Siglent SSA3000X and SSA3000X-Plus Spectrum Analyzers

[KE5FX]

So I made a return loss bridge to use with the Siglent analyzer in VSWR mode. Some of the more difficult parts to find is very thin 50Ohm quality coax cable to make the balun to couple the bridge to the measurement port. Actually I found a seller for 1.2mm PTFE coax but then I came across an offer for adapters that connect an SMA socket to one of those ultra-tiny PCB clip-on connectors (for WiFi or cell phone equipment). But the best about this was that it contained about half a meter 0.8mm PTFE coax and cost close to nothing (at least compared to the amounts to be spent for a "proper" RLB -- it must have been arond 4€). So I quickly found some scrap double sided FR4 (ages old but the photoresist still working), mixed some developer and etching solution and tried if I still can "cook" PCBs "the old way"... 

Very nice!  I think that's the best-performing homebrew return loss bridge I've seen, or at least it has the potential to be.

Can you be more specific about where you got the 0.8mm coax?  It's incredibly useful stuff.  I've been getting it by cutting up Murata JSC jumpers from DigiKey, but that process only yields 200 mm at a time.

[Bicurico]

What I really think needs improvement on the SSA is how "frozen" traces are processed. They scale properly if level-related adjustments are done, but if frequency-related parameters are changed, they just stay where/as they are. No stretching/shrinking or shifting at all so they are completely useless if they are supposed to be compared to a newly recorded trace that (partly) covers a similar span. On the DSA815 this is solved much better!

Funny, as this is exactly the same problem I have with my SMA Simple Spectrum Analyzer software! I implemented traces (and set each to WRITE, MIN, MAX, AVG permanently on later versions). When you change amplitude, the stored values are correctly shifted in Y direction. But if you change frequencies, I simply delete what is stored in each trace.

The reason is simple: imagine I have trace 2 in VIEW mode for a span of 10MHz. Now the user decides that the span should be 100MHz. How should I draw trace 2? In order to do it correctly, I would have to actually ignore the span setting internally and record the full span, so that I could show whatever part of the span the user selects after having VIEW activated. This would require a really fast spectrum analyzer!

Alternatively I could just draw whatever bit of trace is stored, like a curve that starts and ends somewhere in the middle of the screen.

Finally, a cheat could be done, where the VIEW mode is temporarily interrupted to gather new data and then set again.

Now that you mention the Rigol, I wonder how they do it?

Regards,
Vitor

[hendorog]

What I really think needs improvement on the SSA is how "frozen" traces are processed. They scale properly if level-related adjustments are done, but if frequency-related parameters are changed, they just stay where/as they are. No stretching/shrinking or shifting at all so they are completely useless if they are supposed to be compared to a newly recorded trace that (partly) covers a similar span. On the DSA815 this is solved much better!

Funny, as this is exactly the same problem I have with my SMA Simple Spectrum Analyzer software! I implemented traces (and set each to WRITE, MIN, MAX, AVG permanently on later versions). When you change amplitude, the stored values are correctly shifted in Y direction. But if you change frequencies, I simply delete what is stored in each trace.

IIRC the Signalhound SA124B removes the trace if you change pretty much anything - and it doesn't return if you change the setting back.

Alternatively I could just draw whatever bit of trace is stored, like a curve that starts and ends somewhere in the middle of the screen.

This behaviour would be the one I would choose - just draw what you have in the correct location when any setting is changed.

[papo]

TurboTom

Nice work on the bridge. I'm also wondering about your source for the thin coax -- this looks like a fun project indeed.

Since you and rf-loop had a good look at it... Question on the RBW limit when the TG is on: How would you characterize a filter that has a very deep notch, say 70 to 80 dB? The 30 kHz RBW limit seems to be a problem in that case since it's not possible to further reduce the displayed noise level. Or am I mistaken? I have to admit that while I feel confident about having understood how spectrum analyzers work, I don't think I have fully understood what's going on with the tracking generator in this instrument.

[rf-loop]

TurboTom

Nice work on the bridge. I'm also wondering about your source for the thin coax -- this looks like a fun project indeed.

Since you and rf-loop had a good look at it... Question on the RBW limit when the TG is on: How would you characterize a filter that has a very deep notch, say 70 to 80 dB? The 30 kHz RBW limit seems to be a problem in that case since it's not possible to further reduce the displayed noise level. Or am I mistaken? I have to admit that while I feel confident about having understood how spectrum analyzers work, I don't think I have fully understood what's going on with the tracking generator in this instrument.

I think this picture tell something about what SSA3000X is capable of using TG. (if talk about "dynamic range")  With external sweep generator, of course, can reach much wider range.  But, then we perhaps meet also SA own phase noise, depending filter width and other things.

Here is simply cascaded two Mini circuits SHA800  High pass filers. (poor method but stop band is well below so that only limiting factor is SA.

[rf-loop]

National Instruments  LabVIEW driver for Siglent SSA3000X series.

[TurboTom]

So let's start with the possible sources for very small coaxial cable: One source for 1.13mm PTFE / FEP cable is UCTRONICS: http://www.uctronics.com/1m-meter-rf1-13mm-silver-plated-audio-video-signal-cable-32awg-black-red-gray.html. I didn't find a source for "naked" even thinner cable so for the 0.81mm variety, I reverted to salvaging pre-configured cable assemblies (from the "bay"): http://www.ebay.de/itm/171651868955. Make sure to chose the maximum length available (50cm/20" in this case). One such cable was exacly sufficient for one of my RLBs (around 48.5cm required). I'll post a BOM and design information on my RLB later and probably in a separate thread.

@papo -- In general, characterizing a high resolution filter with the 30kHz RBW of the SSA3000X when the TG is in use isn't a problem. You could actually do that even without any filtering at the input (i.e. with a broad-band RF power meter), only scanning the frequency with the generator. On the SSA, the VBW setting affects the increments at which the TG scans the span, so this works perfectly well for filters / components that slope in the range of about 80db/10kHz. For whatever reason, components with steeper slopes (i.e. the high precision 1MHz crystal that I used in several of my tests which has a series->parallel resonsnce slope of about 80dB/1kHz), the SSA3000X produces artefacts. This hasn't got anything to do with the measurement principle, I rather guess there's a problem with the settling of the TG / gating of the SA relationship. I'm sure with some modifications in the SSA's firmware and maybe allowing a little more time for the sweep, the problem can be solved (though manual adjustments of the sweep time don't have any effect).

@Vitor -- On the Rigol DSA the "frozen" traces are stretched/compressed as data is available. But one ugly detail is that the last valid level is used as the missing data to fill the screen (see the attached screenshot, the magenta trace was recorded before and then the span has been increased, compressing the "real" information just to the left of the window). Of course, this doesn't make any sense but maybe the graphics engine requires all traces to span from left to right of the display window and none are premitted to start / stop at an arbitrary point. If this is actaully a requirement, I would prefer to have the level of the "fill-in" data at the minimum that can be displayed (-171dB or whatever the math resolution permits). Of course, displaying the frozen traces in non-vectorized form (or a dashed line if possible) would be best if they are stretched/compressed. This would directly make it obvious to the user that the scale is not matching the record for these traces.

Cheers,
Thomas

[rf-loop]

This image tell better what is dyn range for measure filters using TG.
Green: Normalized with very short cable.
Blue: Base noise level (connection between TG out - SA in OFF)
Yellow: Connection between TG out - SA in OFF-ON-OFF-ON-OFF-Short(leak-peak)-ON-OFF-Short(lleak-peak)*)
*) in this case short peaks because it was my finger what was too close what destroy OFF isolation - just like kid playing - what hapend if....

(OFF isolation in this context "infinite")
(all time same cable, same normalization, same RBW and VBW))

In practice this mean that in practice  range is  max ~70dB - 80dB
(naturally as can see normalization is "mirrored" to noise level.)

[papo]

TurboTom and rf-loop, thanks for the info, clarifications and screen shots. It looks like the 30 kHz limitation is exactly like I originally assumed. I agree that it's not a limiting factor for most measurements -- I have purchased the instrument because I very much liked what I saw. Yet I am wondering about the reason for it. Just out of technical interest, any idea on why the limit is there? I feel like I'm missing something here and I don't like that because with these instruments, it's easy to mis-interpret a result if one doesn't understand how the instrument works.  All possible explanations for the 30 kHz RBW limit that come to mind are somewhat strange...

[Bicurico]

Hi,

I made a small and simple video of the Siglent SSA3021X being remote controlled by EasySpectrum in response to a few people asking for a video showing spectrum analyzer at work.

https://youtu.be/_qphYeB0-TM

Regards,
Vitor

[Siglent America]

I know that many people have asked for videos on the Siglent SSA3000X Spectrum analyzers. Shahriar will be doing a review although he has said that he is running behind on his Signal Path work.

Here is a nice video that one of our new aps engineers just posted to YouTube this morning. I thought it might be interesting to some of you.
https://www.youtube.com/watch?v=QtZPz-OUCSU&feature=youtu.be

Thanks

[Bicurico]

Nice video.

Perhaps you might want to make further videos:

- explaining general and special functionality in more depth (even for existing users)
- showing general performance (which is what I tried to show on my video): people are often just interested in seeing basic things like the sweep rate

Regards

[Siglent America]

Nice video.

Perhaps you might want to make further videos:

- explaining general and special functionality in more depth (even for existing users)
- showing general performance (which is what I tried to show on my video): people are often just interested in seeing basic things like the sweep rate

Regards

Thanks, Bicurico.
We will definitely be making more videos showing different measurements and applications.
Thanks for your comments.
Steve

[Jester]

Let me try to put it like this: The Rigol DSA815 has a very simple but otherwise clearly understandable approach to measuring with its TG in use: Both TG and SA section scan the frequency span in constant increments, defined by the count of displayable pixels on the screen (round about 700 in case of the DSA815) and the span. So the TG outputs the center frequency of the "bin" (defined by the resolution bandwidth) the SA is sampling. Since the bandwidth of the TG signal in each "bin" is much smaller than the minimum 100Hz RBW of the Rigol, you can consider the signal to be picked up after a narrow-bandwith component (crystal for instance) a discrete frequency scan of the component. The video bandwith setting has no effect on the TG output.

The situation with the Siglent is much more complex and so far, I'm not sure if I understood it completely. Since the RBW of the instrument with the TG in use is limited to 30kHz and above, frequency discrimination has to be done by scanning with the TG. And here's the big difference vs. the Rigol SA: The VBW setting affects the discrete frequencies to be swept with the TG, i.e. a lower VBW setting results in the TG discrete frequencies to be more closely spaced while a higher VBW causes them to be further apart. Consequently, this also affects scan speed. Why this method of sweeping the frequency results in the "hills and valleys" may be the result of some "chirping" of the TG when setting the new frequency during a sweep. It should be possible yet to eliminate this problem by temporal "windowing" the measurement after the TG has settled. I guess Siglent should address this issue.

Cheers,
Thomas

My question relates to the TG explanation quoted above:
With an oscilloscope connected to the TG output of my SSA3021, I observe a pretty nasty looking waveform on my scope (see below)
This is with start = 200.0 kHz, stop = 200.1 kHz, frequency step = 1Hz. Sweep time 50 seconds

My first impression was there is something seriously wrong with the TG as I was expecting a nice smooth and clean looking frequency ramp. My next thought was that the TG is actually stepping not sweeping the frequency and perhaps the frequency requires some time to stabilize.  I suppose that if the SA measurement firmware knows the time required for the TG to become stable and then ignores the data while it's stabilizing that might be okay?

Decreasing the sweep time or increasing the VBW makes the really easy to observe.

Is this normal or do I have a bad TG?

[nctnico]

An RF generator (which is what a TG output is) isn't a function generator so in general don't expect a clean frequency sweep. So yes, the spectrum analyser will step the TG along with the frequency it is sampling.

[TurboTom]

@Jester --

that's basically what I found here: https://www.eevblog.com/forum/testgear/hack-of-sigllent-spectrum-analyzer-ssa3021x/msg975408/#msg975408, yet I don't know why Siglent have arranged it like that. May it be unintentional, or is it done like this by purpose? So at that time, I wouldn't consider it normal but apparently it's present on all the TGs of the SSA3000X series of Analyzers. Other SA's TGs are also not the "cleanest" but not as weird as this. When testing high-Q components, the huge frequency spread may actually excite resonances that aren't part of the span one is looking at. As yet, there's no official statement on this behavior from Siglent. And obviously, there's the documented problem with analyzing components with very steep slopes in the frequency domain.

Cheers,
Thomas

[MotoDog]

Is the 2.1 GHZ model in production now stable.  Should I wait to buy one?   Lately it has been all about output picture format.
I would not worry that much.  Is it buggy and has Siglent taken care of any bug fixes?   Are there any game-stoppers?

I watched the eval against the Rigol (which I got), looked impressive.

I want a tracking generator and it looks like the Siglent has a lot better noise
performance and the narrow bandwidth, as well as TG hardware built in?

I am just a experimenter Ham type.  I would be happy if I could retrofit the Rigol with a tracking generator, but they won't do it.
I bought my Rigol used, did not know it could not be retrofitted.

Thanks for any info!
Mike

[Bicurico]

If it is just the tracking generator that you are missing out on, let me recommend this neat device: http://www.roverinstruments.com/prodotti.php?idp=4&camblingua=2&changelan=yes&idprod=214

It is a small wide noise source, reasonably flat and can be powered by *any* USB port. It will consume very little energy, so even a low spec USB battery will last ages. But you can just pop it on the spectrum analyser's USB port (works with my Siglent SSA3021X).

You will get an output power of -56dBm from 4MHz to 2500MHz. I am not sure about the price, but I think it costs around 300 Euro. It comes with an SMA to F-type adapter cable and a T-connector, which is used on Rover Instruments meter's with the reflectometer app.

With this noise source you will effectlively be able to do the same as with a tracking generator.

Another, cheaper, option is this: http://www.ebay.de/itm/381623138109

This is a low cost noise source. Not as flat, precise and robust as the Rover Instruments one, but 10 times cheaper.

Carefull though, as it a) gets really hot and b) does not have any case or shielding: I broke the first one when the PCB touched a cable conector lying on my table...

To sum up: if you already own a Rigol, you don't really need to buy a Siglent just for the tracking generator. However, you might want to sell your Rigol and get the Siglent, for enhanced display resolution, increased bandwidth, lower RWB and more features/options.

Regards,
Vitor

[MotoDog]

Thanks!

Yes, I actually laid out a pcb and built a noise source.  2  RF Amp ICs looking at a Zener Noise source.
I never got it to work.  I guess I gave up.  I even used stip line technique per a data sheet.  Trying to keep 50 ohms.

So, maybe I better go pull out the project from the junk in the garage.
I guess if I could get my Rigol sold for a decent price $500?  It is still over another $1000 for the slick Siglent!

BUT, Rigol just released a cheaper model! Crap.  I should have bought what I wanted in the first place.

OK, go find the noise maker in the junk box.

Mike

[tautech]

Thanks!

Yes, I actually laid out a pcb and built a noise source.  2  RF Amp ICs looking at a Zener Noise source.
I never got it to work.  I guess I gave up.  I even used stip line technique per a data sheet.  Trying to keep 50 ohms.

So, maybe I better go pull out the project from the junk in the garage.
I guess if I could get my Rigol sold for a decent price $500?  It is still over another $1000 for the slick Siglent!

BUT, Rigol just released a cheaper model! Crap.  I should have bought what I wanted in the first place.

OK, go find the noise maker in the junk box.

Mike

@ MotoDog
Here's the link to Ca suppliers:
http://siglentamerica.com/map.aspx?id=1054

Do remember Saelig offers EEVblog members discounts, use this thread to ask for the code to be PM'éd to you.
https://www.eevblog.com/forum/testgear/equipment-discounts-from-saelig/

[bozidarms]

Hi,
some nice curves - SSA3000X pushed to the limit:
1 - CW signal  -127dBm,
2 - CW signal  -143dBm
Regards

[rf-loop]

@Jester --

that's basically what I found here: https://www.eevblog.com/forum/testgear/hack-of-sigllent-spectrum-analyzer-ssa3021x/msg975408/#msg975408, yet I don't know why Siglent have arranged it like that. May it be unintentional, or is it done like this by purpose? So at that time, I wouldn't consider it normal but apparently it's present on all the TGs of the SSA3000X series of Analyzers. Other SA's TGs are also not the "cleanest" but not as weird as this. When testing high-Q components, the huge frequency spread may actually excite resonances that aren't part of the span one is looking at. As yet, there's no official statement on this behavior from Siglent. And obviously, there's the documented problem with analyzing components with very steep slopes in the frequency domain.

Cheers,
Thomas

It is under investigation in Siglent.

[videobruce]

 I know the US distributor is in New York State, but isn't service on these done back in China?

I just starting reading thru this, a quick question, is the general conscious this is better overall than the Rigol?

[tautech]

I know the US distributor is in New York State, but isn't service on these done back in China?

I just starting reading thru this, a quick question, is the general conscious this is better overall than the Rigol?

Ohio actually.
http://www.siglentamerica.com/how_to_buy

Repair of modern instruments is done at module level with faulty modules sent back to the factory if still under warranty. There are instances of members requiring replacement PCB's of other Siglent products and some have dealt directly with Siglent Tech support and others have sourced from their local dealers.


Do some more reading, there's some comparison between the two brands and there's also Dave's comparison vid between them where this Siglent comes out on top.

https://www.eevblog.com/2016/06/20/eevblog-891-siglent-ssa3021x-vs-rigol-dsa815-spectrum-analyser/
https://www.eevblog.com/2016/06/22/eevblog-892-siglent-ssa3021x-spectrum-analyser-teardown/

[bozidarms]

Hi,

some measurements with this extraordinary instrument:
directivity of diy return loss bridge,
reflection for 40m inverted vee and magnetic loop antenna,

regards