NI-9211 Measurement Accuracy

[MarkMoss]

Figures 5 through 9 of the operating instructions for the NI-9211 4-Channel Thermocouple Input Module contains several charts that give typical and maximum measurement error under various conditions and for various types of thermocouples.  These charts cover a very wide range of temperatures.  For example, the Type K chart covers -200 °C to +1400 °C with a division only every 200°C.  This makes it difficult to user these charts to determine the maximum measurement error at a specific temperature.

Is there a forumla or table of values I can use to determine the maximum measurement error for a smaller range of temperatures?  Specifically, I'm interested in determining the maximum measurement error for measurements between -40 °C and 125 °C with a cold-junction temperature of 23 °C ± 5 °C for measurements type J, K, and T thermocouples.  (Not including errors due to the accuracy of the thermocouple itself.)

Thanks,

Mark Moss

Electrical Validation Engineer

GHSP

[J_Thomas]

Hi MarkMoss,

I just wanted to let you know that I have read your post and I am currently trying to find out more information about the data you have requested. I will try to reply back as soon as I found out anything that will be helpful.  Hopefully I will have something for you soon.

[J_Thomas]

Hi MarkMoss,

First of all, let me apologize for the delay in posting back.  Things got really busy these past few weeks with NIWeek 2008. 

The error shown in the figures for the 9211 was derived using a very tightly controlled process in R&D that takes into account both voltage errors and CJC sensor errors.

You could get a rough estimate using the Gain, offset and CJC temperatures.  The math would be a bit challenging, but doable.  Basically you could calculate the voltages for the various temperatures for the specified thermocouples. You would then calculate the gain and error in terms of voltage. You then will need to convert this value to a change in temperature in that range and add the CJC error. This will give you a rough estimate, but will not include the thermocouple error. The graphs in the manual include this thermocouple error.

The NIST defines a voltage vs. temperature curve for each thermocouple type at their website. This information can be found at this link: http://srdata.nist.gov/its90/main/.

Another way to get an estimate would be to get as many data points on the graphs in the manual as you can reasonably estimate and plot them into an Excel Spreadsheet.  Plot them on a graph and then do a curve fit between the data points, essentially recreating the figures.  For most of the thermocouples the error profile is a pretty smooth curve so this method would not introduce to much additional uncertainty into the measurement. Any error estimation is subject to uncertainty, and so even an extremely precise table of data points for the error may not represent the error in your actual test system any better than the approximations you can get from these figures.

[oilyfingers]

Dear Mark et al

I have a similar question but am happy to try and get a rough error estimate from the graphs. However I'm having trouble reading the graphs. Is your interpretation of the 'error' value on the y-axis that it means that the error is +/- E (where E is the value read off the graph), or that 2E is the value read off the graph? (ie. if the graph says 1 K, the uncertainty band is +/- 0.5 K)

Dave

[oilyfingers]

Sorry - just another thing, Jared said "This will give you a rough estimate, but will not include the thermocouple error. The graphs in the manual include this thermocouple error."

This is not true - the 9211 manual explicitly says "The figures account for gain errors, offset errors, differential and integral nonlinearity, quantization errors, noise errors, and
isothermal errors. The figures do not account for the accuracy of the hermocouple itself." (page 15 of http://www.ni.com/pdf/manuals/371566e.pdf)

Dave

[Seth_B.]

Dave,

You are correct that these charts do not include the thermocouple error.  That error is dependent on how well the thermocouple was manufactured.  The error given in the chart is ± E.  Thus, an error of 2K means that you are accurate to ± 2K.