How do you design a photodiode transimpedance amplifier?
Set Rf = Vout ÷ Imax (5 V ÷ 50 µA = 100 kΩ). Then choose Cf: for a bandwidth well inside the op amp’s reach, Cf = 1 ÷ (2π Rf f), which is 159 pF for 10 kHz; for the most bandwidth with a flat response, Cf ≈ √(Cin ÷ (π Rf GBW)), giving f−3dB ≈ √(GBW ÷ (2π Rf Cin)). Cin is the photodiode plus the op amp’s input capacitance.
Transimpedance amplifier formula: gain, bandwidth and stability
The photodiode current flows through Rf, so the output is −I × Rf at low frequency. Cin at the inverting input makes the circuit second order, and Cf across Rf is what keeps it stable.
Rf = Vout ÷ Imax f0 = √(GBW ÷ (2π Rf (Cin + Cf))) Q = √(2π GBW Rf (Cin + Cf)) ÷ (1 + 2π GBW Rf Cf)
- Rf, Cf
- feedback resistor and capacitor
- Cin
- photodiode + op amp common-mode + differential input capacitance
- GBW
- op amp gain bandwidth product in Hz
- f0, Q
- natural frequency and quality factor of the closed loop
That is TI SBOA122’s second-order model with a single-pole op amp. Q = 0.707 is maximally flat, and with Cin much larger than Cf it gives the familiar design equations (TI SBOA122 eq. 12, OPA320 data sheet):
Cf ≈ √(Cin ÷ (π Rf GBW)) f−3dB ≈ √(GBW ÷ (2π Rf Cin)) GBW ≈ 2π f² Rf Cin
TI SBOA055 places the feedback pole on the open-loop gain curve instead, which is faster with Q close to 1:
Cf = [1 + √(1 + 8π Rf Cin GBW)] ÷ (4π Rf GBW)
For a bandwidth below what the op amp can reach, set it with Cf and check the op amp is fast enough (TI SBOA268):
Cf = 1 ÷ (2π Rf f) GBW > (Cin + Cf) ÷ (2π Rf Cf²)
Photodiode transimpedance amplifier design examples
Example 1: 0–50 µA to 0–5 V at 10 kHz (TI SBOA268)
Rf = 5 V ÷ 50 µA = 100 kΩ. For a 10 kHz feedback pole, Cf = 1 ÷ (2π × 100 kΩ × 10 kHz) = 159 pF; TI rounds to 150 pF. With Cin = 6 pF (the OPA170’s 3 pF common-mode plus 3 pF differential), the op amp needs GBW > 11 kHz, far below the OPA170’s 1.2 MHz. The model gives f−3dB = 10.6 kHz with no peaking; TI’s simulation shows 10.57 kHz.
Example 2: a fast front end, OPA380 with a reverse-biased BPW34
At 3 V reverse bias the BPW34 is 25 pF (typical); with the OPA380’s 3 pF and 1.1 pF, Cin = 29.1 pF. 100 µA into 4 V gives Rf = 40 kΩ, rounded down to 39.2 kΩ. The maximally flat Cf is 1.62 pF (3.45 MHz); the next E12 value, 1.8 pF, gives 3.08 MHz with Q = 0.64. TI’s SBOA055 value, 1.17 pF, would reach 4.34 MHz with 1.0 dB of peaking, and at 0 V bias (70 pF) the flat design drops to 2.18 MHz.
Example 3: noise of an OPA320 with a BPW34 at 0 V
10 µA into 4 V gives Rf = 392 kΩ, and with Cin = 79 pF the flat design uses Cf = 1.8 pF for f−3dB = 315 kHz. Rf alone contributes 205 fA/√Hz. Adding the OPA320’s 7 nV/√Hz and 0.6 fA/√Hz in SBOA122 eq. 13, the equivalent input noise over that bandwidth is 665 fA/√Hz, or 374 pA rms: the op amp’s en × 2π F × Cin term dominates (633 fA/√Hz). Full scale is 89 dB above the noise.
Maximally flat bandwidth and Cf for a 10 MHz GBW op amp
Each cell is the −3 dB bandwidth with the exact Q = 0.707 feedback capacitor in brackets, for the Rf in the row and the total input capacitance in the column. A Cf under about 0.2 pF is smaller than a surface-mount resistor’s own stray capacitance (OPA320 data sheet), so those designs are set by layout.
| Rf \ Cin | 5 pF | 20 pF | 100 pF | 500 pF |
|---|
| 10 kΩ | 4.29 MHz (3.66 pF) | 2.39 MHz (7.82 pF) | 1.16 MHz (17.8 pF) | 543 kHz (39.9 pF) |
| 100 kΩ | 1.6 MHz (1.25 pF) | 841 kHz (2.52 pF) | 388 kHz (5.64 pF) | 176 kHz (12.6 pF) |
| 1 MΩ | 543 kHz (0.4 pF) | 277 kHz (0.8 pF) | 125 kHz (1.78 pF) | 56.2 kHz (3.99 pF) |
| 10 MΩ | 176 kHz (0.13 pF) | 88.6 kHz (0.25 pF) | 39.8 kHz (0.56 pF) | 17.8 kHz (1.26 pF) |
Feedback resistor noise at 25 °C
Thermal noise of Rf as an input current, √(4kT/Rf), and as an output voltage density, √(4kT Rf). The current falls as Rf rises, so a larger Rf gives a better signal-to-noise ratio for small photocurrents.
| Rf | Input noise current | Output noise density |
|---|
| 1 kΩ | 4.06 pA/√Hz | 4.06 nV/√Hz |
| 10 kΩ | 1.28 pA/√Hz | 12.8 nV/√Hz |
| 100 kΩ | 406 fA/√Hz | 40.6 nV/√Hz |
| 1 MΩ | 128 fA/√Hz | 128 nV/√Hz |
| 10 MΩ | 40.6 fA/√Hz | 406 nV/√Hz |
| 100 MΩ | 12.8 fA/√Hz | 1.28 µV/√Hz |
What goes wrong on the bench
Cin is more than the photodiode: add the op amp’s common-mode and differential input capacitance from its data sheet and a picofarad or two for the input trace. Cf already includes about 0.2 pF of stray capacitance across a surface-mount Rf, so a calculated Cf under 1 pF is mostly layout; a lower Rf followed by a voltage gain stage is often the cleaner answer. DC errors scale with Rf too: in TI’s SBOA122 example the OPA846’s 19 µA bias current through 310 kΩ puts 5.89 V at the output, so high-Rf designs need FET or CMOS inputs. Because the high-frequency noise gain is 1 + Cin/Cf rather than 1, a decompensated op amp can be used if that noise gain exceeds its minimum stable gain. The model treats the op amp as a single pole, so real phase margin is lower than calculated; confirm with a simulation or a step test.
Glossary
- Transimpedance
- Output voltage per input current, in ohms; Rf at low frequency.
- Cin
- Everything at the inverting input: photodiode junction capacitance plus the op amp’s common-mode and differential input capacitance and layout.
- Noise gain (1/β)
- 1 + Zf/Zin: flat at low frequency, rising from the zero at 1/(2π Rf (Cin + Cf)) and leveling at 1 + Cin/Cf.
- Feedback pole
- 1/(2π Rf Cf), where Cf takes over from Rf in the feedback path.
- Q and peaking
- Q describes the second-order closed loop: 0.707 is maximally flat, higher Q peaks in frequency and overshoots on steps.
- Photovoltaic and photoconductive mode
- The photodiode at 0 V bias (lowest dark current) or reverse biased (lower capacitance, faster).
Frequently asked questions
What is a transimpedance amplifier?
An op amp transimpedance amplifier (TIA) converts an input current into a voltage. The photodiode current flows through the feedback resistor Rf, so the output is −I × Rf at low frequency, and Cf across Rf keeps the loop stable.
What is the transimpedance amplifier gain formula?
Vout = −I × Rf at low frequency, so the transimpedance gain is simply Rf, in ohms (volts per amp). In TI’s example, 5 V ÷ 50 µA gives a gain of 100 kΩ.
How do I choose the feedback resistor of a transimpedance amplifier?
Divide the output swing you can use by the largest photocurrent: Rf = Vout ÷ Imax. TI’s example converts 0–50 µA into 0–5 V with 100 kΩ. Round down to a standard value so full scale stays inside the swing.
How do I calculate the transimpedance amplifier feedback capacitor?
For a maximally flat response with Cin much larger than Cf, Cf ≈ √(Cin ÷ (π Rf GBW)); TI’s SBOA055 formula, Cf = [1 + √(1 + 8π Rf Cin GBW)] ÷ (4π Rf GBW), gives a smaller Cf: in the OPA380 example on this page, 26 % more bandwidth with 1.0 dB of peaking. If you want less bandwidth than that, Cf = 1 ÷ (2π Rf f) sets it directly.
What limits transimpedance amplifier bandwidth?
The op amp’s GBW, Rf and the total input capacitance: the most you get with a flat response is about √(GBW ÷ (2π Rf Cin)). A 10 MHz op amp with 1 MΩ and 20 pF reaches 282 kHz; quadrupling GBW only doubles it.
Why does my transimpedance amplifier oscillate?
Transimpedance amplifier stability depends on Cf: without it, the input capacitance makes the noise gain rise at 20 dB per decade until it meets the op amp’s falling open-loop gain, leaving almost no phase margin. With 1 MΩ, 20 pF and a 10 MHz op amp the model gives Q = 35 and 1.6° of phase margin. Just 0.8 pF across Rf brings Q to 0.707 and the phase margin to 66°.
Should I reverse bias the photodiode?
For speed, yes: a BPW34 drops from 70 pF at 0 V to 25 pF at 3 V (typical), which takes the flat bandwidth of the OPA380 example from 2.18 MHz to 3.45 MHz. The cost is dark current, which adds offset and shot noise; photovoltaic (0 V) operation is better for precision at low frequencies.
What op amp GBW do I need for a TIA?
For a flat response, GBW ≈ 2π f² Rf Cin (TI SBOA122 eq. 12). 1 MHz with 100 kΩ and 20 pF needs 12.6 MHz. Use the GBW needed tab for the exact value, which counts Cf in Cin.
How much noise does the feedback resistor add?
Its thermal noise current is √(4kT ÷ Rf): 128 fA/√Hz for 1 MΩ at 25 °C, and it falls as Rf rises. That is why it pays to take as much gain as the swing allows in the TIA itself rather than in a later stage.
More free PCB tools
See all free electronics tools →
- 4-20 mA calculatorConvert 4-20 mA to percent and engineering units and back, size the receiver resistor, and check loop compliance.
- Wheatstone bridge calculatorBridge output voltage, the balance condition and the unknown resistor of a Wheatstone bridge.
- 10k thermistor table and NTC thermistor calculatorThermistor resistance at any temperature or temperature from resistance with the Beta model, plus divider voltage, ADC code and a lookup table.
- Strain gauge calculatorBridge output for quarter, half and full strain gauge bridges from strain, gauge factor and excitation, or strain from output.
- Steinhart-Hart equation calculatorSteinhart-Hart A, B and C coefficients from three resistance-temperature points, Beta from two, and conversions with both.
- Thermocouple chart and calculatorThermocouple voltage from temperature and temperature from voltage for types K, J, T, E and N, with cold-junction compensation.
Design the whole board, not one number
ProtoFlow is free AI PCB design software: describe a circuit and get a schematic with real parts, PCB layout with an autorouter, SPICE simulation and Gerber files. Native KiCad, Altium and EAGLE import and export. Free on macOS, Windows and Linux.