Channel cross-talk issue on Rigol 1054Z ?

[Earendil]

I'm still not seeing how the previous probing method could affect Ch1 when Ch2 is connected. Could that be due to the return current traveling down the ground of the second probe and reflecting odd the end?

You add a parallel capacitance to the ground by connecting the probe on channel 2. I think this in itself can explain the effect.

That would (maybe) explain why connecting/disconnecting the probe makes a difference but it wouldn't explain why simply turning channel 2 on/off makes such a big difference.

Yes. I was referring to connecting/disconnecting the probe.

I can't reproduce the effect when he turns ch2 on/off. I'm not sure what's behind that.

I can now reproduce turning ch2 on/off effect too. I think turning ch2 on/off changes the waveform slightly (due to the added parasitics) and this causes a triggering issue. If I move the trigger level then this doesn't happen.

[Earendil]

I'm still not seeing how the previous probing method could affect Ch1 when Ch2 is connected. Could that be due to the return current traveling down the ground of the second probe and reflecting odd the end?

You add a parallel capacitance to the ground by connecting the probe on channel 2. I think this in itself can explain the effect.

That would (maybe) explain why connecting/disconnecting the probe makes a difference but it wouldn't explain why simply turning channel 2 on/off makes such a big difference.

Yes. I was referring to connecting/disconnecting the probe.

I can't reproduce the effect when he turns ch2 on/off. I'm not sure what's behind that.

I can now reproduce turning ch2 on/off effect too. I think turning ch2 on/off changes the waveform slightly (due to the added parasitics) and this causes a triggering issue. If I move the trigger level then this doesn't happen.

This is unlikely to be caused by aliasing. I used 400 kHz square wave to test this. By changing the timebase from 500ns to 1us the sampling rate also changed from 1GS/s to 500MS/s. Yet this didn't fix the triggering issue. However by turning ch2 on the signal triggers stable. So I'm pretty sure it's caused by parasitics.

[rf-loop]

This is unlikely to be caused by aliasing. I used 400 kHz square wave to test this. By changing the timebase from 500ns to 1us the sampling rate also changed from 1GS/s to 500MS/s. Yet this didn't fix the triggering issue. However by turning ch2 on the signal triggers stable. So I'm pretty sure it's caused by parasitics.

First you tell that you think there is not aliasing, After then you "proof" it telling you use 400kHz square wave and changing scope samplerate from 1G to 500M. I'm very interested to listen more... about this study. Sorry but why set this strong comment where you try explain there is not aliasing and then you nearly tell use that you need first study basic fundametals about signals and also about sampling and aliasing.  Your square wave 1st harmonic frequency do not tell nearly anything here. I can push 500MSa/s aliasing using 1Hz square wave. Do you know how? Do we really start from basic school level fundamentals. After you are familiar with basic fundamentals of digital scopes and sampling and aliasing: Look these OP images and think!  After you are readu coma agen and tell us you do not see there aliasing. I see there aliasing when I look 0,5s this 500MSa/s image and signal shape. (I see there also some other natural "effects".



About 2 channel image on and trigger.
This can not proof if there is not time jitter in signal or trigger jumps (I suspect last one because... look trigger position and signal shape it match well to time shift with these some acquisitions including knowledge that Rigol trigger performance is <censored>.

Of course IF signal have some high frequency parasitics example far over Oscilloscope nominal BW but nerly Nyquist freq or higher they can produce or sure produce also visible aliasing but still nearly impossible to see aliasing signal itself and aliasing may be sign what tell to user that there is "something" what need inspect more deep. And this need knowledge and experience so that understand what possible things can detect from scope screen.

In OP's images there can see absolutely sure aliasing. If not, then I'm santaclaus.. What cause aliasing is other question. But there is also things so that all differencies between images are not at all only aliasing based.

[Simon]

First you tell that you think there is not aliasing, After then you "proof" it telling you use 400kHz square wave and changing scope samplerate from 1G to 500M. I'm very interested to listen more... about this study. Sorry but why set this strong comment where you try explain there is not aliasing and then you nearly tell use that you need first study basic fundametals about signals and also about sampling and aliasing.  Your square wave 1st harmonic frequency do not tell nearly anything here. I can push 500MSa/s aliasing using 1Hz square wave. Do you know how? Do we really start from basic school level fundamentals. After you are familiar with basic fundamentals of digital scopes and sampling and aliasing: Look these OP images and think!  After you are readu coma agen and tell us you do not see there aliasing. I see there aliasing when I look 0,5s this 500MSa/s image and signal shape. (I see there also some other natural "effects".

Calm down a bit. I gather english is not your first language but you are coming across a little agressive.

[Earendil]

First you tell that you think there is not aliasing, After then you "proof" it telling you use 400kHz square wave and changing scope samplerate from 1G to 500M. I'm very interested to listen more... about this study. Sorry but why set this strong comment where you try explain there is not aliasing and then you nearly tell use that you need first study basic fundametals about signals and also about sampling and aliasing.  Your square wave 1st harmonic frequency do not tell nearly anything here. I can push 500MSa/s aliasing using 1Hz square wave. Do you know how?

I'm not sure what's your problem but by attacking me personally instead of what I'm saying surely isn't helpful and not a very constructive approach.
About aliasing: Normally I don't suspect aliasing if my signal's frequency content is much lower than the sampling rate. Even if I consider the 5th harmonic of a 400 Khz square wave that's still only 2 Mhz which is much lower than 500MS/s sampling rate. I set the memory depth to 12K that's why the sampling rate changes at 1us. I wrote the message very late in the night (and it's much too early for me now) so I might missed something. Still, I didn't expect to be lectured in such a harsh way.
Also if you're so knowledgeable than instead of criticizing me personally you should explain how a 1 Hz square wave sampled at 500MS/s can cause aliasing. I admit that I haven't heard about such thing. I suspect that I'm not the only one though. Even the Keysight document's figure 4 you linked only mentions aliasing happens when the 2nd, 3rd, etc. harmonics are over the Nyquist frequency. That is not the case here.

[rf-loop]

First you tell that you think there is not aliasing, After then you "proof" it telling you use 400kHz square wave and changing scope samplerate from 1G to 500M. I'm very interested to listen more... about this study. Sorry but why set this strong comment where you try explain there is not aliasing and then you nearly tell use that you need first study basic fundametals about signals and also about sampling and aliasing.  Your square wave 1st harmonic frequency do not tell nearly anything here. I can push 500MSa/s aliasing using 1Hz square wave. Do you know how?

I'm not sure what's your problem but by attacking me personally instead of what I'm saying surely isn't helpful and not a very constructive approach.
About aliasing: Normally I don't suspect aliasing if my signal's frequency content is much lower than the sampling rate. Even if I consider the 5th harmonic of a 400 Khz square wave that's still only 2 Mhz which is much lower than 500MS/s sampling rate. I set the memory depth to 12K that's why the sampling rate changes at 1us. I wrote the message very late in the night (and it's much too early for me now) so I might missed something. Still, I didn't expect to be lectured in such a harsh way.
Also if you're so knowledgeable than instead of criticizing me personally you should explain how a 1 Hz square wave sampled at 500MS/s can cause aliasing. I admit that I haven't heard about such thing. I suspect that I'm not the only one though. Even the Keysight document's figure 4 you linked only mentions aliasing happens when the 2nd, 3rd, etc. harmonics are over the Nyquist frequency. That is not the case here.

I do not try attack person at all but if I see there is reason to suspect you do not understand theory I say it. I do not understand at all why you take 400kHz and then talk some few harmonics about it when OP signal is really totally different where exist far far over 100MHz components what leads to aliasing when samplerate drops from 1G to 500M.
If still something unclear I can demonstrate it with other scopes and known signal. 
 
Did you not note at all original message images?  OP image tell where he is.
And, there is aliasing due to low samplerate related to signal frequency components.
If need I can demonstrate this aliasing with many different scopes. Math is math and when we have right theory it match also with practice. If practice do not match theory, then we have wrong theory.

I try agen with OP image where 2 channels in in use. Look carefully signal changes related to time and used sampling interval (marked also in image)

There may be what ever reason in signal, parasitic oscillations and what ever but it do not violate any kind of fact when I say that there is digital oscilloscope aliasing visible and really clearly and without any doupt. Please look image and calculate / estimate your self.

[Earendil]

I can push 500MSa/s aliasing using 1Hz square wave. Do you know how? Do we really start from basic school level fundamentals. After you are familiar with basic fundamentals of digital scopes and sampling and aliasing: Look these OP images and think!  After you are readu coma agen and tell us you do not see there aliasing.

I do not try attack person at all but if I see there is reason to suspect you do not understand theory I say it.

I wonder what happens if do want to attack my person.

I do not understand at all why you take 400kHz and then talk some few harmonics about it

Because as I mentioned I reproduced the same issue using a signal generator.

If need I can demonstrate this aliasing with many different scopes. Math is math and when we have right theory it match also with practice. If practice do not match theory, then we have wrong theory.

You've just claimed that I don't understand theory. And now you say that theory is wrong? Maybe this theory only exists in your own little mind?

There may be what ever reason in signal, parasitic oscillations and what ever but it do not violate any kind of fact when I say that there is digital oscilloscope aliasing visible and really clearly and without any doupt.

I wrote "This is unlikely to be caused by aliasing" referring to waveform changing on the scope by turning ch2 on/off.
I never claimed that there's no aliasing (corner wobbling) visible on the OP's scope.

I can push 500MSa/s aliasing using 1Hz square wave. Do you know how?

You still haven't explained how this is possible. Since you claim that I don't understand sampling theory please do explain. If you can't then I'll need to conclude that it is you who don't understand sampling theory.

[Simon]

Please can we stop this before I lock the topic.

[rf-loop]

Here is other example about aliasing.

Here ~400kHz freq.

In all images signal parameters same.
I forget one image. If I adjust signal risetime (harmonics) less fast of course aliasing disappear, of course.
Here signal itself measured risetime is somewhere around near 1ns but here transmission and scope input destroy it and also produce some ringing due to impedance mismatch. (Scope input 50ohm resistance is not same as impedance)  (HP specification for risetime is = or < 1.4ns  (in this gen, rise and fall time is also adjustable))

500MSa/s image there is aliasing and it is (in this case "corners wobbling") visible very clearly (compare this with Rigol display  there is somehow similar visible but of course different due to different signal and due to poor scope front end and fake Sinc function and also digfferent more poor interpolation between sample points for positioning AND we do not know what all this unknown signal include what we can not see directly in OP Rigol scope image)

And if you think this theory is only in my mind. Please read this previously linked HP/agilent/Keyshit  paper and find it there.

1GSa/s     nearly no aliasing visible
2GSa/s     no aliasing visible.

[EI6JA]

This is unlikely to be caused by aliasing. I used 400 kHz square wave to test this. By changing the timebase from 500ns to 1us the sampling rate also changed from 1GS/s to 500MS/s.

What is the rise time of your 400 kHz square wave. In this case we are looking at aliasing around the trigger point of our signal. If your 400kHz signal had a very good rise time its bandwidth would be much higher than you expect. For example a 1 nS risetime would have a bandwidth of 350 MHz using the formula Rise time = 0.35 / Bandwidth (assuming Gaussian response), in this case a 500 MS/s would result in aliasing of the rising edge.
The 400 kHz square wave would not suffer from aliasing as it is well bellow Nyquist sample rate but the rising edge being aliased by the trigger circuit may be a different story.

[Earendil]

This is unlikely to be caused by aliasing. I used 400 kHz square wave to test this. By changing the timebase from 500ns to 1us the sampling rate also changed from 1GS/s to 500MS/s.

What is the rise time of your 400 kHz square wave. In this case we are looking at aliasing around the trigger point of our signal. If your 400kHz signal had a very good rise time its bandwidth would be much higher than you expect. For example a 1 nS risetime would have a bandwidth of 350 MHz using the formula Rise time = 0.35 / Bandwidth (assuming Gaussian response), in this case a 500 MS/s would result in aliasing of the rising edge.
The 400 kHz square wave would not suffer from aliasing as it is well bellow Nyquist sample rate but the rising edge being aliased by the trigger circuit may be a different story.

I expected that explanation from rf-loop. You're right of course. I generate the signal using a Siglent SGD1025. I've measured the rise time using a 500Mhz real time scope. It's about 7.5ns. (The coax -3db point is about 1.5Ghz). So assuming Gaussian response that's about 46.6 Mhz.

[rf-loop]

Now also this need keep in mind. (bit sidestep but it is also so near in this particular case what OP show)
In many scopes Sinc can not turn off at all or just in very limuted cases. (what heck these scope designers have thinked). Including this Rigol model with these settings, Sinc forced on.. (more far sidejump: Also it need know that different scopes Sinc fidelity is different)

Because OP's Rigol use Sin(x)/x  imitation I use also Sin(x)/x ON in this example image with semi fast edge pulses.
(also OP image have quite fast details related to sample rate)

With Sinc need be some times careful and  also keep it in mind specially if can not turn it off and look what is visible with lines (Sinc OFF)

Here what old Tek basics tell (about Sinc with "undersampling"):

[Earendil]

Here is other example about aliasing.

Here ~400kHz freq.

In all images signal parameters same.
I forget one image. If I adjust signal risetime (harmonics) less fast of course aliasing disappear, of course.
Here signal itself measured risetime is somewhere around near 1ns but here transmission and scope input destroy it and also produce some ringing due to impedance mismatch. (Scope input 50ohm resistance is not same as impedance)  (HP specification for risetime is = or < 1.4ns  (in this gen, rise and fall time is also adjustable))

500MSa/s image there is aliasing and it is (in this case "corners wobbling") visible very clearly (compare this with Rigol display  there is somehow similar visible but of course different due to different signal and due to poor scope front end and fake Sinc function and also digfferent more poor interpolation between sample points for positioning AND we do not know what all this unknown signal include what we can not see directly in OP Rigol scope image)

And if you think this theory is only in my mind. Please read this previously linked HP/agilent/Keyshit  paper and find it there.

1GSa/s     nearly no aliasing visible
2GSa/s     no aliasing visible.

Fine and dandy. But as I said previously I never questioned the existence of corner wobbling.
What I'm saying is that in OP's case the effect is caused by ground parasitics and (very probably) trigger instability. I can set the trigger level in such way that connecting the probe on ch2 causes trigger instability. Also turning ch2 on/off can cause trigger instability. If I move the trigger level around than the scope triggers properly with no corner wobbling visible. Also changing the timebase from 500ns to 1us (with memory set to 12K) doesn't impact the waveform shape or triggering.

Why don't you just reproduce the setup and report back what you see? It's not that difficult and would be far more productive.

[rf-loop]

Why don't you just reproduce the setup and report back what you see? It's not that difficult and would be far more productive.

Because I do not have OP's signal available and also today I have not anymore Rigol 1000Z (what I know also well because I have owned one and used it)

But, what I see OP image and as far as I know, where ever he set trigger level there is aliasing due to low samplerate related to signal details. This is my opinion and based also to knowledge and lot of lab experience. I stay behind my words. If I get original OP's signal and this oscilloscope I proof it. Without it, I keep my opinion based to theory and mu experience.

Of course there can exist also other reasons and so on as you explain and I can well accept also these may happend and even exist in this case also. This do not disable clear aliasing what there is sure with 500MSa.

Trigger instability is other case. (I mean these visible acquisitions what are horizontally different location (randomly trigged) This may be in singnal under test but because in image it can clearly see what is in signal just around trigger level my suspect is that this disappear with small change in trig level.

But then other case.
Topic name is: Channel cross-talk issue on Rigol 1054Z ?   

(Rigol can here forget because it exist in all scope what I know, more or less.)

Attached: This is oscilloscope channels cross talk.
(look V/div scales ratio (1:500)  and estimate cross talk ratio)
CH3 Signal is 5V ~5ns pulses. CH4 open with BNC cap for reduce external EMI.

[Earendil]

Okay, I reproduced the setup and made some pictures. Hopefully this will help.

Here's the setup:



25 Mhz, 3Vpp, square wave (load set: 50 ohm). Rise time about 7ns - 7.5ns (as measured previously).
The probe visible on the picture is connected to Ch1.

Scope setup:
Ch1 and Ch2 AC coupled. Trigger: DC coupled, falling edge.

First I only enable Ch1:



If I enable Ch2 too then trigger is unstable (trigger level is the same as previously):



If I move the trigger level then it triggers stable:



Also triggers stable with only Ch1 enabled:

[rf-loop]

Okay, I reproduced the setup and made some pictures. Hopefully this will help.

Here's the setup:



25 Mhz, 3Vpp, square wave (load set: 50 ohm). Rise time about 7ns - 7.5ns (as measured previously).
The probe visible on the picture is connected to Ch1.

Scope setup:
Ch1 and Ch2 AC coupled. Trigger: DC coupled, falling edge.

First I only enable Ch1:



If I enable Ch2 too then trigger is unstable (trigger level is the same as previously):



If I move the trigger level then it triggers stable:



Also triggers stable with only Ch1 enabled:

Signal shape ok and without signs about aliasing. This is also expected with using SDG1025 (we know there is not enough hf components in signal for aliasing with  500MSa/s.)

Trigger: as can expect. Normal. (this same situation exist also in OP image)

(also as can see and estimate your signal do not have so high f components as OP signal but as we know there is also Rigol bit weird "sinc" imitation doing its own effects. (and of course trigger is not derived from this signal image what you see in display)

[Mechatrommer]

just my 2cnts observing the OP and discussion around here. this is not 1 problem issue. it consists of several problems...

1) poor probing technique, long uncoupled loop (long gnd wire, and long signal wire), a slight change to the setup will change the probe impedance, loading and ringing characteristic of the hi-freq elements of the signal...
2) crosstalk between dso channels. as demonstrated by the OP in 0:53. and discussed before... https://www.eevblog.com/forum/testgear/ds1054z-issue/msg886447/#msg886447 in that thread there is "soft crosstalk" so i guess there must be some hardware crosstalk resulting the effect... this crosstalk will change signal loading esp the hi freq harmonics.
3) dso trigger threshold. ds1000z is known for its weakness in this regards... https://www.eevblog.com/forum/testgear/rigol-ds1054z-unable-to-trigger-cal-signal-at-lt-0-6v/msg855558/#msg855558
4) and last but not least, the so much disputed "sample aliasing" resulting in different sinc output seen on the screen...

all of these and probably more, contribute to the effect... my 2cnts...

[Earendil]

just my 2cnts observing the OP and discussion around here. this is not 1 problem issue. it consists of several problems...

1) poor probing technique, long uncoupled loop (long gnd wire, and long signal wire), a slight change to the setup will change the probe impedance, loading and ringing characteristic of the hi-freq elements of the signal...
2) crosstalk between dso channels. as demonstrated by the OP in 0:53. and discussed before... https://www.eevblog.com/forum/testgear/ds1054z-issue/msg886447/#msg886447 in that thread there is "soft crosstalk" so i guess there must be some hardware crosstalk resulting the effect... this crosstalk will change signal loading esp the hi freq harmonics.
3) dso trigger threshold. ds1000z is known for its weakness in this regards... https://www.eevblog.com/forum/testgear/rigol-ds1054z-unable-to-trigger-cal-signal-at-lt-0-6v/msg855558/#msg855558
4) and last but not least, the so much disputed "sample aliasing" resulting in different sinc output seen on the screen...

all of these and probably more, contribute to the effect... my 2cnts...

Let's not start this again from the beginning, please. Read the thread more carefully.

[mrnuke]

Thank you everyone for your help. This was a very interesting case where a bunch of factors contributed to what seemed like a truly odd phenomenon. I just happened to have my trigger set to a level where even very slight variations in the signal would cause dramatic effects. Combine that with the sub-optimal probing method I was initially using, and I was asking for trouble.

I've learned a lot from this thread, and it has helped me continue troubleshooting the 3325A I showed in the original video. Thank you all!