For a thermistor in a car, it probably won't make any difference. But if it did? How would you know? One way you might find out is you might get an opportunity to rebuild your engine because the ECU did something stupid with an erroneous reading. (We know how well that's working at the moment for Boeing's MAX aircraft, don't we?)
bit of an extreme example. as far as knowing if it makes a difference, i can record the voltage to the ecu pin with a scope and lift the probe from arduino a couple times to see if voltage changes at all. i'll see what kind of risk i have for an engine rebuild lol. before i wire up though, was looking for comments whether the arduino circuit might have any obvious load to bias voltage which it sounds like it does not.
Yes, it won't be accurate.
The normal thermistor circuit on the Arduino is fully ratiometric, which depends on the resistive divider
being powered from the Arduino's Vcc, which is also the default reference for the ADC.
Using a separate supply voltage from another device will mean a different voltage will be used for
the resistive divider than the Arduino's reference.
5V supplies on computers have only about +/-5% accuracy and often considerable noise too.
You could use a 1:1 resistor divider powered by the computer to allow the Arduino to sense its
voltage (relative to the Arduino's 5V), and thus correct for the difference, but this won't eliminate
noise and errors due to noise.
BTW its not a pull-up resistor, its one arm of the resistive divider (or half-bridge).
Pull-ups are for logic signals and don't need to be accurate, this is an analog circuit where
the resistor accuracy is vital.
Thanks Mark. Great comment about measuring 1:1 divider and correcting, had not thought of that.
The whole tolerance stack is really not that critical. I'm going to give it a try. I'll report back as a matter of interest or any help to others that want to share a sensor with ECU.
That RC filter worked very well when I had this setup with it's own sensor. Before and after pic of the voltage reading. Scope set to AC so I could zoom in on voltage around 0v. It's within 1 adc step with the filter and I poll an avg.
The interesting part will be testing any change to voltage to ECU with scope when pin is lifted. I'll also be able to read ECU reading next to Arduino reading so I'll see any temp differences, however ECU only outputs 8 bit data in 1°c granularity through OBD so I can't really split hairs which is not a bother. Main idea was to read it direct so I can run fan PID algo using higher resolution by tapping into it direct instead of pulling from OBD (and get rid of a dedicated sensor). On Arduino, each ADC step is 0.1-0.2°c depending how close the sensor resistance is to the 'one-arm of the resistive divider'
Reading sensor directly and calculating inputs for PID with 0.1c resolution instead 1.0c resolution from OBD helps the fan PID stay steady by quite a bit.