Measurement circuit for electro-chemical oxygen sensor

[pman92]

I'm trying to accurately measure the output of an R22 electro-chemical oxygen sensor for a project with a microcontroller.

Essentially these sensors output ~10mV at atmospheric oxygen levels (21% PP02), and include an inbuilt resistor network for temperature compensation. The output varies linearly (within limits) with oxygen partial pressure. They are commonly used in diving equipment (don't worry, that's not what I'm using them for). There's an excellent write up about how they work here: https://www.advanceddivermagazine.com/articles/sensors/sensors.html

This is my first attempt at a circuit:

The idea was for the TL431 to generate a 2.5v reference voltage so we are working within the input common mode range of the INA333 instrumentation amplifier. The sensor is connected across load resistor R14. R15 and C29 provide some noise filtering. R16 (505R 0.1%) should set the amplifier up for a gain of x100. The voltage out to the ADC would be 100 times the input (eg. 1 volt at 10mv input) so it can be measured more accurately by an ADC.

However my circuit doesn't work very well.
When I measure across the sensor in-circuit with my multimeter, I find 9.8mV. When I measure with my oscilloscope, I find a DC average of 12mV.
When I measure the sensor out of circuit, both my multimeter and oscilloscope agree on almost exactly 10mV.
The voltage over the TL431 is 2.481v.
I shorted R15, and still find 12mV with the oscilloscope.

The output voltage of the INA333 is also way off what I was expecting, 1.88v (IE not a gain of x100, more like x180).

Everything from the ADC onwards works as it should. The ADC reading is agreeing on the 1.8v I measured on its input. Its the circuit trying to amplify the small 10mV signal to something meaningful that I can measure accurately that is causing the problem.

Can anyone offer guidance / recommendations? At this point I am getting a bit frustrated, and I'm would be more than glad to scrap the whole lot and go back to the drawing board. What is the correct way to amplify the 10mv signal accurately?

Thanks in advance

[tszaboo]

TL431 doesn't like capacitors.
Also, please take 5 minutes to rearrange your circuit. Power supply up, ground down. Inputs left outputs right. Draw the INA as an actual amplifier not a box.

[CaptDon]

I haven't looked at your exact sensor datasheet but isn't the 100K resistor a bit low in value? The sensors we used at the hospital worked into FET opamps with a very high input resistance and I think the 'load impedance resistor' was more like 1 or 10 megaohms. Be aware these sensors are basically operating like a battery and oxidizing/consuming a substance. The richer the oxygen content the faster the device ages!!! When new the sensors are sealed. They will consume the very small amount of oxygen within the sealed area and the aging process will slow down drastically. The output voltage drops off as the sensor ages. We replaced them yearly at the hospital and tested them monthly for accuracy at higher and lower oxygen levels to observe accuracy across the measurement band.

[Terry Bites]

Sorry to point out the obvious short circuit!

[Kleinstein]

The "short" may be OK, as the TL431 seems to be there to produce a kind of virtual ground. 2.5 V from the ref. could be OK here.

Most OP-amps and INAs don't like 10 nF at the output and tend to oscillate. So not just the 2.5 V ref. but also the amplifier with too much capacitance.
There is quite some load resistor to the cell aready inside the sensor - it really depends on the sensor model how much additional load resistance is correct. The resistor would effect the temperature compensation and the scale factor.

Overall the circuit is rather odd and complicated. A simple 1 OP-amp amplifier should also be sufficient.