What Vpp accuracy to expect from oscilloscope at 500mV scale,and Auto or Single?

[alexg]

Hi All,
briefly: I have purchased my first oscilloscope yesterday, RIGOL DS1054Z, before that I had friend's SIGLENT SDS1102CML for about a month, now I am trying to switch to using mine and there are some differences in measured numbers when it comes to measuring same circuit (max and min voltages as well as Vpp), I am not sure now which oscilloscope has better accuracy, if this is the accuracy I should expect at all from an oscilloscope of this class and whether I even measure it properly.

now in detail:
For self-education purposes and out of curiosity I am testing different 3.3V voltage regulators that I built based on different chips, different input/output capacitors types and values, different inductors, etc... Then I test them in the same circuit under same load and observe waveform fluctuation, noise.
While I was using my friend's Siglent one regulator was showing only 60mV Vpp, when I measure it with Rigol it shows 120mV Vpp. The measuring parameters are the same, 1mS time scale , 500mV voltage scale. They also detect MAX and MIN voltages differently, siglent showing one numbers while Rigol showing slightly different numbers. I switched probes between scopes but result still the same.
(you can see the illustration below)
So my question is, is this normal to have slightly different readings between two scopes, how do I determine which one is showing more correct measurements? Does this seems correct to you at all?

Also, accuracy of measurement with Rigol seems to improve when I switch it to Single mode, what I do, is I switch it to single mode and push "force" button, so scope capture one waveform and then Vpp is less, is this how you suppose to take your measurements? Asking because normally I would just keep it in Auto and simply watch numbers change or push Stop when needed, but I am wondering if that's not the proper way to capture data?

[wraper]

 
Do you understand that you are measuring nothing useful, mostly ADC noise? Vpp on Rigol is higher because there are more waveforms captured.
Set AC coupling and use  lower voltage scale so you can see the actual signal you want to measure. Or at least lower voltage scale and set DC offset so the actual signal / AC noise on DC signal becomes visible on the screen. What do you expect from 8 Bit ADC / measuring signal not visible on the screen (only DC component of it)?

[alexg]

Do you understand that you are measuring nothing useful, mostly ADC noise? Vpp on Rigol is higher because there are more waveforms captured.
Set AC coupling and use  lower voltage scale so you can see the actual signal you want to measure. Or at least lower voltage scale and set DC offset so the actual signal / AC noise on DC signal becomes visible on the screen. What do you expect from 8 Bit ADC / measuring signal not visible on the screen (only DC component of it)?

To answer first line in your reply: no I honestly do not really understand yet what I am measuring  I think i mentioned I am a beginner, or didn't I? Uops, sorry, yes I am a beginner with very little knowledge about some things but I figure things easier this way, I do something wrong, then figure out why and then how to do it right. Will do exactly that you saying about AC coupling and lowering scale. Thanks.

[Gyro]

A bit harsh to a newbie, wraper!

@alexg:
Yes, the issue you have (probably the same on the Siglent) with these things is that they do only have 8 bit ADCs (256 steps) to cover the whole dynamic range of the input. That limits your measurement accuracy to about 0.5% of the waveform voltage when it is the maximum for that range. What you are doing (assuming that you have the input set to DC coupled) is artificially focussing in on a very small portion of the signal, using very few bits of ADC resolution to take your measurements. This means that your measurements could be several % or maybe even 10s of % out. As wraper says, if you change the input coupling to 'AC' then you can lose the DC component increase the input sensitivity (down to a few mV/div) and just study the amplitude of the noise.

This obviously prevents you from displaying a measurement of the DC voltage at the same time but if you want to measure that then you ought to use a  DMM to get decent accuracy anyway.

[alsetalokin4017]

You can display/measure the DC component and the AC ripple if you use two channels, connected by their probes to the same point in the DUT. The DC component should be displayed with DC input coupling and at a vertical range sensitive enough so that the signal covers half the screen or more from baseline to signal trace for voltage accuracy. The AC ripple should be displayed with AC input coupling and again, a vertical range that will produce a decent displayed amplitude on the screen. Use the 10x probe and channel settings whenever possible.

The Rigol uses just the screen displayed data for all of its left-menu measurements. The better picture you have on the screen (vertical size, number of periods), the better the measurements will be.

Yes, the DMM will generally give you better accuracy on a DC voltage measurement.

[danadak]

One way of measuring noise is to use scope on infinite persistence and capture
a large number of sweeps. This will capture pk-pk noise quite effectively. Try
that on both scopes, then you should get same answer if both channels caled
on their probes.


Regards, Dana.

[w2aew]

Averaging is a useful technique to remove the random noise component in a signal you're trying to measure.

Also, make absolutely sure that you've properly compensated your 10x probes - on the scope you're using!  It is certainly possible (likely) that the compensation adjustment will be different for the two different scopes.  This can make a BIG difference in the magnitude of the signals that are above several kHz in frequency.

This video might help:

[danadak]

If you are trying to measure pk-pk noise then do NOT use averaging, as that
effectively removes uncorrelated noise from the signal, the information that you
are seeking to measure.

This might help -

http://www.pearl-hifi.com/06_Lit_Archive/14_Books_Tech_Papers/Motchenbacher_Connelly/Low-noise_Electronic_Design.pdf


Regards, Dana.

[alexg]

Thanks everybody for good suggestions and pointing me into right direction, I learned alot since I posted that question and Am very familiar with my scope now. I also realized very quickly that getting 4 channel scope was a really good idea.

I have another question though: Let's say I am measuring 12V regulator's output on one channel and VPP noise is 240mV, when I connect second channel that noise is 320mV now, so apparently other channels couple noise to other channels, is there a trick how to make different channels not to couple noise to one-another?

[rx8pilot]

I have another question though: Let's say I am measuring 12V regulator's output on one channel and VPP noise is 240mV, when I connect second channel that noise is 320mV now, so apparently other channels couple noise to other channels, is there a trick how to make different channels not to couple noise to one-another?

Your grounding/probing technique will influence this a lot. My biggest oscilloscope lessons have been on probing technique. My intuition was to focus on the buttons, switches, and internal signal path. Search around for some papers and videos on probing technique and do some experiments. When frequencies are high and/or voltage is very low - it becomes harder and harder to get the signal to the front end of the scope without changing it.

w2aew has a fantastic library of videos - many of them touch on this topic.