SamoL It looks like EAF has a bug as this appears on all my EAFs with external temperature sensor (four of them, each with it's own sensor) and with two different AsiAir Plus.
The bug is in the EAF firmware. The bug has been there for a while now, and I had bought a 10 turn potentiometer a while ago just to measure it (at this point I have 8 EAFs).
The following are readings from a recently purchased EAF (July 11, 2022) that I just measured.
The jig looks like this:
The blue barrel is a 10-turn 100 kΩ wirewound pot with a turns counter, in series with a 47 kΩ resistor (you can barely see the SMD resistor). This gives me a range of about 50 kΩ to 150 kΩ. The other end of the cable is another 3.5mm stereo pulg that is alternately plugged into a multimeter and the EAF (the photo shows it connected to the meter).
I stepped the pot in 1 kΩ steps, and at each step, plugged the end of the cable into the EAF to take a reading from ASIstudio (macOS).
The EAF readings look like this:
The firmware behaves normally at above 26ºC -- the EAF assumes a 100 kΩ NTP (negative temperature coefficient) thermistor centered at 25.0ºF. A typical table can be found here:
Notice that starting at 93 kΩ, the EAF is reading properly. Until the resistor reaches past 98 kΩ (EAF reads 25.2ºC). It then suddenly jumps (at 99 kΩ) down to 17.7ºC and continues to fall to the minimum at 101 kΩ (EAF reads 16.3ºC). The EAF recovers at about 104 kΩ, reading 23.5ºC and continues to fall as the NTP curve.
Exacerbating this problem is that the EAF is reading the resistance with a cheap 8-bit DAC in a microprocessor (it itself has a temperature dependency) and this is what is causing the ocscillatory readings that you folks are seeing: each reading is sampled as a different resistance. The Fluke 76 meter is good to 0.4% + LSB, by the way.
Until this is fixed in firmware (and heaven help you if, like me, you only use macOS and Linux), your only recourse is to add a 10 kΩ resistor in series with the thermistor to bias the readings above and away from the 100 kΩ region of the thermistor. The slope of the curve in not too far off that you can still judge ∆T sufficiently well.
With a 10 kΩ in series, 26.3ºC (94 kΩ thermistor) will appear as 23.5ºC (about a 3ºC difference). The error will then move from the 25ºC region up to the 28ºC region. If you have warmer nights, bias it with a 15 kΩ resistor or something, but any addiional bias will introduce a larger slope error.
Just get two 3.5mm stereo TRS (tip-ring-sleeve) jacks. Connect the sleeve of the two jacks together, and bridge the tip of the two jacks with the 10 kΩ resistor. Thats is; if you just need ∆T readings. Don't wire anything to the ring of the TRS jack -- it is used by the EAF handcontroller.