Photodiode Transimpedance Amplifier Layout

A photodiode produces a current rather than a voltage, and that current is small: a few nanoamperes in a low light application and a few microamperes in a bright one. The transimpedance amplifier converts that current into a usable voltage, and because the signal is so small, the layout around the summing node is as important as the choice of components.

How a Photodiode Produces Current

Light absorbed in the depletion region of a reverse biased diode generates electron hole pairs that are swept out by the field, producing a current proportional to the incident optical power. The relationship is linear over many decades, which is what makes the photodiode useful for measurement rather than just detection, and the response time is set by the junction capacitance and the load impedance.

The diode also behaves as a capacitor of a few picofarads and as a source of leakage current that doubles roughly every ten degrees of temperature rise. Both matter in the amplifier design: the capacitance sets the stability limit of the loop, and the leakage sets the offset that appears at the output and looks exactly like a slowly changing light level.

The bias voltage across the diode is a design choice. Operating at zero bias removes the dark current contribution and reduces noise at the cost of speed, while a reverse bias of a few volts widens the depletion region, lowers the capacitance and speeds the response at the expense of additional leakage. The choice follows from whether the application measures a steady level or a fast pulse.

Why the Current Is Converted, Not Amplified

A photocurrent cannot simply be fed into a voltage amplifier with a large gain, because the amplifier input impedance would develop a voltage that depends on the diode capacitance and on the frequency. The transimpedance configuration holds the summing node at a fixed potential and forces the current through a feedback resistor, so the output is the current multiplied by that resistance.

The value of the feedback resistor sets the gain directly, and the range is wide. A 1 megohm resistor gives one volt per microampere, which suits a measurement application, while a 1 kilohm resistor gives one millivolt per microampere and suits a fast receiver. The resistance also contributes thermal noise, which falls as the square root of the resistance, so a larger resistor gives a better signal to noise ratio at low frequencies.

The summing node is the most sensitive point in the circuit. It sits at the amplifier input, it is a high impedance node, and it is where any injected noise, leakage or capacitance has the greatest effect. Protecting that node with clean layout and, where necessary, a guard ring is the central task of the board design.

Feedback Capacitance and Stability

The diode capacitance appears in parallel with the amplifier input, and together with the feedback resistor it forms a pole that erodes the phase margin. If nothing is done, the amplifier will overshoot or oscillate, and the oscillation may appear only at one light level, which makes it difficult to diagnose.

A small feedback capacitance is the standard fix. It introduces a zero that compensates the phase lag from the diode capacitance and the input capacitance, and it also limits the bandwidth, which reduces the integrated noise. The value is usually a few picofarads and is best found by calculation from the diode capacitance, then confirmed by looking at the step response.

Too much feedback capacitance slows the response more than necessary, while too little leaves the loop marginal. The optimum produces a small overshoot on a step of light; a critically damped response is slightly slower but more robust across production spread. Measuring the step response with a modulated light source is the only reliable way to confirm the design.

Choosing the Amplifier

input bias current is the first specification to check, because it flows through the feedback resistor and appears as an output offset. An amplifier with a bias current of a few picoamperes suits a high impedance design, while a bipolar amplifier with nanoampere bias current would produce an offset of millivolts with a 1 megohm feedback resistor.

The input capacitance of the amplifier adds to the diode capacitance and lowers the stability limit, so a part with a small input capacitance allows a larger feedback resistor for the same bandwidth. The gain bandwidth product sets how much transimpedance is achievable, and the voltage noise matters when the feedback resistor is small, while the current noise matters when it is large.

Rail to rail operation is convenient but not essential. What does matter is the input common mode range, which has to include the bias voltage applied to the diode, and the output swing, which has to cover the range that the converter needs. Check both against the actual circuit rather than the headline specification.

Bandwidth, Noise and Gain Trade-offs

Bandwidth and transimpedance are linked by the amplifier gain bandwidth product. For a given amplifier, the achievable bandwidth falls as the feedback resistor rises, which is the opposite of what a designer might expect. Where both a large signal and a fast response are needed, the answer is a faster amplifier rather than a larger resistor.

Noise follows the same rule. At low frequency the noise is dominated by the thermal noise of the feedback resistor and by the current noise of the amplifier flowing through it, while at high frequency the voltage noise of the amplifier multiplied by the total input capacitance dominates. The crossover between the two regions sets the optimum feedback resistance for a given bandwidth.

Practical photodiode designs often sit near that optimum, with a feedback resistor between 100 kilohms and 10 megohms and a bandwidth between a few kilohertz and a few hundred kilohertz. Pushing beyond those values requires a smaller diode, a lower capacitance package and a guard ring to control leakage, all of which are design choices made together rather than separately.

Layout and Guarding

Keep the summing node as small as possible. The amplifier input pin, the feedback components and the diode cathode should meet in a few square millimetres of board area with no other trace in the vicinity. Every additional millimetre of trace at that node adds capacitance that reduces the stability margin and picks up more interference.

A guard ring around the summing node, driven to the same potential, interrupts the leakage path across the board surface. Where the amplifier has a low impedance pin at the same potential, the ring can be driven from it; otherwise the ring is connected to the reference potential of the node. A guard ring that is left floating does nothing, and one connected to the wrong potential is worse than none.

Cleanliness matters more here than on most boards. Flux residue and moisture form a leakage path with a resistance that varies with humidity, and the resulting offset appears as a slow drift that nothing in the circuit can correct. Coating the area after assembly, using the practices described in conformal coating, keeps the surface resistance high and stable for the life of the product.

Ground and power follow the usual analogue rules. Use a solid plane under the amplifier, decouple the supplies locally, and keep switching regulators and digital lines away from the photodiode area. The partition between the analogue and digital sections of a mixed board is described in mixed signal board design, and a photodiode front end belongs on the quiet side of that partition.

Testing and Common Problems

Cover the diode and measure the output. The result is the dark output, and it should be close to the expected value from the dark current and the input bias current. If it drifts upwards over a few minutes, the board is warming up or there is a leakage path from a nearby trace. If it moves when a hand approaches the board, the summing node is picking up interference.

Shine a known light level on the diode and check that the output follows linearly across several decades. Departure from linearity at the low end indicates leakage or an offset problem, while departure at the high end indicates that the amplifier is running out of output swing. Both are easy to see on a log plot and both point at a specific part of the circuit.

An oscillation that appears only at certain light levels is the classic symptom of an under compensated loop. Increasing the feedback capacitance slightly usually cures it, at the cost of bandwidth. An oscillation that appears only when the enclosure is closed points instead to interference or to a change in the stray capacitance of the assembly, and the fix is mechanical or a better screen rather than a component change.

Our article on EMI suppression design describes how to find the path that interference takes into a sensitive node. On a photodiode front end that path is almost always the capacitance of a nearby conductor or the leakage of an unclean surface, and both are visible with a simple test.

FAQ

How do I choose the feedback resistor? Start from the largest photocurrent you expect and the output swing you need. Then check the resulting bandwidth against the requirement, and adjust the diode or the amplifier rather than pushing the resistor beyond what the loop can support.

Can I use a general purpose op amp? For a large diode and a modest bandwidth, yes. For a small diode with a large feedback resistor, the input bias current and the input capacitance of a general purpose part usually make it unusable.

Why does my reading change with humidity? Surface leakage across the board is the usual cause. Clean the assembly, add a guard ring and coat the summing node area so the leakage remains constant rather than varying with the weather.