I am making a medical device that requires to sense very very low intensity blue light. I have attached my circuit design which is a 2-stage amplifier(transimpedance and differential amplifier), but I still need to increase the total gain of the circuit. Can someone please tell me how should I make modifications to my design to have a much higher gain. Should I add another 2-stage voltage amplifier?

Currently in my circuit (from the diagram), Rf=1Mohm, R2=1Mohm. If I increase these values, my dark voltage increases, which I do not want. It should give zero in dark which I am getting with these values. Photodiode being used is hamamtsu S1133 and opamp being used is TI OPA2381

Thanks. enter image description here

  • 2
    \$\begingroup\$ It's very unusual to have the ground pins of Opamps "up in the air, and the 5V supply on the lower side. I don't know your schematic capture program, but I'm hopeful it has a "flip symbol" operation so you can fix that, and have supply and the (+) non-inverting input on the upper side of the Opamp. Can you explain why you chose exactly that Opamp? I'm sure there's a good reason, but understanding why you chose it will make it easier for you to recommend improvements, I promise :) \$\endgroup\$ Aug 31, 2017 at 21:01
  • \$\begingroup\$ For example: You're using an output filter with ca 1.6 Hz cutoff frequency (why?), but you use an Opamp that has a gain bandwidth (and thus, WILL amplify noise in that bandwidth) of 18 MHz. Find the discrepancy! \$\endgroup\$ Aug 31, 2017 at 21:03
  • \$\begingroup\$ @MarcusMüller Output filter cutoff is ~18Hz. The transimpedance amplifier has a cutoff of ~280Hz due to Cf. \$\endgroup\$ Aug 31, 2017 at 21:14
  • \$\begingroup\$ @SpehroPefhany ah, yeah, my in-head-square-rooting mixed up the number of zeros; sorry, Kashish! \$\endgroup\$ Aug 31, 2017 at 21:17
  • 1
    \$\begingroup\$ Do you have access to the blue light source? Can you modulate it? \$\endgroup\$ Aug 31, 2017 at 23:35

3 Answers 3


Since you can modulate the blue LED which is your light source, you should look into building a lock-in amplifier. A good starting point would be the AD630, although there may be more modern ICs available which do the same job.

Basically, you modulate the LED with (let's say) a 1 KHz square wave. Then you demodulate the TIA output synchronously with the modulation, using a fairly low bandwith low-pass filter, something like 1 Hz lowpass. The result is that the DC component of the TIA output, including offsets, gets transformed to AC at 1 KHz, and is discarded by the LPF. The LED signal, meanwhile, shows up as a DC component and is recovered.

This is only possible if you know the phase of the modulating signal, and since you are driving the LED, you get this for free.


It is best to increase gain at the TIA because it improves SNR whereas increasing gain at the voltage amplifier mostly just changes the signal level. You will need to decrease the feedback capacitance to maintain the same bandwidth. Your maximum possible bandwidth will decrease since you are increasing the high frequency noise gain (1 + Cd/Cf). Increase the feedback resistance and decrease the feedback capacitance as much as possible while achieving your target bandwidth, then increase the gain of the voltage amplifier if needed.

To minimize DC offsets and read 0V in dark conditions:

1.) Match the impedance of the inverting and non-inverting inputs of the TIA. Choose an op amp with low input offset current. If you are using very large feedback resistors, be careful that the (input bias current) * (feedback resistance) is within the common mode input range of the op amp.

2.) Choose op amps with low offset voltages: the TIA is less sensitive to offset voltage because Vos will experience a gain of ~1 if your photodiode has a large shunt resistance. Vos matters more on the voltage amplifier, especially if it has significant gain.

  • \$\begingroup\$ Thanks for your answer. Isn't feedback capacitance inversely related to freqeuncy/bandwidth? Because if that is so then won't decreasing the capacitance increase the bandwidth and in turn decrease the gain? \$\endgroup\$
    – Kashish
    Sep 1, 2017 at 22:04
  • \$\begingroup\$ BW = 1/(2*pi*R*C) so if you want to increase R for more gain you have to decrease C to maintain the same bandwidth. \$\endgroup\$
    – DavidG25
    Sep 1, 2017 at 22:07
  • \$\begingroup\$ But in data sheets they refer to the term Gain-bandwidth and not bandwidth, which is a constant term. So if gain=R and BW=1/(2*piRC) , then isn't GBW product =1/(2*pi*C). Its a little confusing. Could you please clarify? \$\endgroup\$
    – Kashish
    Sep 1, 2017 at 23:05
  • 1
    \$\begingroup\$ Gain-bandwidth product (GBW) refers to op amps that have an open loop gain with a single pole. The concept you are thinking of is more applicable to voltage amplifiers with a flat frequency response, such as a basic inverting amplifier. The bandwidth of a transimpedance amplifier is set by the feedback elements. It is limited by the op amps open loop gain, but it is not as straightforward as an inverting amplifier for example. \$\endgroup\$
    – DavidG25
    Sep 2, 2017 at 0:53

If you have an offset in the front end, it will be amplified by subsequent stages.

Since your op-amp has a bias current of only 3pA the typical output offset due to the bias current is only about +/-30mV. Input offset voltage of +/-7uV typically results in +/-70mV at the output.

Increasing Rf (and likely decreasing Cf proportionally) is probably the best approach, as it will increase the former but not the latter. You could go to 10M (or 100M and decrease the gain in the differential amplifier by 10:1).

  • \$\begingroup\$ But, if the bias current is the critical thing, why not optimize that? Wouldn't it be worth reducing that current by a factor of 10, simply by going for a lower-offset Opamp (LMP771* comes up, but others, too)? This is a very honest question, because I feel that a JFET input stage transimpedance amp might sound like a better idea for someone with more experience than I have (I'd have gone for any CMOS low offset voltage/bias current opamp), but I don't have any guidelines on selecting the right opamp type. \$\endgroup\$ Aug 31, 2017 at 21:24
  • \$\begingroup\$ @MarcusMüller It might be but the offset voltage is much worse on that amplifier (150uV max/25uV typ with a large drift). So if OP finds an amplifier with much lower Ib then they can decrease the gain of the subsequent stage and increase Rf. It's not clear to me why 30mV (typ) or even 300mV is an issue, although I would like to see some bias in the output amplifier so that it won't end up on the wrong side of ground. \$\endgroup\$ Aug 31, 2017 at 21:29
  • \$\begingroup\$ aargh, I mixed up the units; it's picoamperes on the LMP7716's input bias current, but millivolt on the offset. Sorry. \$\endgroup\$ Aug 31, 2017 at 21:33
  • \$\begingroup\$ @MarcusMüller no worries. \$\endgroup\$ Aug 31, 2017 at 21:41
  • \$\begingroup\$ I think you would need some expensive PCB dielectric to pull off 100M feedback resistance. \$\endgroup\$
    – DavidG25
    Aug 31, 2017 at 22:08

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.