Using DMMs to read average voltages / currents

[MarkE]

Poynt99 did some really great work investigating DMM average voltage readings versus fast oscilloscope average readings.  If you have not watched his videos they are linked here:

http://www.youtube.com/watch?v=-2KhGpmXPjc
http://www.youtube.com/watch?v=pXTbcToC5T4
http://www.youtube.com/watch?v=70sPnpG2JO4

In addition to Poynt99's work, Steve Weir and I have performed a number of tests done a bit differently to evaluate the same question:  "How faithful are DMM average voltage / current readings when looking at signals that are much faster than the DMM sample rate of a couple to a few dozen samples per second?"

The answer is that for signals that are 50X faster or more than the sample rate, they are very good.  Here is the test set-ups that we used, the data and the plots using a very expensive oscilloscope, and DMMs that range from a 6.5 digit bench 8846A to good handhelds, to $10. eBay specials:  XL830s.  The conclusion is that even without using an external anti-aliasing filter, the nature of the integrating A/D converters in DMMs makes them good to 1% or better across wide frequency bands. 

The tests that we conducted with a DC Block in place prevented the signal generator from introducing any DC offset.  There was a small offset of about 3.4mV introduced by leakage current from the oscilloscope input amplifier.  The raw data shows that all of the DMMs correctly averaged to within 3mV of each other at all frequencies.  The Fluke 87-III shows the greatest deviation in the 100kHz to 1MHz band.  Applying 1.99V of bias to the signal changed very little except for one of the $10. meters which appears to be suffering some RF rectification effects above 1MHz.  Removing the DC block allows us to go to much lower frequencies.  Below about 50Hz the meters exhibit sampling artifacts as expected from sampling theory.  The last test with the 10K/10uF low pass filter as expected yields the best results.  The very tightly matched results confirm the idea that common rejection and RF rectification were the causes of the Fluke 87-III meter suffering in the 100kHz - 1MHz band and the cheapy $10. XL830 meter falling off rapidly above 1MHz.  One final note:  The oscilloscope has an 8 bit A/D front end.  Any performance better than +/-0.2% is just chance.  The oscilloscope was triggered by the square wave sync output of the function generator, and the cycle average was used.

There is some observable common mode rejection degradation in the 100kHz - 1MHz region for the Fluke hand held meters we tested, and above 1MHz for one of the cheapy $10. meters.  We were able to pretty much eliminate the CMRR by using a 10K Ohm resistor in series with a 10uF capacitor.  We built that filter using surface mount parts between to SMA connectors that we then wrapped in copper tape to shield out RFI.

What all of this means is that if you can arrange to measure your voltage where it is a stable DC value, then you can measure the current using just a current sense resistor and a DMM, and you will be able to get average power measurements even if the current has high frequency spikes.  If both your voltage and current are AC, then this technique will not work for you.  In that case you will need to sample the voltage and current at a high enough rate to capture the signal content, multiply the instant voltage and current sample values, and then integrate / average those products.

[TinselKoala]

Nicely done.

Have you also done any tests with really wild signals like spiky pulse trains such as are produced by the Ainslie kludges? It would be nice to see how the DMMs compare to the scope's onboard integration of such signals.

[MarkE]

Nicely done.

Have you also done any tests with really wild signals like spiky pulse trains such as are produced by the Ainslie kludges? It would be nice to see how the DMMs compare to the scope's onboard integration of such signals.

Steve did tests almost a year ago using square waves with < 1ns rising and falling edges.  We have talked about redoing those tests now that he's just recently bought those cheapy XL830 meters as well.  If one believes in Fourier, then the matched results from 50Hz to 20MHz covers rise times down to about 18ns.

[TinselKoala]

Steve caused me to be sent one of the XL830L meters (dropshipped). Believe it or not, it is actually much _worse_ in its response to RF than the "standard ElCheepo" Cen-Tech DMMs from Harbor Freight. As you know I use two of the Cen-Techs as input monitors to many experiments, one as a voltmeter and another as an inline ammeter. I put the new XL830L as ammeter in series with the Cen-Tech and compared both on the same Ainslie test trials. With the strong spikes produced by using the Gate Boost circuit, all the meters can be made to go "crazy", including the digital ones back in the main bench PSU even though I have a big filter cap in parallel. But the XL830L goes crazy first and by the most off-value. It is also much more sensitive to position wrt the rest of the circuit than the Cen-Tech meters.

It might be interesting to see what kind of results you guys can get if you can generate the same kinds of spikes that are seen in the Ainslie trials with the "plain Q17" setup. The "Shifting Paradigms" board with its good filtering does not cause this deviation, as far as I can tell.

I haven't opened the XL830L yet, but I do note that the Cen-Techs have some RF shielding inside, a foil plane on the removable case half that is contacted by a spring on the main board.
When the RF level is low, though, the XL830 tracks the other meters within a few milliamps/millivolts usually.

The XL830L is nice, though, I like the "pseudoFluke" styling and the LCD backlight, which has an autofade feature. Also...it seems to have come with an extra 9V battery!

[MarkE]

We used high frequency coax everywhere right up to the SMA to Banana plug adapter that plugged into the DMMs.  So stray RF was not a factor.  There may well not be any shielding in the XL830s.  With a slow enough oscillator they wouldn't need it to meet EMC regulations.  My guess is that the issues you had were more to do with RF in the environment than the signal content on the signal leads passing through the meter.  That guess could be wrong if the current sense wiring acts as an efficient antenna inside the XL830 case. 

Enough stray RF will eventually pollute even heavily shielded instruments.  I would be inclined to put any noisy assembly in a shielded enclosure and then bring out the measurement signals.  That means using a CSR in the shielded box for current measurements.  I'll talk to Steve and see if he is interested in tests with big spikes on unshielded wiring.  He has talked about doing experiments showing current measurements by different means.  Maybe he can be convinced to do something like this instead.