Author Topic: Siglent - 11/20 - New SDS1104X-U, 4 channel 100MHz, 1Gsa/s economy oscilloscope  (Read 16687 times)

0 Members and 1 Guest are viewing this topic.

Offline 2N3055

  • Super Contributor
  • ***
  • Posts: 3639
  • Country: hr
Scope timebases on Keysight 3000T range from 5E-10 to 50 seconds per div.. It is hard doing tricks like that with 11 orders of magnitude of time span.
Also, you forget about measurements, those have to be optimized for speed and for that you want streamlined data formats and flow.
Also you cannot retrigger while capturing unless you have datapath that will change where it writes data all the time..

Capturing goes into circular buffer of certain size, all the time. Trigger engine only leaves the marker where trigger happened, and then you rollover and keep capturing until you reach trigger point from the other side. Than you start dumping data on other memory location, leaving last data buffer for display/measurement engine to process it. That is also your history buffers.

Or you can do what Keysight does, always getting absolute minimum of data points and reconfigure to single on STOP. And you shuffle and do tricks inside Megazoom chipset. That are possible only because very small memory. If you were to make 10-MEgazoom or 100-Megazoom chip, it would have to work differently. It's literally 100 times more processing than what 1-Megazoom chip has to do. Not to mention fixed low res display window compared to higher resolution displays...
 

Online kcbrown

  • Frequent Contributor
  • **
  • Posts: 625
  • Country: us
Scope timebases on Keysight 3000T range from 5E-10 to 50 seconds per div.. It is hard doing tricks like that with 11 orders of magnitude of time span.

Why?  The timebase, combined with the memory depth, ultimately determines the sampling rate.  The sampling rate is set so as to ensure that the time covered by the buffer is never less than the time represented by the screen.  The algorithm I described is independent of the sample rate.


Quote
Also, you forget about measurements, those have to be optimized for speed and for that you want streamlined data formats and flow.

Sure, and that would place some additional demands on at least some of the data.  I fully expect that the sampled data would have to be decimated and processed in real time by the FPGA.  But these are separate processing paths.

I admit that if you're using DRAM, then a memory controller that makes possible the kinds of parallel processing of data needed here might be quite a challenge to design.  Decimation into a separate buffer (likely internal to the FPGA) would have to be able to keep up with the fastest sampling rate the scope is capable of.   Fast SRAM is more expensive but might be needed for something like this if the sample rate is high enough.  You might even incorporate SRAM into your DRAM memory controller design and use that as a write-through cache to satisfy reads when possible.

Quote
Also you cannot retrigger while capturing unless you have datapath that will change where it writes data all the time..

Why is that?  As you note, a trigger event merely sets a marker.  That happens while capturing.  The implementation I described requires only a single trigger event location pointer, which always points at where the last trigger event occurred.

Quote
Capturing goes into circular buffer of certain size, all the time. Trigger engine only leaves the marker where trigger happened, and then you rollover and keep capturing until you reach trigger point from the other side. Than you start dumping data on other memory location, leaving last data buffer for display/measurement engine to process it. That is also your history buffers.

That's exactly what I was describing above.  The primary difference is that what I described would re-arm the trigger after the acquisition had passed the location boundary of the display, rather than the end of the buffer, and whether or not the buffer location for the next acquisition would be reset to the start of the current buffer section of N points would depend on whether a trigger event occurred within the last N points (if no event occurred within the next N points then it would reset the acquisition location to the start of the existing buffer area because the rest of the buffer area might contain a trigger event and that would need to be preserved).

Think of the implementation I described as a conditional circular buffer implemented within another circular buffer that's twice the size.  The point of it is to ensure that you always preserve a memory region of N points that contains the last seen trigger event within it, as long as at least one trigger event was seen at all, as well as to ensure that the trigger fires, and the resulting processing happens, as often as possible.

My question is: how is what I described so different from what scopes currently do that what I described is not possible to implement with the hardware used by low-end scopes?   What mandates that the trigger must fire at most once per amount of time represented by the memory depth and sample rate combination?
« Last Edit: February 04, 2021, 07:40:56 pm by kcbrown »
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 21241
  • Country: nl
    • NCT Developments
My question is: how is what I described so different from what scopes currently do that what I described is not possible to implement with the hardware used by low-end scopes?   What mandates that the trigger must fire at most once per amount of time represented by the memory depth and sample rate combination?
Simplicity and low return on investment. See how much discussion already is sparked by relatively basic acquisition schemes. Besides that the search function found on many oscilloscopes nowadays can look for trigger events in the acquired data (either in a single record or segments). This is similar to what you are proposing.
« Last Edit: February 04, 2021, 08:47:59 pm by nctnico »
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Online kcbrown

  • Frequent Contributor
  • **
  • Posts: 625
  • Country: us
Simplicity and low return on investment. See how much discussion already is sparked by relatively basic acquisition schemes. Besides that the search function found on many oscilloscopes nowadays can look for trigger events in the acquired record. This is similar to what you are proposing.

I guess so.  That's especially true if it requires designing a custom DRAM controller.
 

Online tautech

  • Super Contributor
  • ***
  • Posts: 21651
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
Talk about cutting it fine  :phew: .....finally got around to checking the SDS1104X-U BW after spotting Defpom reporting is as only a little over its rated 100 MHz BW and indeed it just barely makes spec.
Source SDG6022X

Sweep 0-200 MHz



-3dB point ~103 MHz
Avid Rabid Hobbyist
 
The following users thanked this post: dmulligan

Online tautech

  • Super Contributor
  • ***
  • Posts: 21651
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
New firmware for SDS1104X-U.

Version: V1.1.5R6
8.3MB
https://int.siglent.com/upload_file/zip/firmware/Oscilloscope/SDS1104X-U_1.1.5R6_EN.zip

Release notes
Support EasyScopeX
Avid Rabid Hobbyist
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 3581
  • Country: cn
  • Born in Finland with DLL21 in hand
Talk about cutting it fine  :phew: .....finally got around to checking the SDS1104X-U BW after spotting Defpom reporting is as only a little over its rated 100 MHz BW and indeed it just barely makes spec.
Source SDG6022X

Sweep 0-200 MHz



-3dB point ~103 MHz


Because oscilloscopes are individuals and also test setups. Here other  same kind of test. Note that BW shape looks bit better.
-3dB >110MHz  But -1dB point clearly higher.
Used short RG316 and 50ohm feed thru and signal source also some SDG6kX
It is very good BW is rejected to just over 100MHz because fNyq. 125MHz and oscilloscope type of Sinc interpolation need some "free air" before Nyquist wall. Less aliasing = better trusted measurements with unknown signals. Even -6dB point is far over fNyquist leading strong aliasing. But, this is entry level cheap scope and this happen only if 3 or 4 channels simultneously in use. With 2 channels in use fNyq. is 250MHz and 0.8fNyq is 200MHz where attennuation is ok in this class. But still user need know FFT may display quite strong aliases. If need avoid these, then need use external LPF. (and for digital filters fanboys, digital filter do not know if signal after ADC is alias or not, they do not have any label "hey I am alias"... only good (and expensive) place for filter is before ADC. ) 


(you see small unsymmetry so -3dB is not just as cursor 11.52 (115MHz) reading is. True is bit below.
Sidenote. Why I use this trigger mode and settings. Because it trigs rock solid even when I attenuate signal down to -12dB from sweep start reference level. So I do not need adjust trigger when I change level in wide range.
If this scope BW -3dB point is even bit higher I recommend modify its front end for more narrow BW or Siglent implement to FW some warning message about possible aliasing when 3-4 channels is in use, because its analog BW violates Nyquist-Shannon.

« Last Edit: May 12, 2021, 11:45:27 am by rf-loop »
If practice and theory is not equal it tells that used application of theory is wrong or the theory itself is wrong.
-
Huawei HarmonyOS 2.0  |  ArcFox Alpha S
 

Online mawyatt

  • Frequent Contributor
  • **
  • Posts: 910
  • Country: us
If need avoid these, then need use external LPF. (and for digital filters fanboys, digital filter do not know if signal after ADC is alias or not, they do not have any label "hey I am alias"... only good (and expensive) place for filter is before ADC. ) 

This is conventional thinking, requiring an analog antialiasing filter before ADC which can get quite involved depending on how close to Nyquist and what post ADC signal degradation is acceptable. A good example of these analog filters was in the original Sony CD players before oversampling was later introduced. This analog filter had to suppress almost 100dB at 1.1 band transition (22KHz/20KHz) which later oversampling relaxed considerably. Recall analyzing this Sony analog anti-aliasing filter, which used polystyrene precision caps, bobbin wound air-core inductors, and complex active circuitry to keep the aliasing and distortion over 100dB down!! Our interest wasn't for audio, but we were dealing with signals that needed to be preserved with over 100dB fidelity, and aliasing and distortion posed a similar issues. The Sony filter was indeed a work of engineering art :-+ 

Over a decade ago advanced research work at USC produced a completely new and different type ADC where the waveform being observed creates it's own Nyquist limit. This ADC was called Non-Uniform Sampling (NUS) ADC and had the unique feature of sampling the input waveform in both amplitude and time, which produced a result that was in fact post processed with antialiasing filters in the digital domain, so the waveform did sort of say "hey I am alias"...   :)

This was all part of an advanced research project that involved break-thrus in multiple disciplines, and back in ~2012 seeing a 1GHz signal sampled by the NUS ADC with stunningly good results and thinking, "this would make a superb ADC for a RTSA or DSO". Keysight was very interested in this NUS ADC, so maybe we may see these trickle down to our instrumentation world soon ::)

Best,

 

Research is like a treasure hunt, you don't know where to look or what you'll find!
~Wyatt Labs by Mike~
 

Offline tv84

  • Super Contributor
  • ***
  • Posts: 2236
  • Country: pt
New firmware for SDS1104X-U.

Version: V1.1.5R6
8.3MB
https://int.siglent.com/upload_file/zip/firmware/Oscilloscope/SDS1104X-U_1.1.5R6_EN.zip

Release notes
Support EasyScopeX

This is PRODUCT_ID 17001.
 

Offline QuitButton

  • Contributor
  • Posts: 34
  • Country: gb
Have I got this right or am I missing something here?  Timebase of 2ns, I only get 28 samples. (or 14 or 7 with other channels on)

Is that the limit or anyway of increasing the number?
 

Offline tttonyyy

  • Regular Contributor
  • *
  • Posts: 54
  • Country: gb
Have I got this right or am I missing something here?  Timebase of 2ns, I only get 28 samples. (or 14 or 7 with other channels on)

Is that the limit or anyway of increasing the number?

14 horizontal squares of 2ns = 28ns across the screen width.   So you have one sample per ns which is 1GSa/s, which is right.

The number is only increased by opening the wallet wider :)
« Last Edit: June 18, 2021, 02:30:49 pm by tttonyyy »
 

Offline QuitButton

  • Contributor
  • Posts: 34
  • Country: gb
I'm probably not wording things right. My old Hantek somehow manages to display 800 dots from the 40 samples it takes at 2ns timebase. I guess its averaging even though it shouldn't be, or there's some basic difference in the operation between these models that I'm not understanding yet.

[attach=1]

It also has an Equivalent-Time capture that the Siglent seems to be missing, which allows the creation of relatively useless captures like this one:

[attach=2]

Edit: No, the 800 display dots is not just the width of the display...
« Last Edit: June 18, 2021, 03:47:33 pm by QuitButton »
 

Offline QuitButton

  • Contributor
  • Posts: 34
  • Country: gb
Can someone confirm what I'm seeing here isn't me going bonkers and its just "how things are"?

On my Hantek, at 2ns in Dots display I get this:

[attachimg=1]

That's obviously a lot more than the ~28 samples it should be at 1GS/s so I'm guessing that either a) Hantek fill in the gaps (its not really "dots"  is it?), or b)  Hantek do more than one sample per single shot, or c) Something is happening that I don't understand.

Siglent give me this sort of display tt the same timebase:

[attachimg=2]

I can though, get something close the Hantek's display by using Sequence capture on the Siglent, thus:

[attachimg=3]

Is that what Hantek do?



 

Online bdunham7

  • Super Contributor
  • ***
  • Posts: 2642
  • Country: us
If you have more dots than mathematically possible during a single capture, the scope is overlaying multiple captures. 
A 3.5 digit 4.5 digit 5 digit 5.5 digit 6.5 digit 7.5 digit DMM is good enough for most people.
 

Offline tttonyyy

  • Regular Contributor
  • *
  • Posts: 54
  • Country: gb
That's the answer.  This only works with repetitive signals.  If you single trigger the scope you'll see the true reflection of number of sample points on screen.

Edit: there's a great explaination here:
https://uk.tek.com/document/application-note/real-time-versus-equivalent-time-sampling

Which is why one of your captures shows apparent noisey edges.
« Last Edit: June 19, 2021, 06:36:26 pm by tttonyyy »
 

Online tautech

  • Super Contributor
  • ***
  • Posts: 21651
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
Can someone confirm what I'm seeing here isn't me going bonkers and its just "how things are"?

Siglent give me this sort of display tt the same timebase:


If you have more dots than mathematically possible during a single capture, the scope is overlaying multiple captures. 
If you single trigger the scope you'll see the true reflection of number of sample points on screen.
Both these ^^
However it's evident QuitButton's capture was manually instigated whereas if it was done with a Single trigger one sample point will be aligned with both the horizontal and vertical trigger points.
Avid Rabid Hobbyist
 

Online bdunham7

  • Super Contributor
  • ***
  • Posts: 2642
  • Country: us
However it's evident QuitButton's capture was manually instigated whereas if it was done with a Single trigger one sample point will be aligned with both the horizontal and vertical trigger points.

I believe that is not true--and with an interpolated trigger where the trigger does not in any way control the sampling clock, how could it be true?  I'm not sure what you mean by 'manually instigated', but you can see that there are two dots that when interpolated by sin(x)/x would plausibly be right at the trigger point, but there's no sample there.
A 3.5 digit 4.5 digit 5 digit 5.5 digit 6.5 digit 7.5 digit DMM is good enough for most people.
 

Online tautech

  • Super Contributor
  • ***
  • Posts: 21651
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
However it's evident QuitButton's capture was manually instigated whereas if it was done with a Single trigger one sample point will be aligned with both the horizontal and vertical trigger points.

I believe that is not true--and with an interpolated trigger where the trigger does not in any way control the sampling clock, how could it be true?  I'm not sure what you mean by 'manually instigated', but you can see that there are two dots that when interpolated by sin(x)/x would plausibly be right at the trigger point, but there's no sample there.
Run/Stop vs Single.
Avid Rabid Hobbyist
 

Online bdunham7

  • Super Contributor
  • ***
  • Posts: 2642
  • Country: us
Run/Stop vs Single.

The only difference I see between Run/Stop and Single is that the segmented mode only works in Single.  Either way, at 1ns/div I get 1 sample per division (except with segmented mode I get multiples at random locations relative to each other) and the samples can be anywhere.
A 3.5 digit 4.5 digit 5 digit 5.5 digit 6.5 digit 7.5 digit DMM is good enough for most people.
 

Offline tttonyyy

  • Regular Contributor
  • *
  • Posts: 54
  • Country: gb
I'd not really thought about this before. I would guess that the scope firmware just sees that two successive samples cross the trigger threshold, and linearly (or sin(X)/X) interpolates where that crossing point would be in time and shifts the display of all sample positions appropriately to make the zero point line up like we see above.
 

Online tautech

  • Super Contributor
  • ***
  • Posts: 21651
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
I'd not really thought about this before. I would guess that the scope firmware just sees that two successive samples cross the trigger threshold, and linearly (or sin(X)/X) interpolates where that crossing point would be in time and shifts the display of all sample positions appropriately to make the zero point line up like we see above.
Good, so some more for you to ponder on.
These scopes (all DSO's) covert the input analog signal to an entirely digital stream (data points) where in modern Siglents an entirely digital trigger awaits a data point that meets its threshold before interpolation is applied to reconstruct the waveform into something we all understand and expect to see.
For the most part you can use these entirely in Dot mode although for screenshot captures and Single shot it's better to be in Vector mode so to have interpolation applied.
Avid Rabid Hobbyist
 

Offline QuitButton

  • Contributor
  • Posts: 34
  • Country: gb
This only works with repetitive signals.  If you single trigger the scope you'll see the true reflection of number of sample points on screen.

That's just it, this IS (supposed to be) a single shot capture. Which doesn't make sense why Hantek shows the result it does.
 

Offline StillTrying

  • Super Contributor
  • ***
  • Posts: 2714
  • Country: se
  • Country: Broken Britain
I'd not really thought about this before. I would guess that the scope firmware just sees that two successive samples cross the trigger threshold, and linearly (or sin(X)/X) interpolates where that crossing point would be in time and shifts the display of all sample positions appropriately to make the zero point line up like we see above.

I've looked at loads of scope shots where there's only 2 or 3 full ADC speed samples per division trying to determine if the trigger interpolation is linear or sinx, most of the time the X timing looks too good for the trigger interpolation to be just a linear line between the 2 samples.

It's difficult to tell much when the trigger is in the middle of a fast edge where any interpolation is going to be very nearly a straight line between the 2 nearest samples anyway.

We need someone to experiment with dots + persistence ~2 samples division and the trigger nearer the false pre and over shoot that sinx creates when there's not enough samples on the edge. If the middle of the fast edge now wobbles we'll know the trigger interpolation is exactly the same as used for the displayed trace, - I think. :)
CML+  That took much longer than I thought it would.
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 3581
  • Country: cn
  • Born in Finland with DLL21 in hand
Here is 100MHz sine in CH4. Also other channels are on for reduce samplerate to 250MSa/s. (Ch 1-3 trace display hidden)



Attached image 1.
First one shot with Sinc interpolation. Persistence infinite.





Attached image 2.
Then same acquistion but turned display for dots only. (note that Siglent do not produce fake dots as many other scopes do in this case)
Now it still keep this persistence but its intensity is now reduced and real ADC sample dots are visible.
If trig is based to linear interpolation there can see where trigger position is in this case. And if this interpolated trig position is now placed to trigger position we can see really high horizontal positioning jitter because samples are randomly in different positions in every sequential acquisitions. But as next image show trigger is rock solid.




Attached image 3.
All same but now just running with infinite persistence and trigger position is rock solid and as can see in image 2 samples can be where ever.



--------
This last example (SARI) is old, made with SDS1104X-E , (it works also in X-U but with bit reduced performance due to different trigger system, due to single ADC system)
Also when I read this thread... one tiny tip. In Siglent there is two different acquisition modes. Fast and slow.
Slow mode works like conventional DSO.  one acquisition and display... one acquisition and display...
Fast mode... working as DPO (Siglent name SPO) depending timebase and signal, trigger etc... it acquire as many acquistion as it can before need display. Display interval is normally 40ms. In some cases it can acquire up to thousands acquisition in one display interval and they are displayed overlaid in this one display frame.

 

Note in image what is sampling speed and what is signal. This is NOT  Equal Time. This is Siglent SARI (Sequential Acquistion Random Inteleaving) what is also not LeCroy RIS mode but somehow tiny tiny bit its cousin. Siglent have not advertised it in any place, and also no need advertise. But experienced user of course can use it because he also then understand its all limits without ranting this and that. But this is useful in some special cases when Sinc may produce Gibbs ears when working with samplerates.
« Last Edit: June 21, 2021, 01:03:35 pm by rf-loop »
If practice and theory is not equal it tells that used application of theory is wrong or the theory itself is wrong.
-
Huawei HarmonyOS 2.0  |  ArcFox Alpha S
 
The following users thanked this post: StillTrying

Online bdunham7

  • Super Contributor
  • ***
  • Posts: 2642
  • Country: us
Note in image what is sampling speed and what is signal. This is NOT  Equal Time. This is Siglent SARI (Sequential Acquistion Random Inteleaving) what is also not LeCroy RIS mode but somehow tiny tiny bit its cousin. Siglent have not advertised it in any place, and also no need advertise. But experienced user of course can use it because he also then understand its all limits without ranting this and that. But this is useful in some special cases when Sinc may produce Gibbs ears when working with samplerates.

You can get the same result with slow mode and persistence, but either case the usefulness seems quite limited.  Your sample rate still has to be pretty close to the desired bandwidth otherwise this too falls apart.

I take it that this example demonstrates that when reducing the sample rate to fit in a smaller memory space, the scope does the sinc trigger interpolation before decimation? 

Your statement that Siglent doesn't intentionally lie about dots seems true as far as I can tell--but the reason the scope can't do something like this example on aliased signals is because it can't properly place the dots in the correct horizontal position because it is relying on a sinc interpolation of an aliased signal.  So it is precise but incorrect about the horizontal position, resulting in a very clean but wrong aliased signal, dots or vectors don't matter.  In this example, it can place the dots correctly even though they are over Nyquist for the decimated sample rate--and when overlaid, the represent the signal correctly. 
A 3.5 digit 4.5 digit 5 digit 5.5 digit 6.5 digit 7.5 digit DMM is good enough for most people.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf