DMM 4071 accuracy @ 4.5 digits

[JLAnderson]

The NI 4071 DMM DC accuracy is only specified @ 7.5 digits.  The Navy makes measurements @ 4.5 digits so I need to know that accuracy in order to calculate test accuracy ratios (TAR).  I also need the same information for ohms.

John Anderson

[KatieC]

John, 

Unfortunately, there isn't an easy answer as it depends on your voltage range, when it was last calibrated, and what the temperature difference is since last calibration. I will walk you through an example...

You are going to want to start off by determining the aperture time for your 4071. You can find the default aperture times by looking at the DMM Help (NI Digital Multimeters Help >> Devices >> NI 4071 >> DMM Measurements >> DMM Measurement Cycle >> Aperture Time).

We can see that for 4 1/2 digits on the NI 4071 the aperture time is 50us.

Next, we will need to add in additional noise in error by looking at the graph on page 3 of the specifications and manual.  In order to relate the aperture time to the graph, we will need to convert it into Power Line Cycles (PLC). A quick search of the NI DMM Help yields a conversion (Selecting Aperture Time for DC Measurements): 1 PLC equals 16.67 ms for 60 Hz powerline or 20 ms for 50 Hz

We will assume it is a 60Hz powerline, so 1 PLC = 16.67ms. Converting 50us to PLC = 0.003.

We now look at the graph on page three:

Notice that both scales are logarithmic. 0.003PLC is less than 1ppm of the range, but for simplicity I am going to use 1. 

We now have 1ppm of the range, but we need to find the correct multiplier. The table is below the graph on page 3:

Let's assume we are interested in the 100V range. Our multiplier is 6, so we now have 6ppm of the range. We are also interested in accounting for the peak-to-peak noise error, so we will multiply the rms noise by 6 to get 36ppm of the range.

Lastly, if you view the accuracy table for DC voltage on page 2, find the appropriate coefficients for your application:

For this example, let's say the device was calibrated between 90 days and 2 years ago and is within 1C from the temperature at which the external calibration occurred at. Our coefficients would be 18 + 2. We will now add the 36ppm of the range to the 2ppm of range to get coefficients of 38ppm of range. Our equation now reads: DC Voltage +/- (18ppm of reading + 38ppm of range). Our range is 100V, so you would need to add in what your reading. 

I know this was complicated! Unfortunately, we only provide accuracy specifications for 7 1/2 digits, so we have to calculate for other modes. 

Katie

[KatieC]

John, 

There are two errors in my above post. I said:

"Our multiplier is 6, so we now have 6ppm of the range. We are also interested in accounting for the peak-to-peak noise error, so we will multiply the rms noise by 6 to get 36ppm of the range."

This is only partly correct. That is for pk-pk, but since we do accuracy as + or - we only need half. It would then be 36ppm/2 = 18ppm. 

The other error: I was a column off on the accuracy chart. It would actually be 20 + 2 for our coefficients for the 100V range, 2 year calibration, and within 1C:

Therefore, we add the 18ppm of the range to the 2ppm of the range specified for a total of 20ppm of range. The final equation would be DC Voltage +/- (20ppm of reading +/- 20ppm of range). 

My apologies!

Katie

[JLAnderson]

A few more questions:

1. Is there a formula available for the Additional Noise Curve?

2. The 7.5 digit entry in the Aperture table states 4 samples are averaged.  Does using other averaging sizes affect the accuracy in the table?

3. I assume the same method applies to ohms?

John Anderson

[KatieC]

John, 

1. The graph is based on test results, so there is no formula available. 

2. In order to get 7 1/2 digits, you have to average 4 samples. If you notice, for 6 1/2 digits, it has the same aperture time without the requirement for averaging. The other digits of resolution assume only 1 sample is taken. If you do averaging, it will improve your accuracy but we do not spec by how much. 

3. Correct!

Katie

[JLAnderson]

Spoiler

I have some follow-up questions:

1. The aperture table does not include ohms.  Should I use DCV entries?

2. The ohms RMS multiplier table has no entry for 30 M ohms

3. There is no 3.5 (or 3) digit aperture entries in the default table.  What should it be?

4. The NI GUI supports 3.5 digits.  Does it use 12 bits (3.6 digits) instead of 10 bits (3.0 digits)?

5. For the 7.5 digit range DCV, does the instrument default to 4 samples averaging?

6. What is the rational for using Peak noise instead of RMS?  For a sine wave it is 1.414 times instead of 3 times.

7. Why doesn't NI publish tables of aperture vs. RMS noise instead of using a chart?  Extrapolating for 0.003 PLC is difficult.

John Anderson 

[Izzzzz]

Hello John,

1. Yes since a resistance measurment requires a voltage measrument. I will double check with some colleagues to ensure this is correct.
2. 30MOhm is not listed in the Help documentation or accessible in the test panel for this device. Although I am as perplexed as you are in seeing it appear in this chart below. I do not think this is a valid range for the 4071, but this is something I will need to do more research on.
3. When the Sampling rate is at 3 or 31/2 digits then the DMM is in digitizer mode and sampling at 1.8 MS/ seconds. The minimum aperture time for this method of acquisition is 8.89 seconds and maximum is 149s. The aperture time you select is based on the number of samples. More information about this can be found in the DMM Help (NI Digital Multimeters Help >> Devices >> NI 4071 >> Waveform Acquisitions >> Sample Rate >> Aperture Time).
4. Could you point me towards the NI GUI? I am not familiar with this name so I cannot provide an answer.
5. Yes
6. To my understanding the Peak to Peak noise is a better representation of the uncertainty in the measurement that a user could experience due to its overestimation. That I believe is the rational for using this method.
7. PLC and Aperture time are related. You could select to conduct your acquisitions based upon PLC so that you could better utilize the graph presented. As Katie previously mentioned the graph is based on experimental data and a table would yield the same information.

Image Referred to in Answer 2 

Regards,

Izzy O.

Applications Engineer

National Instruments 

[JLAnderson]

Note: I made another post on the incorrect aperture times elsewhere also addressing some of these issues.

1. I confirmed the aperture times by writing a program to read them.  DCV and ohms are the same.

2. We both should have read note 11.  There is no 30M range but it only applies to 2-wire ohms below 30M on 100M range.

3. The Test Panel (GUI) does not use digitizing for 3.5 digits  It uses the standard call to make one measurement at a time.  I determined the aperture time to be 20us for both 3.5 and 4.5 digit readings

4. The standard Test Panel that come with installing the software package.

5. I confirmed that in the help documentation but have not been able to measure it yet.  I always get -1 or 1.

6. There is a significant difference in the accuracy for peak (x3) and peak-peak (x6).  Since p-p is two sided and peak one sided, one sided should be used to match one side of a +/- accuracy tolerance.  Katie stated it should be peak and I agree.  But I would still like to see formal detailed explanation.

7. I am dealing with contractor written software and NI Test Panel that I cannot change.  I have attached a spreadsheet to match the RMS noise graph with extrapolation below 0.01 NPLC

John Anderson

[Brandon_G.]

Hey John,
Let me chime in on some of the points you and Izzy have been discussing...

5.  A value of -1 indicates that the aperture time is configured automatically based on the number of digits selected. As a reassurance that the number of averages is configured appropriately, you can call NI-DMM Get Measurement Period.vi and you will see a direct increase in Measurement time when you change the number of averages.

6.  Attached are graphs from an experiment I ran today at 3 1/2 digits in the 10 V range with a shorting bar on the input,  From what you can see from the noise in the measurement, the RMS value multiplied by 3 gives us a good peak noise estimate and x6 gives us a good pk-pk noise. The x3 gives us a confidence interval of 3 sigma or 99.73%.  If using the RMS noise in our accuracy spec, our measurements would only be 1 sigma wide or 68.2% confidence interval.  Therefore a multiplier of 3 is more appropriate for pk noise. 

7. The graph you've provided looks to be a good conservative estimate based on the measurement I've performed above.  At 3 1/2 digits we're at 20 us aperture which is ~.001 PLCs @ 60Hz.  From your graph, the noise looks like 30 ppm RMS (or 300 uV in the 10 V range) at .001 PLC, whereas in my measurement, I'm seeing only 103 uV RMS.  Perhaps you can use the attached VI in the measurement function/range you will use and see how the noise performs with your particular settings?

Thanks!

Brandon G

[JLAnderson]

Brandon,

Thanks for the info but I don't have LabVIEW.  I have been using VS 2010 C# with the C library.  Maybe you can convert it to a pdf for me.  The 3.5 digit PLC is estimated at 0.0012 or 23.2 ppm, not 30.  That data is extrapolated so it could be off by a factor of 10 since it is a log curve.  I initially set the 0.001 PLC to 20 ppm and the curve looked fine but looked better at 30 ppm.  Also our particular hardware has a very noisy PC power supply that another engineer discussed with an NI engineer about a year ago.  Thanks for the input.

John Anderson