PT-1000, temperature sensor

Hi people

I have purchased a PT-1000 sensor to measure temperature. It's a RTD (resistance with temperature dependency). At 0 Celsius it's exactly 1000 ohms, and it raises its resistance about 4 ohms with each degree.

To measure the resistance I made a usual current divisor with a resistance of 1k. After reading the voltage I get the temperature from a look-up table. It works really well when it's calibrated, but the problem is the power supply has to be super stable, because if instead of 5.0 volts I got 4.98, the error is big.

I have to make a design independent from the power supply because I can't calibrate it each day. Any experience? What do you recommend: Wheatstone bridge, voltage reference, current source, ...?

Cheers!

I'm thinking that you want to have a local supply regulator that is fed 5V that regulates down to say between 3.3 and 4V with constant current control. This way, dips and spikes in 5V source get regulated out.

Using a TL431 as a voltage reference (cheaper) you could get +/-2% accuracy and 90mv Dropout versus a 7805 with nearly a 2V dropout. (You would need 7V in to get regulated 5V out)

The current used by the PT-1000 or similar KTY84 is so low that a TL431 would be sensible. (Untested by me)

You could use a simple JFET as a constant current source: http://www.allaboutcircuits.com/vol_6/chpt_5/15.html