Siglent SDS2000X Plus

[Performa01]

Yes, the SDS1004X-E provides up to 54Mpts per channel pair, the SDS5000X has up to 433Mpts per channel pair - and both have come as a surprise, since I don't think it has ever been specified that way.

Yet the SDS2000X+ is actually limited to <200Mpts per channel pair with the current pre-release FW.

I think this is still very acceptable for its class and price range - there is just no nice surprise in the form of a hidden present in this regard.

[edigi]

Keysight offers 70, 100 and 200MHz models, R&S has 70, 100, 200 and 300MHz models. There are quite some below 200MHz, right?

I didn't mean Siglent only for this kind of low bandwidth models but it's an ancient old marketing trick that no vendor can resist: Look we have this sky high feature and the prices of some of our models start only this low and then they expect people to connect these 2 things. Surely it works for many people but does it work for people who do engineering stuff (where logical thinking and good memory is basic pre-requisite)? Anyhow, unless these models can be hacked (mainly an option for hobbyist only) they probably sell quite badly.
Maybe these stretched models sell well in some countries, I don't know, time will tell.

I don't like (and I'm probably not alone with this) shared controls no matter of the vendor. A well implemented touch screen maybe masks this a bit but it's still to be seen.

[jemangedeslolos]

Hello,

I have to say that Im a little bit disappointed by this 2000X+
I had a 2204X and I'm waiting patiently for his replacement.
The 10 inch touch screen are a very great improvement. I have a MSO5000 at home and I use the touch screen all the time.

I'm sure there will be plenty of great features but the shared control 
Maybe a big touch screen + shared control are not so annoying.
And I was hoping a bit more in terms of sample rate.

I think Siglent has made some choices not to walk over on the SDS5000X market.

[Performa01]

I didn't mean Siglent only for this kind of low bandwidth models but it's an ancient old marketing trick that no vendor can resist: Look we have this sky high feature and the prices of some of our models start only this low and then they expect people to connect these 2 things. Surely it works for many people but does it work for people who do engineering stuff (where logical thinking and good memory is basic pre-requisite)? Anyhow, unless these models can be hacked (mainly an option for hobbyist only) they probably sell quite badly.

Well, it obviously works for all the so called "A-brands", where your criticism is certainly justified. Enterprises can negotiate some (artificially) expensive options for free and feel like a winner.

Siglent of all brands, appears to have a mostly reasonable marketing strategy. Price premium for higher bandwidth is bearable except for the SDS5000X, where the higher bandwidth models actually have a more costly frontend compared to the 350MHz model.

With regard to options, the Siglent scopes are fairly complete from the outset. There are no memory options and almost all software features are standard, only exception would be some exotic packages like Power Analysis. All the popular serial triggers and decoders are included for free and the rest is reasonably priced, even on the SDS5000X. This is also true for the MSO option.

Even enterprises don’t have unlimited budgets for their T&M requirements anymore (some never had) and this is even true for the aerospace industry. Even big companies do not need high end scopes for all their departments, all the more so as today even entry level instruments offer performance and features only found in top end devices not too long ago. Given your statements, I guess you would be flabbergasted if I you knew the percentage in the Siglent sales statistics which are enterprise purchases these days…

[e0ne199]

I didn't mean Siglent only for this kind of low bandwidth models but it's an ancient old marketing trick that no vendor can resist: Look we have this sky high feature and the prices of some of our models start only this low and then they expect people to connect these 2 things. Surely it works for many people but does it work for people who do engineering stuff (where logical thinking and good memory is basic pre-requisite)? Anyhow, unless these models can be hacked (mainly an option for hobbyist only) they probably sell quite badly.

Well, it obviously works for all the so called "A-brands", where your criticism is certainly justified. Enterprises can negotiate some (artificially) expensive options for free and feel like a winner.

Siglent of all brands, appears to have a mostly reasonable marketing strategy. Price premium for higher bandwidth is bearable except for the SDS5000X, where the higher bandwidth models actually have a more costly frontend compared to the 350MHz model.

With regard to options, the Siglent scopes are fairly complete from the outset. There are no memory options and almost all software features are standard, only exception would be some exotic packages like Power Analysis. All the popular serial triggers and decoders are included for free and the rest is reasonably priced, even on the SDS5000X. This is also true for the MSO option.

Even enterprises don’t have unlimited budgets for their T&M requirements anymore (some never had) and this is even true for the aerospace industry. Even big companies do not need high end scopes for all their departments, all the more so as today even entry level instruments offer performance and features only found in top end devices not too long ago. Given your statements, I guess you would be flabbergasted if I you knew the percentage in the Siglent sales statistics which are enterprise purchases these days…

maybe would you like to share some informations about top companies which buy siglent products for their R&D?

[Performa01]

I'm sure there will be plenty of great features but the shared control 
 

I’m sorry that the multiplexed controls disappoint you.

But this is just following the LeCroy concept. Vertical controls are not only for the input channels 1-4, but also for the other buttons around: Math, Ref and Digital. The currently selected (and backlit) button indicates the what is controlled (scale and position). For instance, in Math mode you can also control the vertical scale and reference position of the FFT with it.

For selecting a math or reference channel, just touch the trace on the screen or the corresponding info tab at the bottom.

This is also how vertical zoom works. For the selected channel, the vertical controls will work for either the main or zoom window, depending on which one is selected by either pushing the horizontal control or just a simple touch on the corresponding screen area (selection will be indicated by a dashed frame around it). In the zoom window, the vertical gain can be increased and the trace position shifted.

I think it is a nice concept once you’ve gotten used to it.

[Performa01]

For those who are looking for a scope with reasonable performance that does not slow down to snail pace and get unresponsive as soon as it has something a little more demanding to do, just have a look at the attached video.

Please note that this is a screen recording from the web interface which is limited to less than 10 frames per second. The SDS2000X+ display is both faster and significantly smoother, but I still think you can get a first impression out of the video.

Since Siglent scopes always make use of their full memory and analog record length was just 20Mpts, we can enter history anytime to see the past (nine in this case) acquisitions:


SDS2354+_Performance_Demo_I2C_History

[Performa01]

10Bit Acquisition mode and vertical Zoom:

This is the genuine 8bit acquisition which yields 30 LSB per vertical division. To illustrate this, the main window has a fine adjusted gain of 300mV/div and the zoom window is set to 10mV/div (30x zoom).
In dots display mode, we can see exactly one sample per division as expected:


SDS2354X+_30xZoom_8bit

Plese note the tiny black rectangle in the gray shaded main window – this is the frame for the zoom window.

In 10bits mode, we expect 4 samples per division, corresponding to 120 LSB per division in the main window:


SDS2354X+_30xZoom_10bit

As stated before, bandwidth is limited to 100MHz in 10bit mode.

This is the actual frequency response up to 1GHz:


SDS2354Xplus 2GSa 10bit 1GHz

[nctnico]

Given your statements, I guess you would be flabbergasted if I you knew the percentage in the Siglent sales statistics which are enterprise purchases these days…

You have to split that per type of equipment. I have no problem recommending the simpler Siglent gear (if it has mature firmware) to customers but oscilloscopes... no. There is just no way to smooth over / explain the limitations on the oscilloscopes I outlined earlier. The people I deal with have used other equipment so they know how an oscilloscope is supposed to work.

[Performa01]

Pulse response and bandwidth for optional 500MHz in interleaved channel mode

The 500MHz option is available for all the ones who would like to measure that high occasionally, but don’t need it on all channels at the same time.

Here is the frequency response:


SDS2354Xplus 2GSa 8bit 1GHz

As you can see, my pre-production sample does a fairly decent job:

-1.0dB @ 430MHz
-1.7dB @ 500MHz
-3.0dB @ 570MHz


Here’s the pulse response to a 60MHz squarewave with fast edges:


SDS2354X+_2GSa_Pulse_60MHz_D

In this screenshot, we can spot a number of nice features:

•   The 7-digit frequency counter (how many other 2000 series DSOs have that?) right below the Siglent logo. The actual signal frequency has been measured as 59.9999459MHz, so the deviation is only 0.265ppm. Not bad for a DSO in this class.
•   A nice display with working intensity grading
•   Stable measurements (except overshoot), indicated by low standard deviation in the statistics.
•   Optional histograms for the measurement statistics that can be enlarged for further analysis if so desired.
•   Rise time measurement of ~720ps average, which would hint on a bandwidth just below 500MHz for an input channel with Gaussian filter shape, but that’s certainly not a valid method to determine the bandwidth of a DSO. We already know it’s actually 570MHz.

Finally let’s see the frequency domain of that signal:


SDS2354X+_2GSa_Pulse_60MHz_FFT

We can see that the harmonics of the signal extend well beyond 1GHz and it is not a perfect rectangle, as the even harmonics are still quite strong.

[Performa01]

Given your statements, I guess you would be flabbergasted if I you knew the percentage in the Siglent sales statistics which are enterprise purchases these days…

You have to split that per type of equipment.

And what makes you think I haven't done that?
Of course I was talking about oscilloscopes exclusively here.

Apart from genuine Siglent, what do you think who the primary customers for Teledyne Test Tools are? Guess what the T3DSO1000 and T3DSO2000 series is?

When I was talking to application engineers from LeCroy, no one has told me that Siglent DSOs don't work as expected. But no wonder, as LeCroy might have more options to control the record length and/or sample rate, but essentially follows the same principle otherwise. Likewise, Pico Technology have decades of experience, are the world market leader in USB instruments and their PicoScopes work exactly this way.

I have no problem recommending the simpler Siglent gear (if it has mature firmware) to customers but oscilloscopes... no. There is just no way to smooth over / explain the limitations on the oscilloscopes I outlined earlier. The people I deal with have used other equipment so they know how an oscilloscope is supposed to work.

I'm sorry to hear that, but then again, I do not recommend certain instruments either – with the tiny difference that I could not be bothered (and don't have the time) to post my complaints in just about every thread about brands and products that I neither like nor use.

The topic has been discussed ad nauseam already and no one else has been so fanatic about this. A relevant and comprehensive contribution has been made here in reply #30:

https://www.eevblog.com/forum/blog/eevblog-1220-siglent-sds5000x-1ghz-oscilloscope-review/msg2489091/#msg2489091

The memory selection on the Siglent DSOs allows to intentionally lower the sample rate – only at slower timebase settings of course, because the minimum is 20kpts in interleaved mode on the SDS2000X+. The main purpose of this is to control the maximum number of segments for history and sequence recording (FFT has its own length selection by now).

Of course, offering everything at once and having it all configurable to meet everyone’s taste and habits would be the ultimate solution (if only to see what comes next) – provided it does not require a major redesign of the acquisition engine.

I might eventually discuss this topic with fellow consultants/testers as well as Siglent R&D. If it turns out that it would be easy to implement without risks, then it could become a (low priority) request.

[Performa01]

XY-Mode

It is often said that DSOs suck at XY – and while they cannot compete with their analog grandparents indeed, we can still get decent performance even from a DSO nowadays.

I would like to show how far Siglent got with this – you can expect a similar performance on all current instruments based on the new platform (SDS1000X-E, SDS2000X-E, SDS5000X).

Here’s a screenshot, showing an 8QAM I/Q signal with rather weird settings, providing a rather complex rotating figure. Thanks to the intensity grading we almost get something like a smeared rotating constellation diagram


SDS2354X+_XY_1Mpts_IQ_8QAM_1Hz_1MS

Please note that the record length is 1Mpts in this example and the acquisition length is 1ms, hence signals down to 1kHz can be properly displayed.

To demonstrate the performance, I had to use an old camera since the limited framerate of the web interface would give a false impression of the performance. Please see the attached video for a demo.

[nctnico]

The memory selection on the Siglent DSOs allows to intentionally lower the sample rate – only at slower timebase settings of course, because the minimum is 20kpts in interleaved mode on the SDS2000X+. The main purpose of this is to control the maximum number of segments for history and sequence recording (FFT has its own length selection by now).

But what is the purpose of history mode? I've seen this on a Yokogawa oscilloscope and R&S has it as well but I don't really see a particular use for having a history. If you want to acquire segments then you might as well turn on segmented recording.

Of course, offering everything at once and having it all configurable to meet everyone’s taste and habits would be the ultimate solution

Which is why the A-brands (except for Keysight which always uses maximum memory) all have a selection to set the memory to a fixed length. And I'm not writing this because I want to bash Siglent; I want Siglent to become better at making oscilloscopes in order to get a more competitive oscilloscope market. BTW Picoscope may be different but they are only serving a niche market (which they dominate) with specific requirements.

[Pinkus]

Siglents product page (in Chinese, use Google translate) including pricing:
https://www.siglent.com/products-overview/sds2000xp/#navs

[Performa01]

But what is the purpose of history mode? I've seen this on a Yokogawa oscilloscope and R&S has it as well but I don't really see a particular use for having a history. If you want to acquire segments then you might as well turn on segmented recording.

Every serious DSO has history mode – not only Yokogawa and R&S, but of course also Keysight and LeCroy, the latter even have WaveScan, a very powerful tool that can do much more than a simple search.

Segmented recording is a special mode for optimizing memory usage and minimizing blind time for events with irregular occurrence. It is set up on purpose and has no real time screen update.

History is running in the background just in case we need it. This is the big advantage: it is always there, we can never forget to turn it on and we need not plan anything in advance. There can be many situations where we might want have a look at previous acquisitions again.

During a debugging process we might spot some unexpected event, and by the time we hit the stop button the screen has already been overwritten by new acquisitions. We might even have an idea how to trigger this particular event, but would have to set that trigger (and hope that it actually works) and then wait until the same thing happens again (maybe for a long time, especially when we have the wrong trigger). With history we can look at that very event immediately. We can try various search conditions or just find the event visually during playback or even single stepping (which is of course not an option if the history contains thousands of frames).

History can replace sequence recording for regular events or in general, whenever minimum blind time is not paramount. Infrequent serial messages would be an example for this. We choose the record length so that we always catch a command/response pair (assuming that they are closely spaced) and can use zoom if we really want to look at some details, in short: just work as usual. But we can also stop the acquisition anytime by entering the history, where we can closely inspect not only the current record, but also all the command/response pairs from the past.

[egonotto]

Hello,

how many bits is a point in SDS2000X Plus and SDS5000X if you save to a file?
And if you use 10 bit on SDS2000X Plus does it change the memory size in points?

Best regards
egonotto
 

[Performa01]

how many bits is a point in SDS2000X Plus and SDS5000X if you save to a file?
And if you use 10 bit on SDS2000X Plus does it change the memory size in points?

SDS5000X and SDS2000X+ save 8bit data. Full scale on the sreen is equivalent to 240 LSB.

SDS2000X+ in 10bit mode saves 16bit data, but of course only 10 bits resolution. Full scale on the sreen is equivalent to 960 LSB.

The memory depth in points does not change when you toggle between 8 and 10 bit acquisition as long as the record does not use the max. memory depth. The binary file size doubles in 10bit mode, CSV data size does not change of course.

The maximum memory is halved in 10bit mode. So with all channels active, you get a max. record length of normally 100Mpts but only 50Mpts in 10bit mode. In this case the record length would change accordingly when you toggle 10bit mode.

[egonotto]

Hello,

thanks Performa01.

If you use hi res mod (Eres) is than a point more than 8 bits?
And if yes, is max memory than smaller?

And if you use the Digital Channels affected this the memory size of the analog channels?

Best regards
egonotto

[Performa01]

If you use hi res mod (Eres) is than a point more than 8 bits?
And if yes, is max memory than smaller?

And if you use the Digital Channels affected this the memory size of the analog channels?

ERES is not an acquisition mode anymore, but a math function (LeCroy send their regards!).
This has many advantages. You still have the original data available and filter just the channel(s) where you really need it, even with different settings, without compromising the other channels. Furthermore you can apply ERES to a math expression.

EDIT: Ah, and yes, the ERES math trace has the higher resolution of course.

Because of this, the memory doesn't get smaller as it has been the case with ERES as acquisition mode. SDS5000X for instance is limited to 25Mpts per channel pair in ERES acquisition mode.

Analog memory depth is not affected by activation of the digital channels. You still have 200Mpts per channel pair available.

[egonotto]

Hello,

thanks Performa01.

Can it be, that the SDS5000X has 1GB of memory overall?

Best regards
egonotto

[Performa01]

Hi Egonotto,

this might be true for the acquisition memory, as this would be 866MB for the analog channels (including history) plus (at least!) 125MB for the digital channels, that makes a total of 991MB, hence close to 1GB.

But apart from that the scope needs a lot of memory for the operating system and all the processing - measurements, math, serial decoders, mask editor, power analysis, web server, display management...

[egonotto]

Hi Performa01,

perhaps the memory for the processor is a separate memory.

Best regards
egonotto

[nctnico]

But what is the purpose of history mode? I've seen this on a Yokogawa oscilloscope and R&S has it as well but I don't really see a particular use for having a history. If you want to acquire segments then you might as well turn on segmented recording.

Every serious DSO has history mode – not only Yokogawa and R&S, but of course also Keysight and LeCroy, the latter even have WaveScan, a very powerful tool that can do much more than a simple search.

Segmented recording is a special mode for optimizing memory usage and minimizing blind time for events with irregular occurrence. It is set up on purpose and has no real time screen update.

No. The HP/Agilent/Keysight scopes I have owned & used so far didn't have history mode. By default they use full memory. And it isn't true segmented mode is purely intended to offer fast acquisition of repeating events. For that deep memory is equally useful because you won't save much memory anyway. Some DSOs even have a fast & slow segmented recording mode. In slow mode the progress is shown on screen. In fast mode you don't see anything until the number of segments if captured or press 'stop'.

History can replace sequence recording for regular events or in general, whenever minimum blind time is not paramount. Infrequent serial messages would be an example for this. We choose the record length so that we always catch a command/response pair (assuming that they are closely spaced) and can use zoom if we really want to look at some details, in short: just work as usual. But we can also stop the acquisition anytime by entering the history, where we can closely inspect not only the current record, but also all the command/response pairs from the past.

But do you ever use history mode? Does anyone else feel it is important? IMHO history mode is pretty much useless to look back at 'accidental' signals. In most cases you'll have so many triggers per second that by the time you press the 'stop' button the signal is gone from the buffer anyway. And how to make sure that what you look at is related to a known circumstance?

When I use segmented recording I want the set of acquisitions to be related to a specific measurement condition. I rarely use segmented recording though; deep memory is much easier. Just scroll left/right. I have this with logic analysers too. The first logic analyser I owned had 1kpts per channel. Capturing specific events was very hard because it required setting up very specific triggers. My second logic analyser had (IIRC) 128kpts per channel and an even more advanced trigger system. However I used advanced triggering less. My current logic analyser has 32Mpts and with that amount of memory it doesn't even matter much when to trigger. Sometime I just start it manually because I know it will capture an entire cycle anyway (for example data to a TFT display). After that I can analyse the signals and the best thing is that all the data is related to the same circumstance. So if I see something weird at one point, I can look at the data preceding the event because it is all there (100% related to each other). I use a DSO in the same way.

[Performa01]

I never claimed that long memory isn’t useful, quite the contrary.
I was talking about irregular events, but should have better said low duty cycle events, because it actually does not matter if the spacing is constant or not.
In these cases, deep memory is not enough (and inconvenient) and segmented memory is all the more important for DSOs with a lack of deep memory, like the current Keysight midrange instruments.
On a DSO with deep memory and 90000 history entries, you have all the time in this world to stop and look back.

[nctnico]

Now you are contradicting yourself because your DSO with history has a small amount of memory depth on purpose and enables a crutch to make up for small memory by default. That just doesn't make sense. And looking back requires a lot more button pushes and it likely isn't possible to put the recordings side-by-side horizontally so you can scroll through a contiguous signal. Having one big segment and scrolling left/right is much easier to work with. Besides that the history mode doesn't work if the trigger condition is not met for the signal so you will lose whatever interesting things are happening before or after the trigger event if it happens to be just outside the screen.

But this discussion is going in circles. The bottom line is that Siglent really should add fixed memory length settings instead of always using automatic memory length which is just enough for the screen width. This should be at the top of the list with items to fix because without it they are behind on the competition. There is no working-around that.