Verifying Calibration Automatically - QuantAsylum/QA40x GitHub Wiki

With release 1.221 of the software, the ability to automatically verify calibration has been provided. This will require a DVM with 6.5 digits of accuracy, and in order for the test results to make sense, you will need a recent calibration certificate on your DVM.

Understanding DVM Accuracy

Benchtop DVMs have specifications that specify accuracy as a percentage of the reading and the % of the range. In the plot below, you can see the AC characteristics for the Rigol DM3068. Note that for a 1 kHz tone in the 2V range, we can see the error is specified as "0.06 + 0.03" for measurements made within 1 year of calibration and measurements made within +/-5C of the calibration temperature. This means that for a DVM calibrated up to a year ago, we can expect the DVM error to be ±0.06% of the reading and ±0.03% of the range. So, if we're on the 2Vrms range, and we're measuring 1Vrms, we might have an error of ±600uV related to the reading and ±600uV related to the range. Combined, the error could be as much as ±1.2mV.

A 5.5 digit DVM will not have the accuracy needed to complete the calibration in a meaningful way. For example, a 5.5 digit DVM might have an error specified as "0.9 + 0.05", which means the error is 0.9% of reading. This is a measurement error 15X higher than the 6.5 digit DVM.

DVMs tend to have ranges that based on multiples of either 1's or 2's. That is, some DVMs will have AC ranges that are 0.1V, 1.0V, 10V, while others will be powers of 2, such as 0.2V, 2V, 20V.

The range choice matters, and so make sure you are clear on the your DVM type.

Starting the Test

In the 1.221 and later software releases, you'll see a new menu item in Help, and that is "Verify Calibration". Before running, make sure your DVM and QA40x hardware are both acclimated to room temperature, and that the DVM is within the specified limits in your most recent calibration certificate for that instrument.

When you select the Verify Calibration menu item, you'll see a window as below:

Note that you need to select the range type appropriate for your DVM. And as you change the range type, you'll notice the specified reading and range errors will change. Carefully copy the manufacturer's accuracy specifications for AC readings into the fields. You can also enter the DVM details, room temperature and analyzer serial number. The test frequency you select will be the test at which everything is run. Typically, this will be 1000 Hz.

After entering the test details, you can click "Start Verification" and you'll see the task window populate with tasks you must complete. For example, the first task we see instructs us to apply shorting blocks to L- IN and R- IN.

Complete the task, and then click Next. You'll see the completed task go dim and get pushed downward, and a new task appear at the top.

Work your way through the tasks. You can scroll through to see completed tasks. The first few tasks will use your benchtop DVM to verify the output ranges of the QA40x hardware. Once the errors are known for the L+ output, the DVM is no longer needed and we can then use the L+ out to drive the 4 inputs (L+ IN, L- IN, R+ IN, R- IN). And once we know the error of those inputs relative to the L+ out (for which we know the DVM error) we can then measure the 4 outputs (L+ OUT, L- OUT, R+ OUT, R- OUT). This means all measurements can be made and traced back to the initial DVM errors we made.

There are 13 tasks total, and the process should take about 10-15 minutes. Once it has completed, you'll see the following (the graph will build as you complete the steps).

What this lets us see is the relative IO linearity of the analyzer, from -100 dBV to 18 dBV. Here we can see make the statement that for -20 dBV output, we can see the error is inside the ±0.05 dB error bars. Note that this graph doesn't reflect any of the absolute DVM measurements. This is just relative errors.

When the last task is completed, you should also see a HTML page open in your browser. You can then use your browser to print the HTML report to PDF. Use printer scaling if needed to get it all to fit a single page. You can see a PDF test report HERE

The test report is where you'll see relative and absolute errors combined, and a summary. For the unit tested above, the last paragraph of the summary gives the information we need: "If we conservatively add the worst-case linearity deviation to the worst-case absolute error from the DVM checks, we obtain an overall upper bound on level accuracy of roughly ±0.39% over the region from about 50 dB below full-scale input up to full-scale. In practice, typical errors will be smaller than this bound."

This means we can expect our worst-case error to be about ±0.4% over the measured range. Note that the analyzer can only verify itself up to 18 dBV (about 8Vrms) and so we can't really know what the error might be beyond that, however, we can be reasonably confident it is similar because the input attenuators are resistive, and we'd not expect to see a shift in resistors at higher input voltages.