Action of an Avalanche Photodiode (APD)

When I first got introduced to optical receivers, I remember being confused about why some designs used a plain PIN photodiode while others insisted on something called an Avalanche Photodiode, or APD. It took me a while — and a few frustrating afternoons with datasheets — to really understand why an APD exists, what makes it “avalanche,” and why it matters so much in long-haul fiber optic links, LIDAR systems, and low-light sensing applications. In this article, I want to walk you through everything I’ve learned about APDs: how they work at the physics level, the math behind their gain, how to use them practically in a circuit, and the mistakes I made (and see others make) when working with them.

What Is an Avalanche Photodiode?

An Avalanche Photodiode is a specialized type of photodiode that converts light into electrical current, just like a standard PN or PIN photodiode, but with one critical difference: it has internal gain. A regular photodiode produces roughly one electron-hole pair for every photon absorbed (assuming ideal quantum efficiency). An APD, on the other hand, is engineered to multiply that initial photocurrent internally, often by a factor of 10 to over 100, before the signal even leaves the device and hits your amplifier.

I like to think of an APD as a photodiode with a built-in pre-amplifier baked directly into the silicon or III-V semiconductor structure. That internal amplification happens through a physical phenomenon called impact ionization, and understanding that phenomenon is the key to understanding everything else about how APDs behave.

The Physics: How Avalanche Multiplication Works

To understand APD action, I have to start with the basic photodiode process. When a photon with sufficient energy strikes the semiconductor material, it can excite an electron from the valence band into the conduction band, leaving behind a hole. This creates an electron-hole pair. In a standard photodiode, that pair is simply swept apart by the electric field in the depletion region and collected at the electrodes, producing a small photocurrent.

In an APD, the device is engineered with a much higher reverse bias voltage — often 100V to well over 400V depending on the design — which creates an extremely strong electric field in a specific region of the device called the multiplication region (sometimes called the avalanche region). When the initially generated electron (or hole) is accelerated through this high-field region, it gains enough kinetic energy that when it collides with the crystal lattice, it can knock loose another electron-hole pair. This is impact ionization.

Here’s where the “avalanche” name comes from: that newly freed electron is also accelerated by the field, and it too can create another electron-hole pair through a further collision. This cascading, self-multiplying process is very much like an avalanche of snow — one small disturbance triggers a much larger cascade. Each new carrier pair has the potential to generate more pairs, and this chain reaction happens in picoseconds, all within the multiplication region of the device.

The result is that a single absorbed photon can generate not one, but dozens or even hundreds of electron-hole pairs by the time the current reaches the external circuit. This internal amplification is what gives the APD its defining characteristic: gain.

Device Structure of an APD

Most practical APDs use what’s called a “reach-through” structure. I’ve found it helpful to break this down into layers:

1. P+ layer (or N+ depending on polarity convention): A heavily doped contact layer.
2. Absorption region (π or intrinsic layer): This is where most of the incoming light is absorbed and photocarriers are generated. It’s typically a lightly doped or intrinsic region designed to have a large depletion width so it can absorb light efficiently, especially at longer wavelengths like 1310nm or 1550nm used in telecom.
3. Charge control layer: This layer helps shape the electric field profile so the field is high enough in the multiplication region but not so high in the absorption region that it triggers premature breakdown.
4. Multiplication region: A thin, high-field region where the impact ionization cascade actually occurs.
5. Substrate/contact layer: Completes the circuit.

The separation of the absorption and multiplication functions into different regions is called a SAM (Separate Absorption and Multiplication) structure, and it’s one of the most important innovations in APD design because it lets engineers optimize each function independently — good light absorption in one region, controlled and efficient carrier multiplication in another.

Gain (Multiplication Factor, M)

The single most important parameter that distinguishes an APD from a regular photodiode is the multiplication factor, denoted M. It’s defined simply as:

$$M = \frac{I_{APD}}{I_{primary}}$$

Where I_APD is the total output current after avalanche multiplication, and I_primary is the initial photocurrent that would have been generated without multiplication (i.e., what a plain photodiode would produce under the same illumination).

Typical APD gain values range from about 10 to 100, though some specialized devices (like those used in single-photon counting, called SPADs — Single Photon Avalanche Diodes — operating in Geiger mode) can have effectively infinite gain for a single photon event.

Gain is strongly dependent on the reverse bias voltage. As you increase the reverse bias toward the breakdown voltage ($V_{BR}$), the electric field in the multiplication region increases, and $M$ increases — often exponentially near breakdown. This relationship is commonly approximated by an empirical formula:

$$M = \frac{1}{1 – \left(\dfrac{V_R}{V_{BR}}\right)^n}$$

Where $V_R$ is the applied reverse voltage, $V_{BR}$ is the breakdown voltage, and $n$ is an empirical constant (typically between 3 and 6, depending on the material and structure).

This formula tells me something practically important: gain is extremely sensitive near breakdown. A small change in reverse voltage — even a fraction of a volt — can cause a large swing in gain when you’re operating close to $V_{BR}$. This is why APD biasing circuits require very stable, well-regulated high-voltage supplies, often with temperature compensation, because breakdown voltage itself is temperature-dependent (it typically increases with temperature).

Responsivity and Effective Sensitivity

The responsivity of a photodiode (R, in A/W) tells you how much current you get per watt of incident optical power. For a standard photodiode:

$$I_{primary} = R_0 \times P_{optical}$$

For an APD, the effective responsivity is boosted by the multiplication factor:

$$R_{APD} = M \times R_0$$

So if a photodiode has an intrinsic responsivity of 0.9 A/W at 1550nm, and the APD is biased to achieve a gain of $M = 20$, the effective responsivity becomes 18 A/W. This is a massive sensitivity improvement, which is exactly why APDs are used in situations where the received optical signal is very weak — long fiber spans, free-space optical links, LIDAR return signals, and low-light imaging.

Excess Noise Factor: The Trade-off

Gain doesn’t come for free. The avalanche process is statistical in nature — not every carrier pair triggers exactly the same number of secondary pairs every time. This randomness introduces additional noise beyond the standard shot noise you’d get from a regular photodiode. This extra noise is captured by the excess noise factor, F(M), which increases with gain:

$$F(M) = k_{eff} \times M + (1 – k_{eff}) \times \left(2 – \frac{1}{M}\right)$$

Where $k_{eff}$ is the ionization coefficient ratio (the ratio of hole ionization rate to electron ionization rate, or vice versa, depending on convention). Materials with a low k_eff (ideally close to 0) produce lower excess noise for a given gain — this is one reason InGaAs/InP APDs are engineered carefully, since silicon APDs (used at shorter wavelengths like 800-900nm) generally have much lower excess noise than InGaAs APDs used at 1310/1550nm.

This is the central design trade-off of an APD: increasing gain increases signal, but it also increases noise, and beyond a certain optimal gain point, the signal-to-noise ratio (SNR) actually starts to degrade rather than improve. Every practical APD receiver design involves finding this optimal operating gain.

APD in a Receiver Circuit

In a real optical receiver, the APD is typically reverse-biased at a high voltage (supplied by a dedicated, low-noise high-voltage source) and connected to a transimpedance amplifier (TIA). The TIA converts the APD’s output current into a voltage, which is then further amplified and processed by the rest of the receiver chain (limiting amplifier, clock/data recovery, etc., in a digital communication system).

A simplified signal path looks like this:

Incoming Light → APD (photon absorption + avalanche multiplication) → Photocurrent ($I = M \times R_0 \times P$) → Transimpedance Amplifier → Voltage signal → Further signal processing

Because the bias voltage directly sets the gain, and gain directly affects both signal strength and noise, many APD receiver modules include an automatic gain control (AGC) loop that dynamically adjusts the bias voltage to maintain optimal performance as the received optical power varies (for example, as a fiber link’s attenuation changes with temperature or aging).

Real-World Applications

I’ve come across APDs in a surprisingly wide range of systems:

• Long-haul and metro fiber optic communication: Where signal loss over distance means the receiver needs extra sensitivity to detect weak optical signals reliably.
• LIDAR systems: Automotive and industrial LIDAR relies heavily on APDs (and their single-photon cousins, SPADs) to detect very weak reflected laser pulses over long distances.
• Low-light imaging and scientific instrumentation: Photon-counting applications in physics and biology labs.
• Free-space optical communication: Where atmospheric attenuation and scintillation reduce received power significantly.
• Medical imaging: Positron Emission Tomography (PET) scanners often use APD arrays to detect scintillation light from radiation detectors.

Common Mistakes When Working With APDs

From my own experience and from watching others troubleshoot APD circuits, here are the errors I see most often:

1. Ignoring temperature compensation. Because breakdown voltage shifts with temperature, a fixed bias voltage that works perfectly at 25°C can push the device into excessive gain (and excess noise, or even damage) at higher temperatures, or insufficient gain at lower temperatures. Proper designs use a temperature sensor and adjust bias accordingly.
2. Operating too close to breakdown. It’s tempting to crank up the gain for maximum sensitivity, but operating too near V_BR increases excess noise disproportionately and risks device damage from thermal runaway.
3. Poor bias supply filtering. Because gain is exponentially sensitive to bias voltage near breakdown, any ripple or noise on the high-voltage supply directly translates into gain fluctuations and added receiver noise.
4. Overlooking dark current growth with gain. Just like the signal photocurrent, the APD’s dark current (leakage current present even without light) is also multiplied by the avalanche gain. At high gain settings, multiplied dark current can become a significant noise contributor, especially in low-light applications.
5. Assuming APDs are always better than PIN photodiodes. For high received optical power, a simple PIN photodiode combined with a good low-noise amplifier can actually outperform an APD, because the APD’s excess noise factor becomes the dominant noise source rather than helping.

Troubleshooting Tips

If an APD-based receiver isn’t performing as expected, I usually check things in this order:

• Verify the bias voltage is stable and at the correct value for the desired gain (cross-check against the datasheet’s M-vs-V_R curve).
• Confirm temperature compensation is active and functioning if the device operates over a wide temperature range.
• Measure dark current at the operating bias — if it’s much higher than the datasheet spec, the device may be degraded or damaged.
• Check for excessive noise on the high-voltage bias supply using an oscilloscope with sufficient bandwidth.
• Verify the transimpedance amplifier’s input capacitance and bandwidth are compatible with the APD’s junction capacitance, since a mismatch here can cause peaking, ringing, or bandwidth limitations in the received signal.
• If sensitivity is poor, confirm whether the operating gain is actually near the calculated optimal gain for minimum noise, rather than just maxed out.

Key Takeaways

An Avalanche Photodiode is, at its core, a photodiode with built-in signal amplification achieved through impact ionization in a high-field multiplication region. This internal gain (M) dramatically improves sensitivity compared to a standard photodiode, which is why APDs are the detector of choice in applications where the received optical signal is weak — long-distance fiber links, LIDAR, and low-light sensing.

But that gain comes with a cost: excess noise that grows with gain, dark current that’s also multiplied, and extreme sensitivity to bias voltage stability and temperature. Getting the most out of an APD means understanding this trade-off deeply and designing the bias and amplification circuitry with real care — stable high-voltage supplies, temperature compensation, and operating gain chosen to minimize noise rather than just maximize signal.

Once I understood the avalanche multiplication process and the excess noise trade-off, APD circuit design stopped feeling like a black box and started making a lot more sense. I hope this breakdown does the same for you.