The rectangular waveform can behave quite a bit different for the AC circuit. One point is a possble slew rate limit that may apply.
The other point can be a time delay in detecting the polarity reversal in some analog RMS converters.
A 3rd point is a possibly limited bandwidth - some of the power will be outside the BW of the DMM. This would not be very much for the higher end bench DMMs, but it can be significant for handheld ones.
So the rectangulator wareform is a relatively special case. It is still one of the easier ones to generate.
A little low pass filtering may be a good idea to get at least rid of the slew rate limit and reduce the delay effect. If only filtering out the higher frequencies (e.g. > 50 kHz) the effect of the filter should not depend that much in the parts accuracy and effect of the DMM input.
This is exactly what you DON'T WANT To DO with the low frequency squarewave since you are now placing an additional uncertainty on the slow edge of the LPF result and the source squarewave source impedance. If the slower edges aren't very very close in rise and fall and shape they will introduce and error in both RMS and DC readings, whereas with a much faster unfiltered edges the result is much less influenced by the edge since the edge period is so much smaller than the squarewave period.
Edit: A quick experiment or a little Fourier Analysis can verify the LPF is NOT what you want to do. Just ran a quick setup and added a simple 5us (32KHz) RC (500 ohm and 10nF (9.96nF actually)) low pass and the result was additional error of -2.5mv!!! Kept the series 500 ohm R in place since it's forming a low pass with the DMM leads and input capacitance, then just added the shunt C . So keeping those edges as fast as possible is what you want to do
BTW the earlier comment about the decision delay regarding the polarity under AC True RMS, a slow waveform edge speed would introduce more uncertainty and thus more potential error.
With a slower slope the timing error from noise will be larger, but there is less consequence, as the wrong sign for a signal near zero has essentially no effect. This also still applies to a longer delay from slower reaction. With the square wave the timing error is multiplied with the full voltage.
There is nothing wrong with filtering the square wave waveform a little to round the edges - the effect of the fitler is not easy, but it can be calculated and if not so low, the exact value of the capacitor is also not that critical. With the RMS part it is only about the amplitudes, the phase shifts can distort the waveform, but it does not effect the RMS.
If a meter fails on the square wave test, it can still work pretty well with a sine. An even if a meter reads fine with the square, this does not guarantee that it works OK with a sine. It is a different and relatively extreme test pattern. With the usualy AD637 or similar chips it would not detect a failing averaging capacitor, or the effect at low frequencies where the capacitor is at the low end.
If interested, we can send the geber files & partial BOM for the shown PCB, if I can find them Not sure how to post them here tho, so PM.
A simple thought experiment will show this is thinking is flawed. If you consider slowing the edges down, the extreme case shows the eventual waveform will approach a triangle wave which we all know has an RMS value of 1/(sqrt(3)) not unity as the normalized squarewave. That's a significant reduction in RMS value and shows a strong relationship between RMS and edge slowing the edges down reduces the squarewave harmonic content which reduces the energy available from the waveform and thus reducing the RMS reading.
Capturing the harmonic content is another reason to keep the squarewave frequency low as well as a fast edge.
A simple thought experiment will show this is thinking is flawed. If you consider slowing the edges down, the extreme case shows the eventual waveform will approach a triangle wave which we all know has an RMS value of 1/(sqrt(3)) not unity as the normalized squarewave. That's a significant reduction in RMS value and shows a strong relationship between RMS and edge slowing the edges down reduces the squarewave harmonic content which reduces the energy available from the waveform and thus reducing the RMS reading.What's the problem? Say you built a circuit to produce a triangle wave with an amplitude that's very close to a DC level Vdc. You can compare the RMS reading of the triangle wave to Vdc/sqrt(3). Why is this worse than comparing the reading to Vdc/1?
That makes a lot of sense when testing instruments made for wide-band signals like a scope. I'm not sure if it's the most realistic test for DMMs. Generally the configuration of a DMM, like filter settings and reading rates, are tuned to measure signals with a particular, known, frequency and quite limited bandwidth.
Kind of cool!
Just thinking out loud and imagining a little here: One could try measuring signals with a duty cycle other than 50% with this kind of precise amplitude and see how the meters deal with it... so, for example, the error at e.g. 1% PWM might tell you something about how a particular meter responds to fast edges, which could then possibly be subtracted out from the 50% PWM test later... hmmm...
One can use any waveform they desire for RMS measurements, but doubt that anything other than DC is as simple to create and easy to verify, and has identical AC RMS and Average DC levels!! Please if you know of any waveform other than DC that has these properties please enlighten us!!!
Think you are missing the point of using the squarewave, never said it was the most realistic or best test, just simply the easiest to implement and verify. Doesn't cost much either
Regarding particular frequencies DMMs are "tuned" too, this would be the mains and maybe a few harmonics that come to mind for obvious reasons. Regarding reading rates, some of the newer DMMs like the KS34465A and DMM6500 seem to have higher digitizing rates, although the old HP3458 also had high digitizing rates I believe.
One can use any waveform they desire for RMS measurements, but doubt that anything other than DC is as simple to create and easy to verify, and has identical AC RMS and Average DC levels!! Please if you know of any waveform other than DC that has these properties please enlighten us!!!You keep talking about this unity factor as if it's something magical that is exact by definition. Say you take your chopped DC circuit, run in through a slew-rate limiter. Now say you do the math, and figure out that for the slew rate, frequency and amplitude, the RMS value is (0.95 +/- 0.01)*Vdc. How is this any less accurate than comparing to (1.00 +/- 0.01)*Vdc? Are you doing a bridge circuit with a thermal RMS converter where you are directly comparing the analogue RMS value of two signals without digitizing them, like with the old Fluke 540B? Or are we not talking about a unity factor, but about a very precisely defined factor that may be pretty much any value within a factor of two or so of unity, just to stay in the same DMM range?
One can use any waveform they desire for RMS measurements, but doubt that anything other than DC is as simple to create and easy to verify, and has identical AC RMS and Average DC levels!! Please if you know of any waveform other than DC that has these properties please enlighten us!!!
Having a waveform that has the exact same RMS and Average DC value is a big advantage. One can take a DMM, measure the DC Reference voltage, then measure the Average DC value of the waveform using a simple RC LP using the same DMM. This ratio should be very close to 1/2 or somethings wrong!! Then measure the AC RMS waveform value, it should also be very close to the Average DC value and 1/2 the DC Reference measurement, unless something is wrong or the DMM can't handle the waveform.