DS1000Z periodic "noise"

[c4757p]

Does everyone else have such terrible periodic interference on their DS1000Z? Random I can understand, but this is periodic - clearly synchronous to the acquisition clock, as it displays nicely with the trigger unlocked (as shown in the screenshot). Makes 1mV/div a bit of a joke, IMO.

[Fungus]

I wouldn't have said 2mV of noise at 200MHz was anything to worry about.

Makes 1mV/div a bit of a joke, IMO.

Seriously? What do you expect from a $400 'scope with four channels?

[c4757p]

Hey, I didn't say the scope was a joke, just that particular range. Every scope I've ever had has had at least one feature that was made somewhat useless by its drawbacks...

[Fungus]

Hey, I didn't say the scope was a joke, just that particular range. Every scope I've ever had has had at least one feature that was made somewhat useless by its drawbacks...

1mV of noise isn't really a drawback of the scope, it's a problem with the physical universe.

Have a word with the designer of that...

[c4757p]

Tell that to my other DSO

I'm just wondering if that's consistent with other people's DS1kZ units, nobody needs to complain at me for bashing them...

[MarkL]

Looks like interleave offset in the ADC (HMCAD1511).

Hard to know if this is within spec or not since we don't know what the full scale of the ADC input is in relation to what's being shown on the screen.

[Fungus]

Tell that to my other DSO

I bet it cost more money.

I'm just wondering if that's consistent with other people's DS1kZ units, nobody needs to complain at me for bashing them...

I don't have mine handy but I'd be surprised if it didn't have a couple of mV of noise.

[c4757p]

Tell that to my other DSO

I bet it cost more money.

Indeed, hence why it's not just a "problem with the physical universe", and why I'm curious about it...

[miguelvp]

I count 15 cycles in 6 divisions at 20ns per division.

That will make that signal 125MHz on the dot.

[c4757p]

Yeah, so half the sample rate. Looks like each sample is alternately high or low. Interesting - that would make me think it was inherent in the ADC itself - maybe a bit of clock coupling to the data - but that would make it the same amplitude on different V/div settings, or at least not a direct function of the V/div. This behaves as if it's coupled in before the gain control; it scales with the signal.

[JohnPen]

I get similar results on my scope. However my CH 1 appears to be about half the others noise level.  I notice C4757p's display shows CH 2 better.  Incidentally switching channels off seems to alter the noise level on those still being displayed not that it makes a lot of difference.  Perfection in the Physical world is hard to achieve and is progressively more expensive!

[Maxlor]

My scope seems to behave similar to yours, maybe slightly better.

The second image is with the probe ground clip connected to the probe tip for reference. I'm not worried about the GND noise

[alsetalokin4017]

Pretty much the same here, except that it's my CH3 that is the least noisy.

Also I noticed that with only CH1 on, there is no change in the amplitude of the displayed noise when going from 10 mV/div to 20 mV/div.  With other channels on the displayed amplitude changes as expected.

[fivefish]

Here's my result. High Res Mode, 6M Mem Depth

AC Coupling


GND Coupling

[Marcos]

Same here too.
Quick question: is freeze bug present too?

[Fungus]

Tell that to my other DSO

I bet it cost more money.

Indeed, hence why it's not just a "problem with the physical universe", and why I'm curious about it...

Where does noise come from?

[alsetalokin4017]

Here's my result. High Res Mode, 6M Mem Depth

AC Coupling


GND Coupling

That's great... but what do you see when you actually use _the same settings_ as the OP?   

[alsetalokin4017]

Where does noise come from?

My next-door-neighbor's car stereos, usually.     

[c4757p]

That's great... but what do you see when you actually use _the same settings_ as the OP?   

An interesting point, because his screenshots look better than mine (the noise has about the same amplitude, but it doesn't look as periodic) - mine had looked about the same on almost all the settings I tried, but I hadn't tried that particular configuration, so I did, and mine looked better too.

After a bit of experimenting, I realized I completely overlooked something - whether or not I was looking at an intensity-graded combination of multiple acquisitions. Single-shot capture attached - it doesn't look periodic that way. The 'wobbles' in the signal have a similar spacing by appearance, but if you actually try to line up cursors with them, they're totally different frequencies.

The periodicity only becomes apparent when multiple acquisitions are involved. Second screenshot is the same exact thing, but a screen-freeze of a continuous capture instead of single-shot.

Iiiiinteresting. Something to do with the timing when overlaying multiple acquisitions?

[MarkL]

Yeah, so half the sample rate. Looks like each sample is alternately high or low. Interesting - that would make me think it was inherent in the ADC itself - maybe a bit of clock coupling to the data - but that would make it the same amplitude on different V/div settings, or at least not a direct function of the V/div. This behaves as if it's coupled in before the gain control; it scales with the signal.

Isn't the 1mV range a magnified version of 5mV?

  https://www.eevblog.com/forum/testgear/new-rigol-ds1054z-oscilloscope/msg752786/#msg752786

If so, a 1-bit (or so) offset might disappear when you get to 5mV.  I suppose you could dump the raw 8-bit channel data and figure it out from that.

What do you get with the same settings except only using 2 channels?  JohnPen tried it, but no screen shot.

As far as periodicity is concerned, a single shot capture may be mixing in noise with the offset, making it appear more random.  The scope is also probably applying some sin(x)/x, making it look even stranger.

(I can't say for sure it is interleave offset, just working on a theory...)

[vk6zgo]

To quote the excuse used by car dealers----"They all do it!"

[Mark_O]

Has anyone considered the fact that you're asking the scope to take 60 actual samples, and fill in the remaining 90% of the trace with interpolated data?  Thus turning 60 -> 600.  What assumptions should it use to try and fit such a 'curve'?

Does anyone happen to have sinx/x enabled?  If so, at a 250MHz sample rate, how surprising would it be for it to then fit random noise data to a nice, periodic 125MHz sine wave?  That just 'happens' to be synchronous to the sample clock?  And then have the integrative aspect of persistent mode "reveal" that underlying structure so clearly... that may be slightly less so on a simple one-shot acquisition.

So you're starting out with a low level of random noise, and under-sampling it in both frequency and amplitude.  Then asking the scope to interpret it, and make some sense out of it.  That's going to be rather difficult, when what you started out with was non-sense.     Of course, the scope will try it's best to do what you asked.  And leave the understanding up to the operator.

[c4757p]

I probably should have mentioned that I'm aware of this, and that the behavior is the same with interpolation changed to linear or just completely disabled (dot mode).

[Tim F]

Mine is about the same as everyone elses, however I found a related bug. If I used the autoset function with a 2mVpp 10MHz sinewave connected to one of the inputs, the noise becomes much worse. Upon enabling a second channel the noise improves. After disabling the second channel the noise on the first remains at about the same improved level. The thing is, I can't work out how to replicate the issue again - I was messing around the enabling/disabling channels, antialiasing and sinx/x and after rebooting the scope it doesn't do it anymore 

Auto set, 2mVpp 10MHz sine connected at CH1:


Channel 2 then enabled:


Channel 2 then disabled:

[c4757p]

Ooh, I have seen the same thing a couple times. And I also have no clue how to repro the bug. What a strange one!