I have several transimpedance amplifiers on my board, and I want to avoid the need for a negative power supply, but I can't see any way around it (except for running the current through a resistor but I dont want to load the circuit like that).

At the moment I've been using negative potential to avoid the need for two op amps per channel, but this is also not desirable, is there any way to have a TIA setup without a negative supply?


EDIT: I should elaborate, the elements being read are thermistors. Each thermistor is excited by a buffer, the current through the thermistor is then read out by the TIA.

  • \$\begingroup\$ Any problem in having DC bias? \$\endgroup\$
    – awjlogan
    May 17, 2018 at 11:49
  • 2
    \$\begingroup\$ @Himmel I don't know how many times I've heard lame excuses for not posting a schematic. \$\endgroup\$
    – Andy aka
    May 17, 2018 at 12:16
  • 7
    \$\begingroup\$ wait what? 200kHz bandwidth for a thermistor suggests something highly unusual you're not telling us ... or some wildly unnecessary specs. \$\endgroup\$ May 17, 2018 at 12:17
  • 3
    \$\begingroup\$ @BrianDrummond Have you never tried to keep micro-gram metal objects stable to the degree of mili-Kelvin before? It's very normal practise in mainland Europe to do that with thermistors with non-zero mass. :-P \$\endgroup\$
    – Asmyldof
    May 17, 2018 at 12:21
  • 4
    \$\begingroup\$ @Himmel, to avoid IP issues you might need to re-draw just the relevant parts of your circuit, leaving out details that are proprietary. If proprietary details are necessary to answering the question, you'll either have to reveal them or you won't get a good answer. (Usually the real proprietary information is stuff like how much you pay for a part, who you sell it to, etc., and not the design of the circuit itself, especially if you're asking the internet to help you design the circuit) \$\endgroup\$
    – The Photon
    May 17, 2018 at 15:46

1 Answer 1


I'd do this as a comment, but it will be too long, so bear with me.

First, what you are doing is (at heart) a resistance measurement. Restricting yourself to thinking in terms of current is a potential source of confusion. We call this an XY problem. That is, you want to do X, and you think you MUST do it via Y, and so you ask for help with Y. Which is fine as long as you really must do Y, and not so great otherwise.

Now, about temperature. You are apparently committed to thermistors. The first thing you need to realize about temperature is that it changes very, very slowly (seconds timescale). There are exceptions, but these usually involve things like high-temperature ablation of heat shields and such. Most materials have sufficiently high thermal capacity and low thermal propagation rates that if you want to change the core temperature quickly, and needing 200 kHz measurement bandwidth counts as very quickly indeed, then you have to apply so much thermal power to the surface that you'll burn it away. Look - heat just doesn't propagate that quickly through most materials, even good thermal conductors. The fact that you seem to want to measure temperature at 200 kHz suggests very strongly that you have hold of the wrong end of an XY problem.

Now, let's say you are thinking about thermistors. And I do mean thinking. What sort of temperature range do you need? Thermistors have a wildly nonlinear response, so they are incapable of both sensitivity and wide temperature range. That is, assuming a NTC thermistor, as temperature rises the resistance rises even more quickly, and the current drops appropriately, so at the high end of the range you'll get current response much smaller per degree than at the low end.

Except for high-precision measurements, which are of necessity restricted to a fairly small temperature range, the standard thermistor circuit looks like


simulate this circuit – Schematic created using CircuitLab

VREF can be something as simple as a low-voltage zener diode, but the details are determined by your system accuracy requirements.

R1 is normally set to approximately the thermistor resistance at the bottom of the temperature range.

C1 is used for noise filtering, and since thermistors are usually used in the kilohm range it can be quite large, and the resulting low bandwidth can often handle quite large amounts of noise, so long as it occurs at higher frequencies.

Also, while not shown, if the temperature sensor is mounted off the pcb and connected via wires, a good idea is to put a resistor (say 1k to 10k) between the thermistor and the op amp/capacitor. This will tend to protect the op amp in case somebody plugs the wrong wire into the wrong hole while installing the external thermistor. Oh yes, and did I mention the virtues of connecting with twisted pair wiring?

If you feed the output to an ADC, your temperature resolution will be determined by the ADC resolution (and accuracy) as well as your temperature range. It will be very good at low temperatures and get progressively worse as the temperature goes up. The tradeoff is up to you.

If very high precision is needed, you should look into platinum RTDs, which you can think of as a special variety of PTC thermistors. They are very stable and accurate, and they can be used over large temperature ranges. Of course, nothing comes free or easy, and the cost is that they are not very sensitive, so you need to take care with your conditioning circuitry.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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