Siglent's new product- MSO/SDS2000 Series

[tautech]

Here attached new waveform update rate table.

Note settings. (measurements, statistics, and cursors on)
FW version is 1.1.1.26.5

If someone want measure this, I can tell that it need be really careful know how to measure trig out frequency! Specially with some settings most of cheap counters are in big troubles and leads to wrong result.

Thank you rf-loop.
This table supersedes the one I made then posted and Carrington attached.
https://www.eevblog.com/forum/testgear/siglent's-new-product-msosds2000-series/msg355782/#msg355782

Many of my results were correct but as you have stated:

Specially with some settings most of cheap counters are in big troubles and leads to wrong result.

[rf-loop]

Here more about Trig Out jitter and delay.

Before it was trigged using CH1 signal (Trig source CH1) and monitore trig out signal with CH2.

Here trigger source is EXT Trig.

Signal from signal generators ( pulses are around 5ns pulse every 0.1ms from HP8161A programmable pulse generator. Sweep is just for show and also proof that this rising edge is really trig source)

Signal is splitted so that it have around same travel time to CH1 input and to EXT Trig input.
Trig Out signal is connected to CH1 with around 5ns delay. (1.0m RG223)

Trig Out delay is far less than if use main channel as trigger source. (this was expeted)
Jitter itself is around same. (yes there are some "sweet pots" where jitter is significantly less, but these are sweet pots and in practice what is here in pictures is  normal case)

[marmad]

Trig Out delay is far less than if use main channel as trigger source. (this was expeted)
Jitter itself is around same. (yes there are some "sweet pots" where jitter is significantly less, but these are sweet pots and in practice what is here in pictures is  normal case)

Thanks, rf-loop. Very nice - that's quite a short delay.

Is the jitter always in precise increments? On the Rigol, the jitter increments are spaced at 250ps (double the max. sample rate), which I assume is being derived from the master clock.

[rf-loop]

marmad: Jitter looks random. (EXT trig used for trig and Trig out monitored using CH1 alone (2GSa/s)  (btw, have you checked this using any other tool exept Rigol? Of course it is possible but jitter 250ps interval is "too clean" so I have not strong but light suspicion.)

-------------------

Sequence mode waveform acquire speed Not acquired and realtime displayed wfms/s speed.
As told before with poor quick test, speed is up to over 400000 segments/s

Here is some amount better made tests for test waveform capture speed in Sequence mode aka Segmented memory.

Speeds are not once occured highest peak. I have repeated this test enough so that I can say it is average speed using this kind of signal and settings. Signal affect some amount of speed. There may find some freq and other variables in signal what give even some amount more speed and of course slow speed waveforms give also slow speed. (example it need find trigger event and then fineadjust signal horizontally for exact trigger etc.)
In table is also segment sizes and available max segments for every time/div settings (what change segment size until it reach user selected maximum memory (memory setting value). After this max, when go to slower horiz speed it start reduce sampling speed. Every segment is exactly 14div. (there is not out of display blind time also in normal acquire mode)

Of course after scope is stopped waveform (segment) can zoom in (and out). Segments can zoom using zoom function or changing normal window horizontal t/div. Also, if it was acquired using sin(x)x off and dots. In stop mode vectors mode can turn on and also sin(x)/x

This table is obsolete.

[rf-loop]

Some tiny test about frequency response and risetime.

Of course I know this old thumb rule for relation with frequency response and risetime.
But it is true only for pure Gaussian freq response.
Example this Siglent is not pure Gaussian or "flat response" but some kind of response between them.
http://cp.literature.agilent.com/litweb/pdf/5988-8008EN.pdf

Siglent specifications do not tell risetime. Only frequency range.
(attached image)
With this measurement it is very well inside specifications limits.

1. Attached part of specifications.

2. Attached image about response.
Signal generator HP8644B  and checked that level accuracy is much better than its specified limits.
I have used 10MHz steps. First freg is 1MHz, next 11, and then 21 etc. (10MHz steps)
-3dB points is roughly 350MHz  and flatness is not only just ok, but good.
In picture most left wide yellow
is 1MHz, then next around half div wide is 11MHz, next is 21, then 31, 41,51 etc.
Middle is 251MHz. (around 50MHz/div)


3. Attached image about risetime.
Signal from Tektronix 285 pulser. It give around 70ps rising edges.
If do measurement so that do not take overshoot at all for measurement it is around 1.16ns.
(10% and 90% points from flat bottom and flat top.)
Automatic mesurement tell that average is 1.24ns.
It is possible that TEk 284 itself have small overshoot also. But it is nearly impossible to measure with my equipments so that result can trust.

Siglent overshoot is specified <10%. If Tek284 and connection to scope input do not produce any overshoot then it is slightly over 10%.

If we look this Agilent paper and look what is typical for flat response scopes it still looks very natural.  Also 70ps is quite fast for this kind of front end.  If I connect 0.5ns or 1ns risetime signal I predict this overhoot is less.

So or so, these tiny tests show these results.

[Hydrawerk]

I might have asked before, I am not sure...
Is there an english manual for SDS2000 scope? I could not find it on their website. http://www.siglent.com/en/product/detail.aspx?id=100000008529687&nodecode=119008001
But I found it here. http://gigaplus.makeshop.jp/labros/pdf/SDS2000UserManual_V1.0.pdf
It seems that there is no Wave Inspector-like feature as we know from Tektronix. And there is no search function.

[Electro Fan]

I might have asked before, I am not sure...
Is there an english manual for SDS2000 scope? I could not find it on their website. http://www.siglent.com/en/product/detail.aspx?id=100000008529687&nodecode=119008001
But I found it here. http://gigaplus.makeshop.jp/labros/pdf/SDS2000UserManual_V1.0.pdf
It seems that there is no Wave Inspector-like feature as we know from Tektronix. And there is no search function.

Yes, it's surprising that the other manufacturers haven't followed Tektronix's path on the Wave Inspector; it looks like a very good user interface that could/should be included with almost any oscilloscope.  I think if a manufacturer added 8 digital channels or more, plus a Wave Inspector-like feature on an entry level (less than $1k) MSO with 4 analog channels we would have a big winner.

[Mark_O]

Some tiny test about frequency response and risetime.

Those are very respectable results.  Good to see the performance level confirmed.  No 'games' being played to boost their numbers.

[rf-loop]

Time difference (skev) between channels.

CH1 and CH2 are equal with each others. (group A)
CH3 and CH4 are equal with each others. (group B)

Between group A and B there is some time difference.

Later more fun around this with zoom function window what looks very different. It looks like zoom function somehow delete time errors between groups. But this need investigate some more before I can write my conclusion about it. Also even if zoom from 1ms/div (top window) to 1ns/div bottom. No meaninful jitter and more surprice, time skev disappear. Overall Siglent trigger is very stabile. I have also tested it with adding some small ~200ps delay line. It do not null external delay. It only nuu internal delay differencies between group A and B. Strange.


In these attached pictures signals to both inputs are equal splitted from same source and without any significant time error. Very accurate equal signal pathway to both inputs. (also (of course)checked by swap signals)

First CH1 and CH2  Trig in CH1.
Second image, signal to CH1 and CH3. Trig CH1.
Acquire: Normal. Trig mode: Rising Edge, Normal. Display Dots. Persistence 10s.

[tautech]

Interestingly I have just had latest pricing from Siglent and the MSO option as it has been described throughout this thread is now described as LA not MSO.
So new list of options are FG, DC, PA and LA.
Function Generator
Decoding
Power Analysis
LA Logic Analyzer 8 Ch 500 MSa/s

[rf-loop]

I'm wondering: has this oscilloscope  'roll mode' with peak-detect enabled?

Roll mode can switch on only between 100ms/div to 50s/div
In this FW version what I have:
Peak mode is not available between 100ms/div to 50s/div including also roll mode.
Peak mode is available with 50ms/div or faster and (of course due to its principle)  is effective only when samplerate is below ADC maximum in used mode.

I hope they add peak  function also to  100ms/div - 50s/div range.

Also traces transparency need think. (transparency setting related to other traces is one thing and other thing is transparency related to screen grid lines in cases where lot of area is filled by signal trace.

It is allready nice that user can select what is display overlay priority for different channels traces and it works (just push CH1,2,3,4 buttons using wanted bottom to top order)

[jimjam]

I haven't read through all the pages, so forgive me if this has been posted before.

Has someone done a comparison, or can I compare similar numbers of the samples/sec vs time base etc between SDS2000 vs Agilent DSOX2000 and DSOX3000 series? Does the SDS2000 sit right between the two agilent series?

[tautech]

I haven't read through all the pages, so forgive me if this has been posted before.

Has someone done a comparison, or can I compare similar numbers of the samples/sec vs time base etc between SDS2000 vs Agilent DSOX2000 and DSOX3000 series? Does the SDS2000 sit right between the two agilent series?

I have this from Siglent and no doubt there will be comments to its accuracy. 

Main Data Comparison

                              SIGLENT SDS2000 series                      AGILENT 2000X series
Analog channel                      2/4 ?                                                                2/4 ?
MSO                       8 integrated channels ?                           8 integrated channels ?
Bandwidth     70MHz,100MHz,200MHz,300MHz ?                     70MHz,100MHz,200MHz X
Max realtime sampling                 2GSa/s ?                                                 2GSa/s ?
Max storage depth       14M(single channel)                                50K(single channel)
                               28M?half channel? ?                              100K?half channel? X
Segmented Memory      sequence function ?                                                 support ?
Waveform recording and
analysis                         history function ?                                                     None X
Power analysis                          support ?                                                   support ?
Max waveform refresh
rate                                         110000 ?                                                     50000 X
Built-in function/
arbitrary wave generator
      1uHz?25MHz Support arbitrary wave ?          0.1Hz?20MHz Support arbitrary wave X
Waveform operation    Add, subtract, multiply, divide,
      FFT, integral,Differential, square root,?                      Add, subtract, multiply, FFT X
Statistical
measurement            Current value, average,
                   minimum,maximum,Standard deviation,
                                  statistical times?                                                        None X
Automatic measurement
parameter                              32 kinds ?                                                   25 kinds X
Trigger mode
             9 kinds?Edge, Pulse Width,                             4 kinds?Edge, Pulse Width?                                                                                           
       Video,Slope, Interval,Runt, Window,                                            Pattern?Video X
                                 Pattern, Dropout ?
Serial decoding/trigger
  Support I2C, SPI, UART/RS232,CAN,LIN ?                                       Support I2C, SPI, ?
Input impedance                    1M?/50? ?                                                          1M? X
Display sizes and
resolution                8 inches?800*480 ?                                    8.5 inches?800*480 ?
Display range                  8*14 division ?                                               8*10 division X
Price Good price ?                                                       8000~24000 RMB ?highprice? X

E&OE.

Edit
? = tick
Copy & paste did not copy ticks. 

[rf-loop]

I haven't read through all the pages, so forgive me if this has been posted before.

Has someone done a comparison, or can I compare similar numbers of the samples/sec vs time base etc between SDS2000 vs Agilent DSOX2000 and DSOX3000 series? Does the SDS2000 sit right between the two agilent series?

I have not seen tables where is 1:1 compared these.
Information need collect to many sources and then is question... are these numbers comparable.

Also, wfms/s  tell sometimes only half truth. Other important is visible time and not visible time ratio. (blind time) It need also care that it is calculated using realtime visible part of waveform. (in Siglent captured lenght is equal with visible area of TFT and same in many other scopes in fast mode. But not at all scopes use this. In some cases TFT area may be only very very tiny time slice from captured waveform. In this case wfms/s speed may be also quite slow but even more bad visible blind time is really hopeless (explabnation *)  )

It need also be careful with these "up to" wfms/s values. Sometimes they are told as maximum peak speed without perfect information how it is measured. (input signal, settings, and also important - what is time period) 
Example peak value of Siglent is well over 110kwfms/s. (if I tell example that "typical" maximum value is 114kwfms/s I do not lie.  I have not even looked what is maximum peak value in short time. JUst because it is not useful if think real use)
But in my table is used one second average. (or longer time average for low t/div speeds)
We can measure between these limits: Average over infinite minus some time or  based to shortest time interval between two trigger event.  It is because measuring method is not "standardized". 

Here you find some tables for Agilent 2000x and 3000x.
(3000x is far more fast than Siglent SDS2000)
http://cp.literature.agilent.com/litweb/pdf/5989-7885EN.pdf

But when look Agilent papers, they are famous about biiased papers and there need be careful when they compare competitors.  Also there can not find exact information how tests were arranged for these tables.  Do they try find and use best settings for competitors but is is sure they find just best settings and signals what set they own scopes to as good position as it can.
This can find also when Tektronix do same.

Example after autosetup different scopes use different "defaiults" and perhaps settings have different approach about what user hope.

Example (imagined)  if one scope set default and autosetup in fast timebases vectors off and other use vectors... or in some cases same for sin(x)/x. Then I make test between these without any lies.
I tell what is signal and I tell both scopes are set for factory default, then pressed autosetup and adjusted after then so that timebase is equal.  I do it because I know my scope use optimal settings for fastest wfms/s and other scope default do not but use better settings for some other things than max wfms/s . With optimal individual settings both may have same speed but I did not lie when I write competitor speed was only barely half.

There is so much possibilities.... 

----
(explanation *) )
If scope have example 10div display horizontally and sampling speed is 1GSa/s and timebase is set for 5ns/div. Then we know that scope memory is 4k (fixed).
Visible time is 50ns. It means also 50 sample points. But every horizontal "sweep is 4ksample points. It is 4000 sample points and 4000ns (4us).  Now if we look that it have 2500 wfms/s it means that every 4000us it capture one waveform. Waveform itself is 4us and it is "blind" 3996ns.
4000:4 ratio. (1000:1)

But wait a moment... when I watch display... I can see only 50ns slice and rest parts are out of display area and interval is 4000us. I can see 50ns part and then I'm blind 3999950ns.
4000000:50 ratio (80000:1)

Now it depends what we are doing. If I'm doing real time visual watching and try find some rare exeptions in signal. I'm just interest about this not visible/visible time ratio.

If you see blind time numbers be careful to look what it really means.
Mostly they are right but it is better to check.

[nctnico]

I'm wondering: has this oscilloscope  'roll mode' with peak-detect enabled?

Roll mode can switch on only between 100ms/div to 50s/div
In this FW version what I have:
Peak mode is not available between 100ms/div to 50s/div including also roll mode.
Peak mode is available with 50ms/div or faster and (of course due to its principle)  is effective only when samplerate is below ADC maximum in used mode.

I hope they add peak  function also to  100ms/div - 50s/div range.

Did you ask Siglent already? Roll mode without peak detect is quite useless in most cases. The beauty of roll mode with peak detect is that you get a scope with ZERO deadtime and infinite waveforms/second.

It is allready nice that user can select what is display overlay priority for different channels traces and it works (just push CH1,2,3,4 buttons using wanted bottom to top order)

My Tektronix TDS744A from 1994 can do that too. The selected trace gets slightly brighter.

[rf-loop]

Did you ask Siglent already?

Why? I do not need ask Siglent what I see exactly on my table with my own eyes.
Also as I wrote: "I hope they add peak  function also to  100ms/div - 50s/div range."

You know, this is not question. I hope they do it (or repair it if it is bug).

[marmad]

The beauty of roll mode with peak detect is that you get a scope with ZERO deadtime and infinite waveforms/second.

Interesting thought... albeit obviously a physical and mathematically impossibility. Roll mode, just like every other mode on a DSO, has a finite sample rate and thus a fixed wfrm/s rate and blind time.

[Hydrawerk]

On Agilent scopes you usually cannot set the acquisition length. For me it is not a problem, but most other scope manufacturers allow you to set the length (Rigol, Tek, GW Instek, Siglent, LeCroy...)

[nctnico]

The beauty of roll mode with peak detect is that you get a scope with ZERO deadtime and infinite waveforms/second.

Interesting thought... albeit obviously a physical and mathematically impossibility. Roll mode, just like every other mode on a DSO, has a finite sample rate and thus a fixed wfrm/s rate and blind time.

Peak detect works at the maximum samplerate so you'd see every spike which is within the bandwidth limit of the scope. When using roll-mode you'd see every spike which comes along without blind time.

[nctnico]

Did you ask Siglent already?

Why? I do not need ask Siglent what I see exactly on my table with my own eyes.
Also as I wrote: "I hope they add peak  function also to  100ms/div - 50s/div range."

You know, this is not question. I hope they do it (or repair it if it is bug).

Sorry; I wasn't clear. I meant: did you tell Siglent that roll-mode without peak detect is not very useful?
You did some test on early firmware so I suppose Siglent listens to your feedback.

[marmad]

Peak detect works at the maximum samplerate so you'd see every spike which is within the bandwidth limit of the scope. When using roll-mode you'd see every spike which comes along without blind time.

You really believe the DSO doesn't have any blind time when in roll-mode? You want to post a link to your source for that information? But I won't hold my breath...    Rather like infinite wfrm/s.

[rf-loop]

Peak detect works at the maximum samplerate so you'd see every spike which is within the bandwidth limit of the scope. When using roll-mode you'd see every spike which comes along without blind time.

The claim is incorrect.



Although there is not same principle of blind time what is typically discussed related to wfms/s speeds. But there is other kind of "blindness".
Example, you see one glitch (or spike) and you believe there is one. There may be ten or thousand glitches but it is blind for these.

Simply for clarify.
Example. (and this is very simplified) Some very slow horizontal speed and peak mode on.
Scope may show example 100 sample/second "sampling speed" ("sample" after every 10ms).  Still typically ADC works with its nominal speed behind this and drops enough samples out and take only one sample after every 1/100s. Most of samples just dropped out. 
In peak  it watch  data stream continuously but still plot to display after every 10ms . Say example it have just plotted min-max result of one time point. Then next 10ms it agen watch full speed sampled data and collect min max in this new 10ms "frame" and plot result to screen. Then start next 10ms frame.... and so on.  Now, say example in time point 10ns after one 10ms timepoint occurs 10ns peak to 5V and then rts. What happend next. Time flows and there is still left 9980ns to next time point. Now there is other 10ns 5V peak. But min-max do not change. Then next peak, say example 50ns 4.5V starting in time position 5000ns after last 10ms time point. Still result is same. There is 0 and 5V in min max memory. Time flows and under 5V peaks occurs. We arrive to next 10ms time point and plot 0 and 5V dots. It can detect only one max and min inside these 10ms time frames. 

It can think that there is "box" or better say rectangle where left and right walls are these 10ms time points and bottom of rectangle is detected lowest value and top of box is highest detected value. Now you can print black over this box. It is totally blind for all other things what occurs inside this box including also of course time when this min-max occurs inside this box..
(of course there may be more or less different solutions how it is exactly done by different manufacturers.)

You can test it.  Set scope where is roll mode with peak detect working... set time scale so that it tell example 100sample / s.

Do 5 x 10ns (10ns high and 10ns low) pulses where every pulse is different level, say example 5V:4V:3V:2V:1V,  bursts and burst interval example 100ms. if your scope is 1GSa/s and 100MHz model.

Look what you see. Stop, zoom or what ever to find what you can tell about signal..
 
After this lesson I ask: Is this claim true or false: "you'd see every spike which is within the bandwidth limit of the scope"

[rf-loop]

On Agilent scopes you usually cannot set the acquisition length. For me it is not a problem, but most other scope manufacturers allow you to set the length (Rigol, Tek, GW Instek, Siglent, LeCroy...)

Siglent can do some selections but there is many limits.
Memory selection menu, choices are 7k, 14k, 70k, 140k, 700k, 1,4M, 7M, 14M  (and if in use is one channel in one input group these values are multiplied by 2.)

For some simplification I use in this example situation where is selected two channels in one channel group. (so CH1 and CH2 is on)

Now 1ns/div speed it use 14 sample points memory.
when adjust to slower horizontal speeds it take more and more memory to use.
Now if user have selected 14M memory depth.
From 1ns/div 14ponts, 2ns/div 28 points, 5ns/div 70 points, 10ns 140 points, 20ns 280points, 50ns 700points and so on... until with 1ms/div it reach its maximum 14Mpoints.  In whole this range it use 1GSa/s for both channels.

 When go to over this to slower speeds

In this point system change to sampling speed reduce (due to max limit of memory)
Now if go slower speed then with 2ms/div it use still 14Mpoints but 500MSa/s. Now it is area where is start reduce samplerate when go more and more slow speeds but keep selected memory depth. Until reach 50ms/div where is now sampling speed 20MSa/s and more slow until it reach limit 50s/div with 14M memory and 20kSa/s.
Sarting from 100ms/div and slower there can select also roll mode. Roll mode have same sampling speed and memory depth what can use also with normal mode.

When work in this blue range, user can use all  these 7k, 14k, 70k, 140k, 700k, 1,4M, 7M, 14M  settings.

In brown range, if scope use example 140k memory (it happends at 10us/div speed with CH1 and 2 selected) Now user can reject memory to 70k, 14k or 7k.
If user do this, example reject it with this speed to 14k then scope reduce samplerate to 100MSa/s (of course).
But if user want more memory than this 140k with this speed setting, it is not possible.
One waveform is always one screen and outside of screen is not invisible part of waveform.

With different selections what channels are in use there is some differencies. (memory amount and in point where it change from this example blue and brown range)


Good or bad? It depends what people like or is previously adapted. If reason is previous learned routines, then if can learn away old practice perhaps it feels better. This is case always when in lab people change from other equipment to other equipment where is different using logic.

If need more long capture then can use example more slow speed (in this brown range with full samplerate) and zoom in for details. Zoom works only using splitted window. (and due to very good stable low jitter digital trig, this zoom is really useable (Siglent claim that zoom is HW based - what it means I do not exactly know)
Pity there is not selection where zoomed window can switch to full screen mode. (and this is where Owon solution for zoom is (partially) nice (same as in some old HP). But not so nice because there is not splitted window zoom (as also true in this one old HP))

(this is true with FW version 1.1.1.26.5)

[nctnico]

Peak detect works at the maximum samplerate so you'd see every spike which is within the bandwidth limit of the scope. When using roll-mode you'd see every spike which comes along without blind time.

The claim is incorrect.

Look what you see. Stop, zoom or what ever to find what you can tell about signal..
 
After this lesson I ask: Is this claim true or false: "you'd see every spike which is within the bandwidth limit of the scope"

Let's not go into grammatical nitpicking. When hunting for spikes you want to know if they are there. Roll mode with peak detect showing spikes will tell you 2 things: there are spikes and what trigger level to use. From there you go further into detail to see if it is one spike or more etc.

@marmad: if roll mode has blind time then it is not roll mode. It's simple as that.

[marmad]

@marmad: if roll mode has blind time then it is not roll mode. It's simple as that.

Blind time means any time when a DSO is processing /displaying samples and not acquiring - and I've yet to read documentation which claims that a DSO has zero blind time in any particular mode. So backup your claims with documentation or quit making these pronouncements - they just sound silly.

Let's not go into grammatical nitpicking.

Then I think you should stop writing things "... you get a scope with ZERO deadtime and infinite waveforms/second." which is just begging to be nitpicked.