Siglent SDS1000X and (MSO) SDS1000X+ series oscilloscopes

[rf-loop]

SDS1000X+ MSO

If use LA together with analog signals it is also important to null time difference between analog signals and digital signals.
Many times there is different signal propagation delay to scope inputs and also internal skew between analog channels and digital channels.

Deskew can do quite easy. Connect enough digital lines and both analog channels to same physical point what have suitable signal, example square wave with well under 10ns risetime and example level around 5Vpp. (example 0 - 5V).
Adjust digital bus 1 and 2 User defined threshold to 2.5V and CH1 as trigger source, edge, rising and level 2.5V.
(in image cursor is set so that it crosses around trigger position. (you do not need cursor at all, here it is only for viasualization)
Adjust CH1 Deskew until Ch1 rising edge  is around middle of digital channels rising positions.
After then adjust Ch2 deskew so that skew is minimized. Average you can watch example using automatic measurement function "Skew" with statistics as in image.

After you go back to normal iscolloscope use remember zero these deskew settings for avoid confusions.

After deskew you can use also selected digital channel  for trig of course. Skew to analog is now ok. If you do important time measurements between analog and digital. Repeat deskew when something have changed and if you are doing important time measurements.  With very low speed MSO deskew setting is perhaps insignificant.  Skew between digital channels can not adjust and it is specified as  + 1 digital channel sample period.


Deskew.




Just for fun. (nothing tested, just one random display example)

[vpv]

New FW arrived!
http://www.siglent.com/ENs/gjjrj-xq.aspx?id=1726&tid=15

SDS1000X/X+_firmware_Update
ReleaseDate 2016/11/7 17:24:33
Version: P13.03

For SDS1000X/X+ Series
1. Increased FFT to 16K points
2. Supported saving self-calibration data.
3. Supported auto setup DC signal or signal with DC offset.
4. Fixed bug: Scope can not trigger correctly while take serial trigger mode and MSO is trigger source.
5. Fixed bug: Scope can only measure once or decode once while burst two signal.
6. Fixed bug: The self calibration info is not correct in Russian Language
7. Fixed some other bugs

Update instructions  (very important!)
It need to update twice to update the machine from 1.1.1.2.6 (or below this version) to 1.1.1.2.13R3 (or above). The first time is from 1.1.1.2.6 to 1.1.1.2.13R3, the second time is from 1.1.1.2.13R3 to 1.1.1.2.13R3 itself.

[diodak]

How can I understand this update from 1.1.1.2.13R3 1.1.1.2.13R3? Upload two times the same file?

[vpv]

How can I understand this update from 1.1.1.2.13R3 1.1.1.2.13R3? Upload two times the same file?

Yes. I did the same.

[rf-loop]

How can I understand this update from 1.1.1.2.13R3 1.1.1.2.13R3? Upload two times the same file?

In pdf inside update .rar package:

It is very clear:

8. It needs to do update twice for upgrading. Please have a look at update instructions below.

Update instructions
? Very important!
It need to update twice to update the machine from 1.1.1.2.6 (or below this version) to 1.1.1.2.13R3 (or above).
The first   time is    from   1.1.1.2.6    to     1.1.1.2.13R3,
   
the second time is from 1.1.1.2.13R3   to   1.1.1.2.13R3 itself.

Do 1st update to 13R3.

After it tell it is ready, turn power Off and then On again.

And then next round and this IS mandatory.

(your version is now 13R3 if you read info screen but update is not fully ready. You need now run update 13R3 again over this same 13R3.)

Do 2nd update using same  13R3.

After it tell it is ready, turn power Off and then again On.

Now update is ready and scope is ready for normal use.

Do not ask why two times. Just do it.

[rf-loop]

Some very simple examples about MSO display for better imagine how it looks in use.




LA example Pattern trig with Analog and Digital channels
Pattern trigger.   AND: CH1 H,  CH2 X,  D0 - D7 LLLLXXXX,  D8 - D15 XXXXXXXX






LA example Pattern trig with Digital channels
Pattern trigger.   AND: CH1 X,  CH2 X,  D0 - D7 HLLLLLLL,  D8 - D15 XXXXXXXX 
(Analog signal CH1 is from old Wavetek 23 function generator Signal output and Digital channels D0-D7 are connected  its LF (<1kHz) generation  DAC 8 data lines. Yes it have 8bit ADC. DAC0800.)






LA example Pattern trig with Digital channels. (Window zoom can use)






Waveform History buffer works also for mixed signals, Analog and Digital together.
After normal run scope stopped pushing "History" button (can stop also using Stop and after then push History for look history.  Then searched buffer and looked number 7 acquisition (most high number is latest and most low number is oldest.






Zoom works of course also when look History buffer in stop mode.






Zoomed more. (not max)






Window Zoom works also in history view. If measurements, in Window zoom mode measurements come from bottom window!






Digital channels vertical position can set "High", "Middle" and "Low"   if 8 channels is in use all these can set, if 16 channels, only High and Middle can use. (middle fills from bottom to top when 16 channel in use)
In this image selection is High.






In this image setting is Middle





Here setting is Low.

Also LA channels order can change and also can select only these channels what is in use.
Bus data display can show Hexa, Binary and Decimal. There can select what digital channels are connected to Bus data display.

Note: With this FW (13R3) when use Pattern trigger for digital channels there need be least one Analog channel on. Instead of if you use it or not.  If not use it can move out from image and leave it for "don't care" (X) state in pattern settings.

[tautech]

Some fiddles with the new FFT interface provided in recent FW.

Signal source SDG1062X 1 KHz 50mV 50
Direct BNC connection, 50 SDS1kX internal termination

There are 3 pages in the UI for FFT settings inculding 3 options for FFT display: Full screen, Split screen or Exclusive FFT display.
Settings other than are shown in the following shots: Center Frequency 1 KHz and horizontal 500 Hz/div

Full screen.
SDS1kX FS



Exclusive FFT display
SDS1kX Ex



Edit
The Horizontal Trigger position indicator (blue down pointer) hides the Center frequency as set by the user, scrolling it sideways reveals the settings made by the user.

[tautech]

So FW 13R3 seems to have been pulled and a new version 13R5 is now available:

http://www.siglentamerica.com/USA_website_2014/Firmware&Software/firmware/SDS1000X%20P13.03R5.rar

From the Changelog and update instructions:

1. Fixed bug: the wrong UART decode
2. Fixed bug: the wrong MSO threshold that case MSO function unusual

Update instructions  (Very important ! )

It need to Update twice to Update the machine from 1.1.1.2.6 (or below this version) to 1.1.1.2.13R5 (or above).The first time is from 1.1.1.2.6 to 1.1.1.2.13R5, the second time is from 1.1.1.2.13R5 to 1.1.1.2.13R5 itself.

[rf-loop]

Here some tiny fast check after this "MrWolf" test ( https://www.eevblog.com/forum/testgear/rigol-ds1000z-series-(ds1054z-ds1074z-ds1104z-and-s-models)-bugswish-list/msg1089111/#msg1089111 )

I want look what SDS1000X do in same situation.

Siglent auto measurement values added. If min max are same then only one value.

Two screenshots for visualize how "clever" oscilloscope use is when using 1ms/div for measure 10ns rise time. (10ns rise time adjusted by generator).
If I do first image measurement for rise time or period I feel like just like kids playing...  1ms/div and try measure 30us period or 10ns rise time.
It take 1second  to turn TB knob for right time base and much less time for check rise time just pushing TB knob for  zoom and get more right value independent of if it is stop or run mode.
Of course in non zoomed 1ms/div and this kind of signal, perhaps it is better Siglent show <200 ns or just ---
(it need also note that what ever memory/capture length is this whole capture length ( 14 * t/div) is  on the screen, no overlap)





1.  "MrWolf" test with Rigol and added with Siglent SDS1000X data.



2. Just for imagine what this signal is looking when 1ms/div



3. As 2.   but zoomed

[MrWolf]

It can be derived from the data that "screen sampling" formula or "end user resolution" for SDS1000X+ is 1 / (Timebase / 833.3...) 1 / (Timebase / 1666.666...) for the bulk of the data (50us - 1ms ranges). Meaning it's not doing "pixel math" and has some larger memory buffer. However it would be wise for system to display "<=" not "=" to hint operator may have to zoom. Operator has no indication that measurement accuracy is not reflected by sampling rate / memory depth. In the 1ms timebase case at 1GS/s and 14Mpts would be pretty logical to assume +-1ns precision unless manual states otherwise (does it?).

[rf-loop]

It can be derived from the data that "screen sampling" formula for SDS1000X+ is 1 / (Timebase / 833.3...) for the bulk of the data. Meaning it's not doing "pixel math" and has some larger memory buffer. However it would be wise for system to display "<=" not "=" to hint operator may have to zoom. Operator has no indication that measurement accuracy is not reflected by sampling rate / memory depth. In the 1ms timebase case at 1GS/s and 14Mpts would be pretty logical to assume +-1ns precision unless manual states otherwise (does it?).

Fully agree that there must be some warning about highly reduced accuracy. Just example <(value) or some other indication for user.

I do not know if there is sampling buffer then secondary buffer and finally image map. But one is quite sure, Siglent do not flush out (decimate) real sampled points, all is still mapped to display width, all max 14M sample points. But interesting (least for me)  is how it read true sampling buffer for measurements (I can see there some strange things in some tests) and how it then map all to 700 horizontal columns on display.

Add: After many tests using different known signals (using different pulse widths, different rise and fall times etc.. for detect maximum depth of measurement "buffer".  It looks like there is 70k.  If acquisition length is 70 there is one sample in one measurement horizontal "block". If acquisition length is over 70k samples are in same sample buffer column (or block). Example if length is 700k  then 10 samples are in same sample buffer column. When samplerate is 1GSa/s and memory is 70k (and same for meeasurement) it can clearly detect that data used for measurements have 1ns resolution (but in this case it do not have fine interpolation between real sample points. Using some more fast time bases, it use also interpolation data between true sample points (as it use oversampling / fine interpolation also of course for fine positioning for minimize trig jitter.  Also every sample is mapped to display buffer but its length is 700 points. (display waveform length (pixels)). In same case, 700k, same display column have 1000 samples (with different levels and overlayed, what also is part of intensity gradation process)  I do not have any idea if there is real separate measurement buffer (HW) or if this all is how system read sample buffer. (also it need note that even with 14Mpts all points are still always mapped to display)
 
Why they do this for auto measurements? One reason is perhaps compromise between speed and available quite limited processing power for handle data for measurements.  If there is not real fixed hardware buffer for measurement data, then there may be possible to do full resolution automatic measurements (if someone make new FW for this) even with max acq depth but with very reduced wfm/s speed due to lack of brute force for data handling. Case with PC scopes are very different. (no name)Acquisition Box only collect data and there is full power of PC for handle this captured data.

[rf-loop]

As told previously. It looks like for automatic measurements SDS1000X/X+ use 70k data.  (tests made using SDS1102X+)

If 1 channel is in use real time one shot samplerate is 1GSa/s if t/div is 1ms/div or more fast.
If in this case t/div is 5us/div sample buffer is 70k aand every separate samples have own place in buffer.  Display waveform length is always on the screen 700 pixel. Independent of time base.
If t/div is 1ms/div sample buffer length is 14Mpts. Whole buffer length is (always) same as display screen widht. All sample points are mapped to display. It do not never drop out samples. But when there is more than 700 samples then group of samples are collected to one time position on the screen pixel map. In attached example image there is 70k sample buffer. Display have  700 pixel colums (in wfm area) so there is 100 sample for one display column.

If sample buffer length is 700k then 1000 samples and so on. In this case,Measurement Buffer Legth is still 70k. What happend now. Now need push 10 samples to one place in measurement buffer. And how it do this - I do not know and I have no evidence if it just do raw decimation or something more clever. When time base is more slow, sample buffer lenght is more (if user have not set limit) This means of course that horizontal (time) resolution for measurements drops. For 700k memory 1GSa/s effective resolution for automatic measurements is 10ns.

In image there is used dot mode and slow acquisition so that every screen refresh have only one wfm. This is because I want see rising and falling edges separate sample points.

Signal from SDG5082
Pulse, period 60ns and exactly 50% duty.  Pulse 10/90 risetime and fall time is adjusted for 10ns (note: due to EasyPulse system adjustable rise and fall times produce some small jitter in risetime so it explain part of risetime s.d.

(more about this later)


(trace in image is single shot)

[MrWolf]

I do not know and I have no evidence if it just do raw decimation or something more clever.

I did put your test into proper table now. It might me "more clever" because end-user resolution seems to exceed the case with primitive decimation.
At least with Rigol it's quite straight forward Timebase / 25 (1200/4 for whole screen). But here I get Timebase / 1666.666.... Gives 93333,333.../4 for whole screen following Rigol logic. But it does not make sense hardware wise so math must be bit more clever than in Rigol case. But I know absolutely nothing about formal theory how to calculate this stuff. I just observe nature so to speak...

[analog_user]

Hi,
does anyone know if it is possible to visualize the generated waves at the same time than the input channels in Siglent oscilloscopes with integrated AWG?

[tautech]

Hi,
does anyone know if it is possible to visualize the generated waves at the same time than the input channels in Siglent oscilloscopes with integrated AWG?

Only if they are fed into a spare channel, unlike a standalone AWG where an indicative waveform is shown on the display.

[Siglent America]

I thought some people might get a kick out of this YouTube video.
The user is demonstrating a SDS1202X with a music synthesizer.

[znroot]

Hi,
I observe a very curious bug on my SDS1102X scope.
As you can see on the attached images, at startup, for the first minutes both channel have a small DC ground (about 20mV) offset, that goes to zero once the scope is working from more than 10 minutes.
The problem persist also after calibration (I calibrate it and leave it off for two hours, turn it on and it still have the DC offset), with the default Siglent passive probe disconnected or passive probe tip shorted, no matter of the channels' configuration.
Anyone have the same problem?

[tautech]

I don't know this for sure but it would seem normal behaviour as equipment warms to operating temperatures.
As the manual states a Self Cal should not be performed until the unit has been on for ~20 and this is NO different to any other scope and service manuals of all brands since the early days insisted no internal adjustments are to be made until a scope has been operating for a period of time.

Of course at a 5mV/div setting this cold offset is quite apparent and good reason why you should let equipment reach operating temp before making measurements that require higher levels of accuracy.

Nice that in Roll mode you can capture the offset decreasing as the scope warms up.   

Thanks for sharing.

Edit
Probably a good subject for Dave to do a vid on.

[znroot]

Yes, It's a very curious behavior of the instruments but it's limited because it vanish once the scope warm up.

[rf-loop]

Hi,
I observe a very curious bug on my SDS1102X scope.
As you can see on the attached images, at startup, for the first minutes both channel have a small DC ground (about 20mV) offset, that goes to zero once the scope is working from more than 10 minutes.
The problem persist also after calibration (I calibrate it and leave it off for two hours, turn it on and it still have the DC offset), with the default Siglent passive probe disconnected or passive probe tip shorted, no matter of the channels' configuration.
Anyone have the same problem?

In this case your scope drift is not ~20mV, it is ~2mV.  You have there probe multiplier 10x in use.

It need note that in images real setting is 500uV/div !
(but probe multiplier is 10 so there is displayed  5mV/div.)

There is not so many scope models where  is available true 5mV/div sensitivity using probe multiplier 10.

Using example Rigol DS1000Z  most low real sensitivity is 50mV/div if probe set for 10x.
Most sensitive 10mV/div and 20mV/div they are only zoomed 50mV/div.  Also in DS2000 there is not 5mV/div real sensitivity with probe multiplier 10x  because this is just digital zoom from 10mV/div.


This is normal. Some individual units have more and some have  less thermal drift but it is normal in this unit.
 
I have tested every single SDS1000X  what I have sold and this is normal. There is differencies between individual units, some may have even very low drift and also drift direction may vary.
It looks that your individual unit is around like median.

Also there is not  "Quick-Cal" as there is in SDS2k series  what operates automatically for compensate some drift depending time and temperature changes and  selected V/div range. (if user have set this function on, in SDS1000X there is not this function)

In SDS1000X  vertical shift adjust DAC resolution is around 250uV.  Self cal have no more resolution for adjust DC offset.

[rf-loop]

So many times there is claims that Siglent decode only from display.
If people is not familiar with oscilloscope what use:
What ever acquisition memory length is, it is always same as displayed waveform length on the screen.
Siglent do not never hide part of acquisition length outside of display window.

When Siglent decode it decode active display width, only. Yeas only. But dispplay width is whole length of acquisition memory. If it is 14M or 1.4M or what ever, whole memory length is mapped to display.

Sorry these images are by finnish language. But perhaps they can give some help.


Noname usual DSO.

Note what is blind time and real user visible blind time.
With some settings visual blind time may be lot of, due to fact that diplay width is very very narrow related to true acquisition length.
A = Display window
B = Acquisition length
C = Really blind time (user can not see, oscilloscope can not see)
D = User visual blind time is same as really blind time.
E = visible part of signal is same as whole signal what also oscilloscope can see.
G = This random peak oscilloscope can not detect at all
F = This random peak is also in displayed area.

       I
Siglent SDS1000X, 2000X


In Siglent A=B=E  and  C=D
A = Display window
B = Acquisition length
C = Really blind time (user can not see, oscilloscope can not see)
D = User visual blind time
E = visible part of signal
G = This random peak oscilloscope can not detect at all
F = This random peak is captured but user can see only if he stop and scroll this area to display
       (if you have good luck it exist in this acquisition when you stop, typically these scopes also
        do not have history buffer)

Example if sample memory length is only 1kSa  and 500MSa/s is in use and horizontally 2ns/div.
If display is 10 div width there is 10 sample period on the display and 990 sample points invisible outside of scope screen and what ever there is you do not see it. You are blid for there. Also between every aacquisition there is time what oscilloscope do not acquire and do not even wait trigge. This is real blind time if think oscilloscope. User visual blind time is this + this part of signal what is not displayed.

With Siglent you always see all what is captured. Even if memory length is 140M it is mapped to display width. There is not any reason for overlapp signal outside display. Hw based Zoom is for it.

With this kind of scope user need lear bit different practice. If you want long acquisition but still want look narrow window to some signal detail. Just use Zoom. Using Zoom window scroll function you can easy and fast move to different places in whole acquisition. Just push forward or backward button and toggle for speed. (FW 13R5)



Here example (using 1.4Mpts record lenght and SPI decoding)
(With SDS1000X/X+ maximum decode length is 1.4M.  Memory selection 14M is not in use.)



First image, with 1.4M memory. Full memory length is decoded, in decode table selecter byte near end of memory.  (as can see I have used quite slow SPI, because signal used is not native SPI signal. This is somekind of emulation using 3 channels function generators and MOSI, MISO are fake... (noise)  just for easy demonstration.




Here same signal (possible some small adjustments (no meaning) changed between this and first image. But same basics.
Zoomed and Zoom window near end of whole acquisition memory length.

[nctnico]

Now try to figure out how message number 139 relates to message number 7 AND/OR try to look at a signal edge or spike in the middle of a multi-byte SPI, I2C or CAN message without loosing the decoding. Once you understand these problems then you'll also see why 'decoding what is on screen' is so useless for any real work.

[kcbrown]

So many times there is claims that Siglent decode only from display.
If people is not familiar with oscilloscope what use:
What ever acquisition memory length is, it is always same as displayed waveform length on the screen.

Wait.  Just to be clear, you're saying that if you set your time base to 100ns per division, and you have 1GS/s acquisition rate available to you, then your memory length is automatically set to 1400 points and there's nothing you can do to increase it save for setting a different timebase??

Even if you set it up to stop after triggering?

Siglent do not never hide part of acquisition length outside of display window.

So if you hit the "stop" button and then zoom out, you get nothing??


That's, er, interesting if that's the case.

[rf-loop]

So many times there is claims that Siglent decode only from display.
If people is not familiar with oscilloscope what use:
What ever acquisition memory length is, it is always same as displayed waveform length on the screen.

Wait.  Just to be clear, you're saying that if you set your time base to 100ns per division, and you have 1GS/s acquisition rate available to you, then your memory length is automatically set to 1400 points and there's nothing you can do to increase it save for setting a different timebase??

Even if you set it up to stop after triggering?

Siglent do not never hide part of acquisition length outside of display window.

So if you hit the "stop" button and then zoom out, you get nothing??


That's, er, interesting if that's the case.

Yes.

Why need stop scope for look whole captured length using zoom out.   In my view, it is a bit clumsy way.
Stop and zoom out, stop and zoom out...  I have so limited time for live so I do not want spend it like this.

If I want full window then just stop and zoom to details using window or full screen horizontal and vertical zoom as want. But also when I stop, I can select if I look current last acquisition or previous in history buffer. (and zoom etc just as need).

Why do not look whole capture length in live without need stop scope  and if need details then view also these together with whole acquisition length.

If I need long memory I select needed amount of memory and if it do not show enough details on the display I can use zoom.   Or if I really want of course I can stop and then zoom in/out using window zoom or zoom for details using full window.

But I'm bit unhappy if I have scope what can  not show full capture length before I stop it and zoom out.

I remember time when I first meet this working principle. My mind was adapted to work with scopes what only show small portion from acquisition memory.

Both principles can do same work without any doupt. Both have pros and cons. But both need bit different driving style. Also front-wheel drive and rear-wheel drive car need bit different driving style some times. First, the user must learn and adapt to it. Same we can see in many other things also, example between RPN and arithmetic  calculator. Who have adapted for arithmetic calculator say that RPN is just terrible and useless.


In this image 14Mpts memory, whole memory visible and then your example 100ns/div. No need stop and zoom out.



Just 14Mpts memory and zoomed in more for details. Still aall time visible in real time whole acquistion length.

 
And more, If i really want stop and zoom in full window. Just if I stop here with this image. I can turn after then window zoom off and zoom in for deepest details or then zoom out. Where ever I can also turn window zoom on. In stop mode I get even more.  I can look also previous 14M acquisition and do same things. (because there is always backround working waveform history buffer).

This automatic memory length is also good for maximize amount of waveforms in waveform history buffer.
If take this 100ns/div for example. One wfm length is 1400 ADC data  points.   When I stop scope I can look 29140 last captured waveforms and they are not image frames, they have full ADC raw data. (user can change sinc or linear interpolation or dots on when use. After stop these can free select (all these are postprocess.) 

Perhaps (?) Siglent can add later also manual "force memory length" function - (if there is real need for this)
Personally I have not missed this feature anymore.

[kcbrown]

So many times there is claims that Siglent decode only from display.
If people is not familiar with oscilloscope what use:
What ever acquisition memory length is, it is always same as displayed waveform length on the screen.

Wait.  Just to be clear, you're saying that if you set your time base to 100ns per division, and you have 1GS/s acquisition rate available to you, then your memory length is automatically set to 1400 points and there's nothing you can do to increase it save for setting a different timebase??

Even if you set it up to stop after triggering?

Siglent do not never hide part of acquisition length outside of display window.

So if you hit the "stop" button and then zoom out, you get nothing??


That's, er, interesting if that's the case.

Yes.

Why need stop scope for look whole captured length using zoom out.   In my view, it is a bit clumsy way.
Stop and zoom out, stop and zoom out...  I have so limited time for live so I do not want spend it like this.

No less clumsy than selecting a 50,000 foot view because you're forced to do so in order to see the surrounding context, and then zooming in to see the actual bit of waveform that you triggered on in the first place, i.e. the method that the Siglent forces you to use.

I selected the example I did in order to make a very specific point: the Siglent won't let you zoom out even when there is no additional detail to be gained by zooming in -- it minimizes the amount of memory available to you in a given capture.  And while the method the Siglent uses gets you the maximum number of waveforms possible in the history buffer, that presumes that what you're triggering on is highly repeatable.


The best memory configuration and waveform view is dependent upon what you're trying to do, and those two things (memory configuration and waveform view) are not always best joined at the hip.

Both principles can do same work without any doupt. Both have pros and cons. But both need bit different driving style. Also front-wheel drive and rear-wheel drive car need bit different driving style some times. First, the user must learn and adapt to it. Same we can see in many other things also, example between RPN and arithmetic  calculator. Who have adapted for arithmetic calculator say that RPN is just terrible and useless.

Pretty much.  But the point here is that when you can't select the configuration yourself, you're forced to use a driving style that isn't necessarily optimized for your use case, and indeed may prove quite clumsy.

Perhaps (?) Siglent can add later also manual "force memory length" function - (if there is real need for this)
Personally I have not missed this feature anymore.

I suspect some might be surprised by Siglent's approach here.  The ability to specify the capture length wouldn't be the worst feature Siglent could add.  Indeed, there are some novel ways to specify it that they could add that might prove more intuitive or more appropriate for certain use cases, e.g. rather than directly specifying the capture length in terms of samples, you might instead specify the maximum number of waveforms to capture in the history buffer, or you might specify the capture length in terms of a multiple of the timebase or a multiple of the window size.