RTD/NTC self heating compensation.

[miro123]

Hello,
I was struggling with this subject in the past. During the Measurement meeting 2024 Dr. Pyte briefly mention that he use sqrt(0.5) current to determine the self heating effect. Later on somebody from chat session ask similar question.

1. Here are my findings on this subject
  - Simply switching the test current and waiting for a wile does not work optimal for the most application.
  - if you want long the changing conditions influence the result - most application measure non constant temperature. E.g after changing the current and wait I see 0.2K difference. Is this self heating error or measured temperature changed over time?
  -  0.707 ration was good to realize via hardware but not optimal to solve the problem. That said it gives some improvement - I can describe it as pickin up of low hanging fruit solution.
 2. My best solution so far.   
I use the following model The thermal probe is represented with thermal resistance and thermal mass.
Both parameters are variable over time since everything depend on testing environment , type of materials, fluid and fluid speed. Even simple fluid like air changes over time - Rh% atmospheric pressure, or somebody decide to open window or door
I use the RC electrical equivalent of the probe - Thermal mass is capacitance. Thermal resistance R.
I use two test current I1 and I2. I constantly switch them I1, I2 , I1, I2....
The temperature change is exponential curve. and I determine the exp function coefficients.
The algorithm has two tuning parameters are - 1. I1/I2 ratio   2. Test current switching time.
Higher current ratio allow faster and more accurate determine of exp function coefficients, but add HW complexity and HW limitations e.g. extra components, complex circuit,  limited SNR
I1,I2 chopper time determine in general the BW of my measurements.  Some feed forward component can be added to improve the BW

My question is are there any better solution on this subject? How do you determine the self heating effect in your application?

Lessons learned - Building something that it works is easy, making work for specific application is difficult but duable, making general purpose measurement equipment seems impossible for me.

BR,
MIro

[Andreas]

Hello,

my strategies are:
- reduce the measurement current as much as possible (dependant on ADC input impedance) (e.g. using 33K NTCs instead of 10K standard value).
- further reduction can be done by switching off (e.g. per Port-Pin) during those times where the ADC does no reading

See also 5Vs (switched) in AD587LW cirquit: https://www.eevblog.com/forum/metrology/ad587lw-10v-precision-travel-standard/msg1449488/#msg1449488

with best regards

Andreas