Continue to Site

# Calibration of temperature Sensors

Status
Not open for further replies.

#### urnuj

##### Junior Member level 3
Hi

I am working with 2 different temperature sensors. One of them is a themistor and the other is a RTD. I put them in the same place and the temperature that they both measured differed in 0.1degrees celcius which is good enough I think.

However, when the temperature rises to 50degrees then to 80 degrees, the difference in the temperature measures widens and it came to be around 10 degrees of difference.

How do I make the 2 sensors be more accurate and stable and have a matched reading?

Regards

What kind of controller you used?

I am using a pic33ep512MU814.

Regards

You can try electrically matching them with non-linear amplifiers but by far the easiest way is to do it in software. Try plotting their measurements across the entire temperature range you want them to work over. Then see if a mathematical formula can be used to match the values. For example something like:
"if(sensor2 > 50) Sensor2 *= 1.1;"
If that isn't possible, you could consider a look-up table to convert actual to expected values.

Brian.

Points: 2

### nop90

Points: 2
Hi,

id consult the datasheet. Is the difference within the tolerance?

If not, change the defective sensor.

Klaus

Given detail is not enough for checking about defective sensor.
You should to match the values of sensors output values to their datasheet.

Hi

I am working with 2 different temperature sensors. One of them is a themistor and the other is a RTD. I put them in the same place and the temperature that they both measured differed in 0.1degrees celcius which is good enough I think.

However, when the temperature rises to 50degrees then to 80 degrees, the difference in the temperature measures widens and it came to be around 10 degrees of difference.

How do I make the 2 sensors be more accurate and stable and have a matched reading?

Regards

1st: please type exact name of temperature sensors!

The 2 sensors are HEL-705-U-1-12-00 and 192-103LET-A01, RTD and thermistor.

Right now I have not gone over a temperature of 100degrees celcius so it should not be out of the tolerance.

gain errors may be your passive tolerance stackup and Vref error.

you need a good reference like ice water and boiling water or better

Hi,

Code:
Right now I have not gone over a temperature of 100degrees celcius so it should not be out of the tolerance.`

I don't see what the absolute temperature has to do with tolerance of the sensors...

Klaus

Ok. So I have to use a good reference for the calibration to offset the variation from the accurate temperature right?

This could be a little difficult as the sensors are currently being installed into the structure.

I would have to see what can be done then.

Regards

10K temperature measurement error with sensors that can achieve better than 1 K accuracy indicates either an inapproriate measurement circuit or wrong sensor characteristic used in your software. We don't know anything about these points.

I'm presuming so far that the sensors are actually exposed to the same temperature. Although temperature measurement sounds like a simple problem, it's not impossible to achieve 10 K error if you ignore the heat conduction effect of sensor wires.

Hi,

if they are installed in the structure, how do you know they have the same temperature?

If the sensore can´t be exposed to dedicated temperature, then maybe use identical sensors to calibrate yourhardware.
(But this can´t calibrate sensor deviations ... but maybe worth a try)

Klaus

Do you mean the loss of heat through the wires that can result in the inaccuracy in the reading?

Yes the sensors are actually planted to measure the temperature of a heat source and they should measure the same temperature or very close. The difference in temperature should only be due to the slight variation in positioning of the sensors.

Regards

Do you mean the loss of heat through the wires that can result in the inaccuracy in the reading?
Surely. There shouldn't be a problem with the Pt1000 packaged in ceramic cylinder, which achieves good heat transfer if it's placed in an exactly drilled hole in a metal block.

But the small NTC has no "natural" heat transfer interface. It should be embedded in thermal grease or another heat conductive compound to achieve similar heat transfer. Or the lead wires must be fixed to the measured object to achieve thermal equilibration.

I see. That is to say that pt1000 should give a more accurate reading of the heat source right?

It's easier to achieve less temperature error with the the Pt1000, if the measured object is something like a metal block. There are however thermistors with better suited interface for this situation.

Hi,

if you measure high temperauture (about 100 °C) then i´d say the sensor that shows the higher current has better thermal conductivity.
Does this meet your measurement values?

Klaus

At a higher temperature, wouldnt a RTD be having a higher resistance and that would mean a smaller current?

Status
Not open for further replies.