Transimpedance Amplifier Peaking With Photodiode Cable
A photodiode preamp uses a 1 MΩ transimpedance gain and must achieve 100 kHz bandwidth. The photodiode capacitance is 10 pF, the op-amp input capacitance is 5 pF, and the selected op-amp has 10 MHz gain-bandwidth. The board has a guard ring and a short feedback trace. On the bench, the output shows 6 dB peaking at about 180 kHz and occasional oscillation when a long cable is connected to the photodiode. The feedback capacitor currently fitted is 0.2 pF, but the layout adds roughly 0.3 pF stray. The supply is ±5 V, and the photodiode is reverse-biased at 2 V. The customer needs either 100 kHz bandwidth with at least 45° phase margin or a documented reason to change the op-amp.
1. Estimate the feedback capacitance needed for about 45° phase margin and compare it with the installed value.
Read the worked answer
Use the total input capacitance Cin = 10 pF photodiode + 5 pF op-amp = 15 pF. The input pole is f_p = 1/(2π Rf Cin) = 1/(2π × 1 MΩ × 15 pF) ≈ 10.6 kHz. The op-amp gain-bandwidth is f_u = 10 MHz. For roughly 45° phase margin, place the feedback zero near the geometric mean: f_z ≈ sqrt(f_p f_u) = sqrt(10.6 kHz × 10 MHz) ≈ 325 kHz. Then Cf = 1/(2π Rf f_z) = 1/(2π × 1 MΩ × 325 kHz) ≈ 0.49 pF. The installed 0.2 pF plus about 0.3 pF stray is nominally 0.5 pF, but that is marginal because cable capacitance raises Cin and moves the required value upward. The observed peaking and oscillation are consistent with insufficient effective compensation.
2. Why does adding a long cable make the oscillation worse, and what does that imply about the 0.2 pF capacitor?
Read the worked answer
The cable adds capacitance at the summing node, increasing Cin. That lowers f_p and increases the noise-gain plateau, so the feedback zero must move lower in frequency to preserve phase margin. A fixed 0.2 pF feedback capacitor therefore becomes less effective as cable capacitance grows. The 0.3 pF layout stray is not a reliable compensation element because it is uncontrolled and may vary board to board. The result is more peaking and possible oscillation. The installed 0.2 pF should not be treated as adequate; either shorten/guard the cable, increase Cf to a controlled value, or reduce Rf and add a second gain stage if 1 MΩ is not essential.
3. Recommend a practical fix that preserves 100 kHz bandwidth and explain the tradeoff.
Read the worked answer
Fit a controlled 1 pF feedback capacitor and verify the response with the cable attached. That gives f_z = 1/(2π × 1 MΩ × 1 pF) ≈ 159 kHz, which is above the 100 kHz target and should provide margin against the input pole. If peaking remains, reduce Rf to 500 kΩ, use Cf ≈ 2 pF, and add a second stage for gain; this lowers the transimpedance but makes compensation more repeatable. Do not rely on 0.2 pF plus stray. Measure the closed-loop output with a small-signal photodiode stimulus or injected current, check phase margin by step response overshoot, and confirm the long-cable case does not oscillate over temperature and supply variation.