# Op amp to supply current to 3 thermistors and an ADC voltage reference input

I have 3 thermistors in parallel that I need to measure. These thermistors will be fast changing - think up to 100C in under 1 second.

I want to do this ratiometric measurement, so I want to supply the ADC reference from the same voltage which the thermistors are excited by.

Is it OK to do this using 1 opamp follower from the voltage reference? The LT1499 says it can supply up to 30mA, and there will be lets say max 2.5mA for each thermistor when it heats up (total 7.5mA), and a few 100um for the ADC reference pin.

Is this a normal technique? It would mean I can use just a quad package of opamps (+3 for buffering the thermistor output to the ADC input channels)

These thermistors are fast moving temperature, so will go from 10k to 500 Ohm in under 1 second, and with the current needed to maintain the 2.5V going up 20x accordingly. So this is effectively a variable load for the op amp? and I may have to put some 10n or 100n caps in parallel with the thermistors

Well, you have 800 ohm in series with each thermistor so, at the minimum thermistor resistance, the op-amp output sees 3 parallel resistors of 1300 ohms = 433.33 ohms and, if as you imply, the reference voltage is 2.5 volts then, the total current to be supplied by the op-amp output is 5.8 mA. I don't think this will be a problem to most op-amps providing that the 2.5 volts isn't close to their lowest output voltage when the negative supply is at GND.

The LT1499 can supply that.

• Would it preferable to use the voltage reference IC directly and put a 100n or 1u cap on the output? It is possible to buy ones that claim 30mA output current as well, like ADR441 But I need the quad op amp package anyway, and just feel like it is not normal to run variable loads with several mA from a reference IC Commented Jun 9, 2020 at 22:01
• You might be able to use it directly. Some references can supply tens of mA but I would only consider using one unbuffered if it can deliver the goods effectively AND the added offset error from the op-amp was intolerable. Commented Jun 9, 2020 at 22:31

No, the resistance seen by the opamp for each thermistor only ranges from 10.8k down to 1.3k, an 8:1 change in current.

The total maximum current from the opamp for all three is 5.8 mA, well within its capabilities.

And changes over 1 s are glacially slow as far as the opamp is concerned.

The reference amp output current changes 5.075 mA in under 1 second. Lets call it 6 mA per second. That's pretty slow. Except for the noise, offsets, and headroom issues, a 741 could do it. Same for the three buffers - 1.35 V in under 1 second, rounded up to 1.5 V or even 2.0 V per second, still is very slow.

Focus on the opamp's input errors and their drifts. If the chip is seeing even a fraction of the same temperature swings as the sensors, thou shalt have wandering error terms in the output voltages.

Caps in parallel with the thermistors will slow things even more.

Update: In case it isn't clear, I am not recommending you use an LM741 in this application.