Pt100 oddities

[RocketScienceSmurf]

This maybe isn't the right place to ask but I'll give it a shot anyway since there is a lot of knowledge here in a lot of different fields.

I was helping a friend a couple of months ago trying to calibrate cheap ebay 4-20mA transmitters for Pt100 temperature sensors. A while back I found ultra precision resistors (0.01% 2ppm) of 100 and 120 Ohm. If we convert these resistor values to Pt100 temperatues it is 0,00 degrees and 51,57 degrees Celsius +-0,03 degrees or thereabout.

We connected each resistor to a 4-20mA transmitter and applied voltage. The transmitters were connected to shunt resistors on an Arduino Uno. After letting the system stabilize for 24 hours we wrote down the A/D values and switched resistors between the transmitters to get two calibration points and repeated the procedure.

After this was finished we assumed we had a fairly good compensation for any miscalibration in the cheap transmitters.

Then we connected two cheap Pt100 sensors also bought from China on ebay. They showed two different temperatures that were nowhere near the actual room temperature so we decided to make an ice bath calibration to get the offset for each sensor. We assumed this offset would be fairly linear and that once we subtracted or added the offset the sensor would show roughly the correct temperature in the 0-50 degree range but the temperature reading was of by ~3 degrees at regular room temperature.

Since calibration equipment is expensive I have no way of verifying that the sensors are close to the specs they SHOULD have. Should we have calibrated in a different manner? The resistors are brand new but has few years on them so there might be some ageing but not this much I think. Also, two Pt100 devices bought from the same seller at the same time should be at least within on degree from each other I think.

[T3sl4co1l]

I don't get it, Pt100 sensors are usually used for reasonably high precision applications, and the "ultra precision resistors" sound like confirmation.  But then you've got an unstated 4-20mA transmitter, an Arduino (what, 12 bits at best..?), and... floor sweepings from China?

Idunno.  What were you expecting?  What do you actually want?

Within a few percent, after calibration, from random Chinese parts, that could've been plain old carbon film resistors on leads, or something?  Heck, turned out better than I would've expected...

Tim

[RocketScienceSmurf]

From what I can remember the A/D value was exactly the same when we reverted to the fixed resistors. In my opinion this confirms that the method can be used to compensate for miscalibration in the transmitters and suggests that the problem is the actual Pt100 sensors but I could be wrong.

Our goal/hope was to get an accuracy of roughly +-0,5 degrees. I haven't done the math with the atmega328, voltage references, shunt resistors, etc etc but I feel that it might be possible to reach 1% (+-0,5 degrees in the range 0-50 degrees). The most important thing for this project was not absolute accuracy but rather that difference between the two temperatures measured was accurate and constant. The intended application was to measure forward and returning water temperature in a radiator/convector circuit and maybe also add extra sensors for outdoor and indoor temperature.

[johnf]

All of this depends on many things
are you using 4 wire or 2 wire did these things (PT100s)cost 10$s each?
What voltage are you using across the PT100 (self heating)?

what actual grade where the PT100s 5% 1% 0.1% 0.001%?

[davekni]

If I'm understanding correctly, you are down to calibrating the cheap Pt100 sensors, as your electronics are repeatable.  If so, a scale factor is more appropriate than an offset.  Being cheap, these are likely thin-film RTDs, perhaps rejects where the deposited platinum was too thin so couldn't be laser-trimmed to 100 ohms at 0C.  Measure the actual resistance at 0C.  Calculate the required sale factor to get to 100 ohms.  Then multiply all resistance readings by that scale factor prior to converting to temperature.

I think platinum resistance/temperature curve extrapolates to 0 ohms at -260C.  So, if your 4-20mA results are already in degree C units, add 260, multiply by the scale factor, then subtract 260.