SDS800X HD Bug Reports + Firmware

[Performa01]

All these comparisons analog vs digital are often a bit shortsighted. People seem to believe that analog means infinite resolution, while Sync reconstruction in digital sampling systems is some sort of voodoo, but it’s far from that.

In the analog world, quantum steps ultimately define the maximum resolution, so it is not infinite because of this alone. Other effects will destroy the Illusion of infinite resolution long before quantum mechanics takes its toll.

It’s easy to determine that in electricity – for Electronic Engineers at least:

As an example, thermal noise for a 1 MΩ source resistance at room temperature (23 °C) and only 10 kHz bandwidth would already be 12.78 µVrms. The peak-to-peak value of that noise is usually 10 times the RMS value, hence some 128 µVpp. Even an ideal analog instrument which adds zero noise (physically impossible) could only have an ENOB of less than 6 bits at 8 mV full scale (like an oscilloscope at 1 mV/div). This would increase to about 9.3 bits for a full scale value of 80 mV (10 mV/div). In this scenario, an 8-10 bit DSO would be more than adequate already.

In practice, the situation is even worse, because there is no such thing like a noise-free instrument. Of course you can find other examples, with lower source resistance, even narrower bandwidth and maybe also lower temperatures, which will result in better ENOB figures. But there you can get into impractical territory pretty quickly.

A more practical scenario would be 50 Ω source impedance, but certainly not limited to 10 kHz Bandwidth. If we assume at least 20 MHz bandwidth, then we get ~4 µVrms noise, equivalent to about 40 µVpp. A completely noise-free oscilloscope at 1 mV/div would have an ENOB of 7.64 bit and 10.97 bit at 10 mV/div. Once again, we do not even start to exploit the resolution of a 12-bit sampling system.

In practice, the noise figure (NF) of an oscilloscope can be significant, due to 1/f noise of the semiconductors and the elevated LF-noise of the split path input buffer architecture – and this is absolutely the same for analog or digital oscilloscopes. That means, that 12 bits already require special conditions like heavily bandwidth limited signals or massive averaging to provide a real benefit. But then, digital filters and averaging increase the number of bits as well, so that we could get a similar effect as the real 12-bit ADC.

Any analogy to photography is severely flawed as well. Analog (chemical) film does have a certain granularity, known as grain. This quickly gets a significant issue with high-speed films like ISO 400 or higher and gets better with specialized low speed “fine grain / high definition” films at ISO 50 and lower. Then there is the effect of the camera lens, which acts like a lowpass filter and ultimately limits the resolution as well. It’s the same as with oscilloscopes; sample rate and ADC resolution become meaningless as soon as we exceed the bandwidth of the analog frontend.

Back to the analog film: for a 35 mm low speed high-definition film and a quality lens (like Canon L-series), a full format DSLR with a 16 MP sensor will extract all the detail that you can get from this particular lens/film combination. Zoom in any further on the DSLR image, i.e. blow it up to more than 16 MP during post processing, and you’ll see interpolated data. Try to scan an analog 35 mm exposure to more than 16 MP, you’ll end up with a crisp and clear picture of the film grain. There you can use post processing to eliminate the grain, but without gaining any additional information for the original exposure.

It's not by chance that there have been large photo negative formats like 4 x 5” or even larger to overcome the limits of analog film grain. Size, weight and cost increase significantly when using such equipment. And meanwhile, there seem to exist large sensor high resolution digital systems as well.

Practical comparison tests indicate, that modern DSLR systems – even “only” 35 mm ones - can now play in the same ballpark as 4 x 5” analog systems for professional photography:

https://www.mountainphotography.com/gallery/4x5-film-vs-digital-resolution-comparison/

Fazit: Be it Oscilloscopes, Audio or Video: people tend to compare the high-end analog systems from the old days either to early and immature digital systems, that clearly leave a lot to be desired, or with modern cheap junk. One should rather compare the old analog systems to the same class of digital systems of today.

Hint: even the cheapest modern digital camera will easily win any resolution contest against a (comparatively cheap back then) old Polaroid camera and will run circles around it in just about every other aspect.

[mawyatt]

Digital Imaging chips benefit from the physical pixel area, the larger pixel produces a better SNR at the expensive of lower resolution for a given Imaging chip size.

A few decades ago while attending a meeting at MIT Lincoln Labs we saw the "chips" used in one of the Hubble Telescope Cameras (MITLL made the telescope chips). These CCD "chips" were huge, 100mm square with only 4M pixels and 4 "chips" were used to make the particular 16M pixel camera sensor which was 200mm square overall.

Best

[hfleming]

Heck, I can just remember from my old days trying to do astrophotography with analog film. The faster the film, the more grain you got, and using special techniques ( pushing the film, both before taking a pic) and post-photo were 🤮🤮🤮🤮. Heck, even my 17-year old Canon DSLR (Canon EOS-7D) can take much better astronomical images than what I could ever do with film (including hours/days of pre-sensitizing films and post-processing)… what the latest digital cameras can do is beyond my wildest imagination…. Same with electronic test equipment. I can still remember fondly those analog VTVM’s (1-3% accuracy), and yes, for purely sentimental reasons I still use it, but compared to even a cheap modern DMM, it leaves a lot to be desired. As far as oscilloscopes (and VNA’s) are concerned, in the beginning, digital oscilloscopes were a novelty (except for some very specific purposes). Even until around 15 years ago, I preferred analog scopes, but the latest generations of affordable digital scopes can do so much more than what those very expensive scopes that costed more than a house (and weighed nearly as much) could do… heck, I could remember working on a project, needing a very special spectrum analyzer, and by the time the budget for it was approved, the project was finished… nowadays a NanoVNA can do 99.9% of that job…

[mawyatt]

Yes it's come a long way from Film. Early on Tektronix CCDs were used, then Kodak CMOS (fabbed at IBM), then Nikon CCD, then Canon CMOS with correlated noise reduction, then Sony CMOS and now BSI CMOS (Back Side Illuminated). BSI reduced the photons lost to top side metallization, chip reverse mounted. Then of course cooling of all these sorts!!

Best

[Conrad Hoffman]

What I miss from my old Tek boat anchors has nothing to do with resolution. It's the specialized plug-ins. Want a high gain differential that goes down to 10 uV/cm with adjustable high and low pass filters? 1A7A. Want to test diodes for recovery time? Get a Type S plug-in. And about a dozen others for the job at hand. The later 5000 and 7000 series had 'em too.

Fortunately, I can accomplish a huge amount with my Siglent. It even makes a dandy LCR meter if you measure current, phase & frequency. Tek never had a plug-in for that, or bode plots. Their spectrum analyzer plug-ins were nothing great, a PITA to use and you were stuck with a limited linear sweep. If I really need a high gain differential input I can build a preamp or buy an SRS- https://www.thinksrs.com/products/preamp.html

[awakephd]

All these comparisons analog vs digital are often a bit shortsighted. People seem to believe that analog means infinite resolution, while Sync reconstruction in digital sampling systems is some sort of voodoo, but it’s far from that.

In the analog world, quantum steps ultimately define the maximum resolution, so it is not infinite because of this alone. Other effects will destroy the Illusion of infinite resolution long before quantum mechanics takes its toll.

It’s easy to determine that in electricity – for Electronic Engineers at least:

As an example, thermal noise for a 1 MΩ source resistance at room temperature (23 °C) and only 10 kHz bandwidth would already be 12.78 µVrms. The peak-to-peak value of that noise is usually 10 times the RMS value, hence some 128 µVpp. Even an ideal analog instrument which adds zero noise (physically impossible) could only have an ENOB of less than 6 bits at 8 mV full scale (like an oscilloscope at 1 mV/div). This would increase to about 9.3 bits for a full scale value of 80 mV (10 mV/div). In this scenario, an 8-10 bit DSO would be more than adequate already.

In practice, the situation is even worse, because there is no such thing like a noise-free instrument. Of course you can find other examples, with lower source resistance, even narrower bandwidth and maybe also lower temperatures, which will result in better ENOB figures. But there you can get into impractical territory pretty quickly.

A more practical scenario would be 50 Ω source impedance, but certainly not limited to 10 kHz Bandwidth. If we assume at least 20 MHz bandwidth, then we get ~4 µVrms noise, equivalent to about 40 µVpp. A completely noise-free oscilloscope at 1 mV/div would have an ENOB of 7.64 bit and 10.97 bit at 10 mV/div. Once again, we do not even start to exploit the resolution of a 12-bit sampling system.

In practice, the noise figure (NF) of an oscilloscope can be significant, due to 1/f noise of the semiconductors and the elevated LF-noise of the split path input buffer architecture – and this is absolutely the same for analog or digital oscilloscopes. That means, that 12 bits already require special conditions like heavily bandwidth limited signals or massive averaging to provide a real benefit. But then, digital filters and averaging increase the number of bits as well, so that we could get a similar effect as the real 12-bit ADC.

Any analogy to photography is severely flawed as well. Analog (chemical) film does have a certain granularity, known as grain. This quickly gets a significant issue with high-speed films like ISO 400 or higher and gets better with specialized low speed “fine grain / high definition” films at ISO 50 and lower. Then there is the effect of the camera lens, which acts like a lowpass filter and ultimately limits the resolution as well. It’s the same as with oscilloscopes; sample rate and ADC resolution become meaningless as soon as we exceed the bandwidth of the analog frontend.

Back to the analog film: for a 35 mm low speed high-definition film and a quality lens (like Canon L-series), a full format DSLR with a 16 MP sensor will extract all the detail that you can get from this particular lens/film combination. Zoom in any further on the DSLR image, i.e. blow it up to more than 16 MP during post processing, and you’ll see interpolated data. Try to scan an analog 35 mm exposure to more than 16 MP, you’ll end up with a crisp and clear picture of the film grain. There you can use post processing to eliminate the grain, but without gaining any additional information for the original exposure.

It's not by chance that there have been large photo negative formats like 4 x 5” or even larger to overcome the limits of analog film grain. Size, weight and cost increase significantly when using such equipment. And meanwhile, there seem to exist large sensor high resolution digital systems as well.

Practical comparison tests indicate, that modern DSLR systems – even “only” 35 mm ones - can now play in the same ballpark as 4 x 5” analog systems for professional photography:

https://www.mountainphotography.com/gallery/4x5-film-vs-digital-resolution-comparison/

Fazit: Be it Oscilloscopes, Audio or Video: people tend to compare the high-end analog systems from the old days either to early and immature digital systems, that clearly leave a lot to be desired, or with modern cheap junk. One should rather compare the old analog systems to the same class of digital systems of today.

Hint: even the cheapest modern digital camera will easily win any resolution contest against a (comparatively cheap back then) old Polaroid camera and will run circles around it in just about every other aspect.

I stand corrected! As I said, I claim no expertise in analog photography, and may have butchered the analogy (and apparently did). No doubt the analogy would have worked better 20+ years ago, when my first digital camera proudly claimed 1 Megapixel resolution (!!!) - which did not compare well to analog film.

[eTobey]

Does anyone know if there is a new firmware version coming up in the next few months?
Checked a few bugs, but only a tiny fraction has been fixed.

[electr_peter]

Math waveforms vertical scaling exhibits the same behavior in both modes modes of Vertical Ref (Utility > Menu > System Setting > Reference Pos > Vertical Ref Fixed Position/Fixed Offset. It seems to always expand around zero (which is represented by Fixed position), but it does not expand around center of the screen. Other manufactures calls this setting vertical expand around screen center or around zero.

I think there maybe a bug as math expansion around screen center is not possible. Can anyone confirm?

SDS800X HD, FW 1.1.6.5 / 3.8.12 FW

[2N3055]

Math waveforms vertical scaling exhibits the same behavior in both modes modes of Vertical Ref (Utility > Menu > System Setting > Reference Pos > Vertical Ref Fixed Position/Fixed Offset. It seems to always expand around zero (which is represented by Fixed position), but it does not expand around center of the screen. Other manufactures calls this setting vertical expand around screen center or around zero.

I think there maybe a bug as math expansion around screen center is not possible. Can anyone confirm?

SDS800X HD, FW 1.1.6.5 / 3.8.12 FW

Yep you are right.
The math channels always had slightly different control than input CH.
They are virtual and not all intuition for physical channels work all the time.
I agree this is something they should look into.
It is not completely simple though because FFT also lives in math and needs to be figured out how to do it.
But a good suggestion.
I will report to them and see what happens.

[eTobey]

Nothing new... 

Math channel ignores Vertical Ref setting (Fixed offset)
Compare the behaviour of a normal channel to a math channel. (manual page 217)

Are you doing this to piss us off? We have been through this before, about 500 times: Describe what you did, tell us what you expected to see, tell us what you saw instead, attach a screenshot to illustrate. I have no intent to play a game of scavenger hunt, trying to figure out what you are talking about based on little clues.

That would imply that he cares for our opinion. He doesn't. He demands the servants to obey...

[ebastler]

Nothing new... 

But there is a difference between your bug report from last year and electr_peter's -- which probably correlates with the difference in response which the two reports got: Peter actually described the observed behavior and his expectation. In contrast, as expressed in my comment from last year, I could not figure out what you were referring to back then. I assume the same applied to other readers.

[eTobey]

But there is a difference between your bug report from last year and electr_peter's -- which probably correlates with the difference in response which the two reports got: Peter actually described the observed behavior and his expectation. In contrast, as expressed in my comment from last year, I could not figure out what you were referring to back then. I assume the same applied to other readers.

Indeed, but since there is actually a clear instruction of it in the manual, i saw no need but as to give the reference of it.

RTFM 

It is a single simple setting. How hard could it be to understand and try out?

I probably should have created another user, so the perception would have been neutral, and people would think more about what is said, rather then who it said.

[2N3055]

But there is a difference between your bug report from last year and electr_peter's -- which probably correlates with the difference in response which the two reports got: Peter actually described the observed behavior and his expectation. In contrast, as expressed in my comment from last year, I could not figure out what you were referring to back then. I assume the same applied to other readers.

Indeed, but since there is actually a clear instruction of it in the manual, i saw no need but as to give the reference of it.

RTFM 

It is a single simple setting. How hard could it be to understand and try out?

I probably should have created another user, so the perception would have been neutral, and people would think more about what is said, rather then who it said.

Ebastler nicely explained what is the problem.
There are no paranoid reasons.
If you don't care to explain it right, we don't understand.

At those times you wrote something understandable it was read.
Simple as that.

[Sorama]

Hence he’s nickname
To Obey 🤣

(Sorry eTobey)

[ebastler]

Indeed, but since there is actually a clear instruction of it in the manual, i saw no need but as to give the reference of it.

RTFM 

It is a single simple setting. How hard could it be to understand and try out?

I probably should have created another user, so the perception would have been neutral, and people would think more about what is said, rather then who it said.

Look, the point is not whether a reader could have derived the intent of your post with reasonable effort (which may include checking the manual). The point is that you want to get your readers' attention, and gain their support -- in pointing out to Siglent that a certain deficiency in the firmware is important and should be fixed.

And in order to do that, it is helpful if you make an effort to explain what the issue is, such that the reader can see it without doing further homework. And then explain why you consider it important that it gets fixed -- what does it stop you from doing, where can it cause misunderstandings when using the scope, etc.

[eTobey]

And in order to do that, it is helpful if you make an effort to explain what the issue is, such that the reader can see it without doing further homework.

It is also helpful that people would actually turn on the scope and try it (RTFM), dont you think so?

BTW:
I just dropped a kind of a bombshell:
https://www.eevblog.com/forum/chat/exposed-how-a-hobbyist-turned-an-entire-community-into-free-beta-testers/

[MikePie]

Hi,

may I kindly ask for your support:
I came across an unexpected lockup of my newly received SDS804X-HD (Firmware 1.1.6.5).
It would no longer respond to keystrokes, even the RTC display was frozen, also the power key would not respond, so I had to pull the power cord.
Fortunately I had saved the setup not long before, so I had a starting point to try to reproduce my steps.
While I was unable to get to the same lockup condition, I noticed that certain steps lead to failure of AUTO trigger and a very unresponsive user interface (takes several seconds to respond to a keystroke) - a kind of "pre-lockup" imho.
Rebooting the unit restores proper AUTO triggering and responsive user interface.
I also carried out these steps via web browser with results identical to front panel operation.

I am attaching DSO.zip containing the setup file and a textfile with the steps to reproduce the problem.

Is this a bug which will hopefully be taken care of by Siglent?

(background info: I was trying to track down power supply switching noise (ringing) on 1.5V DC output using two probes in a setup which was aptly described as "Poor Men’s Differential Probing" by Performa01 in his reply #6).

Thank you,
MikePie

[2N3055]

Your setup file does confuse the scope.
But it can be that that file is corrupted somehow.

If I setup scope with edge trigger, CH1 and 2 to 10x probe, 50mV/div, create F1 channel with C1-C2 (5mV/div) and then set F2 to FFT from F1 it works fine.
Is this what you wanted to do?

[MikePie]

Thank you for taking the time to investigate.

Your setup file does confuse the scope.
But it can be that that file is corrupted somehow.

If I setup scope with edge trigger, CH1 and 2 to 10x probe, 50mV/div, create F1 channel with C1-C2 (5mV/div) and then set F2 to FFT from F1 it works fine.
Is this what you wanted to do?

Basically yes, but then I went on and changed to Qualified Trigger so to capture an event following the initial trigger taken from a different channel (CH4).
The setup file was saved with this Qualified Trigger active so this is the first thing I try to undo after recalling the setup.
I did not do anything to this setup file other than copy it from the DSO.
Also, that setup file is recalled successfully by the DSO (according to the message shown by the DSO after loading it).
If that file is corrupted, then it was the DSO corrupting it (well, I presume the file copy process via USB stick was not flawed and therefore the file on my DSO and the zipped file in my previous post are identical).
As far as I can remember, before changing to Qualified Trigger I did not encounter any unexpected behaviour in DSO operation.
My assumption is that changing back to Edge Trigger and setting AUTO Trigger mode must return the DSO to "free running" - which it does not, in addition to it being very slow to process user inputs.

[2N3055]

Thank you for taking the time to investigate.

Your setup file does confuse the scope.
But it can be that that file is corrupted somehow.

If I setup scope with edge trigger, CH1 and 2 to 10x probe, 50mV/div, create F1 channel with C1-C2 (5mV/div) and then set F2 to FFT from F1 it works fine.
Is this what you wanted to do?

Basically yes, but then I went on and changed to Qualified Trigger so to capture an event following the initial trigger taken from a different channel (CH4).
The setup file was saved with this Qualified Trigger active so this is the first thing I try to undo after recalling the setup.
I did not do anything to this setup file other than copy it from the DSO.
Also, that setup file is recalled successfully by the DSO (according to the message shown by the DSO after loading it).
If that file is corrupted, then it was the DSO corrupting it (well, I presume the file copy process via USB stick was not flawed and therefore the file on my DSO and the zipped file in my previous post are identical).
As far as I can remember, before changing to Qualified Trigger I did not encounter any unexpected behaviour in DSO operation.
My assumption is that changing back to Edge Trigger and setting AUTO Trigger mode must return the DSO to "free running" - which it does not, in addition to it being very slow to process user inputs.

I am not putting any blame or even having any conclusions yet. That file when loaded does screw things up. Meaning there is something in that file that is hard to swallow.
That is given. Who, what, why is still being researched.  It can be bad stick, but it can be scope saving wrongly, or loading something wrongly.
Something in the file can be wrong in such a way that parser does not reject it but loads some parameter in a funny way.
I don't know yet.

Your further details about how you stumbled upon this is nice info. Thanks. I will play with it more over the weekend and post if I have some news. And report it if there is need.

[2N3055]

Thank you for taking the time to investigate.

Your setup file does confuse the scope.
But it can be that that file is corrupted somehow.

If I setup scope with edge trigger, CH1 and 2 to 10x probe, 50mV/div, create F1 channel with C1-C2 (5mV/div) and then set F2 to FFT from F1 it works fine.
Is this what you wanted to do?

Basically yes, but then I went on and changed to Qualified Trigger so to capture an event following the initial trigger taken from a different channel (CH4).
The setup file was saved with this Qualified Trigger active so this is the first thing I try to undo after recalling the setup.
I did not do anything to this setup file other than copy it from the DSO.
Also, that setup file is recalled successfully by the DSO (according to the message shown by the DSO after loading it).
If that file is corrupted, then it was the DSO corrupting it (well, I presume the file copy process via USB stick was not flawed and therefore the file on my DSO and the zipped file in my previous post are identical).
As far as I can remember, before changing to Qualified Trigger I did not encounter any unexpected behaviour in DSO operation.
My assumption is that changing back to Edge Trigger and setting AUTO Trigger mode must return the DSO to "free running" - which it does not, in addition to it being very slow to process user inputs.

Can you please tell me if that setup file was saved with exactly the same FW version that you have now?

[MikePie]

Thank you for taking the time to investigate.

Your setup file does confuse the scope.
But it can be that that file is corrupted somehow.

If I setup scope with edge trigger, CH1 and 2 to 10x probe, 50mV/div, create F1 channel with C1-C2 (5mV/div) and then set F2 to FFT from F1 it works fine.
Is this what you wanted to do?

Basically yes, but then I went on and changed to Qualified Trigger so to capture an event following the initial trigger taken from a different channel (CH4).
The setup file was saved with this Qualified Trigger active so this is the first thing I try to undo after recalling the setup.
I did not do anything to this setup file other than copy it from the DSO.
Also, that setup file is recalled successfully by the DSO (according to the message shown by the DSO after loading it).
If that file is corrupted, then it was the DSO corrupting it (well, I presume the file copy process via USB stick was not flawed and therefore the file on my DSO and the zipped file in my previous post are identical).
As far as I can remember, before changing to Qualified Trigger I did not encounter any unexpected behaviour in DSO operation.
My assumption is that changing back to Edge Trigger and setting AUTO Trigger mode must return the DSO to "free running" - which it does not, in addition to it being very slow to process user inputs.

Can you please tell me if that setup file was saved with exactly the same FW version that you have now?

Yes, there was no change to the FW.

To reduce the ambiguity caused by the file copy process, I also just now did a file compare between
a.) the setup file copied via SMB to my Windows 7 laptop
b.) the setup file extracted from the the DSO.zip which I posted (the file which had been copied using the USB stick).
Result: no differences found.

[eTobey]

I found that using SCPI can lead to lockup aswell.

Did you have the scope connected to the PC, webinterface or a Wifi dongle when the scope locked up? If yes, try without anything connected to it.

[MikePie]

I found that using SCPI can lead to lockup aswell.

Did you have the scope connected to the PC, webinterface or a Wifi dongle when the scope locked up? If yes, try without anything connected to it.

I can't remember what the LAN connection status was when the DSO locked up.
Also, unfortunately, I was unable to reproduce the full lockup - it only happened once.

I now carried out the same steps (as given in my previously posted zipped textfile) with and without LAN cable connected (I only use Ethernet cable, no Wireless dongle) and found no difference in the resulting impaired DSO behaviour.

I tried some basic SCPI queries using the DSO's web server - no problem here, the SCPI responses appear ok to me, but probably you were using some other SCPI control program.

[eTobey]

I tried some basic SCPI queries using the DSO's web server - no problem here, the SCPI responses appear ok to me, but probably you were using some other SCPI control program.

I cant tell since i used different ways. But i can tell you, that you can unlock feature from some higher class scopes, so you may also introduce issues here. Have you tried a firmware reset? Try to remember what you did, and change things back? I believe the settings for SMB servers or network are not reset via firmware reset.

If nothing works, than stop wasting your time and return it and get it replaced.