Recently I published a simple platinum resistance temperature detector (PRTD) design idea (Reference 1) that was largely inspired by a deviously clever earlier DI by Nick Cornford (Reference 2).
Remarkable and consistently constructive critical commentary of my design immediately followed.
Reader Konstantin Kim suggested that an AP4310A dual op-amp + voltage reference might be a superior substitute for the single amplifier and separate reference I was using. It had the double advantages of lowering both parts count and cost.
Meanwhile VCF pointed out that >0.1 °C self-heating error is likely to result from the multi-milliamp excitation necessary for 1 mV/°C PRTD output from a passive bridge design. He suggested active output amplification because of the lower excitation it would make possible. This would make for better accuracy, particularly when measuring temperatures of still air.
Figure 1 shows the outcome of some serious consideration and quiet contemplation of those, as they turned out to be, terrific ideas.
 |
|
| Figure 1. | Nonlinearity is cancelled by positive feedback to PRTD constant excitation current feedback loop via R8. A2’s 10x gain allows reduced excitation that cuts self-heating error by 100x. |
A1’s built-in 2.5-V precision reference combines with the attached amplifier to form a constant-current excitation feedback loop (more on this to follow). Follow-on amplification allows a tenfold excitation reduction from ~2.5 mA to 250 µA with an associated hundredfold reduction in self-heating from ~1 mW to ~10 µW and a proportionate reduction in the associated measurement error.
The sixfold improvement in expected battery life from the reduced current consumption is nice, too.
The resulting 100 µV/°C PRTD signal is boosted by A2 to the original multimeter-readout compatible 1 mV/°C. R1 provides a 0 °C bridge null adjustment, while R2 calibrates gain at 100 °C. Nick’s DI (Ref. 2) includes a nifty calibration writeup that should work as well here as in his original.
Admittedly the 4310’s general-purpose-grade specifications like its 500-µV typical input offset (equivalent if uncompensated to a 5 °C error) might seem to disqualify it for a precision application like this. But when you adjust R1 to null the bridge, you’re simultaneously nulling A2. So, it’s good enough after all.
An unexpected bonus benefit of the dual-amplifier topology was the easy implementation of a second-order Callendar-Van Dusen nonlinearity correction. Positive feedback via R8 to the excitation loop increases bias by 150 ppm/°C. That’s all that’s needed to linearize the 0 °C to 100 °C response to better than ±0.1 °C.
So, cheaper, simpler, superior power efficiency, and more accurate. Cool! Thanks for the suggestions, guys!
References
- Woodward, Stephen. "The power of practical positive feedback to perfect PRTDs."
- Cornford, Nick. "DIY RTD for a DMM."
