- This topic has 6 replies, 2 voices, and was last updated 4 years, 4 months ago by
.
- Posts
I just tried the mooshimeter on a freshly calibrated 5 volt 100 Hz square wave reference voltage (4.999 volts to be specific). The buffered graph was both clean and correct. The meter part however was off by at least 3% or more depending on sample frequency 8000 Hz was 4.85 VAC, 4000 Hz was 4.802 VAC and 2000 Hz was unstable and hovered around 4.67. While the value at 8000 HZ the value is good enough for the girls I go out with. I would like to understand what is going on. The graph read exactly 5 VAC at peak. It was 50% duty cycle. The rise and fall were not perfect, but as close as I have ever seen.
Hi Owen,
In AC mode, the meter uses zero-crossings to set the boundaries for the RMS calculations (since doing an RMS calculation on an incomplete waveform creates calculation error). If the 5V square wave only oscillates between 0 and 5V without actually crossing 0, I think the RMS calculation window may be off by few samples, which could account for the error you’re seeing.
Does this explanation make sense? Best
~JamesI should add the voltage reference is an equally innovative device for checking meter calibration. I understand that between +5v and -5v there would be no zero crossing were it not for circuit impedance. And I understand that with digital sampling the zero crossing window is very very small. But my cheap averaging meters read 11% high and my RMS meters read very close to 5 volts. (I am a meter junkie) I don’t envy your job of reconciling digital sampling with AC theory.
I wanted to put a link to what I am talking about but it didn’t make it in the last post.
http://www.voltagestandard.com/DMMCheck_Plus.htmlHi Owen,
Ohhh, the plot thickens! Thank you. If it’s bouncing a square wave between -5V and 5V, then the zero crossing detection should work, but this gives me a different theory, and I think it explains why you see better results as you raise the sampling frequency: A square wave has component frequencies much higher than the fundamental (in the case of the DMMCheck plus the fundamental is 100Hz). Since the Mooshimeter is only sampling at 8000/4000/2000 Hz, it drops all frequency information above Nyquist, resulting in lower RMS reading.
If your sampling theory is a little rusty, the intuitive way to think about this would be to see that when the meter is sampling at 1000Hz, every sample represents a smearing together of a 1ms window. If the signal crosses from 5V to -5V within that time, the meter will read some value in between 5 and -5 for that sample. When it does the RMS calculation later, the sample buffer now has a low readings at every transition of the square wave, resulting in the RMS calculation being too low. The effect gets less pronounced as you raise the sampling frequency because every transition of the square wave affects a smaller ratio of samples.
Also I have a DMMCheck (not a DMMCheckPlus) on my desk, big fan of these products :)
Please let me know if this explanation makes sense to you! Best
~JamesHi James,
I understand now. The graph shows a shorter period each time you step the sample frequency down from 8000 Hz. I don’t know if it is possible, but wouldn’t over sampling reduce the problem? I could see a saw tooth or even and lopsided sine-wave doing the same thing.Sometimes in strange answers you can find a lot of the most unusual ideas that can change your decision and bring the most optimal solution to the question.
