Ultra Precise/Accurate RTD sensor measurement.

Hi All. I am developing an oceanographic data instrument that need very accurate (+/- 0.01 degrees C) temperature detection capabilities. In fact +/- 0.001 degrees C would be preferable. I started a design that involved a precision current source, high-quality instrumentation amp. Seeing I am only interested in a narrow temperature range (-5 - 25 degrees C) I followed the instrumentation amp with a difference amplifier to remove the baseline voltage, and scale it up to the full scale of the A/D. I was going to do a 4 point measurement on the RTD to avoid any wire resistance effects. Then I found this: Analog Devices: Analog Dialogue: Transducer/Sensor Excitation and Measurement Techniques (Namely figure 9)

AC-excitation of the RTD seems to have a number of benefits. Does anyone have experience with this? Is this considered on of the "best" ways of measuring an RTD?

What I would do here is reduce the reference resistor so the AD would be full scale over my desired temperature range, and use the built in 6-bit DAC offset to remove the baseline from the RTD.

Any comments or tips would be greatly appreciated.

Thanks,
Sam

I think your dreaming. +/-0.001°C, even a very controlled laboratory environment, is difficult to achieve. +/-0.01°C accuracy in a field instrument is not going to be a simple circuit, even if achievable. We've used these, and they were the best I could find: **broken link removed** and https://www.swstechnology.com/pdfs/brochures/Diver-Series-Brochure-English.pdf Note the resolution is 0.01°C, but the accuracy is only 0.1°C. I suppose you might develop a correction table to get closer to 0.01°C over your range..

Sometimes researchers ask for spec's that would be "nice" for them, but not realistic. If you find a way, device, or circuit that will meet your spec...please share it. I could use it.

Ken

Precision is much more important to us, we can always calibrate it to remove any accuracy issues.... I already know we will need to linearize the data with the Callendar-Van Dusen equation. As long as I can get good precision the accuracy is relatively easy to correct.

Noise, although an issue shouldn't be terribly difficult to take care because I will be able to average over a minute period or so.

Certainly the advantages of AC excitation including no DC offset sensitivity, minimal 1/F noise effects, and minimal parasitic thermocouple effects seem compelling if you are trying to design a very sensitive temperature measurement system. Offset effects can be a killer in sensitive DC systems. Even a soldered wire joint can generate small thermocouple voltages.

Analog Devices is a very good analog design company and I would trust their recommendations. The example using their AD7730 chip may very well do close to what you want with minimal extra circuitry.