How to amplify thermocouple readings

[LesHammer]

> NI has a therocouple input on some of their DAQ products I think,

> but they use a temperature measurement at the terminal board

> to "fake" a cold junction (read: this is sleezy).

There's nothing "fake" or sleezy about it.  Cold doesn't have to be 0 degrees C.

As long as you know what the cold temperature is you can compute the

thermocouple voltages.  Agilent Technologies does it this way too.

Unless you want to carry an ice bath with you, it's the way to go.

Les.Hammer@SmartSensorSystems.com

[Ryan_V[DE]]

Sheldon,

We do use a precision temperature sensor to "fake" cold junction on our terminal blocks, but it does significatly increase the accuracy. Additionaly, there would be no way to support multiple thermocouple types without doing this. You always have the option of implementing your own cold junction compensation dependent upon the thermocouple composition.

Regards,

[Sheldon_Stokes]

Sorry if I offended anyone with the "fake" and "sleazy" editorial comments regarding the thermocouple inputs on the NI DAQ products. But it is not exactly a robust or highly accurate method of handling thermocouple signals. There's an assumption that the terminals are at the same actual temperature as the junction, which is probably reasonable. Also, your thermocouples aren't isolated like a dedicated pre-amp typically is. For example, take a look at the AD part for thermocouple signal conditioning:

http://www.analog.com/en/prod/0,2877,7B27,00.html

They also use an external thermistor as part of the signal processing. but the output is isolated from the input, and is already low pass filtered to allow slow data collection speeds without aliasing.

But the cold junction compensation scheme is really not an issue, because if anybody is interested in highly accurate or precise temperature measurements, they shouldn't be using thermocouples in the first place.


Sheldon

[UCLA-engineer]

I am attempting to read thermocouple values from a line that is physically located next to a 500kHz line.  Unfortunately, there is no way to physically separate the two, although each line is electrically isolated.  Due to their close proximity I read erronious temperatures when the 500kHz line is energized.  I have read the NI threads regarding signal conditioning and have not been able to filter the noise (which makes one TC think the temperature dropped and the other TC think the temperature rose).  I am using a SCXI-1102 with a SCXI-1303 32-Channel Isothermal Terminal Block to read the temperatures from the two TCs.  Once again, the TCs read normally when the 500kHz signal is not on but once it turns on the TCs go crazy.  Is there anything I can do to fix this problem?

[LesHammer]

> I am attempting to read thermocouple values from a line that is physically located next to a 500kHz line. 

Meaning that opposing peaks are 1 micro second apart.  (or odd multiples of 1us)

> I am using a SCXI-1102

333kSamples/Second => 3 microseconds per reading.

>  Is there anything I can do to fix this problem?

Take 2 readings, 3 uSeconds apart and average. 

Les Hamer

[angelo]

>... if anybody is interested in highly accurate or precise temperature measurements,
> they shouldn't be using thermocouples in the first place.


Of course that depends on what, exactly, is meant by "highly accurate". Thermocouples are more than "accurate enough" for the vast majority of applications. Also, thermocouples are extremely reliable (robust).

I just felt like I needed to defend the lowly thermocouple!

[Ryan_V[DE]]

UCLA-engineer,

As Les recommended, if analog filtering does not help, I would suggest oversampling and averaging. You may want to even acquire a large number of samples and average them, rather than only 2 samples.

Hope this helps,

[UCLA-engineer]

I will try the oversampling method.  I will be using the LabVIEW Express VI DAQ Assistant, should I just tell it to sample X samples at 333kHz?  I am not sure how accurate the timing is.  Also, I did previously try to use "continuous" as my sampling method using the DAQ Assistant and that didn't reveal noise when I graphed the output.

[Ryan_V[DE]]

UCLA-engineer,

As long as you select a finite acquisition, that will acquire the number of samples you specify with very accurate timing. I'm not sure why a continuous acquisition did not reveal the noise. On second thought, how were you able to observe the noise if not using a continuous acquisition?

Regards,

[UCLA-engineer]

The "noise" I was reading was visible in the following way.  With the 500kHz source powered down the TCs would read similar temperatures, but when the 500kHz source was powered up (to be used as a heating element) one TC would increase in temperature and the other would decrease.  The increase in temperature would often be far higher than is realistically possible, and obviously there is no way the temperature could drop upon application of power.