Temperature Sensor - A mixed Signal-Problem!

Hello all,

Actually I work on the design of a miniaturized board with many analog and digital parts on it. But my task is not fully done I still need to integrate an temperature sensor on it. But there is not enough space on this board for a voltage reference and an temperature sensor (i.e. LM20) PIC with ADU is still on the board. (A acurate way)
So I decided to use an NTC (or PT10K) an Reference-Resistor and an Capacitator. The processor charges and discharges the capacitator once over Reference-Resitor and ones over NTC. I just count the time for both charge cycles and calculate the temperature from the differences this counters. I read this in an articel somewhere, its very small (3x0603) and I can use 3 digital I/O's !

My problem is: How acurate could this measurement be.
If you think of an real capacitator and the behavior of this, way input-capacitance of the IO-Ports and so on?

Maybe someone has a good hint for me to think about!

Kuehnsen

There are several sources of error.

1. Putting current through the resistor will raise its temperature.

2. The capacitor has a temperature coefficient.

3. The current from the PIC has a temperature coefficient.

All in all I would suspect poor performance. You should look into squeezing the parts on the board closer together and use a real sensor like the LM20.