Thermocouple accuracy near room temperature

Most thermocouples appear to have a very wide temperature range, this seems less than ideal when attempting to measure accurately in the 20-50°C range.

Does anyone have any insight into thermocouples that would be accurate to a fraction of a degree near room temperature?

Any comments on accuracy of the various thermocouples out there? (cheap ebay vs. Fluke or Fisher Scientific)

You might want to look at a different sensor if you're only concerned with room temp. Maybe an RTD.

It is not the limitation of the thermocouple, we need to focus on the measurement of small potentials more accurately.

You can increase the sensitivity arbitrarily by putting 10 or 100 thermocouples in series. You of course lose on the response time.

Have to agree with Barry here.
There are many limitations to accuracy that include the accuracy of cold junction correction, plus the purity of the metal alloys in the thermocouple, plus the accuracy and calibration of the electronic amplifier.

Thermocouples have a great many advantages, but absolute accuracy is not one of them.

Your best bet might be a platinum RTD. The platinum is pretty pure, and very well behaved so it gives a good reliable temperature reading.

But even that, will take great care to get anywhere near the type of accuracy you are after. Problems of thermal gradients and Seebeck effect can lead to errors. Its not at all an easy thing to do.

0.1K resolution (corresponding to 4 µV with popular type K TC) can be achieved with precision amplifiers. 0.1K accurate cold junction compensation can be a problem but isn't impossible. Temperature differences can be measured down to 0.01 - 0.02 K.

Hi,

A thermocouple measurement is a "relative" measurement method.
It measures the temperature difference between the thermocouple "sensor point" and the "cold junction point".
To get an absolute temperature reading you need a sensor for "absolute temperature" at the cold junction point.

Example:
If you have a handheld thermocouple type "K" (about 41uV/°C near room temperature) measurement device. Let's say the temperature of the device is 25°C.
And you want to measure the temperature of an object with 26°C...
Then the voltage at cold junction is about 25°C x 41uV = 1025 uV (you can't measure this voltage)
The voltage at the sensor point is 26°C x 41uV = 1066uV ( you can't measure this also)
You measure the difference: 1066uV - 1025uV = 41uV.

Now you could say "why not simply take the difference of the temperatures... and multiply them with 41uV?
My answer is: Because your concern is "precision" and the 41uV are only true near room temperature. If you want to measure very high or very low temperature you need to take care about this. You need to use different factors for cold junction and measurement point.

Back to room temperatures:
You urgently need type K wires if you want to extend the sensor wires. Exactely to the cold junction point.
For high precision around room temperature you need a cold junction block, maybe aluminum, around the cold junction points and the absolute temperature sensor.
Then you need an amplifer with very low offset voltage (chopper stabilized) and a good layout to be sure not to introduce additiinal offset errors.
I'd say the practical limit will be around 0.1°C temperature difference.
And for sure you need an absolute temperature sensor with the same precision: 0.1°C.
The key to succes is a good schematic, (noise shielding), high quality thermocouple sensors, high quality extension wires, a good pcb layout and good amplifiers...a lot of things to take care.

Klaus

- - - Updated - - -

Added:

You can increase the sensitivity arbitrarily by putting 10 or 100 thermocouples in series. You of course lose on the response time.

Click to expand...

You can not simply connect thermocouples in series: you need to wire each thermocouple back to the cold junction point and there connect one thermocouple in series.

Klaus

KlausST said:

You can not simply connect thermocouples in series: you need to wire each thermocouple back to the cold junction point and there connect one thermocouple in series.

Click to expand...

Sorry I forgot to mention that. Yes, you need to insulate all the welds, bundle the cold (and the hot) junctions together (cold and hot junctions separately) and and make sure that the cold junctions are all at the same temperature (put them on a decent heatsink).

Temperature, I know it sounds funny, is somewhat like the potential (yes, temp is nothing but a thermal potential) - only differences are significant and absolute temperature exists only in abstract thermodynamics. A Beckmann thermometer easily measures a millidegree without a fuss.

The big advantage of the thermocouples is that they are usually very silent (not noisy) because they have usually rather low resistance.