Siglent SDS1104X-E and SDS1204X-E Mixed Signal Oscilloscopes

[kcbrown]

Have you tried it with more wide slope steep. Example 1ms / 2ms. Do it change situation.

I altered the time range to be 0.5ms to 1.5ms, and kept the voltage thresholds the same.  It made no difference -- the problem reproduced easily.

[kcbrown]

The issue also reproduces with acquisition in "slow" mode, i.e. fast or slow makes no difference here.

[rf-loop]

Have you tried same with analog side BW reject.

So far this issue has *not* reproduced with the 20MHz bandwidth limiter enabled on the probe.  Looks to me like whatever's causing this, it's very high frequency noise that somehow doesn't appear in the capture.

I'll try some of the other things next.

This is very strange if it do not appear in capture or captured raw ADC data. It must be there, infinitely. Without it, it do not trig. But all is possible. Also some bug in FW what cause this reason is hiding. 
Do you have FULL  .CSV or Binary data. Screen shot png and paired with this full saved sample data. Example 3 set of these pairs so that these pairs are somehow marked so that can not mix and with perfectly explained trigger setting and scope other settings (norm/Auto, possible trigger coupling filters and, trigger setup exactly and so on. (display dots/interpolatin line or sinc do not affect any thing, they are post processed.)
Because if there is really no raw ADC data what explain wrong/failed trig, it need solve what make it. These can bit look and forward to Siglent for further analysis. It is not lottery machine and it do not have its own mind what some times do irrationally something like human..

[kcbrown]

This is very strange if it do not appear in capture or captured raw ADC data. It must be there, infinitely. Without it, it do not trig. But all is possible. Also some bug in FW what cause this reason is hiding. 
Do you have FULL  .CSV or Binary data.

Yes.  Well, I *had* it but had to reboot my computer, and didn't have them stored in a location that would survive the reboot.  But seeing how I can reproduce this essentially at will, getting the data isn't a problem at all.

Screen shot png and paired with this full saved sample data. Example 3 set of these pairs so that these pairs are somehow marked so that can not mix and with perfectly explained trigger setting and scope other settings (norm/Auto, possible trigger coupling filters and, trigger setup exactly and so on. (display dots/interpolatin line or sinc do not affect any thing, they are post processed.)

Sinc versus dots versus linear doesn't affect triggering??

OK, that raises a question: does the triggering mechanism *always* use sinc?  Or straight point-to-point?  Or something else?  It has to somehow detect that a transition has happened between acquired points, clearly, so it has to be doing some kind of interpolation for that, no?

Because if there is really no raw ADC data what explain wrong/failed trig, it need solve what make it. These can bit look and forward to Siglent for further analysis. It is not lottery machine and it do not have its own mind what some times do irrationally something like human..

Agreed.  OK, so the question then is: how do you want me to upload the data?  To where?  The resulting 7-zip archive of 14Mpoints of CSV data is far too large for uploads to this forum.

[rf-loop]

Have you tried same with analog side BW reject.

Limiting the bandwidth helps.  I'm running a test now to see if the issue reproduces at all with a 20 MHz bandwidth limit on the input. But limiting the bandwidth shouldn't be necessary.  What if you're trying to find high-frequency noise in the signal?  The point here is that if noise causes the trigger to fire and the trigger mechanism exclusively uses ADC output, that noise should be present in the capture!   But it wasn't, and isn't.

Of course. I think you have missed this point what I try.

I try analyze this problem. Not make guide how you can use it for something.
Primary purpose when I ask you do some like this is for isolate problem ana analyze because I know bit about how it works. So I may have some logic behind example this question. But I do not tell these my internal speculations behind these for avoid some mess.

Now I know after your words many things what you have tried and no effect. Now I also know that BW limit have effect. First time we hit something what have even some effect. So... this finding is Important but what all need think due to this... but also about this finding I do not want start speculate here what it may mean or not.  But it is important finding together with these findings about settings changes what have no effect. 

 

[rf-loop]

This is very strange if it do not appear in capture or captured raw ADC data. It must be there, infinitely. Without it, it do not trig. But all is possible. Also some bug in FW what cause this reason is hiding. 
Do you have FULL  .CSV or Binary data.

Yes.  Well, I *had* it but had to reboot my computer, and didn't have them stored in a location that would survive the reboot.  But seeing how I can reproduce this essentially at will, getting the data isn't a problem at all.

Screen shot png and paired with this full saved sample data. Example 3 set of these pairs so that these pairs are somehow marked so that can not mix and with perfectly explained trigger setting and scope other settings (norm/Auto, possible trigger coupling filters and, trigger setup exactly and so on. (display dots/interpolatin line or sinc do not affect any thing, they are post processed.)

Sinc versus dots versus linear doesn't affect triggering??

OK, that raises a question: does the triggering mechanism *always* use sinc?  Or straight point-to-point?  Or something else?  It has to somehow detect that a transition has happened between acquired points, clearly, so it has to be doing some kind of interpolation for that, no?

Because if there is really no raw ADC data what explain wrong/failed trig, it need solve what make it. These can bit look and forward to Siglent for further analysis. It is not lottery machine and it do not have its own mind what some times do irrationally something like human..

Agreed.  OK, so the question then is: how do you want me to upload the data?  To where?  The resulting 7-zip archive of 14Mpoints of CSV data is far too large for uploads to this forum.

It need think. Also it need take account that what if...  and so on..  Do you have binary data also from same acquisition what you have CSV and I do not mean small part of acq. but of course every byte from start to end.
How big your zips are for both separately.

[kcbrown]

It need think. Also it need take account that what if...  and so on..  Do you have binary data also from same acquisition what you have CSV and I do not mean small part of acq. but of course every byte from start to end.
How big your zips are for both separately.

Well, "zip" and "7-zip" are actually two different things.  "7-zip" does a *much* better job of compressing.  While "zip" of the 14Mpoint CSV gets me a 30 megabyte zip file, "7-zip" gets me a roughly 9 megabyte "7z" file.  Both are archive formats that can store multiple files within the archive, so they're both good for the same thing.  You can get "7-zip" from https://www.7-zip.org.

I don't know how much space an archive containing the binary format file will take.  I'll let you know.

Right now I'm running the test with the trigger bandwidth limiter enabled, and it hasn't reproduced yet (it's been running for about 30 minutes so far, with about 15K frames analyzed -- I'm using the mask function to detect a violation).  I'll let that run a bit longer before reverting back to full bandwidth on everything for the purpose of gathering the data.

The real question is where I should upload the resulting (rather large) archive files to.  Any suggestions?

[bdunham7]

You can try a free Box account (not Dropbox) as I don't think they are banned in China...yet...while Dropbox is.

[rf-loop]

I know this all may be bit frustrating. If I have here now this scope I can try repeat this error using noises etc... but I do not have.

In this image there is 1GSa/s  1ms/div 14M memory.

There was some other image where was 500MSa/s and 2ms/div

Both show this error.

But I have not seen any test (or my far over best before date eyes have not hit) test where is native 500MSa/s sampling speed.
In this 2ms 500MSa image it is decimated samplerate.

So, can you do also test just like this image:


But change one thing. Turn just this ADC from interleaved mode to non interleaved mode so it have real 500MSa/s.
This you can do more easy than easy.
Keep 1ms and this channel. Turn Ch1  ON.  Now you see samplerate is 500MSa/s now ADC is not inteerleaved mode but normal 2 channel 500MSa mode.
As told some previous image also have 500MSa/s but is is decimated from 1GSa interleaved sampling.

It may have some difference depending how trigger engine listen these 4 x 250MHz 8 bit bus from ADC or 2 x 250MHz bus in non interleaved mode...
If it give real big difference.. it is very extremely important finding, together with this previous finding about BW reject (and case is that before BW reject in ADC data can not see any reason for wrong trig. )

Imho, both are important finding, it this change affect or if not. These can imagine are like road crosses and  how to go forward, right or left or straight.

[rf-loop]

It need think. Also it need take account that what if...  and so on..  Do you have binary data also from same acquisition what you have CSV and I do not mean small part of acq. but of course every byte from start to end.
How big your zips are for both separately.

Well, "zip" and "7-zip" are actually two different things.  "7-zip" does a *much* better job of compressing.  While "zip" of the 14Mpoint CSV gets me a 30 megabyte zip file, "7-zip" gets me a roughly 9 megabyte "7z" file.  Both are archive formats that can store multiple files within the archive, so they're both good for the same thing.  You can get "7-zip" from https://www.7-zip.org.

I don't know how much space an archive containing the binary format file will take.  I'll let you know.

Right now I'm running the test with the trigger bandwidth limiter enabled, and it hasn't reproduced yet (it's been running for about 30 minutes so far, with about 15K frames analyzed -- I'm using the mask function to detect a violation).  I'll let that run a bit longer before reverting back to full bandwidth on everything for the purpose of gathering the data.

The real question is where I should upload the resulting (rather large) archive files to.  Any suggestions?

I am not just born. (ref zips) 
I have used many many kind of packaging things as there have been personal computers far before IBM compatible shits.

It is fun how different result these different methods/algorithms give some times.

[rf-loop]

It is explained bit more but unfortunately "scrambled by natural enigma". Our "enigma" is our Finnish language what is said is second difficult after Chinese in world. 
In some images there is some english/finglish explanations. But more deep text explanations just with Finnish.
https://siglent.fi/oskilloskooppi-tietoa-sds1004x-e--wfm-speed.html

I am slowly reading through this with help from Google translate.  It's very helpful, thank you for writing it!

PS: there is a typo in the second (all yellow) table shown in Figure 1, in the line with t/div= 1ms.  Column 6 contains 1G sa/s. It should read 1M sa/s.

Typing mistake corrected, thank you.

[kcbrown]

I know this all may be bit frustrating. If I have here now this scope I can try repeat this error using noises etc... but I do not have.

In this image there is 1GSa/s  1ms/div 14M memory.

There was some other image where was 500MSa/s and 2ms/div

Both show this error.

But I have not seen any test (or my far over best before date eyes have not hit) test where is native 500MSa/s sampling speed.
In this 2ms 500MSa image it is decimated samplerate.

So, can you do also test just like this image:


But change one thing. Turn just this ADC from interleaved mode to non interleaved mode so it have real 500MSa/s.
This you can do more easy than easy.
Keep 1ms and this channel. Turn Ch1  ON.  Now you see samplerate is 500MSa/s now ADC is not inteerleaved mode but normal 2 channel 500MSa mode.
As told some previous image also have 500MSa/s but is is decimated from 1GSa interleaved sampling.

It may have some difference depending how trigger engine listen these 4 x 250MHz 8 bit bus from ADC or 2 x 250MHz bus in non interleaved mode...
If it give real big difference.. it is very extremely important finding, together with this previous finding about BW reject (and case is that before BW reject in ADC data can not see any reason for wrong trig. )

Imho, both are important finding, it this change affect or if not. These can imagine are like road crosses and  how to go forward, right or left or straight.

The issue reproduces just as easily under these conditions as it does at 1GS/s.   I've captured the data for this, so I'll include it in the data dump.

I also found the original CSV from the single shot reproduction (I'd forgotten it was also still on the USB stick).

[rf-loop]

If you have box.com account you can PM me download link. Please name files or mark somehow other way so I know what is what.

[HendriXML]

Have you tried same with analog side BW reject.

So far this issue has *not* reproduced with the 20MHz bandwidth limiter enabled on the probe.  Looks to me like whatever's causing this, it's very high frequency noise that somehow doesn't appear in the capture.

I'll try some of the other things next.

This is very strange if it do not appear in capture or captured raw ADC data. It must be there, infinitely. Without it, it do not trig. But all is possible. Also some bug in FW what cause this reason is hiding. 
Do you have FULL  .CSV or Binary data. Screen shot png and paired with this full saved sample data. Example 3 set of these pairs so that these pairs are somehow marked so that can not mix and with perfectly explained trigger setting and scope other settings (norm/Auto, possible trigger coupling filters and, trigger setup exactly and so on. (display dots/interpolatin line or sinc do not affect any thing, they are post processed.)
Because if there is really no raw ADC data what explain wrong/failed trig, it need solve what make it. These can bit look and forward to Siglent for further analysis. It is not lottery machine and it do not have its own mind what some times do irrationally something like human..

Are we absolutely sure the trigger voltages are converted to adc values, and not let's say analog comparators? The strange thing is that when using scpi those voltages are attributes of each channel. In a complete digital situation, it would be more logical for them to be attributes of the trigger.
It could explain the behavior if they where analog...

[rf-loop]

Have you tried same with analog side BW reject.

So far this issue has *not* reproduced with the 20MHz bandwidth limiter enabled on the probe.  Looks to me like whatever's causing this, it's very high frequency noise that somehow doesn't appear in the capture.

I'll try some of the other things next.

This is very strange if it do not appear in capture or captured raw ADC data. It must be there, infinitely. Without it, it do not trig. But all is possible. Also some bug in FW what cause this reason is hiding. 
Do you have FULL  .CSV or Binary data. Screen shot png and paired with this full saved sample data. Example 3 set of these pairs so that these pairs are somehow marked so that can not mix and with perfectly explained trigger setting and scope other settings (norm/Auto, possible trigger coupling filters and, trigger setup exactly and so on. (display dots/interpolatin line or sinc do not affect any thing, they are post processed.)
Because if there is really no raw ADC data what explain wrong/failed trig, it need solve what make it. These can bit look and forward to Siglent for further analysis. It is not lottery machine and it do not have its own mind what some times do irrationally something like human..

Are we absolutely sure the trigger voltages are converted to adc values, and not let's say analog comparators? The strange thing is that when using scpi those voltages are attributes of each channel. In a complete digital situation, it would be more logical for them to be attributes of the trigger.
It could explain the behavior if they where analog...

How many times it need say: Siglent trigger engine is totally and fully in digital side.
If you do not believe, please take one SDS on table and parts... and try find analog trigger side pathway from it,  starting from BNC and ending to ADC input. If you today find low price segment new DPO and it use analog trigger system just throw it out from window as fast as it arrive to your hand.  It belongs to museum.
Siglent is not Keysight IV machine what still have analog side pathway trigger engine lika in old times mostly all DSO.
Also Siglent OLD DL/CNL/CNL/CFL series have analog side pathway analog comparator trigger system
All X models have  full true digital side trigger system.
Only exception is ExtTrig input in some X/X-? models. It is analog trigger  where trigger level to compatator ís done using DAC.'
If you do not know what is true digital side trigger system you can find example nice Rohde&Schwarz explanation about it.
With these kind of Trigger things what are example in Siglent and many other modern scopes is quite difficult to do with analog trigger using simple circuits.

Of course now this case rise one question. These trigger things are complex and time critical things deep inside system. Always is possible that we hit some unexpected error. These are so comples no one can never test all combinations in product development. If some say it can he do not know what he is talking. So, this possible need keep in mind until further data and tests.
We have now ONE thing what is clear it affect. BW limit. Other things like more wide slope rising time limits did not affect. Now I am not sure if he have tested if trigger filter have effect but if I understand right trigger hysteresis change did not. Also changing ADC mode from interleaved to non interleaved did not affect.
Also just this particular trigger type many conditions must be met, much more complex than some simplest possible edge trig.
If it make error and do not trig..  btw... it have made this error. We have seen it in every  image. So this need keep carefully in thinking logic.
It need think. Every time it have "trigged" wrong place... we really do not even know if it is trigged... but other thing we know sure... every this fail image we see part of signal what meet all needs for trig and it have not trigged  or least positioned it to right place... what is also important to keep in mind.
How we know it is trigged just there where is trigger time position and signal position is wrong. Think again... who ever is thinking this... how we know it IS trigged  just in this position where trigger mark is and we see this signal in wrong position in case it have failed. Who know... I do not know and trigger do not tell what and when she have done things

[HendriXML]

Truth or fact is not about how many times one says something.. 

I don't say its likely, but seeing the scope's behavior..

Also there could be benefits to analog triggers, for example when using slow "time div's", but very fast signals. At the moment I'm experimenting with that, to see how the scope responds.

Its stil running but it managed to capture a pulse of 32 ns each 700s using only 10 Samples/s (100 s/div thus 23 min/screen). That's amazing if it where only digital...

Maybe it sampling faster under water than it says?

If thats true than we have another possible explanation..(I see that the faulty trigger examples are at max sample rate, so probably not)

[2N3055]

Truth or fact is not about how many times one says something.. 

I don't say its likely, but seeing the scope's behavior..

Also there could be benefits to analog triggers, for example when using slow "time div's", but very fast signals. At the moment I'm experimenting with that, to see how the scope responds.

Its stil running but it managed to capture a pulse of 32 ns each 700s using only 10 Samples/s (100 s/div thus 23 min/screen). That's amazing if it where only digital...

Maybe it sampling faster under water than it says?

If thats true than we have another possible explanation..(I see that the faulty trigger examples are at max sample rate, so probably not)

It is true because he knows it for fact. It is fully digital, and triggering engine works with full sample rate, before decimation....

[HendriXML]

On a slow time base it is very possible to have it trigger on a pulse (2.5V, 32 nS) without seeing it happen on screen. Using a very simple edge trigger.

It probably doesn't prove much, because it is propably sampling faster under water.

Edit:
This post crossed the above one... in which my guess is already confirmed

[rf-loop]

Truth or fact is not about how many times one says something.. 

I don't say its likely, but seeing the scope's behavior..

Also there could be benefits to analog triggers, for example when using slow "time div's", but very fast signals. At the moment I'm experimenting with that, to see how the scope responds.

Its stil running but it managed to capture a pulse of 32 ns each 700s using only 10 Samples/s (100 s/div thus 23 min/screen). That's amazing if it where only digital...

Maybe it sampling faster under water than it says?

If thats true than we have another possible explanation..(I see that the faulty trigger examples are at max sample rate, so probably not)

ADC in SDS1000X-E works always full speed. interleaved mode full speed or non interleaved full speed. If we are now talking about SDS1000X-E. Trigger do not care your decimated samplerate at all. No one make serious oscilloscope what trig from decimated data...
If scope display show example 1ksa/s ADC works still full speed! And digital trigger engine is there.

Is now time to take some lesson how digital oscilloscopes overall works behind screen and input connectors. I think it is better to open new thread just for this if need.

[2N3055]

On a slow time base it is very possible to have it trigger on a pulse (2.5V, 32 nS) without seeing it happen on screen. Using a very simple edge trigger.

It probably doesn't prove much, because it is propably sampling faster under water.

That is why there exists peak detect mode. It will show those peaks..

[rf-loop]

It probably doesn't prove much, because it is propably sampling faster under water.

How useful is this talking and wondering if you do not believe or listen anything what is said about it.

[HendriXML]

On a slow time base it is very possible to have it trigger on a pulse (2.5V, 32 nS) without seeing it happen on screen. Using a very simple edge trigger.

It probably doesn't prove much, because it is propably sampling faster under water.

That is why there exists peak detect mode. It will show those peaks..

I just figured that one out by myself as well
Just before I thought it was only a way of displaying underlying data. Now I know why the aquisition needs to be redone.
Now I wonder, does it need twice the memory..? Will look it up, because it is not relevant to the issue.
B.t.w. I know I'm lacking technical knowledge regarding this scope, I see myself more like a user. But the question is how much knowledge will be needed to pinpoint the issue some more. I think experiments are propably more important. I'm also not yet fully engaged in this issue (yet), but I don't dislike to finding "truth" with very little information. For instance I know now why going on a "slower sampling speed" will not induce comparable behavior. Which might together with the non analog trigger already be enough technical info.

[2N3055]

On a slow time base it is very possible to have it trigger on a pulse (2.5V, 32 nS) without seeing it happen on screen. Using a very simple edge trigger.

It probably doesn't prove much, because it is propably sampling faster under water.

That is why there exists peak detect mode. It will show those peaks..

I just figured that one out by myself as well
Just before I thought it was only a way of displaying underlying data. Now I know why the aquisition needs to be redone.
Now I wonder, does it need twice the memory..? Will look it up, because it is not relevant to the issue.
B.t.w. I know I'm lacking technical knowledge regarding this scope, I see myself more like a user. But the question is how much knowledge will be needed to pinpoint the issue some more. I think experiments are propably more important. I'm also not yet fully engaged in this issue (yet), but I don't dislike to finding "thruth" with very little information. For instance I know now why going on a "slower sampling speed" will not induce comparable behavior. Which might together with the non analog trigger already be enough technical info.

No need for twice the memory. Peak detect happens only when decimating data, in which you keep one point, discard next 9.. You just decimated by 10x. If you look at those 10 points and keep largest value from them, you just did peak detect. If you keep min and max, you get envelope detect, that one will have twice the data...

As for facts and beliefs, there are facts that are common sense, there facts that come from people who work with it so they know, and there is information coming from speculations that need to be verified. Not every thing need to be always proved from first principles..

Complex triggers can have bugs that won't show at certain timebases, if for not other reason, they might have temporal conditions, that might have bugs in calculations and scaling...

[HendriXML]

Okay so no negative peaks, like minimals..

About the facts. I might have skipped the digital trigger (vs analog) trigger post.. Got some brain damage, so reading and processing information is a bit impaired.
So my remark:
Truth or fact is not about how many times one says something.. 
Was more a general remark, not a disbelief in what was said (which I didn't read).

But either way: in this situation that fact should be traced to its roots. If rf-loop is that root of knowledge than .

There's more to say but I'll leave it like this.

[2N3055]

Okay so no negative peaks, like minimals..

About the facts. I might have skipped the digital trigger (vs analog) trigger post.. Got some brain damage, so reading and processing information is a bit impaired.
So my remark:
Truth or fact is not about how many times one says something.. 
Was more a general remark, not a disbelief in what was said (which I didn't read).

But either way: in this situation that fact should be traced to its roots. If rf-loop is that root of knowledge than .

There's more to say but I'll leave it like this.

Peak detect mode is extrema, positive and negative. Commercial names for functions are not necessarily mathematically correct..
We all have our problems and coping strategies. Will to do it right is only thing important. So in the end, only time needed is different, but results are same and correct.

For Siglent equipment, users RF-loop and Preforma01 are authoritative source.. Also Tautech (he's Siglent dealer) have lots of information, and what he doesn't know he can usually ask Siglent and get back quality info..
There are also about dozen other users that are very knowledgeable and have various Siglent  equipment that are good source of info..

Regards,