Siglent - 11/20 - New SDS1104X-U, 4 channel 100MHz, 1Gsa/s economy oscilloscope

[switchabl]

probably zero-padded to 16k?

Why would they zero-pad it?

I don't know, and I don't think they should (rather than just having 16k actual data points). But it sure looks like it no? If they didn't, you wouldn't see the that much detail on the side-lobes.

It doesn't look like that FFT uses much more than 4k of data in any case. If there is a way to use more, I didn't find it. This is with a current firmware and using "Memory*" as the source (whatever the * means).

[EEVblog]

Teardown photos:
https://www.flickr.com/photos/eevblog/albums/72157717234695506

Yep, much simpler front end.

[merser]

Definately looks a good alternative to the DS1054Z but for Oz buyers its more like $630 not $399. Price point is similar with the rigol.

[TurboTom]

Just for comparison: Here's a shot of both the FFTs of a DS1000Z and a MSO2000A with a signal as per @rf-loop's example (dual-tone sine 300&350kHz). For every of those screenshots, I've optimized the settings to the best results. Actually, I wouldn't call this "utterly useless", especially DS1000Z with the updated firmware (already dating a few years back) made a lot of difference. But still, the new Siglent (and other's as well) scopes' FFT wins hands down, especially usability-wise. The peak search / table function appears to be a very nice add-on, usually only to be found in SAs.

Edit: Added another FFT of the DS1000Z with the same sampling rate as @rf-loop selected on his SDS1000U. Here, the deficiencies of Rigol's FFT implementation get clearly obvious. Calculations point into the direction that Rigol's sampling-memory based FFT processes at least 50kbytes of memory, assuming that each column on the screen represents one bin of resolution at minimum available kHz/div setting.

[rf-loop]

@TT
Oh you just take this ultra simple example out from context where only purpose was tell that Siglent do not hide things like FFT data length and also user do not need calculate bin resolution because it is also displayed. It was only purpose about this image, and also show there is markers and peak search etc . This was also not my test at all but from Siglent marketing material where just display that hey there is also peak search.

Also here is markers, now for Delta.  And instead of dual tone here is used "triple tone"



In this image split window display. Carrier is over display range but not worry, of course level is there just right. Only adjusted so for get more noise level to bottom.  1MHz, AM modulate 1kHz 20%. (As can see (as can see sidebands levels are -20dBc.)




Here full window display, 1MHz, AM, 1kHz, 100% (as can see  LSB and USB levels -6dBc.)

It is not SDS1x04X-E what have 8 times better resolution with 1048576 points FFT data length when X-U model have 131072 points what give here with 5MSa/s (5MHz)
5000000Hz/2/(131072/2)= 38.15Hz when  X-E model have in same place  4.77Hz.

As can see in images this signal is very easy to SDS1104X-U. This is resolution for 0 - 2.5MHz band. Of course if take 0 - 500kHz FFT like in @TT image then Siglent X-U  FFT bin is  7.63Hz.

This table can also show more, also freq etc but I turn these off and only show delta with just only 3 markers for more clear visibility.


In this image tiny bit more challenging case but still without any problem with SDS1104X-U.
Signal 325kHz sine, carrier level -40dBm. AM modulation with 100Hz. Modulation depth 0.2% using -40dBm carrier  sides need be -100dBm level (-60dBc).  Just like checkbox - done.

Also it tell that input channel noise level is not very bad. Using full BW in same setup rise base noise level bit. If you think rise this persistence yellow noise top to -110dBm line it is just there. (note averaging)

This is what you can try...

ETA:  Table in Reply #65 corrected. There was error in amount of available memory.
Right is: 1 channel alone on, max 14M. When 2, 3 or 4 channels simultaneously on, all channels have max 7M !

[TurboTom]

Thanks @rf-loop for demonstrating this kind of performance. It's not even worth to try to match that with any of the FFT implementations of the "mature" Rigol scopes. I don't know about their more recent designs or other manufacturer's products. But it seems that noise- and FFT-wise, the engineers at Siglent really did their homework! It's quite interesting to watch how the products evolve, especially since hardware-wise, the DS1000Z and the SDS1000U aren't too far apart, except for the main SOC silicon... I guess, the software approach makes most of the difference, of course in combination with the more powerful Zync SOC/FPGA combination in Siglent's product. I'ld actually love to see a very compact scope (like the DS1000Z), equipped with the grunt of an SDS2000X Plus or (probably more realistic) of a DS5000.

[rf-loop]

It is not one or two time when someone ask how to get AM signal trig'd  to modulating frequency. Of course old traditional answer is, adjust Trigger Hold Off  time. Yes it works. But, it is also very poor method. When modulating frequency change bit more adjust again.. adjust again.. some times very frustrating. Also I have seen many times peopled give advice to use trigger holdoff.  All modern digital scopes have lot of more clever things that these traditional things what come from analog oscilloscope era.

With this method here trigger can be rock solid without any adjustment over whole normal audio range (90us in image is too long time except if need trig to very low, well under 10Hz, mod frequencies what are not normally used).

This modulating freq change speed is not constant. This is not at all because scope trigger. This is how I adjust generator what is not very handsome  for this purpose.

Signal is 21.2MHz and modulation stepped 10Hz - 4kHz.



Of course this is not just Siglent SDS1104X-U feature. Most modern scopes have this or other suitable trig functions for this.

[rf-loop]

SDS1104X-U  have also timing inaccuracies as all oscilloscopes have more or less.

It is good to know least roughly how much is scope own time jitter when we try measure some signal jitter what we do not know. Of course if you have oscilloscope in your hand you can check it quite easy, if you have tools. But all do not have. Data sheet tell only roughly how big error there can be in frequency but it do not tell anything about how much it have jitter. Also just this time I do not have signal source what jitter I know sure but with other scope I can compare. If other scope is much better with it I can enough sure that test signal have same or less jitter but it can not be more.
So here is test. Same one signal come from signal source and it is split to both scopes and test is made simultaneously just in same time.
Other scope is SDS2104XPlus what have quite good reference. How ever I try SDS2104X+ can not show more total jitter over 10ms than up to 0.5ns peak to peak over more than 15minute period.
After then I start both scopes with around same settings and infinite persistence.
In image there can see SDS1104X-U result and for compare in same image is SDS2104X+ result.
Is it  much or not. It depends what user is doing. It is ≤ 0.6ppm peak to peak.  So,  1ms something like 0.6ns.
SDS1104X-U most fast time scale is 2ns/div.



Of course this is just one individual scope so do not take this result as "data sheet".
Btw in SDS2kX+ image can also see corners "wobbling" because what is not of course from scope in this case.

[rf-loop]

Images tell.


Only one wonder. In sequence mode it is bit faster than SDS1104X-E if look top speed.



Note, history buffer wfm amount is same as sequence mode maximum amount.



Really this wfm/s thing is not most important thing. Also there may exist more fast burst speeds but I do not note these. Some manufactures may also use these peak speed. These are not peak speed but continuous average wfm/s  speed least over 1s. (sequence speed is different thing)

[tautech]

Apparently one is on the way to me. Wioll be interesting to see how stripped down the design is compared to the existing model.

3 weeks pass and only crickets. 

[rf-loop]

Here is other "fact sheet". 
t/div, max mem limit, (only two selections noted, most low and most high) then background history buffer max wfms count (fifo length) and then max amount of segments in fast sequence mode. Also there is calculated max amount of captured data memory in different cases. Max up to 56M.


I hope I have corrected my mistake in every place where is exist.
I also repeat it here.
Right information is:
If only one Ch is in use:  it have max 14M
If two, three or four channels are on simultaneously: Every channel have max 7M.

[2N3055]

Apparently one is on the way to me. Wioll be interesting to see how stripped down the design is compared to the existing model.

3 weeks pass and only crickets. 

He did a teardown:

 

[tautech]

Apparently one is on the way to me. Wioll be interesting to see how stripped down the design is compared to the existing model.

3 weeks pass and only crickets. 

He did a teardown:

Maybe that's all we'll get from Dave for the X-U. 

Therefore I'll have to get a unit to Defpom for a deeper look when some I have coming arrive next week.

[rf-loop]

Here freq response.

[tautech]

Got my hands on some. 

Dunno why but immediately noticed a little addition in the Save/Recall menu where in the X-U the Bin2CSV convertor application has been added whereas for X-E it's currently only available with a virtual key from the instrument control feature within the webserver.
Maybe Siglent will add this to X-E in a future update. 

X-U


X-E