Here's a eye-opener. I hooked an o-silly-scope up to the thermistor on my THGR122NX sensor to see what is happening when the sensor measures temperature. I used a 10Meg-ohm, 10x probe with 16pF of capacitance. A PNG image with the voltage-vs-time waveform is attached to this post.
Yikes -- they ARE using an AC signal to measure the thermistor's resistance. I've only shown the very beginning of the waveform here (first milli-second or so) but it lasts for about 0.1 seconds and looks like this all the way through. The frequency is about 12kHz.
This explains why cable capacitance creates temperature errors. For those who aren't electronics-savvy, a capacitor looks like a resistor when AC is present -- and the larger the capacitor and the higher the AC frequency, the lower the equivalent resistance presented by the capacitor. This "distorts" the apparent resistance presented by the thermistor and the OS sensor winds up thinking that it's warmer than it really is. For those who are electronics savvy, my apologies -- I've taken great liberties in this analogy in the hopes that it will suffice for the current topic.
The next obvious experiment was to connect various-sized capacitors across the thermistor and see what happens. These measurements were made at about 17C where the thermistor has about 14k-ohms of resistance. The capacitance values and temperature errors I got are as follows:
Capacitance (pF) | Temperature Error (C) | Note |
16 | 0.1 or less | 10x scope probe capacitance |
95 | 0.3 | 1x scope probe capacitance |
350 | 0.7 | |
1000 | 3.2 | |
Someone mentioned using coax to extend the thermistor away from the sensor. I don't know what kind of coax was used, but for example, RG58/U coax has a capacitance of 28pF/foot. Using just a foot of this coax can create small errors at 17C (maybe 0.1 to 0.2C), and using 10 feet would probably create 0.5C or more of error.
Now that the measurement waveform is known, it is safe to conclude that this problem gets worse at low temperatures and better at high.
I also looked at a THGN801 sensor -- measurement waveform is almost identical.
I'd like to thank everyone who posted on this topic -- I never would have figured this out otherwise. I hope you all find this interesting and useful.