New Siglent SDS2304 in some tests.

[rf-loop]

This is SDS2000 series "flag ship"  but many things same as in whole SDS2000 series.

SDS2304
-4 channels
-300MHz
-2GSa/s max per channel if 2 channel in use. 1GSa/s max if all 4 channel in use.
-Up to 28Mpoints full speed sampling buffer.
-Separate Vertical adjustments for each channel (position and V/div (1-2-5 steps) including fine adjust)
-Trig out / Pass-fail out.
-Ext trig

-Signal generator (multi waveform, 1 channel)


Here some tiny test examples.

First tested waveform update rate.

All 4 channels on and 50ns/div it have maximum 65k waveform / s
Channels 1+2  or channels 3+4 in use it reach 110k waveform / s
(CH1+2 on  109.4k;  CH3+4 on 110.4k)
Interesting is that also mesurements do not drop update rates.
Only what really drops update rate is if select vectors instead of dots for display.

Test data in attached image.
Removed as Obsolete

[Hydrawerk]

Yes, this scope looks good, but why is the separate CH1 slower? Any idea?

[rf-loop]

Yes, this scope looks good, but why is the separate CH1 slower? Any idea?

I quess secret is how they handle this double ADC output 2x8bit+2x8bit datastream in single and double Channel mode.

[Herman]

a video test of SDS2102 made by Siglent.the TI ADC08D1020 chip http://www.ti.com/lit/ds/symlink/adc08d1020.pdf

[rf-loop]

Topic name is "New Siglent SDS2304 in some tests."


Perhaps this other thread...

https://www.eevblog.com/forum/testgear/siglent's-new-product-msosds2000-series/135/

...is better for other things related to whole SDS2000 series. This topick is, as there read, for SDS2304 tests.

[echen1024]

a video test of SDS2102 made by Siglent.the TI ADC08D1020 chip http://www.ti.com/lit/ds/symlink/adc08d1020.pdf

What were the settings used on the scope at that time?

[Herman]

Just as what rf-loop setup

[echen1024]

Just as what roof setup

OK. And I'm pretty sure you meant rf-loop

[marmad]

@rf-loop: Thanks for your tests - and, as I've mentioned before, the new DSO from Siglent looks very interesting.

But as I've also mentioned before (in the New Siglent DSO thread), the measured waveform update rates at slower time bases looks slightly strange.

The problem is simply a matter of time - and I don't just mean the time it takes to actually sample 14M or 28M of data - which in itself, is a fairly large chunk of time - but more importantly, the time it takes to decimate and process that many samples to display memory.

According to your Trigger Out readings - the Siglent is very fast at slower time bases - faster than the Rigol DS2000 or Agilent X-2000 or even the 1 million wfrm/s Agilent X-3000 - or ANY other DSO, for that matter, that I've seen measured - if you look at the throughput (sample size * wfrm/s).  Starting at 50us/div, the blind time is considerably lower than all of the competition - but if you consider the Siglent's throughput, it's speeds are better than Agilent's MegaZoom ASIC design even at smaller time bases. Of course, these rates are in Dots mode - while the Agilent is doing anti-aliasing and interpolation - but it's significant nonetheless.

Now of course, it's possible that Siglent has achieved something new with their design - but it's also possible that they might be cheating with the Trigger Out rate. For example, @500us/div, the Siglent supposedly has a rate of 135 wfrm/s with a 14M sample size. Alternatively, the Rigol DS2000 series has a waveform update rate of just 39 wfrm/s when having to capture and decimate a 14M sample size, but a waveform update rate of 140 wfrm/s with a 14k sample size - much closer to the Siglent (and Agilent's) rate.

So, you see my point: this has to either be a new technological advance by Siglent with very-deep memory sampling and decimation speeds - or it has to be something else. Perhaps you can run some tests just to confirm that the DSO is actually capturing AND decimating, for example, 135 wfrm/s of 14M @ 500us/div?

And perhaps you can test the Siglent's wfrm/s speed when you switch between different sample lengths @ a single time base like 500us/div?

[Carrington]

So, you see my point: this has to either be a new technological advance by Siglent with very-deep memory sampling and decimation speeds - or it has to be something else. Perhaps you can run some tests just to confirm that the DSO is actually capturing AND decimating, for example, 135 wfrm/s of 14M @ 500us/div?

Interesting, I am impatient to see the results of these tests.

[rf-loop]

Here check what is signal full vertical range related to display area.
Result. Signal full scale is small amount over 10 div. Displayed is 8 div.
(this is also semi important for note when compare example different oscilloscopes displayed noise levels) Also good to know.

First scope run and signal in and signal here over 10div pp.
Then scope stop and shift up and down to look there is not clipping or bad distortion near f.s.

Small note, in running scope do not draw fake line to bottom or top display.
But, in stop mode it do it. (it must not do this!)

I try keep some tests here out from common SDS2000 series discussing.

Soon I will show something about internal constructions and details.
(it is now disassembled for investigations)

 

[rf-loop]

Here something about freq response.
First test with around 70ps risetime signal from Tek 284.
There is moderate overshoot (not from Tek). It is somehow explained in Agilent paper where is explained things about oscilloscopes risetime and frequency response.
Siglent response is not gaussian. Due to this overshoot amount I hope Siglent do just small  fine adjustment in front end circuit.

Agilent paper can read here. Look rising edge and overshoot in Siglent and compare to Agilent paper Figure 2. http://cp.literature.agilent.com/litweb/pdf/5988-8008EN.pdf

Next pictures: I have set 1MHz sine as "reference".
Signal is coming from Agilent (HP) 8644B signal generator what individual unit have also very good flatness in this freq area. After then good quality around 1M 50 ohm coaxial and oscilloscope input (internal) 50ohm.
In these pictures scope is in Sin(x)/x mode.
-3dB related to 1MHz ref is of course over 300MHz (it was my mistake adjust 320MHz instead of designed 325MHz (I find this mistake after I look images).
It is really good to note that risetime measurement alone is NOT ok for classify freq response or upper -3dB point. (it is also noted in this Agilent paper)

[rf-loop]

Pass/fail function speed.

Siglent advertise it have fast hardware based  pass/fail mask test.

It looks like it is true and it can around same speed as normal full waveform update rate.

Here tiny test. (test made with disassembled (case open etc) oscilloscope what may lead to some extra noise in signal but here it do not have any meaning)

CH1 and CH2 on. 50ns/div.
Display mode = dots.
Signal adjusted so that it give fail continuously.
Also some automatic measurements on. (not affect speed)
Pass/Fail output give pulse every time waveform is captured and it do not pass.

with this setting, ~109400  test fail indicating pulses in second.
Measured with HP53131A counter (reading frequency from Pass/fail output.)
Checked also using Tektronix 2465 analog oscilloscope.

[Mark_O]

rf-loop, thanks for performing all the tests, and sharing the results.

Siglent advertise it have fast hardware based  pass/fail mask test.

It looks like it is true and it can around same speed as normal full waveform update rate.

That's quite impressive.  I'm not sure what other scopes have such a capability?


One negative comment though, and this falls into the "nitpicking" you mentioned should be deferred until the real problems are resolved.  However, I find the text (font) in general a bit difficult to read, and completely impossible in the Pass/Fail stats box.  Especially the Red.  I did a grab & zoom, and now I see why.  It's squeezed so tight that the pixels of adjacent characters are overlapping.  See the attachment.

[rf-loop]

One negative comment though, and this falls into the "nitpicking" you mentioned should be deferred until the real problems are resolved.  However, I find the text (font) in general a bit difficult to read, and completely impossible in the Pass/Fail stats box.  Especially the Red.  I did a grab & zoom, and now I see why.  It's squeezed so tight that the pixels of adjacent characters are overlapping.  See the attachment.

Yes, there is many this kind of things in GUI what need more or less adjust/fix.
I hope also that test mask can draw without signal and also I hope adjustable transparency for test mask and/or if it may be only test limits lines (as example in some older HP).
During pass/fail test on I really want still see whole display area and signal trace/traces there. Also in case that I have selected stop if fail... of course I want also look this failed signal (without turning mask off).

[marmad]

Siglent advertise it have fast hardware based  pass/fail mask test.

It looks like it is true and it can around same speed as normal full waveform update rate.

Very nice.

That's quite impressive.  I'm not sure what other scopes have such a capability?

Apparently, the Agilent X-Series does it in hardware at the same speed as waveform updates. The Rigol's do not - although you can capture segments at waveform update rates and then run Pass/Fail on those  - but of course, it's not a real time comparison and you're limited by the number of segments.

[kebogen]

Pass/fail function speed.

Siglent advertise it have fast hardware based  pass/fail mask test.

It looks like it is true and it can around same speed as normal full waveform update rate.

Here tiny test. (test made with disassembled (case open etc) oscilloscope what may lead to some extra noise in signal but here it do not have any meaning)

CH1 and CH2 on. 50ns/div.
Display mode = dots.
Signal adjusted so that it give fail continuously.
Also some automatic measurements on. (not affect speed)
Pass/Fail output give pulse every time waveform is captured and it do not pass.

with this setting, ~109400  test fail indicating pulses in second.
Measured with HP53131A counter (reading frequency from Pass/fail output.)
Checked also using Tektronix 2465 analog oscilloscope.

The P/F function on my SDS1202CFL is pretty good,but the speed is not  satisfying.
But I use this function very seldom.

[rf-loop]

About inputs 1Mohm / 50ohm.

Of course input circuits capacitance can not remove. No one can do it. But it can build so that input impedance is near 50ohm over some specified frequency range.

I have tested some scopes and here now I do not have comparable results available but, I remember that some scope was so that internal 50ohm give nearly same result as external Tektronix 50ohm (011-0049-01) Feed-through with 1Mohm input.
Note that Siglent input capacitance is quite high for 300MHz scope. (18pF +/-?)
(if they can drop it to example 8 - max 12pF, I like it better))

Here this test with Siglent SDS2304
Signal source Tektronix 284, ~70ps risetime pulse generator.
1m M17/084 coaxial.

All settings in scope exactly same, exept
image a with Siglent input "50ohm"
image b with Siglent input "1M" + Tektronix Feed-through terminator between signal cable and CH1 input.

Difference is big enough for think that there is "something" for better impedance match than just pure resistance over input.  (but as I have write before: I hope Siglent do small fine adjust in front end circuit for better pulse response.)

It need note that this kind of input can not handle 70ps rising edge of course. If I drive it with 1ns risetime signal, situation with internal 50ohm is much better and clearly less overshoot. But still high overshoot if use external terminator and scope set 1M input. (later I can do more tests but now scope is totally disassembled.)