Avalanche Photodiode (APD)
Understanding Avalanche Photodiodes
An APD is a reverse-biased p-n junction operated near its breakdown voltage, where the electric field in a thin depletion region is strong enough to accelerate photogenerated carriers to energies sufficient for impact ionization. A single absorbed photon generates one primary electron-hole pair, which then triggers an avalanche of secondary carriers in the multiplication region. The resulting current gain M (typically 10 to 100 for InGaAs telecom APDs) amplifies the photocurrent before it reaches the transimpedance amplifier, reducing the effective contribution of amplifier thermal noise to the receiver sensitivity.
The critical performance tradeoff in APD design is between gain and excess noise. The avalanche multiplication is a random process: each primary carrier produces a statistically varying number of secondary carriers. This randomness adds shot noise beyond what a deterministic gain process would produce, quantified by the excess noise factor F(M). The optimal operating gain is the value that maximizes the receiver SNR by balancing the benefit of signal amplification against the penalty of excess noise. Beyond this optimum, further gain degrades sensitivity rather than improving it.
APD Gain and Noise
M = 1 / (1 - (Vbias/VBR)n)
Where VBR = breakdown voltage, n = 2 to 6 (material-dependent)
Excess Noise Factor (McIntyre model):
F(M) = k × M + (1 - k) × (2 - 1/M)
Where k = ionization ratio. Si: k ≈ 0.02, InGaAs: k ≈ 0.4-0.7
APD Receiver Sensitivity Improvement:
ΔSensitivity = 10 × log10(M / √F(M)) dB vs. PIN
InGaAs APD at M=10, k=0.5: F ≈ 5.9, improvement ≈ 6.1 dB
Si APD at M=100, k=0.02: F ≈ 3.9, improvement ≈ 14.0 dB
APD vs. PIN Photodiode Comparison
| Parameter | PIN Photodiode | InGaAs APD | Silicon APD |
|---|---|---|---|
| Internal Gain | 1 (unity) | 10 to 30 | 50 to 500 |
| Excess Noise Factor | 1.0 | 5 to 12 | 3 to 5 |
| Bias Voltage | -5 to -15 V | -30 to -80 V | -150 to -400 V |
| Wavelength Range | 0.8 to 1.6 μm (InGaAs) | 1.0 to 1.6 μm | 400 to 1000 nm |
| Bandwidth | Up to 100 GHz | Up to 25 GHz | Up to 1 GHz |
| Sensitivity Advantage | Reference | +5 to 8 dB | +10 to 15 dB |
| Temperature Sensitivity | Low | High (gain varies with T) | Moderate |
| Cost | $ | $$$ | $$ |
Frequently Asked Questions
How does an avalanche photodiode provide gain?
An APD operates with a high reverse bias voltage (20 to 400 V) that creates a strong electric field in a thin multiplication region. When a photon generates an electron-hole pair in the absorption region, the primary carrier drifts into the multiplication region where the field accelerates it to energies above the bandgap. The carrier creates a new electron-hole pair through impact ionization, and these secondary carriers can themselves create additional pairs, producing an avalanche multiplication effect. The gain M ranges from 10 to 100 for InGaAs APDs and 50 to 500 for silicon APDs. Unlike optical amplifiers, APD gain comes with excess noise because the multiplication process is random.
When should you use an APD instead of a PIN photodiode?
Use an APD when the receiver is thermal-noise limited, meaning the signal current from a PIN diode is too small relative to the amplifier noise. An APD multiplies the photocurrent by its gain M before the TIA adds noise, improving electrical SNR by up to 10 dB. The optimal gain is where increasing excess noise exactly equals the decreasing relative contribution of amplifier noise. APDs are preferred for long-haul fiber links without optical amplifiers, free-space optical communication, laser rangefinders, and lidar. PIN diodes are preferred when optical amplifiers (EDFAs) are present, because amplified spontaneous emission noise dominates and APD gain provides no benefit.
What is the excess noise factor and why does it matter?
The excess noise factor F(M) quantifies the additional noise from the random nature of impact ionization. For a deterministic gain process, F would equal 1. For real APDs, F increases with gain M according to the McIntyre model: F(M) = k*M + (1-k)*(2 - 1/M), where k is the ionization ratio. Silicon APDs have k of approximately 0.02, giving F of about 4 at M = 100, which is excellent. InGaAs APDs have k of 0.4 to 0.7, giving F of 8 to 12 at M = 10, which limits useful gain. This is why InGaAs APDs operate at gains of 10 to 20, while silicon APDs can use gains of 100 or more.