η = 0.8 at 1.3 µm. Find its responsivity and the photocurrent for 10 µW incident. Can a silicon diode detect this light?
R = η·λ/1.24 = 0.8 × 1.3/1.24 = 0.839 A/W.I_p = R·P = 0.839 × 10 × 10⁻⁶ = 8.39 µA.- Silicon cut-off
= 1.24/1.12 = 1.11 µm < 1.3 µm, so silicon cannot absorb it. Answer: R ≈ 0.84 A/W, I_p ≈ 8.4 µA; silicon is unusable at 1.3 µm.
Example 2 (GATE level). A receiver with responsivity 0.9 A/W receives 1 µW in a bandwidth of 100 MHz with a 1 kΩ load at 300 K (ignore dark current). Find the SNR for (a) a PIN diode and (b) an APD with M = 10 and x = 0.7 at the same responsivity.
I_p = 0.9 × 10⁻⁶ = 0.9 µA.- Shot noise:
2qI_pB = 2 × 1.602 × 10⁻¹⁹ × 0.9 × 10⁻⁶ × 10⁸ = 2.88 × 10⁻¹⁷ A². - Thermal noise:
4kTB/R_L = 4 × 1.381 × 10⁻²³ × 300 × 10⁸/10³ = 1.66 × 10⁻¹⁵ A². - (a)
SNR = (0.9 × 10⁻⁶)²/(2.88 × 10⁻¹⁷ + 1.66 × 10⁻¹⁵) = 8.1 × 10⁻¹³/1.686 × 10⁻¹⁵ = 480→ 26.8 dB. - (b)
F = 10^0.7 = 5.01; signal(M·I_p)² = 8.1 × 10⁻¹¹ A²; noise= 2.88 × 10⁻¹⁷ × 100 × 5.01 + 1.66 × 10⁻¹⁵ = 1.44 × 10⁻¹⁴ + 0.17 × 10⁻¹⁴ = 1.61 × 10⁻¹⁴ A². SNR = 8.1 × 10⁻¹¹/1.61 × 10⁻¹⁴ = 5030→ 37.0 dB. Answer: about 26.8 dB with the PIN and 37.0 dB with the APD — a 10 dB gain because the PIN receiver was thermal-noise limited.
Example 3 (photon counting). A detector with η = 0.8 receives 5 × 10¹² photons/s. Find the photocurrent.
I_p = q·η·Φ = 1.602 × 10⁻¹⁹ × 0.8 × 5 × 10¹². Answer: I_p ≈ 6.4 × 10⁻⁷ A = 0.64 µA.
Common mistakes
- Using λ in nm in R = ηλ/1.24 (it needs µm).
- Saying responsivity is constant with wavelength; it rises with λ up to cut-off for fixed η.
- Assuming more APD gain always helps; excess noise makes an optimum gain.
- Forgetting that APD noise multiplies by M²F, not M².
- Adding noise amplitudes instead of mean-square values.
- Saying an unbiased photodiode gives much less current; at low frequencies the photocurrent is the same, but the response is slower and less linear at high power.
For GATE IN
- Responsivity, quantum efficiency and cut-off wavelength numericals.
- Photocurrent from optical power or photon flux; APD output current with gain.
- Shot and thermal noise, receiver SNR, and when an APD improves SNR.
- Bandwidth from RC time constant of the junction capacitance and load.
Quick check
- A detector has η = 0.6 at 0.85 µm. What is R?
- What is the cut-off wavelength of germanium (E_g = 0.66 eV)?
- A PIN gives 2 µA; an APD with M = 25 sees the same light. What current does the APD give?
- Which noise does an APD help to overcome? Answers: 1. 0.41 A/W. 2. About 1.88 µm. 3. 50 µA. 4. Thermal (amplifier) noise.
Interview questions
All Communication and Optical Instrumentation interview questionsTry answering each one aloud before you open it.
1.What is a photodiode and how does it work?Concept
A photodiode is a semiconductor device that converts light into an electrical current. It operates by the principle of the photoelectric effect, where photons hitting the semiconductor material generate electron-hole pairs. When a reverse bias is applied, these charge carriers are swept across the junction, creating a current proportional to the light intensity.
2.Explain the working principle of an avalanche photodiode (APD).Concept
An avalanche photodiode (APD) is a highly sensitive semiconductor device that operates by the avalanche multiplication process. When a photon is absorbed, it generates an electron-hole pair. Under high reverse bias, these carriers gain enough kinetic energy to ionize other atoms, creating more carriers and thus amplifying the current. This makes APDs suitable for low-light detection.
3.What are the main differences between a photodiode and an avalanche photodiode?Concept
The main differences are in sensitivity and gain. Photodiodes have a linear response to light and are less sensitive compared to APDs. APDs, on the other hand, have internal gain due to the avalanche effect, making them more sensitive and suitable for low-light applications. However, APDs require higher operating voltages and have more complex circuitry.
4.When are avalanche photodiodes preferred in optical communication systems?Application
An APD's internal gain M multiplies the photocurrent before the amplifier adds its thermal noise, so it improves sensitivity by roughly 5–10 dB when the receiver is thermal-noise limited, as in long unamplified links at high bit rates. The gain process adds excess noise F(M) ≈ M^x, so there is an optimum gain, and APDs need high, temperature-compensated bias. Where an optical amplifier precedes the receiver, or for short links, a cheaper PIN diode is usually preferred.
5.What happens if a photodiode is operated without a reverse bias?Application
At zero bias (photovoltaic mode) the diode still generates a photocurrent; into a short circuit or a transimpedance amplifier it remains linear with light and nearly equal to the reverse-biased value. Because there is no bias, dark current is almost zero, giving the lowest noise for precision, low-frequency measurements. The depletion region is narrower, however, so junction capacitance is higher and more carriers are collected by slow diffusion, making the response much slower. Into a high-resistance load the output becomes a logarithmic open-circuit voltage.
6.How does temperature affect the performance of avalanche photodiodes?Application
As temperature rises, increased lattice (phonon) scattering makes it harder for carriers to gain ionising energy, so the ionisation coefficients fall and the breakdown voltage rises (around 0.1–0.2% per kelvin for silicon). At a fixed bias the gain M therefore drops as temperature rises, and dark current also increases. APD receivers compensate by adjusting the bias with temperature or by thermo-electric cooling to hold M constant.
7.Calculate the photocurrent generated by a photodiode with a responsivity of 0.6 A/W when exposed to a light power of 5 mW.Numerical
The photocurrent (I) can be calculated using the formula I = R × P, where R is the responsivity and P is the light power. Here, R = 0.6 A/W and P = 5 mW = 0.005 W. Therefore, I = 0.6 A/W × 0.005 W = 0.003 A or 3 mA.
8.A photodiode has a dark current of 2 nA. If the photocurrent is measured to be 50 nA, what is the net current due to illumination?Numerical
The net current due to illumination is the photocurrent minus the dark current. Here, the photocurrent is 50 nA and the dark current is 2 nA. Therefore, the net current is 50 nA - 2 nA = 48 nA.
9.Explain the term 'quantum efficiency' in the context of photodiodes.Concept
Quantum efficiency is a measure of a photodiode's effectiveness in converting incident photons into charge carriers. It is defined as the ratio of the number of charge carriers generated to the number of photons incident on the photodiode. High quantum efficiency indicates that the photodiode is effective in converting light into electrical signals.
10.What are the advantages of using silicon photodiodes over other types?Application
Silicon photodiodes offer several advantages, including high quantum efficiency in the visible spectrum, low cost, and ease of integration with electronic circuits. They also have a fast response time and are stable over a wide range of temperatures. These characteristics make them suitable for a variety of applications, including consumer electronics and scientific instruments.
Finished this topic? Mark it so your progress, study plan and readiness keep up.
Stuck on something here?