I'm wondering if someone has a precision temperature measurement front-end circuit they would like to share?
From googling/reading it seems there are a few considerations:
- 4-wire measurement to avoid lead-resistance effects
- ratiometric operation of the ADC to reduce effect of current-fluctuations in the RTD
- AC-excitation to eliminate offset-effects, thermovoltages, etc.
My application has very long time-constants, so the 4.7Hz (or lower) conversion rate of e.g. the AD7195 is sufficient. Maybe there are other converters I should look at?
Also, I am wondering about the role of voltage-reference stability for this application (given that the measurement is ratiometric, and AC-excited). A colleague has worked with the LTZ1000 ultra-precision reference, but I have a feeling it might not be worth the cost&effort and an ADR421 or similar could be used instead (i.e. gain/offset errors are dominated by other parts, not the voltage reference).
Hi Chris, thanks for your quick reply.
I've sketched out something in NI Multisim:
500uA Howland current-source with a ADR425 or similar reference. Could maybe use a matched resistor network here for R1,R7,R8,R9 ?
Not at all sure what op-amp to use here, OP196 was just a random pick
ADG884 switch for switching the direction of the sensing current, e.g. based on ACX output from AD7195.
The simulation shows about 4-5 uA of current is lost in the switch - not sure why or if there is a better part for this.
4-wire connection to pt100. I'm interested in a narrow range of temperature say +20C to +40C.
112 Ohm reference resistor, to center output around the corresponding temperature. We pass the same sensing current through this resistor.
I've used the AD8221 for further amplification (it was suggested on the AD620 datasheet, a part I've used previously).
- The voltages over both the sensing and reference resistors are amplified
- Reference output goes to ADC reference input
- Difference between sensor and reference is further amplified to give the final output.
One problem with this is that IIRC the AD7195 reference input pins (REF- and REF+) do not allow absolute voltages outside GND and AVDD, and as such the ADC_REF output of my circuit might not be useable.