Siglent SDS2000X Plus

[nctnico]

What features should that be that Siglent needs to add in your opinion, considering that Peak Detect and HiRes acquisition modes are apparently already offered in the SDS2kX+?

As I wrote before; two simple things: User selectable memory length settings and full memory decoding.

Reading this thread suggests that both is already implemented.

No it isn't. You need to enable zoom mode as a work-around but that eats away a large part of the vertical space and it needs more tweaking controls & knobs from the user.

Lots of buzz words right there, but funny enough it appears that the most loved (for its easy and quick operation) scope Keysight DSO-X lacks at least the first one (user selectable memory), but still manages to be the most widely sold bench scope Series in its class (the DSOX3k is as abundant as Athlete's Foot).

That is because Keysight uses full memory always but still manages to get high waveform update rates through a clever parallel acquisition system.

As to "Some of the engineers I have spoken to agree that Lecroy has some unique analysis features but if the competition would provide something similar they'd swap the Lecroy scopes in a heartbeat", you left out the part where I asked for the reason of why they would swap LeCroy scopes (which ones btw) "in a heartbeat", and I'd really like to hear why this is.

That didn't come up in the conversation and I didn't ask because that would have diverged from the main topic of the conversation too much. I wasn't there to discuss test equipment.

And you also forgot one major player in the scientific market: National Instruments. They have a very solid combination with their hardware and software.

Since when does NI make bench scopes?

They kind of do if you buy the modules to build a self contained system.

Wavepro 7300A

Looks like you bought the wrong scope for your needs.

No, I didn't. I bought it for high frequency and low trigger jitter measurements. I couldn't resist for $400 and after shipping ($350), fixing and fan upgrades I have around 850 euro in it. It was a deal I could not let pass by me.

Understandable, but being a good deal doesn't change what type of scope this is, nor should it have surprised you that it's not a good general purpose scope.

It didn't surprise me. I just noted that having an abundance of complex analysis features doesn't make for a good bench scope so Siglent shouldn't follow that example IMHO. So basically already writing what you wrote above.

[rf-loop]

...zoom mode as a work-around but that eats away a large part of the vertical space and it needs more tweaking controls & knobs from the user.

It eats vertical space if compare to non zoomed. But display is now bigger and more resolution than example in models what have today example 800x480 TFT and in previous some Siglent models wfm area height is 400pixel and in zoom mode main timebase and second timebase window both same height. This model have 1024 x 600 but now dual timebase display is 1/3 for main timebase and 2/3 for second (details) timebase.
I do not know exact pixel height about these but clearly much more than these some other Siglent models. (personally I do not at all understand this popularity and copycatting this TFT size fashion. These are test and measurements tools not entertainment displays fo movies etc. 4:3 form factor is much better  and perhaps for Siglent even so that long side is vertical. Why need follow others.

[thinkfat]

(personally I do not at all understand this popularity and copycatting this TFT size fashion. These are test and measurements tools not entertainment displays fo movies etc. 4:3 form factor is much better  and perhaps for Siglent even so that long side is vertical. Why need follow others.

Because this is the screen size that you can get cheap'ish now. It used to be 800x480, now it's 1024x600. In this low-cost market you need to penny-pinch if you want to make a profit. I guess this size and aspect ratio is common now for cheap Android or whatever tablets or other now-popular consumer gear so the Rigols and Siglents of the TE industry ride that wave. They don't sell enough to have screens custom made I guess.

[nctnico]

I don't think wide screens are bad. They do fit the form factor of an oscilloscope casing better and having more width is not a bad thing. After all you can fit more signal on the screen. Still a 4:3 screen offers more pixels compared to a wide screen with the same diagonal. Look at the MicSig TO1000 series for example. These have an 800x600 screen which offers 25% more pixels compared to 800x480.

[Martin72]

Could you add how many points were sampled, at which sample rate, for each of the timebase settings?
EDIT: also, would be interesting if the trigger setup has an influence on the update rate, too. Edge trigger is easy, but how about runts or other, more complex setups?

See below, the trigger thing I can test when second scope was here.

[Wuerstchenhund]

Lots of buzz words right there, but funny enough it appears that the most loved (for its easy and quick operation) scope Keysight DSO-X lacks at least the first one (user selectable memory), but still manages to be the most widely sold bench scope Series in its class (the DSOX3k is as abundant as Athlete's Foot).

That is because Keysight uses full memory always but still manages to get high waveform update rates through a clever parallel acquisition system.

No, it doesn't. There is no "clever parallel acquisition system".

During normal operation (i.e. continuous mode), MegaZoom scopes like the current InfiniVision Series only use a small part of the memory (enough to feed the display), and only uses the full memory size in the last acquisition prior to a stop or in Single acquisition. In addition, the whole memory is divided in two, where one half is used to store the data of the current acquisition cycle while the other half is used for display processing. The halves are swapped after each acquisition cycle which avoids the need for the acquisition system to "wait" for the processing to finish. Only when the acquisition is stopped and the processing of the previous cycle data has completed, the whole memory is then used for the last acquisition, which is the only time the full amount of memory is actually used.

In Single mode the scope always uses the full memory for that single acquisition, as there is no need to maintain high update rates.

That's all there is. No magic. No special sauce.

So no, MegaZoom really doesn't use all the memory (for acquisition) all the time.

ANd considering the complexity of the different memory allocations it does make sense that the user doesn't get any memory information whatsoever.

There is a very good description of MegaZoom by HP in the HP Jorunal article describing the 54645D (the first MSO):
https://www.hpl.hp.com/hpjournal/97apr/apr97a3.pdf

Also, MegaZoom is only really fast at slower sample rates (i.e. 5GSa/s), and with increasing sample rates the performance goes down. Which is sufficient for entry-level and lower midrange bench scopes like the DSO-X Series, but it's unable to deal with the high amount of data a high end scope (which samples at rates in excess of 20GSa/s) produces. But then, waveform rates are really only a thing in the lower end of the scope market, so there's that.

As to "Some of the engineers I have spoken to agree that Lecroy has some unique analysis features but if the competition would provide something similar they'd swap the Lecroy scopes in a heartbeat", you left out the part where I asked for the reason of why they would swap LeCroy scopes (which ones btw) "in a heartbeat", and I'd really like to hear why this is.

That didn't come up in the conversation and I didn't ask because that would have diverged from the main topic of the conversation too much. I wasn't there to discuss test equipment.

Fair enough, but in that context the reasoning would have been very interesting.

Since when does NI make bench scopes?

They kind of do if you buy the modules to build a self contained system.

Even if you put NI Digitizers in a NI PXI chassis and use NI LabView, it's still a Digitizer system where one has to write the software for it and not a general purpose bench scope like the SDS2kX+. Aside from costing an insane amount of money.

So I'd say let's just not go there

It didn't surprise me. I just noted that having an abundance of complex analysis features doesn't make for a good bench scope so Siglent shouldn't follow that example IMHO. So basically already writing what you wrote above.

I don't think any of the points which make the WavePro a poor general purpose scope apply to the SDS2kX+ - or any Siglent scope as of today.

[Martin72]

Would be interesting to understand, if, using Zoom, you can actually get the highest sample rate together with the highest memory depth?

Maybe I´ve misunderstood this, but I take the last timebase where 200mpt avaible and fed in a 1Mhz sine, then zooming in...

[thinkfat]

Could you add how many points were sampled, at which sample rate, for each of the timebase settings?
EDIT: also, would be interesting if the trigger setup has an influence on the update rate, too. Edge trigger is easy, but how about runts or other, more complex setups?

See below, the trigger thing I can test when second scope was here.

Thanks a lot for that. I understood now what the memory depth setting actually does. The scope always only stores sample points to fill one sweep (timebase x 10 x samplerate = total samples). This equation will always be satisfied. The memory depth setting allows you to limit the amount of memory. So instead of 200Mpts at the, say, 10ms/div timebase, if I choose a limit of 100Mpts, the sample rate will drop to 1GSa/s.

For such an option to make any sense the waveform update rate must be essentially bounded by the memory fill rate. It's a feature to increase the waveform update rate at the expense of temporal resolution.

[thinkfat]

Would be interesting to understand, if, using Zoom, you can actually get the highest sample rate together with the highest memory depth?

Maybe I´ve misunderstood this, but I take the last timebase where 200mpt avaible and fed in a 1Mhz sine, then zooming in...

Yes, that's what I was asking. The table you posted already answered the question. Using the zoom this way was the proposed alternative to having the scope obey the memory depth setting and using sample memory for off-screen samples. I was just curious if such a combination existed. In hindsight, why shouldn't it? Siglent aren't a bunch of idiots.

[Martin72]

Hi,
Test it with even 1sek, no Problem...
Next, 1Mhz in Dot-Mode, 20Mhz in Dot-Mode.

[nctnico]

Lots of buzz words right there, but funny enough it appears that the most loved (for its easy and quick operation) scope Keysight DSO-X lacks at least the first one (user selectable memory), but still manages to be the most widely sold bench scope Series in its class (the DSOX3k is as abundant as Athlete's Foot).

That is because Keysight uses full memory always but still manages to get high waveform update rates through a clever parallel acquisition system.

No, it doesn't. There is no "clever parallel acquisition system".

which avoids the need for the acquisition system to "wait" for the processing to finish. Only when the acquisition is stopped and the processing of the previous cycle data has completed, the whole memory is then used for the last acquisition, which is the only time the full amount of memory is actually used.

ANd considering the complexity of the different memory allocations it does make sense that the user doesn't get any memory information whatsoever.

Yes it does use all the memory from the user's point of view so what happens under the hood is just semantics. And there is definitely something clever going on. Think about the situation where a new trigger comes before the entire memory is filled in combination with needing to have pre-trigger data as well. The PDF you linked to is a simplified explaination for what is really going on under the hood. But for the user it doesn't really matter.

Since when does NI make bench scopes?

They kind of do if you buy the modules to build a self contained system.

Even if you put NI Digitizers in a NI PXI chassis and use NI LabView, it's still a Digitizer system where one has to write the software for it and not a general purpose bench scope like the SDS2kX+. Aside from costing an insane amount of money.

So I'd say let's just not go there

Well, you brought up scientific use cases and I disagreed Lecroy didn't have any competition in that market.

I don't think any of the points which make the WavePro a poor general purpose scope apply to the SDS2kX+ - or any Siglent scope as of today.

Currently the way the memory is managed. I was just curious on why (how Siglent got the idea) the SDS2000X doesn't allow to record outside the screen and much to my surprise (yes, that did surprise me) the Wavepro 7200A works in the same way. Since all other DSOs I have come across so far do allow recording outside the screen without needing work-arounds I take the liberty to assume that this is something Siglent copied from Lecroy.

[rf-loop]

Could you add how many points were sampled, at which sample rate, for each of the timebase settings?
EDIT: also, would be interesting if the trigger setup has an influence on the update rate, too. Edge trigger is easy, but how about runts or other, more complex setups?

See below, the trigger thing I can test when second scope was here.

Thanks a lot for that. I understood now what the memory depth setting actually does. The scope always only stores sample points to fill one sweep (timebase x 10 x samplerate = total samples). This equation will always be satisfied. The memory depth setting allows you to limit the amount of memory. So instead of 200Mpts at the, say, 10ms/div timebase, if I choose a limit of 100Mpts, the sample rate will drop to 1GSa/s.

For such an option to make any sense the waveform update rate must be essentially bounded by the memory fill rate. It's a feature to increase the waveform update rate at the expense of temporal resolution.

Obvious but I feel this need still note for some random readers. User selectable memory limit is important - not just for wfm/s speed.

There is this very handy and useful waveform history FIFO. How many last sequential acquisitions it can keep depends highly about memory length of one acquisiton.

Example if we are looking some slowly repeating serial messages we do not mostly need full samplerate but one acquisition we need least as long time as one message last. Then we may want capture many messages to buffer for analyze. 
We can use 10k, 100k, 1M, 10M or 100M when non interleaved mode. Smaller acq length more acquisitions history FIFO can keep.
 

[thinkfat]

Could you add how many points were sampled, at which sample rate, for each of the timebase settings?
EDIT: also, would be interesting if the trigger setup has an influence on the update rate, too. Edge trigger is easy, but how about runts or other, more complex setups?

See below, the trigger thing I can test when second scope was here.

Thanks a lot for that. I understood now what the memory depth setting actually does. The scope always only stores sample points to fill one sweep (timebase x 10 x samplerate = total samples). This equation will always be satisfied. The memory depth setting allows you to limit the amount of memory. So instead of 200Mpts at the, say, 10ms/div timebase, if I choose a limit of 100Mpts, the sample rate will drop to 1GSa/s.

For such an option to make any sense the waveform update rate must be essentially bounded by the memory fill rate. It's a feature to increase the waveform update rate at the expense of temporal resolution.

Obvious but I feel this need still note for some random readers. User selectable memory limit is important - not just for wfm/s speed.

There is this very handy and useful waveform history FIFO. How many last sequential acquisitions it can keep depends highly about memory length of one acquisiton.

Example if we are looking some slowly repeating serial messages we do not mostly need full samplerate but one acquisition we need least as long time as one message last. Then we may want capture many messages to buffer for analyze. 
We can use 10k, 100k, 1M, 10M or 100M when non interleaved mode. Smaller acq length more acquisitions history FIFO can keep.

Yep, that's a perfectly valid reason. Is there a setting to configure how many captured waveforms to keep in the history? Or is it implicit only, depending on the mem depth limit? That would be a pity. The history buffer seems to be a very useful feature, it should not be hidden away.

[TK]

Siglent SDS1000X-E scopes does not fully decode all the captured memory shown in screen.  I run some tests and it seems to stop decoding after 2000-3000 packets in SPI (and I think the same limit is applied to other protocols).  When the "BUG" was reported, they fixed it by adding an error message on the screen saying that a decode limit was reached  If a 5000X or 2000X plus owner can test how decoding works for long captures, it would be very helpful.  If it behaves the same way, it means that there is another limitation that they are not telling about, so I assume if the hidden limit applies to decoding, it can also apply to measurements and other analysis.

https://www.eevblog.com/forum/testgear/impressive-keysight-1000x-series-(edux1002g-modded)-spi-triggering-rate/msg2412543/#msg2412543

[nctnico]

In general (not just for Siglent oscilloscopes) there is a limit to the total number of packets which can be decoded. I don't think I have ever seen this limit appear in a datasheet.

[rf-loop]

Could you add how many points were sampled, at which sample rate, for each of the timebase settings?
EDIT: also, would be interesting if the trigger setup has an influence on the update rate, too. Edge trigger is easy, but how about runts or other, more complex setups?

See below, the trigger thing I can test when second scope was here.

Thanks a lot for that. I understood now what the memory depth setting actually does. The scope always only stores sample points to fill one sweep (timebase x 10 x samplerate = total samples). This equation will always be satisfied. The memory depth setting allows you to limit the amount of memory. So instead of 200Mpts at the, say, 10ms/div timebase, if I choose a limit of 100Mpts, the sample rate will drop to 1GSa/s.

For such an option to make any sense the waveform update rate must be essentially bounded by the memory fill rate. It's a feature to increase the waveform update rate at the expense of temporal resolution.

Obvious but I feel this need still note for some random readers. User selectable memory limit is important - not just for wfm/s speed.

There is this very handy and useful waveform history FIFO. How many last sequential acquisitions it can keep depends highly about memory length of one acquisiton.

Example if we are looking some slowly repeating serial messages we do not mostly need full samplerate but one acquisition we need least as long time as one message last. Then we may want capture many messages to buffer for analyze. 
We can use 10k, 100k, 1M, 10M or 100M when non interleaved mode. Smaller acq length more acquisitions history FIFO can keep.

Yep, that's a perfectly valid reason. Is there a setting to configure how many captured waveforms to keep in the history? Or is it implicit only, depending on the mem depth limit? That would be a pity. The history buffer seems to be a very useful feature, it should not be hidden away.

First, Sure it is limited. Siglent say UP TO MAX 90000 wfms. (it is, afaik based to other models,  same amount as in sequence mode). Think one wfm what length is 200Mpts. 90000 these in FIFO.... no, it is of course limited. I do not know if 200M can be more than 1 ( I can not test because I do not have)  But sure 20pts wfms can be max, and  also some longer... example SDS1kX-E can have 700pts wfms  45000 in FIFO and not only one channel in use.


When it is normally running and background working history FIFO it is just FIFO what have only maximal length with current settings. Capturing speed wfms/s is current speed. User can not limit it or other ways set number of FIFO length exept changing mem length etc but it do not show how many places there is. I do not know numbers in SDS2kXPlus but I assume it works roughly similar other Siglent scopes what have this.
If you start scope and example if FIFO amount is 40000 wfms with some slow wfms speed it may take while when it is full and then oldest start dropped out.  If you stop it before it is full you see less history what is possible. Of course every wfm there are time stamped. What ever you change, vertical, time, delay etc it resets FIFO so all in FIFO what user see with history viewer is sure captured with equal settings.

wfms possible in FIFO with current setings can indirectly check. Turn sequence mode on and set max number of segments. Amount is this least in some other models.

Sequence mode is somehow similar but it can trig even down to 2us interval,  there user can set number on segments. As long as it capture these segnments it is in fast mode and do not update display for make maximal speed possible (up to 500000 segments/s speed). If single segment it stop and displau overlaid every captured segment until you do something. In history viewer it show all segments separately and all segments can measure, change interpolation, use math, including FFT and decode etc.  In this mode it do not work like FIFO. It is storage and after it is full it stop or if mode is repeating it reset all and start new turn.  These same wfm measurements and handling can do also in normal mode stored wfms in FIFO.

It can say that this bacround running history is "slow segmented acquisition" and sequence mode is "fast segmented acquisition"...  Both are in same memory and both can view using same history wiever and functionally these wfms can handle equal with many tools in system.

[Wuerstchenhund]

That is because Keysight uses full memory always but still manages to get high waveform update rates through a clever parallel acquisition system.

No, it doesn't. There is no "clever parallel acquisition system".

which avoids the need for the acquisition system to "wait" for the processing to finish. Only when the acquisition is stopped and the processing of the previous cycle data has completed, the whole memory is then used for the last acquisition, which is the only time the full amount of memory is actually used.

ANd considering the complexity of the different memory allocations it does make sense that the user doesn't get any memory information whatsoever.

Yes it does use all the memory from the user's point of view so what happens under the hood is just semantics.

It's not, it's a technical reality. You can choose to ignore it of course but it's a simple fact. In repetitive mode, only the last acquisition will use the max available memory.

Also, in terms of MegaZoom, the user has no idea how much memory is actually used because the scope shows no information whatsoever.

And there is definitely something clever going on. Think about the situation where a new trigger comes before the entire memory is filled in combination with needing to have pre-trigger data as well.

So what? Nothing really changes except that some part of the sample memory is used for pre-trigger data (as per horizontal delay setting) and the rest for post trigger data, just with the latter being smaller as some part of the memory is used for pre-trigger.

In repetitive mode, the scope will still only use as much memory as needed to fill the screen, and only use the full available memory for its last acquisition before a stop.

You make it sound as if by some magic MegaZoom uses all the memory (for sampling data of the current acquisition) all the time, but this is, plain and simple, wrong. Also, MegaZoom is not magic, it's actually a pretty simple, although clever, idea to maintain very high update rates - something which was valued by people coming from analog scopes.

The PDF you linked to is a simplified explaination for what is really going on under the hood. But for the user it doesn't really matter.

It does matter, because it explains how your tool works, and understanding your tool (and its limitations!) is pretty important for any serious work.

But to some extend I agree, what's going on inside MegaZoom doesn't matter all that much because the user doesn't even have to care about memory, as there is no manual control or even information about memory usage. Which has been the idea behind MegaZoom all the time - designed for users who came from analog scopes, to give them an as much analog-like experience as possible.

But that still doesn't mean the scope is always using full memory (for acquisition data of the current acquisition cycle), and a user should really not assume this to be the case.

Since when does NI make bench scopes?

They kind of do if you buy the modules to build a self contained system.

Even if you put NI Digitizers in a NI PXI chassis and use NI LabView, it's still a Digitizer system where one has to write the software for it and not a general purpose bench scope like the SDS2kX+. Aside from costing an insane amount of money.

So I'd say let's just not go there

Well, you brought up scientific use cases and I disagreed Lecroy didn't have any competition in that market.

Fair enough, but digitizers are not normally used where scopes are, because they are digitizers and not scopes (and even what NI sells as "PXI scopes" are really just PXI digitizers, although with some scope capabilities).

And none of it is competitive with a bench scope for application where a bench scope would suffice. For example, a NI "PXI Scope" which is at least somewhat comparable to the SDS2kX+ in specs (i.e. 8bit ADC, less than 500MHz BW) starts at $7,846 (PXI-5152, 2ch 300MHz, 2GSa/s, 8M/ch). If you need a bench scope then that's a lot of money for the privilege to also have to write your own software (plus to pay for PXI chassis, controller and LabView, which easily lifts this 300Mhz 2ch "scope" over $13k). And it only gets more expensive from there.

Digitizers have their place in specific use cases, for example where very high precision is required (i.e. true 16bit), or where continuous data streams are needed (scopes are generally incapable of streaming data for an extended amount of time). Outside specific niches however, it's real bench scopes hands down.

I told you, just let's not go there

I don't think any of the points which make the WavePro a poor general purpose scope apply to the SDS2kX+ - or any Siglent scope as of today.

Currently the way the memory is managed. I was just curious on why (how Siglent got the idea) the SDS2000X doesn't allow to record outside the screen and much to my surprise (yes, that did surprise me) the Wavepro 7200A works in the same way. Since all other DSOs I have come across so far do allow recording outside the screen without needing work-arounds I take the liberty to assume that this is something Siglent copied from Lecroy.

Just curious, what settings did you use?

But yes, it's quite normal that in repetitive mode and short time bases a DSO will not use all available sample memory.

[Martin72]

If a 5000X or 2000X plus owner can test how decoding works for long captures, it would be very helpful.

Im not really into (serial) decoding things, so could be a stupid question:

Like this perhaps ?

[TK]

If a 5000X or 2000X plus owner can test how decoding works for long captures, it would be very helpful.

Im not really into (serial) decoding things, so could be a stupid question:

Like this perhaps ?

Yes, something like that.  I tested the MSO5000 and at least it decodes everything that is on display, even when you capture over 10,000 SPI packets.  I tested the same on the Siglent SDS1000X-E and it stopped at around 3,000 packets, so I was wondering if anything changed on the 2000X plus.

[Martin72]

OK, will do it on friday or monday next week.
What I must also test:
Math...When I use math, could I hide the source channel….
Lecroy could, the sds1104X-E could partially afaik.
When they copying lecroy, they should also copy this behaviour. 

[nctnico]

You make it sound as if by some magic MegaZoom uses all the memory (for sampling data of the current acquisition) all the time, but this is, plain and simple, wrong. Also, MegaZoom is not magic, it's actually a pretty simple, although clever, idea to maintain very high update rates - something which was valued by people coming from analog scopes.

I never stated that it uses all memory for each displayed trace. However the Megazoom ASIC can never know what is the last acquisition in advance so in order to have the full memory available from the last trigger it has to have pre-trigger data for the current acquisition cycle. It may be that there are no more triggers after a signal burst. Whatever trigger sequence you throw at it, it will use all of the available memory for the last trace and you can scroll through it until there is a new trigger.

You are also wrong that you can't get memory information on a Keysight scope. It is there but not in plain sight. Probably because the memory depth typically is 1/4th (or even less) of what it says on the badge.

But yes, it's quite normal that in repetitive mode and short time bases a DSO will not use all available sample memory.

To me it is not normal at all. The Wavepro 7200A is the first DSO (from many) I have owned which doesn't / can't use the memory depth setting I selected. Contrary to other DSOs it uses enough memory to fill the screen and that's it. This behaviour makes it cumbersome to use all the available memory because you need to configure very specific settings (and likely switch back & forth a lot) in order to use all the memory. It is much easier if an oscilloscope just samples the memory depth you selected (and split the memory 50:50 at the centre of the screen). If you need to debug an unknown event (on a tight schedule as usual) having as much information from a single acquisition with least hassle (need to concentrate on the problem at hand; not the tools) is a major time saver.

[0xdeadbeef]

To try to get back on the topic: I overlooked another difference compared to the SDS5000X: the qualified trigger (A->B) is missing in the SDS2000X Plus. That's a bit of a bummer.

Regarding dot mode, the following statement can also be found in the SDS5000X manual:

There is no difference between the vector and dot display types when the
number of samples of a frame is greater than 1000. When the sample number
is below 1000, there are some differences.

This seems plausible at first since the display for waveforms is probably exactly 1000 pixels (leaving 12dots to the left and right for channel number etc.).
So 1000 pixel and 10 divisions mean 100 pixels per div. So the above statement suggests that visual problems in dot mode are supposed to occur e.g. when decreasing the time per div below 50ns at 2GSa/s:

2GSa/s -> 0.5ns -> 100 pixels = 50ns
1GSa/s ->   1ns -> 100 pixels = 100ns

However, this doesn't consider the vertical step width. E.g. for a PWM signal with steep slopes, it might happen that only very few (or even no) values are captured within the slope. So even when averaging over dozens of frames, the edges will me missing or much too dark in dot mode. So I fear the above statement is a simplification for signals with relatively slow edges (i.e. which don't change by more than one pixel in vertical resolution between two sample points considering the vertical scale).

[Martin72]

What I must also test:
Math...When I use math, could I hide the source channel….

No.
Except FFT, all math-functions are only visible, when the source-channel is also active - That is disappointing….
They should handle it like lecroy does - hopefully, they can manage this.
It´s the first downer..
On the SDS1104X-E, eres was part of the acquision mode, no need to use extra mathchannel for this.
And:
Take some pics, press "print", always the message appear, "saved" - And then only 2 pics was really saved to the usb-stick..
The update-list is getting longer..

[tv84]

Martin, place your images directly in the forum. It's easier and we can be sure that they wont disappear separately from the msg.

[Martin72]

Done.