Siglent SDS800X HD 12 bit DSO's

[markone]

[...] checking if the DSO keep up the analysis in real time at least with 2MSa @ 2MSa/s.

How do you define and measure the "keeping up in real time"? Is your expectation that the waveform acquisition rate, determined by counting Trigger Out pulses, should not drop at all when you activate the Math function?

Hi eblaster, I mean that Picoscope application updates the waveform screen every second  with full math trace computed (with mentioned settings of course), as well as collecting up dome dozens (32 if i remember well).

If I zoom and analyze the math trace I can see that the duty value is updated every square signal cycle, that is what I specifically need.
I'm not saying that something that will not measure every single PWM cycle is totallu useless, but there are cases where this is mandatory.

Maybe the Picoscope can achieve that since acquisition on the external unit and math computations and plotting on the PC happen in parallel. But I see little chance for the SDS800X HD, and probably most other integrated scopes.

My understanding with Picoscope platform is that math function are computed on host PC, dunno if something is happening also at device level with 3000 series.

Edit: Oh, and momentary Duty Cycle is not an available math function on the 800X HD to my knowledge. So if that function is of particular importance to you, this scope is not for you. If it was just an arbitrary example -- care for a different one, say algebra, differentiation, integration of signals?

It's not an arbitrary, it's the fundamental feature that I need for my job with power electronic.

It's not clear to me how Track and Trend functions will work on top of DSO meas system with duty cycle measurement, do you have some specific knowledge about that ?

[ebastler]

It's not clear to me how Track and Trend functions will work on top of DSO meas system with duty cycle measurement, do you have some specific knowledge about that ?

I will let the Siglent experts chime in regarding possible tracking/trending analysis on top of measurements. But since measurements (as opposed to math functions) give one value per sweep, plus statistical values like min/max/stddev, I don't think this would give you the required time resolution for your duty cycle analysis. 

Edit: Learned something new, again! I had totally overlooked "Track" mode, which can follow the course of measurements within a sweep -- was only aware of "Trend". Thanks to Performa01 for putting this right in the post just below! Well, at least the first sentence of my original post was correct...

[Performa01]

Hi folks,

i would need an idea of SDS 800 HD math / measure performance, a reference example could be the duty cycle analysis continously computed on full 2MSa buffers acquired @ 2MSa/s (1 second acquisitions) of 20KHz pulse width modulated signal, of course Track / Trend functions could be of interest if they cover every single pulse cycle.

I currently make use of a Picoscope 3203D with Picoscope 7 TM software, with a modest PC based on Ryzen 5600 the simple math function "duty(A)" works in realtime with acquisition buffers up to 10MS acquired @ 10MS/s, minimal specs are "almost" realtime readings with a 2MS acquisition @ 2MS/s.

I'm happy with Picoscope but there are cases where a stand alone scope would come in handy ... but only if cheap like the SDS800 HD 

I''m providing in attachment screenshots of Picoscope duty cycle analysis of 20KHz square wave pulse signal (sourced to Analog input A) modulated with 5Hz sine wave with two different acquisition settings, 2MSa @ 2MSa/s and  10MSa @ 10MSa/s, dunno if anyone here is able to perform the same task with an SDS800 HD checking if the DSO keep up the analysis in real time at least with 2MSa @ 2MSa/s.

Thanks in advance to anyone will dare to venture in this test 

Any serious analytical scope should be able to do this, yet there are quite some instruments, even very expensive ones, that can not. The special feature of DSOs that I’m regularly recommending, like the ones from Pico and Siglent, is that you get such midrange features even in their low-end devices. Yet this task can be a bit tricky, see the comments below…

Regarding the speed, there are two aspects:

1. Signal Acquisition.

In order to capture enough data to properly analyze a 5 Hz modulation frequency, we need to capture more than 200 ms of the PWM signal. I have chosen one full second (100 ms/div) to capture five full modulation periods; yet I’ll only be able to display 2.5 periods because of some limitations in the Track plot implementation I did not set the correct horizontal track plot position.

2. Horizontal Parameter Measurement.

The number of measurements within a single record (AIM-limit) is 1000 by default and can be set to a maximum of 25000. For my example to work, we need to set it to at least 10000. The time for this is only a small fraction of the total acquisition time.


SDS824X HD_Track_PWM_20kHz_5Hz

The modulation is retrieved quite accurately; the update speed is 0.914 Hz ≈ 1.1 seconds per frame.

The automatic measurements show the statistics count of more than two million, which already hints on the 10000 measurements per record – it is exactly 10000 times the statistic count of the amplitude measurements, since the AIM-parameter was set to 10000.

The pulse amplitude is measured 608 mV (should be 600 mV), hence the combined tolerances of waveform generator and SDS824X HD (and cabling) are 1.33%.

The peak-to-peak measurement of 615 mV hints on negligible over/undershooting at the pulse edges.

The average positive pulse width is measured as 25.0019 µs (should be 25 µs), hence the combined tolerances of waveform generator and SDS824X HD are ~0.0076%.

The average positive duty cycle is measured as 49.99924% (should be 50%), hence the combined tolerances of waveform generator and SDS824X HD are ~0.01%.

From the Track Plot diagram, we can see that the duty cycle deviation is about 96% (should be 96%) and the modulation period is about 200 ms, corresponding to 5 Hz modulation frequency.

EDIT: Track Plot works as expected and displays the full number of modulation periods if set up correctly.

[markone]

It's not clear to me how Track and Trend functions will work on top of DSO meas system with duty cycle measurement, do you have some specific knowledge about that ?

I will let the Siglent experts chime in regarding possible tracking/trending analysis on top of measurements. But since measurements (as opposed to math functions) give one value per sweep, plus statistical values like min/max/stddev, I don't think this would give you the required time resolution for your duty cycle analysis.

The instrument user manual, section "18.5 Track", states :

The measure values VS. time plot of a horizontal parameter (e.g. frequency, rise time) in one frame
can be observed when the Track is enabled.

As I previously stated, the description is not crystal clear but leads to understand that you get a trace of selected measure towards the time on a whole acquisition buffer.

If positive, it remains to understand how the duty cycle measurement is carried out.

[2N3055]

If positive, it remains to understand how the duty cycle measurement is carried out.

It is measured same way as Pico's DeepMeasure:
Scope goes through full captured buffer and calculates duty cycle for every signal period it detects inside, up to AIM parameter number.

If you are looking at 10kHz signal and capture a 1 sec worth of data, inside that data there are 10000 periods. It will find and measure each and single one.
Measurements like P-P return single value for whole buffer.
But many measurements can extract data for every single period, getting that from single capture you can get statistics for up to 25000 periods for measurements that support it.
From single trigger.

[markone]

Hi Performa01,

many thanks for the effort spent for my question 

1. Signal Acquisition.

In order to capture enough data to properly analyze a 5 Hz modulation frequency, we need to capture more than 200 ms of the PWM signal. I have chosen one full second (100 ms/div) to capture five full modulation periods; yet I’ll only be able to display 2.5 periods because of some limitations in the Track plot implementation.

2. Horizontal Parameter Measurement.

The number of measurements within a single record (AIM-limit) is 1000 by default and can be set to a maximum of 25000. For my example to work, we need to set it to at least 10000. The time for this is only a small fraction of the total acquisition time.

the PWM sine modulation frequency of 5Hz was provided only for example, what I usually need is to measure / store / analyze the accurate duty cycle trend among, at least, a full second of continuos signal acquisition, in some cases I need an almost gapless analysis on much longer period.

This trend has, in some transient, a bandwidth much higher than 5Hz, reason why i need almost a duty cycle measure for each PWM cycle.

Picoscope works great in that thanks also to the "Progressive mode", that I guess you know better than me, where the USB 3.0 instrument streams continuously data to host PC maintaining real time traces visualization and math computation, while the program stores successive scans to PC memory.

This functionality alone it's worth the price of the small box, especially if you make also use of MSO input at the same time.

Coming back to the SDS800X, some questions :

1)  if I set 20000 for the "AIM-limit" I would potentially get a duty cycle measure point in track trace for each PWM cycle, provided that the PWM frequency is 20KHz, with continuous display update ?

2) The process will be gapless ?

3) If negative, what could be the "hole" between scans ?

[markone]

If positive, it remains to understand how the duty cycle measurement is carried out.

It is measured same way as Pico's DeepMeasure:
Scope goes through full captured buffer and calculates duty cycle for every signal period it detects inside, up to AIM parameter number.

If you are looking at 10kHz signal and capture a 1 sec worth of data, inside that data there are 10000 periods. It will find and measure each and single one.
Measurements like P-P return single value for whole buffer.
But many measurements can extract data for every single period, getting that from single capture you can get statistics for up to 25000 periods for measurements that support it.
From single trigger.

If this cheap DSO is able to work in gapless way with realtime display with acquisition of 2MSa capured at 2MSa/s and 20000 periods, it will be an instant buy

[2N3055]

If positive, it remains to understand how the duty cycle measurement is carried out.

It is measured same way as Pico's DeepMeasure:
Scope goes through full captured buffer and calculates duty cycle for every signal period it detects inside, up to AIM parameter number.

If you are looking at 10kHz signal and capture a 1 sec worth of data, inside that data there are 10000 periods. It will find and measure each and single one.
Measurements like P-P return single value for whole buffer.
But many measurements can extract data for every single period, getting that from single capture you can get statistics for up to 25000 periods for measurements that support it.
From single trigger.

If this cheap DSO is able to work in gapless way with realtime display with acquisition of 2MSa capured at 2MSa/s and 20000 periods, it will be an instant buy

What does "gapless" mean to you?

Scope works on [1."ready" for trigger] -> [2.trigger and capture] -> [3.process data] -> [4.getting ready for new trigger]  cycle.
Steps 3 and 4 are a gap when scope is not acquiring new data.

[Performa01]

the PWM sine modulation frequency of 5Hz was provided only for example, what I usually need is to measure / store / analyze the accurate duty cycle trend among, at least, a full second of continuos signal acquisition, in some cases I need an almost gapless analysis on much longer period.

This trend has, in some transient, a bandwidth much higher than 5Hz, reason why i need almost a duty cycle measure for each PWM cycle.

Like the better midrange DSOs with analysis capability, the little SDS800X HD uses up to 10 Mpts of data for its math and measurements. For longer records, the internal processing happens on data decimated to 10 Mpts. It should be obvious that this limits the duration of a single acquisition suitable for meaningful measurements at a given sample rate, e.g. one second at 10 MSa/s. For “much longer” periods it gets tricky. We can limit the sample rate to e.g. 2 MSa/s and then get five seconds record length, but that’s the absolute limit of accurate signal processing at that sample rate. And we need to be aware that the sample rate itself limits the accuracy of time measurements quite obviously.

1)  if I set 20000 for the "AIM-limit" I would potentially get a duty cycle measure point in track trace for each PWM cycle, provided that the PWM frequency is 20KHz, with continuous display update ?

You can use the Track Plot to view the duty cycle variation over the entire record. With continuous acquisition, the Track Plot is continuously updated.

2) The process will be gapless ?

As stated before, it will be gapless within one record, which can be up to 5 seconds at 2 MSa/s, but processing of each record takes a little while, so there will be a gap before the next record can be acquired.

3) If negative, what could be the "hole" between scans ?

I have run a test at 500 ms/div for a total acquisition length of 5 seconds at 2 MSa/s with advanced measurements activated. I’ve measured the time it takes until 100 measurements are completed, i.e. 100 records have been processed. It took some 511 seconds. That means 5.11 s per record; minus the inevitable 5 s acquisition time, the gap would be 110 ms.


Here is an extreme example:

500 ms/div time base for a total acquisition span of five seconds at 2 MSa/s. Modulation frequency is now 1 kHz, so I had to zoom into the Track Plot quite a bit in order to show the details. At only 1 ms/div, you can see one complete modulation period per division and you can count exactly 20 measurement points per division, which proves that every single pulse has been measured.


SDS824X HD_Track_PWM_20kHz_1kz_AIM20k

[markone]

I Performa01,

-snip

I have run a test at 500 ms/div for a total acquisition length of 5 seconds at 2 MSa/s with advanced measurements activated. I’ve measured the time it takes until 100 measurements are completed, i.e. 100 records have been processed. It took some 511 seconds. That means 5.11 s per record; minus the inevitable 5 s acquisition time, the gap would be 110 ms.


Here is an extreme example:

500 ms/div time base for a total acquisition span of five seconds at 2 MSa/s. Modulation frequency is now 1 kHz, so I had to zoom into the Track Plot quite a bit in order to show the details. At only 1 ms/div, you can see one complete modulation period per division and you can count exactly 20 measurement points per division, which proves that every single pulse has been measured.


SDS824X HD_Track_PWM_20kHz_1kz_AIM20k

this is exactly the information I was looking for, the 110ms "hole" between 5 seconds segments is more than acceptable for most of my need and to be honest, considered the price, i'm quite impressed by the SDS800X performance.

My last question about this topic : is the SDS1104X HD math / measure subsystem  faster then SDS804X HD ?

[tautech]

My last question about this topic : is the SDS1104X HD math / measure subsystem  faster then SDS804X HD ?

I don't believe so however coming from Picoscope you might find the 1000X HD with its lesser front panel buttons and more mouse/touch oriented UI a better experience.

Some 800X HD functionality was placed on the front panel due to the smaller display.

[markone]

My last question about this topic : is the SDS1104X HD math / measure subsystem  faster then SDS804X HD ?

I don't believe so however coming from Picoscope you might find the 1000X HD with its lesser front panel buttons and more mouse/touch oriented UI a better experience.

Some 800X HD functionality was placed on the front panel due to the smaller display.

My fear is that I could find 7" display a bit uncomfortable being nowaday used to 10" DSO screens, so i was considering also the bigger model, but if they share same math performance i would consider the price gap a bit "steep" or, if you want, we can say that the smaller one is to be considered a real bargain

[markone]

If positive, it remains to understand how the duty cycle measurement is carried out.

It is measured same way as Pico's DeepMeasure:
Scope goes through full captured buffer and calculates duty cycle for every signal period it detects inside, up to AIM parameter number.

If you are looking at 10kHz signal and capture a 1 sec worth of data, inside that data there are 10000 periods. It will find and measure each and single one.
Measurements like P-P return single value for whole buffer.
But many measurements can extract data for every single period, getting that from single capture you can get statistics for up to 25000 periods for measurements that support it.
From single trigger.

If this cheap DSO is able to work in gapless way with realtime display with acquisition of 2MSa capured at 2MSa/s and 20000 periods, it will be an instant buy

What does "gapless" mean to you?

Scope works on [1."ready" for trigger] -> [2.trigger and capture] -> [3.process data] -> [4.getting ready for new trigger]  cycle.
Steps 3 and 4 are a gap when scope is not acquiring new data.

Yes, you are right, I should have said "almost gapless", that is what you obtain with Picoscope and fast PC, where the steps 3/4 are quite fast. 

[2N3055]

we can say that the smaller one is to be considered a real bargain

This is the most accurate....

[2N3055]

If positive, it remains to understand how the duty cycle measurement is carried out.

It is measured same way as Pico's DeepMeasure:
Scope goes through full captured buffer and calculates duty cycle for every signal period it detects inside, up to AIM parameter number.

If you are looking at 10kHz signal and capture a 1 sec worth of data, inside that data there are 10000 periods. It will find and measure each and single one.
Measurements like P-P return single value for whole buffer.
But many measurements can extract data for every single period, getting that from single capture you can get statistics for up to 25000 periods for measurements that support it.
From single trigger.

If this cheap DSO is able to work in gapless way with realtime display with acquisition of 2MSa capured at 2MSa/s and 20000 periods, it will be an instant buy

What does "gapless" mean to you?

Scope works on [1."ready" for trigger] -> [2.trigger and capture] -> [3.process data] -> [4.getting ready for new trigger]  cycle.
Steps 3 and 4 are a gap when scope is not acquiring new data.

Yes, you are right, I should have said "almost gapless", that is what you obtain with Picoscope and fast PC, where the steps 3/4 are quite fast.

That "gap" is actually called "blind time" in scope parlance.
Blind time of Picoscopes is highly variable, depending on many factors.
I know, I have 3 of them.

[KungFuJosh]

My fear is that I could find 7" display a bit uncomfortable being nowaday used to 10" DSO screens, so i was considering also the bigger model, but if they share same math performance i would consider the price gap a bit "steep" or, if you want, we can say that the smaller one is to be considered a real bargain

The 1000 series has an RTC, bigger screen, and 100Mpts per channel (compared to 50/ch on the 800). Only you can decide if that's worth the price difference. Either one is a bargain.

[markone]

That "gap" is actually called "blind time" in scope parlance.
Blind time of Picoscopes is highly variable, depending on many factors.
I know, I have 3 of them.

I have two of them, among all PC that I used I can say that Intel based are the best in terms of "blind time", I also did some Labview programming test with provided  API and latency can be even lower, but i'm far from develop a complete scope application.

[whodiini]

Hi

New Siglent SDS804x user here and am embarassed to say, I cant figure out how to transfer a png saved screenshot from the internal memory of the  Siglent to my computer to include in a post.  I read the manual, searched this blog, played with it on the Siglent and havent figured out how to do it.  I have several fft screens I hit the save button so it is a png file on the computer.  How to move that file to my computer?  I can provide a USB stick which I can see on the Siglent.  But when I try to restore or move, cant get it to go to the USB stick.  Many thanks!

[tautech]

Hi

New Siglent SDS804x user here and am embarassed to say, I cant figure out how to transfer a png saved screenshot from the internal memory of the  Siglent to my computer to include in a post.  I read the manual, searched this blog, played with it on the Siglent and havent figured out how to do it.  I have several fft screens I hit the save button so it is a png file on the computer.  How to move that file to my computer?  I can provide a USB stick which I can see on the Siglent.  But when I try to restore or move, cant get it to go to the USB stick.  Many thanks!

Utility/Save/File Manager click more extras and use Copy/Paste to move files from memory to USB.
Be sure to highlight USB before Paste.

[whodiini]

Thanks!  It worked. Wasnt obvious to me.

[whodiini]

[/quote]


Probe set x1 (no probe connected, input open, 20M  BW)
If calculate these FFT levels to Vrms...  no need even try measure power supply 0.005mVrms... difficult to get any really reliable results.

[/quote]

I compared my baseline noise with yours and they are similar:



So now I can compare a store bought PSU to one I made with a wall wart.  Mine has a noise peak at 60kHz which I dont understand.  But at least I can make measurements and find out why!   Thanks

[whodiini]

Purchased wall wart linear regulated.



Mine

[whodiini]

I am sorry, trying to figure out how to post images, messing stuff up!

[tautech]

So now I can compare a store bought PSU to one I made with a wall wart. Mine has a noise peak at 60kHz which I dont understand.  But at least I can make measurements and find out why!   Thanks

If it's a SMPS normal I would say.
60k is a little high as they are normally 20-30 kHz however technology is always changing so recent designs may be operating at those frequencies IDK.

[EEVblog]

Look what turned up.