Rigol DS1054Z persistance in stop/wait mode

[Prehistoricman]

Today was my first encounter with a 1054Z and I was just scoping some UART and came across this annoying bug (it must be).
Below are a couple of YT vids showing what I experienced: the persistence kinda analog scope effect doesn't disappear when you stop the scope. This also happens in normal trigger (somewhat to be expected, maybe). As soon as you cause a redraw to the screen, it shows the trace that was actually captured. In the two videos, I cause the redraw by changing the display type and by shifting the horizontal axis.

I couldn't find an option to turn this off. I can't believe that the persistence menu's lowest option is "minimum". Surely no persistence is as low as one can get i.e. "none"! I've used Keysights and even an older Rigol that didn't behave like this.

Firmware was 04.04

[TurboTom]

That's completely normal: When pressing "Stop", the scope just freezes the screen, i.e. the superposition of traces of the last screen scan/persistance period. Changing parameters or doing (certain) measurements on a stopped screen will revert it to the last trace. If you want a single trace on the screen right from the start, you can use the single trigger function.

Being able to freeze the screen can be quite beneficial when doing screenshots or measurements on superimposed traces. You will understand what i mean when you gain some more experience with the scope.

Cheers,
Thomas

[Prehistoricman]

I can definitely see the purpose of being able to see that statically but I'd really hope there's a way to turn it off. There's no reason to see this on data signals.

[RoGeorge]

It's definitely a bug.

Not only that it acts other than expected, but also the comportment is not consistent.

With trigger set to "Normal" and persistence set to "Min", if the trigger condition disappears (so no more traces to draw) , then sometimes the screen will remember all the sweeps overlapped (like a photo taken with a camera from an analog oscilloscope), sometimes the screen will remember only one trace (like when the SINGLE button was pressed).

[David Hess]

I have noticed it before in screenshots and video and always assumed it was just because of how the index graded display is produced and not a bug as such.  In run mode, acquisition records are combined to make an index graded display similar to an analog oscilloscope but when stopped or in single shot mode, there is only one acquisition to display.  Something like this has to be done if acquisition rates which are faster than the display refresh rate are to be made visible.

This also explains why display measurements change when the display switches modes.

[Fungus]

The answer is written on the buttons: Just press "SINGLE" instead of "STOP" when you want a single trace.

STOP mode is meant to freeze the screen "as-is". If it didn't do this (eg. if the persistence vanished even though it was set to "20s") then people would probably complain that that was a bug.

The only thing that calls your attention here is that the persistence vanishes when you move the wave around but the 'scope hardware simply isn't capable of storing/recreating it as you change things, eg. what if you scrolled off to the left then came back again, is it supposed to recreate the previous persistence from thin air?

[TurboTom]

I wouldn't consider it a bug but interpretations of that behaviour may differ. At times, the way it works right now may be quite useful as I pointed out before. I guess I never worried much about that since as I understand it, this behaviour is closely related with the way a DSO works internally.

IMO, the manufacturer has three options:

- do it like Rigol did on the DS1000Z (and all their other "mature" DS/MSO sereies scopes - for the recent models I can't tell due to lack of experience) and just stop the acquisition the moment "Stop" is pressed and display what's in the (persistance) buffer at that time. This may result in the mentioned non-reproducible results but I'ld consider it the most "honest" approach.

- discard all the persistance / overlay data of the previous "raw" buffer scans and just use the last sampled trace before "Stop" was pressed. This means losing the intensity grading and is more or less implemented as the "Single Shot" trigger mode. So not much would be gained against this mode.

- After "Stop" was pressed, keep on acquiring until the persitance/overlay buffer is full to be able to provide always a fully "intensity graded" screen. Behaviour would be consistent but the scope wouldn't stop exactly at the moment intended by the user. Maybe not an issue if initiated by a manual pushbutton but if a scope had a "gated" stop mode, things may get relevant.

Other strategies can be considered like a circular raw buffer of which the screen / persitance buffer only uses the last section to represent the image. But this may require too much interaction between the sampling engine and the persistance/display section and may be too complex on an entry level scope.

Anyway, I can pretty well live with the situation as it is right now.

Cheers,
Thomas

[Fungus]

I wouldn't consider it a bug but interpretations of that behaviour may differ. At times, the way it works right now may be quite useful as I pointed out before.

I think we can agree that freezing "as-is" is the correct thing to do. You might have several seconds (or even minutes) of data accumulated on screen that you don't want to lose. If all the persistence disappeared when you press stop then people would definitely be crying "bug!" for that.

After that I really don't see much choice but to make it disappear when you change the parameters. What if you zoom in? Is it supposed to interpolate the persistence to all zoom levels then recreate it perfectly when you zoom out again? It simply can't be done.

I don't see a more sensible choice than to do it the way Rigol is doing it at the moment.

(ie. Press "STOP" to freeze the screen as-is, press "SINGLE" to grab one waveform and show that)

I've used Keysights and even an older Rigol that didn't behave like this.

What do they do, exactly?

[RoGeorge]

I am not talking about pressing the STOP or the SINGLE button.  The screen is left in a limbo state when suddenly there are no more triggering events, and the WAIT icon in the upper right corner is flashing green.

When the trigger condition disappears, while waiting for more trigger events to come, from a screen full of traces you are suddenly left with
- sometimes a lot of traces drawn on the screen
- sometimes a few less traces drawn
- sometimes only one trace on the display, .

If it were to be a circular buffer with the last n traces mixed on the display, you won't see a single trace, you will always see a mix of the last n traces.  But this is not the case.


Steps to reproduce:
- generate a 5Vpp noise with a signal generator and put it on CH1
- set the trigger to "Normal", "Edge", "Rising", "500mV"
- set the acquire Mode to "Normal", Mem Depth to "Auto"
- timebase to 5ns/div

At this moment you should see a live image of the noise (note the triggered T'D green icon in the upper left corner of the display), looking like the first attached picture.



Let it run for a few seconds, then turn off the signal from the signal generator, so the oscilloscope will find no more trigger events.  Do not touch the oscilloscope.

The "WAIT" icon will now appear in the upper left corner and will start to flash green, indicating the oscilloscope is waiting for the next trigger event.  At this point the display is frozen.


What would you prefer to see on the display at this moment?

Some will say they expect the image above, with the many traces overlapped frozen on the display, some will say they prefer to see only one trace, maybe.  On an analog oscilloscope you will see a blank screen.

No matter your preference, if you repeatedly turn the generator on and off a few times (with a few seconds left in between), you will see the display is frozen in an inconsistent (random) state, sometimes with many traces mixed, sometimes with less traces, and sometimes with only one trace frozen, like in the last 3 pictures attached.

I'll say this is a bug in the UltraVision ring buffer that is supposed to average many traces into one displayed image.

What I'm afraid is this is not a bug, but Rigol just cut some corners to reduce computing power and memory requirements, so it will never be fixed.

I assume (my speculation only) that instead of a running average that will require a circular buffer to implement, they are simply adding the last trace to the displayed image (no circular buffer) then after a while clean the display and start again adding new traces on it.

[Fungus]

I'll say this is a bug in the UltraVision ring buffer that is supposed to average many traces into one displayed image.

I dunno, but I think that's unlikely to be how it's implemented in practice because the RAM requirements could be huge for long persistence times and the processing power needed to composite thousands of traces into a single image would also be unrealistic. People would also come on here and complain if update rates dropped towards zero as they increased the persistence time.

PS: How would "infinite" persistence work? Store every waveform ever then composite them all? 

What I'm afraid is this is not a bug, but Rigol just cut some corners to reduce computing power and memory requirements, so it will never be fixed.

A $400 'scope has limited computing power and memory, what are they supposed to do?

$100,000 'scopes also have limits, I bet you can find them if you look for hard enough.

I assume (my speculation only) that instead of a running average that will require a circular buffer to implement, they are simply adding the last trace to the displayed image (no circular buffer) then after a while clean the display and start again adding new traces on it.

This is much more likely as it requires a fixed amount of RAM and processing power for all possible persistence times and screen update rates.

(It wouldn't clean the display though, it would fade the pixels away based on elapsed time).

[RoGeorge]

Infinite persistence and

PS: How would "infinite" persistence work? Store every waveform ever then composite them all? 

Infinite persistence should work like it is working now:  a logical OR for each pixel to compose a background image (of uniform intensity, dimmed) + last trace.  Infinite persistence is the only one working according to intuition.   

All the other persistence duration are made with the same idea of an OR-ed dimmed background + an increasing number of traces, from one trace up to a number of traces, then clean all traces and start over again.

It seems UltraVision is done in hardware, http://www.saelig.com/supplier/Rigol/ultra_vision_070214p.pdf , so a ring buffer of 64x128 bytes and a complicated algorithm to manage all kind of corner cases might be too expensive.

As I wrote before, might not be a bug, might have been designed that way.  At this moment I am pretty sure it was designed to be this way (no ring buffer), so not a bug.  It can be bamboozling with intermittent signals, especially with digital ones, but I can live with it.   

The only thing that still bothers me a little is that they are still doing the fake persistence thing when "Persis. Time" is set to "Min".  Would have been nice to have a "Zero" persistence setting (or disable UltraVision).

[Prehistoricman]

I've used Keysights and even an older Rigol that didn't behave like this.

What do they do, exactly?

I believe they have an option to turn persistence off. I don't want my digital scope to try to function like an analog one for digital signals.

The answer is written on the buttons: Just press "SINGLE" instead of "STOP" when you want a single trace.

STOP mode is meant to freeze the screen "as-is". If it didn't do this (eg. if the persistence vanished even though it was set to "20s") then people would probably complain that that was a bug.

I agree. Of course persistence should do as it says on the tin: persist. What I would like is a strange type of persistence called "none".
I'll have a go pressing single and see if that works. It's intuitive that it should behave like that.

No matter your preference, if you repeatedly turn the generator on and off a few times (with a few seconds left in between), you will see the display is frozen in an inconsistent (random) state, sometimes with many traces mixed, sometimes with less traces, and sometimes with only one trace frozen, like in the last 3 pictures attached.

I'll say this is a bug in the UltraVision ring buffer that is supposed to average many traces into one displayed image.

What I'm afraid is this is not a bug, but Rigol just cut some corners to reduce computing power and memory requirements, so it will never be fixed.

I assume (my speculation only) that instead of a running average that will require a circular buffer to implement, they are simply adding the last trace to the displayed image (no circular buffer) then after a while clean the display and start again adding new traces on it.

I like the sound of this analysis. I also found the appearance of these persistence artifacts to be temperamental.

[Fungus]

I believe they have an option to turn persistence off. I don't want my digital scope to try to function like an analog one for digital signals.

Just set it to "minimum", you'll hardly see it. 

[Prehistoricman]

I believe they have an option to turn persistence off. I don't want my digital scope to try to function like an analog one for digital signals.

Just set it to "minimum", you'll hardly see it. 

That's the problem! I can see it at minimum and it's not what I'd call minimum.
Thankfully I down own this scope. I'd be kinda annoyed if I had.

[TurboTom]

Sorry but I don't understand the point. The DS1000Z series' minimum persitance is probably the update period of the screen. So if the scope is running (i.e. triggering continuously), you will see approximately the same that would be visible on any good old analog scope's screen: A superposition of waveforms, changing much faster than the eye can resolve. Press stop, and the image of the screen gets frozen, similar to a CRT scope with a storage tube, triggering continuously.

If you need a single shot of your signal, just use the corresponding trigger mode and press "Single" - and voila, no superposition. You may even record multiple "single shots" with the segmented memory function, though decoding-wise you're limited here on the DS1000Z in contrary to the larger DS/MSO2000 and 4000.

It's all a question of how you set up your scope. I'm not aware of an entry level DSO that works considerably different.

[David Hess]

The only thing that calls your attention here is that the persistence vanishes when you move the wave around but the 'scope hardware simply isn't capable of storing/recreating it as you change things, eg. what if you scrolled off to the left then came back again, is it supposed to recreate the previous persistence from thin air?

Not all DSOs discard the display record or the information necessary to recreate it as needed.

After that I really don't see much choice but to make it disappear when you change the parameters. What if you zoom in? Is it supposed to interpolate the persistence to all zoom levels then recreate it perfectly when you zoom out again? It simply can't be done.

Sure it can be done; it is just that Rigol does not support it.  Oscilloscopes operating with segmented memory can do it.  True DPO oscilloscopes do it.

But the sacrifice is that it requires a more memory and processing power or a much smaller record length.  The later is fatal to marketing.  Advertising record length size is easy but how do you advertise consistent measurements and a consistent display?

Personally I consider this behavior a defect.  None of my DSOs or the DSOs I have used suffer from it and I would not put up with it by choice.  Sure, it is making the best of a bad situation but marketing created the situation. (1)

(1) This is my tactful way of saying Rigol is a bunch of big fat lairs.

[Fungus]

Sure it can be done; it is just that Rigol does not support it.

You have a 'scope that can store (eg.) 10s of persistence then scroll it off screen, zoom in/out, etc. without losing a single pixel? Which one?

Personally I consider this behavior a defect.  None of my DSOs or the DSOs I have used suffer from it and I would not put up with it by choice.

How much did they cost?

[David Hess]

Sure it can be done; it is just that Rigol does not support it.

You have a 'scope that can store (eg.) 10s of persistence then scroll it off screen, zoom in/out, etc. without losing a single pixel? Which one?

I do not have one in the sense that you are asking but I have used them.  Is there any doubt about how the Rigol works?

What does duration have to do with anything?  Aren't we discussing the difference between the acquisition or processing record (these may or may not be the same) and the display record and preserving the display record?

I gave examples of what to look for in a DSO which work this way.  DPO operation and segmented memory are good indicators.  I'd like to know how the HP "Megazoom" DSOs behave.  I have watched videos of LeCroy DSOs doing it.

Personally I consider this behavior a defect.  None of my DSOs or the DSOs I have used suffer from it and I would not put up with it by choice.

How much did they cost?

Are you being rhetorical?

The DSOs I own are all old and were bought used.  The last new DSO I was ready to buy was from Rigol and then I found problems like these during evaluation.

[Fungus]

I have watched videos of LeCroy DSOs doing it.

Yes, but most people will only ever see a LeCroy in videos. They're not for ordinary people.

[David Hess]

I have watched videos of LeCroy DSOs doing it.

Yes, but most people will only ever see a LeCroy in videos. They're not for ordinary people.

I knew it; LeCroy users are aliens!  It explains so much.