Temperature Reference / Calibration

Hi,

Have several temperature related circuits but like all thermometers you always find variations between different units.

Have been looking for a reference unit for 0 to 50c, so I can check out the whole range and not just 25c.
Most sensors like the LM35, SHT15 while appearing to have a close tollerance of +-0.5c it is actually only at 25c and over the range mentioned this can increase to +- 1.0c or more.

The DS18B20 seems to offer +-0.5C over its full range but wondered if there is a better sensor around, though looking at the pages of sensors in the Farnells catalogue it seems the best...?

thanks.

±0.5C accuracy is probably about as good as you can get without calibration of the individual sensors.

For high accuracy temperature measurement, platinum sensors are used but they are expensive and require accurate circuits to realize their high accuracy.

Do you have a particular requirement for high temperature accuracy?

Hi,

Do you have a particular requirement for high temperature accuracy?

Click to expand...

Its mainly for checking the temperature curve of existing Pic controlled ntc devices over the 0 to 50c range and for checking the actual accuracy of other existing sensors, used mainly in fishkeeping and plant growth rather than the lab enviroment.

Am aware of the PT sensors but have no practical experience of building with them or the amp needed - just been checking some of the specs on them and they do seem very close eg to within 0.15c - however not sure how easy it would be to build the amp and adc input to the Pic to ensure the same level of quality / accuracy ?

If you have any practical advice / circuits or links for the PT100 they would be much appreciated.

Typically a high accuracy current source generates a current through the PT sensor resistance and the resulting voltage is measured with a high accuracy digital voltmeter or ADC to determine the PT temperature. It order to get full accuracy of the PT sensor you would need to calibrate the constant current periodically using a high accuracy reference. Do you have access to such calibration equipment? If not you can calibrate with a high accuracy shunt resistance and an accurate digital voltmeter.

An accurate constant current can be generated by a Howland op amp circuit. All resistors should be stable 1% resistor types. The control voltage would be from a stable voltage reference circuit. The voltage should be made adjustable with a pot to allow for calibration.

The current through the PT sensor must be low enough to avoid any significant self-heating of the sensor to minimize error. The sensor data sheet usually gives info on this heating versus current.

Hi Carl,


Thats all very interesting and I will look into them in more depth - although I don't have such calibration equipment at home some guys at work do have certified Flukes and the like so they will surely help.

thanks again.

OK. My suggestions were somewhat general, but If you need more specific info on a particular approach, just reply in this thread and someone should be able to help you further.

Wp100:
Do you need relative or absolute accuracy? Or both?
Alternately you could go to temp to frequency sensors. Some that are available are quite accurate, but expensive.
Many years ago I had a program written to allow for the variations of LM35s against a "golden" unit. We would bath a bunch of LMs -including the golden one, mounted on a heatsink, in motor oil on the slowly heating-up stove (your mate will love you for that!), measuring output voltage, than mary the LM to the variables in the program. With the Motorola 705's we did get 0.5% accuracy (less that damned +/- 1 bit) from -20 to over 100 Deg. C
Hope this helps, E

ps: Robert Pease: where are you when we need you?!

If I can make a small suggestion, not knowing your budget restraints. I have been using these little units with a variety of thermocouples and PRTs. We buy in quantity of about 20 to 30 at a time but a single unit in the US runs about $100 USD.

I like them because they are fully programmable for input as well as range. The units we use are made by Inor in Sweden but I have seen cousins of them all over including a variety marketed by Omega Engineering who also sell many PRTs as well.

The beauty of using a temperature transmitter with a PRT (or thermocouple) is the current reference work and amplification is done for you. Using about a 24 VDC supply you get a nice 4 to 20 mA output scaled for your chosen temperature range. Run that current loop through a 250 or 500 Ohm precision resistor and you have 1 to 5 volts or 2 to 10 Volts in a nice linear output as the linearization curves are done for you.

Something to think about anyway...

Ron