Confidence level associated with Accuracy details in data sheet

[kepstein]

What is the confidence level associated with the Accuracy details in the data sheet?

Specifically, I'm interested in the 9205 Analog Input.  Data sheet here: http://sine.ni.com/ds/app/doc/p/id/ds-190/lang/en

There is a table about 2/3 through the document called "Accuracy Details", and I'd like to know what the confidence level is for the numbers in the table.  The reason is so I can use them in calculating uncertainty of my measurement.

Thanks.

[Prickly_Pear]

The specs sheet says that the absolute accurancy has a coverage factor of 3 sigma. This means that confidence level is about 99.7%. 

Jeremy P.

[kepstein]

Thanks Jeremy,

The 3 sigma is showing how to use the random noise parameter to calculate NoiseUncertainty, one component of absolute accuracy.  It doesn't explicitly say that all the parameters in the Accuracy Details table have a coverage factor of 3, but it does imply that they do since all the terms in the equation for absolute accuracy should have the same coverage factor.

I want to calculate absolute accuracy using numbers from the accuracy details table with a different TempChangeFromLastExternalCal, different NumberOfReadings, and a coverage factor of 2.  Do I multiply all the parameters from the Accuracy Details table by 2/3 in order to do the calculation with a coverage factor of 2?

[Prickly_Pear]

Okay, now I believe that I am starting to understand your question. You're trying to figure out calculations were used to create the Residual Gain Error, Temperature Coeffieciencts, etc. Is that right? This is actually a pretty tough one to answer.

I believe that the Residual Gain Error and the Offset Error have been calculated with a coverage factor of 3, although I haven't been able to find a document that specifically says it. I assume that it must be 3 because the Noise Uncertainty has a coverage of 3 and all terms in an equation ought to have the same degree of uncertainty. I also assume that these values were taken with the same number of samples (100), but I'm not sure about that. I don't know how the temperature coefficients were calculated.

I'm going to need some more time to look into this. I'll get back to you.

Jeremy P.

[kepstein]

Yes, you understand my question, and your assumptions seem reasonable to me.  Thanks for looking into it.

[Prickly_Pear · Accepted solution]

Now that I have read through quite a lot of a standard textbook for ADC characterization, I believe that the following terms in the Absolute Accuracy table were calculated experimentally with at least 100 data points and a coverage factor of at least 3:

GainTempco · 

ReferenceTempco
OffsetTempco ·

INL_Error
NoiseUncertainty

The rest of the values are either derived from these or are assumed to be values that are conservative to the estimate of the error. For example, the temperature change since the last calibration is assumed to be 70 degrees because that is the maximum supported temperature change from the temperature that the device was calibrated at.

Each term in the equation has at least a coverage factor of 3. Therefore, the Absolute Accuracy of the measurement has a coverage factor of 3.

Jeremy P.