Precision Temperature Sensors

Here is one way to get away from the look up table. I hope this helps.

Basic Temperature Measurement Using NTC Thermistors

http://www.thermistor.com/BasicMeasurement.cfm

Have you tried using an RTD? These products meet the specs you outlined: RTD Elements.

We have found the same problem and have gone to thermocouples. These are more expensive, but their accuracy, reliability and longevity are well worth the effort, plus they are not unlinear like the thermisters.

You could try averaging your signal over some time to get the extra precision you need.

What sort of calculations are you using to generate your table? My guess is that you are just using proportional control. If you introduced differential and integral control you may be able to get a more stable output. If you want more details pleas post your questions here.

We use RTD's at my place of work. We calibrated them to 0.05C

When used in a wheat-bridge configuration, they give a nice differential signal.

I think I am not a technical person for replying this question, but my friend Mr Vishal Upadhyay might be right person for your problem. they manufacture thermocouple contacts and other things at their unit in Mumbai India. under name of Simplicon industries. They recently participated in an industrial fair in Houston, Texas.

Hi:

Try with this products, may help you

http://www.picotech.com/pt100_sensors.html

There are a number of measurement devices that are routinely used to provide the precision you require. These include thermocouples, platinum resistance thermometers (PRTs) and temperature-measuring crystal oscillators. Any of these can be bought with precisions that meet your requirements.

Thermocouples have rectifying properties, so they need to be very well screened. They also require a reference device to be maintained within a known temperature range, and a suitably temperature compensated measurement system (these should be available from companies like Comark). In addition, signal levels are quite small.
PRTs do suffer thermal hysteresis due to mechanical strain, but this should be less than 0.1 Deg C if you buy a suitably packaged and specified device.
Y-cut crystal oscillators make good measurement systems if you already have a reasonable reference frequency available (8-ppm error corresponds to about 0.1-deg C)

An alternative would be to calibrate thermistors and prepackage them with an eprom reference table.

Since you are just using a look up table and not keeping track of time then you are just using proportional control. However you measure the temperature I doubt you will be able to control the temperature to the degree that you stated with only a proportional control loop. Without going into the theory behind it I would suspect that you would at least need to introduce either differential or integral control on top of the proportional control and probably both.

Having look up table , add two point software or hardware calibration (offset and scale). Then

1. You can use special, more traceable thermistors offered by YSI

2. most PRT sensors after two point calibration will do the job.

3. sensors made from some types of transistors after two point calibration will do the job also (0.1deg traceability) (more cheap version, for multipoint thermometers)

You can also make sensors replacable by including hardware calibrating elements in plug.

If questions: czgut@vp.pl