thermocouple noise

[jdsnyder]

I am using a PCI-6221 board (16 bit) with a cable connection to a CB-37F-LP, connected to T-type thermocouple (slope = 52E-6 V/deg C)

Performing the math, my Temperature resolution should be:

voltage resolution = 2V/2^16 = 3.05E-5V

Temperature res = voltage resolution/slope = 0.59C

My temperatures are jumping +/- 7C. 

I obtained the same results with multiple thermocouples and multiple board positions. 

I also obtained the same results when I wired the thermocouple directly to the PCI-6221

Any ideas?

[jdsnyder]

An additional note (all thermocouples - type T):

T1 in ai0, T2 in ai1

T1 +/- 7C

T2 +/- 3C

switch thermocouples:

T2 in ai0, T1 in ai1

T2 +/-7C

T1 +/-3C

Thermocouple attached to USB 6008 device (12 bits, 1/16th the theoretical resolution)

no calibration through MAX

T3 +/- 10C

[Dennis_Knutson]

What kind of compensation (if any) are you doing? How are you doing the scaling to convert from voltage to temperature? A tutorial on using thermocouples can be found here.

[jdsnyder]

I am using a constant value for compensation. I believe this would affect the offset, but not the noise.

I calibrated on MAX from 21C to 70 C.

New information/clues:

I placed the board on a second computer - better but not perfect results:

T1 positioned in ai0 . When placed in a 40C bath, temperature reading varied +/- 1.5C

T2 positioned in ai1. When placed in a 40C bath, temperature reading varied +/- 0.5C  (this is what I was expecting).

Although I'm not an expert, I have been using thermocouples for many years.  My thermocouples are sheathed in stainless steel probes,

the thermocouples are not grounded to the probe - the wire does not toch the probe.

Until last week, I was using Camile (originally developed in-house by Dow Chemical Company) for data acquisition

- I did not experience any problems with any of the thermocouples.  Signals from all experiments was sent to a Camile box first, then

digitized data was routed to individual lab computers.

By the way, thank you very much for looking into this.

Jon

[jschwartz]

Hi Jon,

From your description, it sounds like you are seeing external noise from measuring single ended, meaning that you are measuring between your analog input and ground.  I would suggest changing your wiring from single ended to differential.  This will eliminate external noise from your measurement.  You can find a our Field Wiring Diagram here. Table 1 is located about halfway down the page and is helpful for configuring your measurement for differential. To wire channel 0 differentially, you will wire your positive lead to AI 0 (pin 68) and your negative lead to AI 8 (pin 34) with the corresponding bias resistors shown in the Field Wiring Diagram.  To find a corresponding differential channel on the 6221, you can add 8 to the channel you are on, i.e., AI 0 and AI 8 are a pair and AI 1 and AI 9 and so on.  I would also try decreasing the voltage range of your measurement.

Regards,

Jim Schwartz

Applications Engineer

National Instruments

www.ni.com/ask

[jdsnyder]

Hello Jim,

I previously had the circuit hooked up as a differential measurement, but without bias resisters to ground

(pin 3 according to the pinout).

I looked at the figures and agree with the reasoning for adding bias resisters.

Unfortunately, after adding the resisters, I am getting the same noise as before.

See the attached photos

[jschwartz]

Hey Jon,

From the pictures, the fluctuations you are seeing look very periodic - it looks to be around 200 Hz. You can get rid of this by adding a low pass filter in software.  In LabVIEW, you can right click on your block diagram then select Signal Processing >> Filters or you could add an express VI filter lby right clicking the block diagram then selecting Express >> Signal Analysis >> Filter.

[jdsnyder]

Yes, there is some periodicity in the noise, but the frequency is higher than indicated by the graph which used a low sampling rate.  I also believe that I am observing a "wagon wheel affect".  When I sample at 1K Hz, and graph for 0.1 seconds, I generally observe between 4 to10 peaks for that time-frame.

Yes, I can use a filter for some of my applications: those where I don't require rapid signal response.  However, some of my applications require immediate response, plus I would like to solve the root cause of my problem.  The specs for the board indicate I should expext 13 micro-volts rms of noise.  1 degree C of noise is roughly 100 micro-volts.

See the attached photos for two different channels.  Two of my channels have high noise, and two channels have

very high noise (I haven't looked at the other four channels without a filter in place).

Thanks,

Jon Snyder

[jschwartz]

Hi Jon,

I would like to see what your voltage information looks like, i.e., what kind of noise you are seeing on your lines with the raw data.  If we see that the voltage is fluctuating this wildly, then we can narrow down the problem a little more.

[jdsnyder]

Yes, the noise is in the original voltage signal. 

See the attached Word document.

Thanks,
Jon Snyder