# Help requested for temperature sensor calibration

## Main Question or Discussion Point

Hi all,

My university did not provide any metrology courses as part of my degree, and our lab experiment sessions are not very hands on, unfortunately. Thus, I am a bit out of my depth. I need to calibrate some temperature sensors as part of my mechanical engineering internship.

I have an 8 channel data acquisition unit with 8 assorted sensors attached: SMDs, RTDs, and general purpose probes. I submerged these sensors into a bath of ice water, and later a bath of boiling water in order to set baselines for zero degrees and one hundred degrees Celsius. I will be redoing the ice water bath and the boiling water bath as I am not satisfied that I did these properly the first time around. I have confirmed that all sensors are working by monitoring their real-time temperature readings as I held them between my fingers.

Shown below in Figure 1 is the data output for the ice water submersion. I will ignore the boiling water for the time being.

Figure 1: Logged temperature data from ice water bath.

This is what I think I need to do next: select a time slice near the beginning of the submersion so that I know that the water is still as close to zero degrees as possible (the time slice that I am proposing to use is shown below in Figure 2), and find the mean value of each of the temperature sensors for the example time slice.

Figure 2: Proposed time slice for determining mean values for each sensor. I feel that this time slice is too larger, however. A slice from the 200 second mark to the 300 second mark may be more appropriate?

The mean value will then be an offset that I will need to subtract from my measured data (my measured data being logged temperature measurements after the calibration has been completed). For example: if I measure 20 degrees Celsius on sensor 3 (and the mean value during calibration for the ice water was 3 degrees Celsius) then I would need to do 20 – 3 Celsius to determine the actual temperature is 17 degrees Celsius.

I am pretty uncertain of most of the things that I have proposed above, so I would greatly appreciate any input.

Best regards,

Doc

Related Mechanical Engineering News on Phys.org
Bystander
Homework Helper
Gold Member
Your sensors are being immersed simultaneously in the ice-water bath? At radically different depths? With no stirring?

Your sensors are being immersed simultaneously in the ice-water bath? At radically different depths? With no stirring?
Hi Bystander,

Yes the sensors are being immersed simultaneously in the ice-water bath. The difference in depth would be no more than 1-2 cm. That is correct: no stirring.

I have recently been advised of the importance of stirring the bath. I will be redoing the bath once I go back after the Christmas break.

Is my proposed method in error?

Regards,
Doc

Bystander
Homework Helper
Gold Member
They have to be kept in the slushy part of the bath. Clear water in the bottom of an ice bath is almost impossible to keep below 4 C even with stirring. Once you hit a stable reading, you're done. Leave them in the bath to demonstrate proper stirring and location with a stable reading over whatever time frame is comfortable.

Your sensors are being immersed simultaneously in the ice-water bath? At radically different depths? With no stirring?
Just to confirm, I take it that simultaneous immersion, with a depth difference of no more than 1-2 cm, and constant stirring is an acceptable procedure?

They have to be kept in the slushy part of the bath. Clear water in the bottom of an ice bath is almost impossible to keep below 4 C even with stirring. Once you hit a stable reading, you're done. Leave them in the bath to demonstrate proper stirring and location with a stable reading over whatever time frame is comfortable.
Okay thanks, I will do this.

Is my proposed method - finding the mean temperature offset for each sensor and then subtracting that mean offset from measured values - correct? Thank you, Bystander.

Regards,
Doc

Bystander
Homework Helper
Gold Member
Without actually digging through each specific sensor and mfr.'s data sheets, long as you're just running ice point to boiling point, it shouldn't put you any great distance out of bounds. You will find that stirring and keeping them all in the slushy part of your ice bath will reduce the offsets quite a bit.

Okay sure.
Does that mean that the rest - finding the mean temperature offset for each sensor and then subtracting that mean offset from measured values - is unnecessary?

Regards,
Doc

Bystander
Homework Helper
Gold Member
- is unnecessary?
Every sensor, even of the same type, is slightly different, and has to be calibrated. Manufacturing techniques are such that fairly uniform behavior can be taken for granted, but the calibration gives you specific constants for each individual/specific sensor.
Google "temperature sensor calibration" and you'll find a variety of notes/procedures/cautions regarding calibration procedures for various types and manufacturers.

Bandit127
Gold Member
What accuracy are you required to achieve? ±0.1°C, ±1.0° ..?

Are you required to evaluate the capability of each sensor to meet this specification?

Every sensor, even of the same type, is slightly different, and has to be calibrated. Manufacturing techniques are such that fairly uniform behavior can be taken for granted, but the calibration gives you specific constants for each individual/specific sensor.
Google "temperature sensor calibration" and you'll find a variety of notes/procedures/cautions regarding calibration procedures for various types and manufacturers.
Understood, thanks again Bystander for your time.

Regards,
Doc

What accuracy are you required to achieve? ±0.1°C, ±1.0° ..?
That hasn't actually been specified by anybody, and I suppose it is up to my partner and I. I imagine that ±1.0° is sufficient, but why do you ask, Bandit?

Are you required to evaluate the capability of each sensor to meet this specification?
No, I would think that this would be beyond the scope of our project.

Regards,
Doc

Bandit127
Gold Member
I ask because those questions will determine the amount of care you are going to need to take in order to achieve your calibration. In fact I can't imagine calibrating without a tolerance.

If you were shooting for ±0.1° I think you would have to answer my second question. Doing so might be good experience for you anyway. You would find out that not all temperature sensors are equal.

However, keeping it simple I would suggest ±2.0° should be achievable for all the sensors that you have.

I ask because those questions will determine the amount of care you are going to need to take in order to achieve your calibration. In fact I can't imagine calibrating without a tolerance.
Okay, I'll have a read about tolerancing. Thank you for bringing it up, Bandit.

If you were shooting for ±0.1° I think you would have to answer my second question. Doing so might be good experience for you anyway. You would find out that not all temperature sensors are equal.
I will do some research and try and figure out how to do this. And yes, you're right: it would be a good learning experience.

However, keeping it simple I would suggest ±2.0° should be achievable for all the sensors that you have.
Okay, I will aim for ±2.0° for what we will be measuring. However, for my own interest and at your suggestion I will try for ±0.1°.

Absolutely, I will keep adding to this thread as I go. It will be helpful for my own learning, and for anybody else who might be interested.

Thanks Bandit.

Regards,
Doc

I have an 8 channel data acquisition unit with 8 assorted sensors attached: SMDs, RTDs, and general purpose probes.
Are you confident that you have setup your data acquisition system properly. Electronic devices needs to configured by hardware jumpers or software for type of input to be applied to each of its inputs.
Then you must be confident of characteristics of your sensors, that is, you have configured you data acquisition for the type of sensor whose characteristics you are sure about. For example RTDs are generally pt100 type, but can be pt1000 or even Cu50 type.
For calibration boiling point is a bad choice because boiling depends upon atmospheric pressure. So for general purpose, one point calibration check is mostly sufficient. This usually is ice point.
Ice bath can be easily prepared. Google for it and make sure your ice bath is without "uncertainty"

As it's the easiest to do I would calibrate your RTD input first using the best accuracy resistors you can find, I would then use the best resistance meter you can get and use that to check the RTD in boiling water and a water ice mixture, having satisfied myself that that input is correct I'd use that as the reference for the other channels. Remember that these devices can be fairly non linear so if you want accurate values between zero and one hundred it might be worth renting a temperature bath.

Are you confident that you have setup your data acquisition system properly. Electronic devices needs to configured by hardware jumpers or software for type of input to be applied to each of its inputs.
Fairly confident. The software that came with the DAQ unit let's us look at the output temperature of each sensor in real time. Holding each sensor between the thumb and forefinger results in increasing temperature. They all appear to be functioning.

For calibration boiling point is a bad choice because boiling depends upon atmospheric pressure. So for general purpose, one point calibration check is mostly sufficient. This usually is ice point.
Couldn't I read the atmospheric pressure off of a barometer and then interpolate to find the temperature from a saturated water table? I am guessing that the highest temperature we will record using any of the sensors will be no greater than 35° as this number is closer to 0° than it is to 100°. Would this indicate that it is not necessary to do the boiling water calibration and to just stick with the ice-water bath as you suggest?

Ice bath can be easily prepared. Google for it and make sure your ice bath is without "uncertainty"
Thanks Muti.

Regards,
Doc

As it's the easiest to do I would calibrate your RTD input first using the best accuracy resistors you can find, I would then use the best resistance meter you can get and use that to check the RTD in boiling water and a water ice mixture, having satisfied myself that that input is correct I'd use that as the reference for the other channels. Remember that these devices can be fairly non linear so if you want accurate values between zero and one hundred it might be worth renting a temperature bath
I don't think that the place I am interning at will go to these lengths to be honest. What is your opinion of my proposed method about finding the mean temperature offset and then subtracting that value from the actual measured temperature? Nobody has commented on that, and I have no idea whether or not I am on the right track with that part.

Thanks Jobrag.

Regards,
Doc

Bandit127
Gold Member
Assuming that you have the DAQ correctly configured for each sensor type, then simply adding or subtracting the mean offset for each is correct.

This thread contains some good observations about how much thought should be given to the accurate determination of that offset. It is good to see that you have evaluated those ideas and made rational choices about which ones you will incorporate in to your calibration.

I don't think that the place I am interning at will go to these lengths to be honest. What is your opinion of my proposed method about finding the mean temperature offset and then subtracting that value from the actual measured temperature? Nobody has commented on that, and I have no idea whether or not I am on the right track with that part.

Thanks Jobrag.

Regards,
Doc
Until you have done the 100 Deg calibration you won't no whether the error is the same through the range if you find that the error is plus three deg at zero but minus four at one hundred, you're going to have to apply something more sophisticated. What temperatures are you trying to measure and in what application.

"I have an 8 channel data acquisition unit with 8 assorted sensors attached: SMDs, RTDs, and general purpose probes." Tell the model number and manufacturer of this device.
Give names of all "sensors" attached to each input.
If possible a picture/ schemetic drawing of how all these "sensor" have been connected to the device. Did you connected the "sensors" to electronic device of it was already wired.

Assuming that you have the DAQ correctly configured for each sensor type, then simply adding or subtracting the mean offset for each is correct.
Okay, thanks a lot, Bandit.

This thread contains some good observations about how much thought should be given to the accurate determination of that offset. It is good to see that you have evaluated those ideas and made rational choices about which ones you will incorporate in to your calibration.

Regards,
Doc

Until you have done the 100 Deg calibration you won't no whether the error is the same through the range if you find that the error is plus three deg at zero but minus four at one hundred, you're going to have to apply something more sophisticated.
I will do the 100 Deg calibration when I return to work. I was planning on checking a range of temperatures with a thermometer to double check that the temperature sensor readings are accurate. If the errors are significant, then I will look into the more sophisticated something.

What temperatures are you trying to measure and in what application.
I am assuming that we will be measuring temperatures in the range of 5° to 40°. I am measuring the surface temperature of optical instrumentation at an astronomical observatory, and how those surface temperatures relate to ambient air temperature inside and outside the telescope dome.

Thanks for the continuing input, Jobrag.

Regards,
Doc

Tell the model number and manufacturer of this device.
The manufacturer is Omega. Model number is OM-USB-5203. Here is a link to the device: http://www.omega.com/pptst/OM-USB-TEMP.html

Give names of all "sensors" attached to each input.
Done, shown below.

The yellow numbered stickers correspond to the channel number that the particular sensor is connected to. The tape is used for cable strain relief. The picture is not very good, I don't have a better one, sorry.

If possible a picture/ schemetic drawing of how all these "sensor" have been connected to the device. Did you connected the "sensors" to electronic device of it was already wired.
Shown below is an image of our ice-bath calibration - I need to redo the ice-bath however.

Unfortunately, we did not know to stir the solution. I am also assuming that it might be beneficial to wrap the jug in aluminium foil to reduce heat transfer.

I will upload a proper connection schematic once I get back to work. I connected the sensors; they were not pre-wired.

Thanks Muti.

Regards,
Doc

Very nice, but I want to remove all doubts about your hardware and software configuration. It is written on page 06 of manual that, "
With the OM-USB-5203, you can take measurements from four sensor categories:
 Thermocouple – types J, K, R, S, T, N, E, and B
 Resistance temperature detectors (RTDs) – 2, 3, or 4-wire measurements of 100 _ platinum RTDs
 Thermistors – 2, 3, or 4-wire measurements
 Semiconductor temperature sensors – LM36 or equivalent"
Now can you tell on which channel which sensor from above has been connected.
If you can not determine, then if this "setup" was provided to you pre-wired, then by back tracing wiring as illustrated from page 15 to 18 of manual you may be able to determine type of sensor connected to each channel.

I would be inclined to insulate the jar with bubble wrap and to get the sensors to the middle rather than the outside. I've a little electric milk frother (it looks a bit like a very small outboard motor) something like that would make a good circulating device.