Rigol voltage measurement with 1 channel vs. with second idle channel turned on?

[Electro Fan]

Anyone notice that after you run a calibration with a Rigol scope and then "test" the scope by feeding it some "known" voltage (from a decent function generator) that in addition to channel 1 and channel 2 on the scope reading somewhat differently from the same source that the measurements will also vary based on whether you have just the channel receiving the test signal on by itself, or whether you also have the second (idle) scope channel also turned on?

I've been using 50 ohm termination on both ends and looking at 6Vpp.  I use the same cable to switch between the function gen and the scope inputs.  Then I toggle on just the active scope input (with the second input off), and then I also toggle on the second idle scope input - and the voltage readings change.

I've found that I seem to get the best resolution when using the Vtop measurement and I've also tried HiRes mode in case that matters.  Frequency seems to be consistently accurate within the displayed resolution of the scope counter (a 1 followed by all zeros at 1 MHz).

In some cases (after a fresh calibration) when measuring Vtop I can get the function generator and the scope to agree within about +/-5 millivolts, in other cases it's about +/- 10 millivolts, and occasionally it's about +/-40 millivolts.  I've tried running the calibration routine several times (after having the scope turned on for an hour or more).

I realize that scopes are generally not ideal precision instruments for measuring voltage.  What I'm mostly curious about is to see if other Rigol users (or other brand users) find that whatever values you get when measuring with one channel, if you get a somewhat different measurement simply by turning on a second idle channel.  Ballpark, I'd say that turning on the second idle channel can change the reading by +/-10 millivolts.  In other words, even if I sometimes can get a calibration that shows channel 1 within +/-5 millivolts reading using Vtop (ie, somewhere between 2.995V and 3.005V) based on the expected 6Vpp from the function generator using 1 scope input, when I turn on the second channel (even though it's idle and not connected to anything), I either lose or add 10 millivolts relative to the reading on the active channel by itself.  So, 2.995V to 3.005V could become anywhere from 2.985V to 3.015V.  Turning off the second idle channel in this example would return the reading to 2.995V to 3.005V.  It's not directionally consistent - sometimes turning on the second idle channel creates a plus, and sometimes a minus - but almost always (I think always) turning on a second idle channel changes the measurement reading of the active channel.

It's probably measurement minutiae but I'd like to understand if this is common and what causes it?  (It's possible it's some combination of reasons including pilot error, I realize. )  Thx

[Fungus]

There's been several threads on this topic. It's a thing.

This sort of thing happens when you multiplex/share an ADC between different inputs.

Fun fact: The manual even has a specification for channel-to-channel isolation.

[Hydron]

Assuming it's a DS1000Z series scope, have a look at the datasheet of the ADC (HMCAD1511). It will be using different combinations of the internal ADCs depending on the particular combination of channels you are using. Calibration is probably required on all different combinations, and any small differences between them will show up in the measurement you just did.

There other are bigger issues with the way they're using the ADC:
- The sample rate, anti-aliasing filter and sin(x)/x are inadequate for 100MHz bandwidth when multiple channels are turned on (it's starting to get iffy at 2 channels, using 3-4 make fast risetime signals look completely horrible with artifacts displayed that aren't actually there)
- By only having 2 analogue gain ranges, digital gain in the HMCAD1511 is being used past the point where it's advisable - there are a number of ranges with missing steps and I suspect many of the rest aren't getting close to 8 effective bits

All that said, it's still great value and very useful when used within it's limitations.

[Circlotron]

I know that channel 1 offset changes when channel 2 is turned on and off.
https://youtu.be/6OQOK85-r0c

[Electro Fan]

I know that channel 1 offset changes when channel 2 is turned on and off.
https://youtu.be/6OQOK85-r0c

This looks like another good example of what I was trying to describe (turning on an idle second channel changes the measurement of the first active channel).  Thanks for posting.

[Electro Fan]

Thanks for all the good insights.  I guess as Fungus says we'll have to chalk it up as "a thing."  And as Hydron says "it's still a great value and very useful when used within it's limitations."  (but speaking of limitations, fwiw, a 1 MHz DUT signal doesn't seem like it should be pushing the scope's bandwidth spec real hard - but I guess it's more complicated than simply looking at the frequency of the DUT signal, ie ADC's, gain ranges, and all....)

Just one other possible avenue toward getting more consistent and accurate results.... and sorry if I've missed it, but is there a way to better use the calibration process - such as individually calibrating each channel?  Or maybe that misses the point that the circuitry in the scope is shared across the channels so regardless of calibration when new variables are introduced (like turning on an idle channel) the circuitry is going to yield a new measurement?

[Electro Fan]

Ok, now that I know it's not just me, ....

When channel 1 is the active channel and channel 2 is turned on as the idle channel, the voltage measurement always increases. When 2 is turned off the measurement goes back down.

When channel 2 is the active channel and channel 1 is turned on as the idle channel, the voltage measurement always decreases.  When 1 is turned off the measurement goes back up.

The behavior is directionally consistent.

(I'm using a Rigol MSO2072A.)

New question: is this a necessary limitation of the circuitry used, or is it possible that with improved firmware that this could be partially, substantially, or fully corrected?  ie, could better math used in the firmware fix the problem?

[Electro Fan]

Ok, the plot thickens.

What I said above is correct - with the understanding that the active channel is the trigger channel.

On my Function Generator I have two output channels.  When I cable them both to the scope (channel 1 to channel 1 and channel 2 to channel 2) and send the same signal (1 MHz, 6Vpp) to both scope channels the behavior is as follows:

Situation 1:  Turning on and off the trigger channel does not seem to change the measurement on the non-triggered channel. 

Situation 2:  Turning on and off the non-triggered channel does change the measurement on the trigger channel.

Is it possible that Rigol only tested Situation 1 and simply forgot to test Situation 2 (in which case maybe they could fix the issue)?

[Fungus]

New question: is this a necessary limitation of the circuitry used

Yes. Completely isolated input circuitry and ADC for each channel would put the price up.

One workaround is to always enable all four channels. 

(but speaking of limitations, fwiw, a 1 MHz DUT signal doesn't seem like it should be pushing the scope's bandwidth spec real hard - but I guess it's more complicated than simply looking at the frequency of the DUT signal, ie ADC's, gain ranges, and all....)

Only a perfect 1MHz sine wave has 1MHz bandwidth. A 1MHz square wave has infinite bandwidth. 

or is it possible that with improved firmware that this could be partially, substantially, or fully corrected?  ie, could better math used in the firmware fix the problem?

Anything's possible in theory. Theoretically you could calibrate for every possible combination of enabled channels and voltage ranges. It's unlikely to happen in practice though.

Remember that the voltage accuracy specification for these things is only about 5% to start with. A 40mV change on a 6V signal is nothing.

[Fungus]

On my Function Generator I have two output channels.  When I cable them both to the scope (channel 1 to channel 1 and channel 2 to channel 2) and send the same signal (1 MHz, 6Vpp) to both scope channels the behavior is as follows:

Situation 1:  Turning on and off the trigger channel does not seem to change the measurement on the non-triggered channel.  Situation 2:  Turning on and off the non-triggered channel does change the measurement on the trigger channel.

Does it happen consistently for all possible combinations of cables and channels?

Is it possible that Rigol only tested Situation 1 and simply forgot to test Situation 2 (in which case maybe they could fix the issue)?

No, because there's nothing to test. It's caused by hardware, not software.

[Electro Fan]

New question: is this a necessary limitation of the circuitry used

Yes. Completely isolated input circuitry and ADC for each channel would put the price up.

One workaround is to always enable all four channels. 

(but speaking of limitations, fwiw, a 1 MHz DUT signal doesn't seem like it should be pushing the scope's bandwidth spec real hard - but I guess it's more complicated than simply looking at the frequency of the DUT signal, ie ADC's, gain ranges, and all....)

Only a perfect 1MHz sine wave has 1MHz bandwidth. A 1MHz square wave has infinite bandwidth. 

or is it possible that with improved firmware that this could be partially, substantially, or fully corrected?  ie, could better math used in the firmware fix the problem?

Anything's possible in theory. Theoretically you could calibrate for every possible combination of enabled channels and voltage ranges. It's unlikely to happen in practice though.

Remember that the voltage accuracy specification for these things is only about 5% to start with. A 40mV change on a 6V signal is nothing.

Roger all that.  Thanks Fungus.  Item 2 (about square waves having infinite bandwidth) is something I should have considered.  I kind of knew and understood this about square waves being made up of odd harmonic sine waves but this exercise and your comment have me working on getting my head around it better.  (Does this mean that if I test with sine waves instead of square waves the scope will be able to overcome the issues we're seeing?  I guess I need to try that.)  Thx again.

[Electro Fan]

On my Function Generator I have two output channels.  When I cable them both to the scope (channel 1 to channel 1 and channel 2 to channel 2) and send the same signal (1 MHz, 6Vpp) to both scope channels the behavior is as follows:

Situation 1:  Turning on and off the trigger channel does not seem to change the measurement on the non-triggered channel.  Situation 2:  Turning on and off the non-triggered channel does change the measurement on the trigger channel.

Does it happen consistently for all possible combinations of cables and channels?

Yes, it does seem to be consistent for all possible combinations of cables and channels - but there is always some possibility of pilot error....

[TMM]

Ok, the plot thickens.

What I said above is correct - with the understanding that the active channel is the trigger channel.

On my Function Generator I have two output channels.  When I cable them both to the scope (channel 1 to channel 1 and channel 2 to channel 2) and send the same signal (1 MHz, 6Vpp) to both scope channels the behavior is as follows:

Situation 1:  Turning on and off the trigger channel does not seem to change the measurement on the non-triggered channel. 

Situation 2:  Turning on and off the non-triggered channel does change the measurement on the trigger channel.

Is it possible that Rigol only tested Situation 1 and simply forgot to test Situation 2 (in which case maybe they could fix the issue)?

This is because when you 'turn off' the channel that is being used as trigger it is still enabled and multiplexed into the ADC, just the trace isn't rendered onto the display. The 'issue' is clearly related to when multiple channels are time-division multiplexed into the ADC vs a single channel being connected to the ADC. It could be a hardware limitation. An oscilloscope - at least a low-end 8-bit one - isn't the correct tool to measure absolute voltages to less than a few % accuracy anyway.

[Electro Fan]

I left the scope and the function gen on for several hours (maybe about 12?).  Probably not too surprisingly, the deviations in the various scenarios seemed less pronounced.  Normally I shut the gear off after using it as it can sometimes be days in between but I think if I was using it with the goal of getting the best measurements I'd leave it on 24x7. 

And as noted by TMM it's clear a scope isn't the best tool for measuring voltage.  I'm pretty sure a $40 DMM can measure voltage better than a $400 or $1200 scope.

Having said that, a new question.  Does the ADC impact just voltage measurements or also frequency counter measurements?  I would think that the ADC has to understand time as well amplitude and if something in the sampling process loses track of one it could lose track of the other.  But maybe for some reason the ADC deals better with time than with amplitude?  The counter seems more accurate than the voltage measurements, but maybe it's apples (1 Hz) vs. oranges (tens of millivolts). 

[metrologist]

(I'm using a Rigol MSO2072A.)

Wanted to make sure this point is not overlooked as it might affect some responses. It's not the standard zed model that naturally comes to mind...

[Fungus]

Does the ADC impact just voltage measurements or also frequency counter measurements?

The frequency counter in the DS1054Z is a separate piece of hardware. It doesn't go through the ADC.

[Hydron]

(I'm using a Rigol MSO2072A.)

Wanted to make sure this point is not overlooked as it might affect some responses. It's not the standard zed model that naturally comes to mind...

Yes, this makes a big difference. I have no idea if my original comments are applicable to anything other than the DS1000Z series (though many HMCAD1511 scopes will have similar behaviour), which is why I prefaced my earlier post noting my assumption of which scope.

[JohnPen]

Another wrinkle, which may have been discussed before, is that the Vmax statistic and probably other voltage stats indicate imaginary readings for a 'no signal' input.  For example on 100v range Vmax indicates 12v.  This reduces in value as one alters the V gain and is really only 'accurate' when on the 10mV range.  For real input signals no problem but it could be distracting and should perhaps be blanked for the higher ranges.  The moral is always look at the trace to confirm that stats are genuinely representing what you see.

[MrW0lf]

Heres drift of my 8bit scope on 4mV/div range over 2x 8m20s.
50x vertical zoom, 1Hz LPF.
ChA, ChB terminated by 50 ohm.
Overlapping traces are auto-zeroed ChA, ChB ON at same time, saved as ref.
Blue trace is ChA ON, ChB OFF.
Level change is about 0.14%.
Interesting that auto-zeroing is not perfect when 2ch ON but gets basically perfect with only 1 left ON. Maybe it actually switches other channels OFF for auto-zero. Or it is just lucky strike by drift.



Some time ago directly compared DMMs and scopes:
https://www.eevblog.com/forum/testgear/cheap-chinese-ad584-voltage-reference-legit-cal-data-let's-find-out!/msg1154930/#msg1154930

Scopes can be fairly accurate at DC when pay little attention to zeroing etc.