I am aiming to detect a 1 degree C rise in temperature per minute using a 5K thermistor and an attiny24.
I was taking some observations of the voltage variations at the thermistor output and realized that if I leave the thermistor in a voltage divider arrangement connected to the 5V power supply the thermistor gradually self heats and produces just enough variation that my controller registers a false reading and registers a 1 degree rise in temperature per minute.
I was just wondering if its possible to say turn off and on the 5 volt supply to the thermistor using a relay and turn on the supply just before I do an analogRead(), in this way I can avoid the thermistor from self heating and cause very little error in my readings? would really appreciate if anyone can share their views on this, is there any drawback to using this approach?
Secondly I was just going through the datasheet of my thermistor and it says the thermal coefficient is 6.6mW/C and it is good practice to keep the wattage to around half or around 3.3mW to in order to avoid large variations through self heating. I am attaching a rough schematic of the circuit I am wired up.
power the thermistor from a digital output. turn HIGH - read analog - then LOW.
As a thermistor often is +/- 2..10% off, the curcuit has to calibrated anyway.
Thank you both for your replies, knut_ny I was thinking along the same lines, think I will first try to supply the thermistor through a transistor instead of using a relay and see how that goes.
Well I tried with a bjt and although I have almost completely eliminated the error in my readings using this thermistor circuit on and off approach, I realized that I am getting a voltage drop of around 0.5 volts across the collector-emitter, where as the digital output pin itself is a solid 5 volts, I have made all my calculations using 5volts as the voltage divider supply so I will most likely be testing out the circuit with the thermistor voltage divider driven directly through an output pin.
