Characterizing a Thermistor

[Noggin]

I'm in need of some food grade temperature probes. I purchased some replacement probes for a DTTC thermometer and planed on using them for a project, but I can't seem to figure out their curve. I made an Excel spreadsheet that generates resistance values for different temperatures a while back. The spreadsheet is based on the Steinhart-Hart equations and has worked well in the past for characterizing thermistors, but it chokes on these DTTC probes.

The data I've gathered on these probes is 445000 ohms at 5.6°C, 285000 ohms at 20.6°C and 24500 ohms at 88.9°C.

Are there any thermistors that, for lack of a better word, don't "thermist" in some temperature ranges? Do I need to get temperature/resistance values at higher temperatures?

As an alternative, does anyone have a good, low cost source, for food grade thermistors? I've had a difficult time finding anything from a good source. Coming with a datasheet (or at least a known beta) would be an awesome plus.

 

[MikeMl]

Do you have more than three points? Maybe we could do a least-squares fit through the points. Who cares if it obeys te Steinhart-Hart curve.

 

[Noggin]

I only have 3 points at the moment, I've contemplated just attaching it to one of my thermistors I do know, putting them on a data logger, and then tossing them in an oven. I could just build a whole table that way with a lot of data points. I would prefer to have a formula I could use instead of a lookup table though. But I suppose I could get that whole set of data, get a curve for it and let Excel give me a formula for it. That's probably what I'll have to do. These unknown thermistors were pretty inexpensive but are used on at least a few kitchen thermometers. $6 each as opposed to $12-$20 for most others I've seen.

As for your comment, "Who care if it obeys the Steinhart-Hart curve," does that mean that it isn't uncommon to have a thermistor that doesn't fit in the curve? I had thought that they all could be characterized like that.

 

[MikeMl]

I have seen some that do not fit the Steinhart-Hart curve. If the goal is to get a math function (with a few coefficients) that fits measured resistance vs temp points, there are many ways to do that. I have used the least-square error curve-fitting methods in Excel to obtain a function that behaves like the thermistor for some specified temperature range. This is useful when simulating a circuit containing the thermistor. I use the "varible resistor" model in LTSpice to do that.

As to getting a temperature measuring probe, have you looked at the PT100 based ones. I found a bunch of those in a surplus store once for cheap, and used them for several projects.

 

[Noggin]

The PT100 looks like it uses RTD probes. While I'd have better results with an RTD, I'd prefer to stick to thermistors so I can connect them directly to my ADC. I'm fine with +/- a couple °C error. I've made a Raspberry Pi smoker controller and want to measure a bunch of temperatures with it so I know where the meat is in its cooking cycle. I have 2 probes that I can use right now, I've borrowed 3 from a co-worker, and I've bought 5 which I don't have a curve for just yet. The 5 probes I can use all measured within a degree or so of each other, and within 2 degrees of the thermometer they came with. Assuming the thermometer is accurate (and I don't really have a reason to believe it has been calibrated), then my Raspberry Pi is +/- 3°C. That 3°C resulted in a FINE brisket, chicken, turkey breast, turkey legs and pulled pork yesterday. God dang that was an easy smoke. All I had to do was add wood chips for smoke generation and watch my screen to let me know when it was done.

Thanks Mike!

 

[rumpfy]

I checked my Philips databook for NTC thermistors and found a type 2322 635 01334.
This has a B value of 4200 +/- 5% at 25 C
It shows R25 as 330Kohm. The typical R90 is say 20k. The maximum temp shown is 150 deg C.
There is another 330k type with a maximum temp of 200 C. It is 2322 627 01334. This shows B25 of 4175 deg K. R25 is 330Kohm. R at 90 deg C is again, about 20Kohm. Looks similar to yours given the uncertainty of temperature /resistance measurement of these things.
I suggest your biggest problem will be to get something that is suitably packaged in a robust sheath with the suitable high temperature lead-out wires.
hope this helps.

 

[Noggin]

That looks like a handy book, and those values might work.

What I ended up doing is I took my two known thermistors and connected them to my Raspberry Pi. I then connected my 5 unknown thermistors to my Rasp Pi. I bound all of my thermistors together and tossed them in my kitchen oven. I heated the whole thing up to 400°F and then turned the oven off. I let my RPi log my temperatures from my known probes and log the ADC values from my unknown probes as the temperature dropped from 400°F down to about 100°F over the course of a few hours. I used the LINEST formula in Excel to get a coefficients for a 5th order polynomial and I can use that in my RPi to calculate the temperatures.

My RPi measures my known probes to within 2 degrees of what my thermometer reads (the thermometers that came with the probes) and my new and unknown probes read within 1 degree of the known probes.

These probes are fairly inexpensive, and are intended for food, so I think I'm good to go. If any break, I'll just buy more.

https://plus.google.com/100606526905010626686/posts/UmoaVMjs7G7

 

[KeepItSimpleStupid]

I'm surprised you used an oven. I did a characterization using water for temperatures below boiling. Just kept adding cold water and ice.

For temperatures above boiling, I used oil. Peanut oil works well. See: https://www.smartkitchen.com/resources/peanut-oil DO NOT add water.