EEVblog® Electronics Community Forum
Products => Test Equipment => Topic started by: eTobey on September 29, 2024, 12:17:06 pm
-
Since no one can explicity answer my questions, i just add some links here, that might help others that come here. The following posts were of no help. At least for me that wants to know things more specific.
https://www.tek.com/en/blog/sample-processing-digital-oscilloscope (https://www.tek.com/en/blog/sample-processing-digital-oscilloscope)
https://www.youtube.com/watch?v=oS9PT98m69c (https://www.youtube.com/watch?v=oS9PT98m69c)
https://www.youtube.com/watch?v=036Ems1z-o0 (https://www.youtube.com/watch?v=036Ems1z-o0)
https://www.youtube.com/watch?v=W4twnd-YQQ4 (https://www.youtube.com/watch?v=W4twnd-YQQ4)
https://www.youtube.com/watch?v=_jAOduDrqIU (https://www.youtube.com/watch?v=_jAOduDrqIU)
Original post:
2N3055 demanded that i post this in the bug topic, but i believe there will be a lot of people who would not like that either, so i make a new topic:
Yes, i know, that not every edge can be triggered on, and the rate at which it happens will vary depending on a few things. But the behaviour i observed still seems quite strange to me. (Yes i am again advancing into the world of low sample rates ::) )
Said in a nutshell, it should be slower with more data to process, and vise versa. But it is not!
Its a 90 Hz signal.
https://youtube.com/shorts/XvDNO43fBtk (https://youtube.com/shorts/XvDNO43fBtk)
Discovered seemingly the same issue with seg aqcuire mode: https://youtu.be/J5-ZZc376wQ (https://youtu.be/J5-ZZc376wQ)
Trying to capture every edge for a certain amount of time seems to be quite a task to manage.
Edit:
Made another Measurement:
- timebase 10us
- signal 90Hz square
- 200kSa/s
- segment mode
- trigger: edge rising / no holdoff / trace is filling the screen, trigger level OK
- Vector mode
- no measure or search activated
First test 3 channels on:
For 1000 segments it takes about 11s (measured with stopwatch)
Trigger output triggers about every 10-11ms
Second test only one channel on:
For 1000 segments it takes about 22s (measured with stopwatch)
Trigger output triggers about every 22ms
-
OMG....
Not another one separate topic!!
Put it in common SDS800 bugs topic...
There people will look into it and we'll see if it is a bug.
-
OMG....
Not another one separate topic!!
Put it in common SDS800 bugs topic...
There people will look into it and we'll see if it is a bug.
Not only ... I would like this thread to be used for 2 reasons only:
- when a bug is found, reproducible and documented
- when a new firmware is released
This would allow me to subscribe to the thread and use it as an "alarm" that time to upgrade firmware arrived.
There are other threads for open discussions, opinions ...
Indeed it seems this person just start typing randomly in different threads or even opening a new one.
Frankly speaking a veri annoying behaviour.
Underneath that post: "The following users thanked this post: Performa01, newbrain, 2N3055, KungFuJosh, Martin72, mawyatt, jasonquin".
I dont think its a bug, i just want to understand why that is, and how to predict this behaviour.
-
Why is the trigger rate of my scope (SDS814X HD) changing in an unexpected way?
FTFY
To optimise over all performance at every timebase setting.
Fun learning about this stuff isn't it ?
-
I don't understand, you set the time base to 10us/div and the sampling rate to 200kSa/s, why?
-
OMG....
Not another one separate topic!!
Put it in common SDS800 bugs topic...
There people will look into it and we'll see if it is a bug.
Not only ... I would like this thread to be used for 2 reasons only:
- when a bug is found, reproducible and documented
- when a new firmware is released
This would allow me to subscribe to the thread and use it as an "alarm" that time to upgrade firmware arrived.
There are other threads for open discussions, opinions ...
Indeed it seems this person just start typing randomly in different threads or even opening a new one.
Frankly speaking a veri annoying behaviour.
Underneath that post: "The following users thanked this post: Performa01, newbrain, 2N3055, KungFuJosh, Martin72, mawyatt, jasonquin".
I dont think its a bug, i just want to understand why that is, and how to predict this behaviour.
OK, then.
Triggering frequency is directly connected to how much processing is needed between triggers.
ADC is ALWAYS sampling at 2 GS/s.
When scope sampling is at full capture rate you capture lots of data, and process big buffers.
When scope is at some medium sample rate, you spend little time decimating and less time processing smaller buffer.
When scope is at extremely low sample rate, you need to decimate at factors of a million or more. While final buffer is small and fast to process, such large decimation takes time, since you have to handle large buffers anyways, this time for decimation.
So your scope will have best trigger rates when set automatic memory management and set with max memory that is not too big.
There are too many possible combinations and this specification does not even exist for any scope that I know of.
-
When scope is at extremely low sample rate, you need to decimate at factors of a million or more. While final buffer is small and fast to process, such large decimation takes time, since you have to handle large buffers anyways, this time for decimation.
So your scope will have best trigger rates when set automatic memory management and set with max memory that is not too big.
I am really wondering, how this decimation works, as i think with a pointer like stepping over, it should take that much longer. I may find some information about this.
But why is the trigger rate increasing, when i have not just one, but three channels activated? That still does not make sense to me.
Edit:
I found a great explanation: https://www.tek.com/en/blog/sample-processing-digital-oscilloscope (https://www.tek.com/en/blog/sample-processing-digital-oscilloscope)
From the above link:
In this mode, one sample is saved in memory during each desired waveform sample interval as shown below, and the remaining samples are simply thrown away. The process is also called simple decimation. It is fast and easy to implement, but does result in potential loss of “interesting” samples.
This is also what i would have expected. I have had acquire already set to fast, but its seemingly not fast enough.
The setting normal/peak does not make a difference in the trigger rate. Why? Wouldn peak need some more time to calculate?
I don't understand, you set the time base to 10us/div and the sampling rate to 200kSa/s, why?
For investigation purposes.
-
Data stream is 4 gigabytes of data per second. With trigger running on it. With copy of data to different output buffers for further processing. With digital phosphorus running. With interpolation...
There are dozens of memory configurations happening as you go through different triggers, channel combinations, decimation, buffer sizes..
It all happens in FPGA.
Problem is that you think only one thing is happening and that is simple. It is not. It is very complicated.
-
You evaded all my questions. I am dissapointed. :(
-
You evaded all my questions. I am dissapointed. :(
Which is sole purpose of my being. My raison d'etre. Oh I failed my life's calling... :-[
Internal workings of what exactly scope does inside FPGA is not something that I would share with you or anybody else in the public. This is not some open source or public domain project.
I told you all you need to know. There are Gigabytes per second data rates, very complicated data paths, and lots of real time processing. As scope goes through all possible combinations of sample rates, channels, segments, background history segments, measurements, math, different acquisition modes, it keeps reconfiguring internal state machines for hundreds of different combinations. It is obvious not all of them can be equally optimized or straight forward.
So behaviour you see is what it is. All scopes have the same, Keysight, R&S, Rigol, Magnova.. All of them, in different combinations..
That is all information you will get. Rest is not for the public.
-
Oh I failed my life's calling...
https://www.youtube.com/watch?v=_I1N905fFDw (https://www.youtube.com/watch?v=_I1N905fFDw)
-
You evaded all my questions. I am dissapointed. :(
I told you all you need to know. There are Gigabytes per second data rates, very complicated data paths, and lots of real time processing. As scope goes through all possible combinations of sample rates, channels, segments, background history segments, measurements, math, different acquisition modes, it keeps reconfiguring internal state machines for hundreds of different combinations. It is obvious not all of them can be equally optimized or straight forward.
You told me things that do not match with the manual and my understanding:
RTFM:
In sequence mode, the oscilloscope only acquires and stores data without processing and displaying, until the specified segments are acquired.
Also from the link above:
"The process is also called simple decimation. It is fast and easy to implement, ..."
And then you use terms like digital phosporus, what actually should make things much quicker, not slower.
And still it makes no sense, that - speaking in other words for activating more channels - putting a trailer behind a car, should make it faster.
-
You evaded all my questions. I am dissapointed. :(
I told you all you need to know. There are Gigabytes per second data rates, very complicated data paths, and lots of real time processing. As scope goes through all possible combinations of sample rates, channels, segments, background history segments, measurements, math, different acquisition modes, it keeps reconfiguring internal state machines for hundreds of different combinations. It is obvious not all of them can be equally optimized or straight forward.
You told me things that do not match with the manual and my understanding:
RTFM:
In sequence mode, the oscilloscope only acquires and stores data without processing and displaying, until the specified segments are acquired.
Also from the link above:
"The process is also called simple decimation. It is fast and easy to implement, ..."
And then you use terms like digital phosporus, what actually should make things much quicker, not slower.
And still it makes no sense, that - speaking in other words for activating more channels - putting a trailer behind a car, should make it faster.
In sequence mode scope will not process display and measurements.
There is oodles of other processing still in the core data pump: triggering, interpolation, data duplication...
Why do you ask questions when all the time you simply discard answers and keep repeating your opinion, based on complete lack of understanding how things really work?
You seem to know better everything, even better than companies with thousands of engineers that make actual devices.
I think you should start a company and start making your own scopes.
-
I think you should start a company and start making your own scopes.
And prepare for a crash landing. :scared:
-
Why do you ask questions when all the time you simply discard answers and keep repeating your opinion, based on complete lack of understanding how things really work?
I figured it out! He's the "Guy On Internet." Example:
Study 4 years for degree.
Study 3 more for PhD.
Join lab, start working.
Spend years studying problem.
Form hypothesis, gather evidence.
Test hypothesis, form conclusions.
Report findings, clear peer review.
Findings published, reported in press.
Guy on internet: "Bullshit."
-
i believe there will be a lot of people who would not like that either
Perhaps if you make a thread it should explain its self and stand alone with all the information needed?
This topic expects us to watch videos and take the information out of them. Many members (such as myself) will just ignore that, just as we ignored the previous threads where it was not explained what you were talking about.
-
In sequence mode, the oscilloscope only acquires and stores data without processing and displaying, until the specified segments are acquired.
In sequence mode scope will not process display and measurements.
Why are you repeating things, i already quoted?
Why do you ask questions when all the time you simply discard answers and keep repeating your opinion, based on complete lack of understanding how things really work?
Which answer? Your so called "answer" did not explicitly answer any of my question from that one post. So yes, i discard that answer too! (Seems like nowadays, asking explicit questions and instisting of an explicit answer is like spitting in someones face :palm:)
I think you should start a company and start making your own scopes.
No thanks!
The hassle of communication and discussing with people is not worth the effort. And good people are hard to find, or already sit at Keysight and the like. I am good with how it is now ;-)
So in summary:
Its not possible very hard to investigate glitches / debug circuits on that scope (on other too?), because sequence mode may fail to trigger on those occurences, if you set the sample rate to low, which is something that you would do, to get a higher frame count and a longer acquire duration. (Gosh... does anyone understand this sentence? ;D)
-
So in summary:
Its not possible very hard to investigate glitches / debug circuits on that scope (on other too?), because sequence mode may fail to trigger on those occurences, if you set the sample rate to low, which is something that you would do, to get a higher frame count and a longer acquire duration. (Gosh... does anyone understand this sentence? ;D)
Yep, and the lack of understanding of how to use an analytical DSO to best advantage.
Persist is your first friend/tool, just to see if you might have a problem.
With Sample rate at max, the powerful trigger suite allows for triggering on most anything.
Persist indicates what trigger you should/might use then Search allows you to find the incidences.
-
With Sample rate at max, the powerful trigger suite allows for triggering on most anything.
Yes, but only for one time every n cycles on something you know, but i am looking for glitches and misalignments. But that seems to be to hard to understand for some. Its not s simple kind of steady signal! :palm:
-
I really don't understand why you have vendetta against all the testing equipment that you purchased.
Can you please list one testing equipment that you bought and it works the way you want it?
-
I really don't understand why you have vendetta against all the testing equipment that you purchased.
Remember, it's only paranoia if it's not all out to get him.
-
With Sample rate at max, the powerful trigger suite allows for triggering on most anything.
Yes, but only for one time every n cycles on something you know, but i am looking for glitches and misalignments. But that seems to be to hard to understand for some. Its not s simple kind of steady signal! :palm:
Read the rest of what he said. PERSIST is what you need to see the glitches and misalignments you seek. If you selectively read and won't try to comprehend what people are telling you, you'll never understand.