I have designed a photodiode circuit using the OSD35-5T photodiode from Centronic. I am using the LTC 2054HV as a transimpedance amplifier (TIA) and the LTC2051HV (dual opamp) as the next stage followed by a LTC2052HV (quad opamp) for the final stages.

Previously, for measuring the light signal levels, I put all the gain onto the TIA stage and the next stage was an unity gain inverting amplifier whose output was shared by 5 other amplifiers. Each of these 5 amplifier had the gains of 4, 16, 64, 256 & 1024. The outputs of these amplifiers were connected via a Low pass filter and fed into a microcontroller.

The system runs on +/-5V rails.

The required bandwidth is 100Hz. Looking at the bandwidth graph of the TIA, I can see that for a gain of approx. 102dB, the maximum allowable bandwidth is way lower than required. Hence I decided to split the gain between the TIA and the inverting amplifier. The TIA would have approx. 60dB gain and the inverting amplifier would have approx. 42dB gain.

What I have found is that the output (as read by the microcontroller) is quite noisy compared to my initial design.

Please note that the bandwidth of the inverting amplifier and the other 5 amplifier stages and the LPF at the end is set to 400Hz so that there is reduced attenuation of 100Hz signal.

Any idea on how I could get rid of the noise?

Furthermore, can you please let me know if there is any problem if I set the bandwidth of the first stage more than what I can get for a certain gain? Will it cause oscillatory or instability problems? If so, how?

Please see the circuit below.

Full Circuit


I reverted the circuit back to its original form where the first stage gain is concentrated fully at the TIA stage and the following amplifier is an unity gain inverter. This seemed to have fixed this noise issue as the output is now quiet. This begs the question as to why the cascaded amplifier design didn't work. I cannot get y head around it.

Any ideas or explanations?

  • \$\begingroup\$ Can this be explained by Friis's formula? en.wikipedia.org/wiki/Friis_formulas_for_noise \$\endgroup\$
    – T Andersen
    Apr 19, 2023 at 9:09
  • \$\begingroup\$ Any amplifier will have noise. Perhaps the input current noise of your transimpedance is killing you, as it's developing an equivalent input voltage noise source that is amplified towards the output of your op-amp. Also, have you made sure that your subsequent gain stages are adding more gain than they're adding noise? All of this is accounted by the signal to noise ratio, perhaps you could simulate that at your output. \$\endgroup\$
    – Designalog
    Apr 19, 2023 at 9:22
  • 1
    \$\begingroup\$ Your first stage is a relatively noisy amplifier with not a lot of gain. The later stages get that slightly amplified signal and all that noise and make both bigger. Pick a different first stage amp and then put more gain in it. \$\endgroup\$ Apr 19, 2023 at 11:39
  • \$\begingroup\$ The reason why I reduced the gain was to get the required bandwidth for the required gain as per the datasheet. If I put the whole gain onto the first stage, according to the datasheet, the bandwidth of the circuit will be below 10Hz (approx.). Therefore, I split the gain between the first and the second stage so that I get the required bandwidth \$\endgroup\$
    – LabMat
    Apr 19, 2023 at 16:50
  • \$\begingroup\$ The reason you have that tradeoff between gain and bandwidth is that you picked an amplifier with extremely low gain bandwidth product that is unsuitable for the application. Go back to the catalog and pick out one that's 50-100 times faster. \$\endgroup\$ Apr 19, 2023 at 19:01

1 Answer 1


"Zero drift" amplifiers (like LTC2054) are typically chopper stabilized. Which means a small AC signal is deliberately introduced to measure (and then correct) any drift. That makes them relatively noixy, and usually not a good choice for an optical receiver. Possibly you can find the chopper frequency and filter it out in a later stage but an easier solution is just choose a different op-amp for your first stage.

Drift is not usually a critical issue for TIA configurations because the voltage offset (what the zero-drift design corrects) sees only a gain of 1 at the output. In any case if you are concerned about the absolute accuracy of your receiver as an optical power meter, you will need to correct for the dark current of your photodiode, and this correction will also correct the op-amp drift.

  • \$\begingroup\$ Hi, thanks for the information. The reason why I chose the LTC2054HVC was because of its ultra low Offset voltage and Bias currents. Also I required a SOT23-Package. I did read in the datasheet that there is a clock feedthrough ripple caused due to the offset correction. However, the amplifier is bandlimited to 100Hz. This is much lower and the 1kHz correction signal. So am I right in thinking that this would be filtered out due to the bandwidth limit of the circuit? Thanks \$\endgroup\$
    – LabMat
    Apr 19, 2023 at 16:34
  • \$\begingroup\$ @LabMat, as I said in my posted answer, I recommend you find a more suitable op-amp for this application rather than try to use a bad op-amp and fix it with filtering. \$\endgroup\$
    – The Photon
    Apr 19, 2023 at 18:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.