rigol DS1054Z wrong rms measurement on long waveform

[MTong]

Hello,

I have a brand new beautiful DS1054Z.

If I measure a sinusoid from a funct generator (500mVrms, 60Hz) the measurement looks correct on the scope if I have a time scale of 10ms/div.

expanding the time scale to 500ms/div, the rms measured changes to 725mV (frequency measured also changes to 100Hz)

I tried at 12M points, 6M and 60k. Same result.  On the expanded time scale, if I stop the scope and enable the delayed sweep time base (pressing on the horizontal scale knob) and zoom in enough, the rms measurement will be correct again. This means that the scope has in memory the correct waveform, with enough details (no aliasing or other problems).

So how is it calculating the wrong rms on the long time scale? or is it downsampling to do the calculations? (also the period is wrong, so it is missing the zero crossings)

Also I noticed that with so many points, the rms is still calculated very quickly (too quickly?) while a tektronics DPO3000 would take several seconds to display rms of a long waveform at 5M points.

Is there anything I am missing?

Thanks a lot

Marco

[Sailor]

Just a guess, but based on a (very) little bit of experience with a Rigol 'scope (MSO4000), some operations/measurements were performed **only on what was displayed on the screen**. So if you have some non-integer number of cycles displayed, maybe it's calculating the rms value of what you can see. It might also explain the wrong frequency. It might also explain the 'too quick' rms calculation.

Or I might be completely wrong I understand they fixed the FW in the MSO for the problems that I saw, and one would like to think that they learn from their mistakes...

[MTong]

thank you, but I do not think it is that.

I am looking at the entire waveform (many tens of cycles). It should in no way affect the measurement so badly, even if it is not an integer number

If it was that, with less cycles I should get the error, but it is instead precise

thanks

[MTong]

you might be partially right. Since the display is showing already an under-sampled set, that is probably being used to calculate the rms, instead of the data from the memory. So this is not related to the non-integer number of cycles.

The function becomes completely useless in case you are looking at a switching waveform of a PFC as you need to read the rms of a switching waveform over a 60Hz period.

[LabVIEWguru]

I'm not familiar with your scope, but it may be using math to calculate RMS based on the screen contents rather than what is actually measured. I use a really high-end Agilent at work and you can select RMS or RMS (full screen). I've not actually looked it up, but the RMS (Full Screen) is probably extrapolated and I know when you change the screen, the data the function uses is going to change.

Just a thought, I'll bet Sailor is right.

[MTong]

I am having a good dialog with one of the reps from Rigol. Their way of doing measurements (all of them, also FFT apparently based on another post) is based on the shown points (not shown amount of time).
This is apparently a design decision that was made without considering slow varying switching waveforms

Not sure if I will be able to convince them about changing the way calculations are done to match what tektronix and agilent do, but I'll try