EEVblog #892 - Siglent SSA3021X Spectrum Analyser Teardown

[EEVblog]

A very detailed look inside the new Siglent SSA3021X 3.2GHz Spectrum Analyser.
The entire RF section is broken down and analysed in depth with a block diagram overlay.

[jancumps]

You may also have broken down youtube. There are playback errors.

The error is very representative for a teardown video error thoug.

[Pinkus]

Very nice. Thank you.
May I ask for some high-res photos also from TG and main processor board?
Interesting to see on the upper right sight: an SD card connector and an unpopulated FPC connector with a missing TSSPO chip. My guess: the latter could be for a planned and/or optional (though resistive) touch panel cable with touch screen controller?
Thx

[EEVblog]

May I ask for some high-res photos also from TG and main processor board?

Yes, will upload tomorrow.

[System Error Message]

Would really like to see this in action, do you think you could do a video of it?

edit: nevermind just found it in a comparison video. Its just unboxing and teardown videos without any action have become too common.

[retrolefty]

The teardown was great. Sure seems like a whole lot of SA for the price point. I don't need a SA but I've always dreamed of owning one. 

[Laertes]

I gotta say, the format you chose for this teardown is great! It seems you took a bit of a page from Shahriars/The Signal Paths book there, with the block diagrams and the photos with overlay, each individual block explained in depth etc.
Excellent work

Though I do imagine it's a lot more work than just dicking around with your pointer straight on the actual board...

[tautech]

Would really like to see this in action, do you think you could do a video of it?

edit: nevermind just found it in a comparison video. Its just unboxing and teardown videos without any action have become too common.

Relevant threads:
https://www.eevblog.com/forum/blog/eevblog-891-siglent-ssa3021x-vs-rigol-dsa815-spectrum-analyser/
https://www.eevblog.com/forum/testgear/comparison-rigol-dsa815-tg-with-siglent-ssa3021x-spectrum-analyzer/
https://www.eevblog.com/forum/testgear/siglent-ssa3000x-spectrum-analyzers/
https://www.eevblog.com/forum/testgear/hack-of-sigllent-spectrum-analyzer-ssa3021x/

Manufacturers website:
http://www.siglent.com/ENs/pdxx.aspx?id=1299&T=2&tid=18

[G0HZU]

At 23:32 in the video the function of the first RF switch U1 is to switch in the overload dummy load R1 (via C9) if there is an overload.

i.e. if there is a large overload the analyser can simply divert the RF input to the big 49.9R resistor R1.

The input blocking cap C10 is likely to be a big value possibly as big as 1uF. This can obviously store a lot of energy (for a given charge voltage) compared to a regular RF cap down in the nF range so I would expect to see something to protect the front end from transients from this cap if DC is ever applied/removed at the analyser input. Maybe those clamping diodes help here but it's hard to see how they are configured.

[G0HZU]

The second SOT23-6 RF switch U2 at 24:06 looks like it allows an internal test signal to be switched in to the front end of the analyser. Possibly for self testing/calibration purposes. The opamp IC to the right of it is probably something to do with the RF overload detection. Possibly a crude diode detector/comparator/trip to control that first RF switch U1 that switches in the (overload) dummy load resistor R1.

The input switches U1 and U2 look quite feeble in the SOT23-6 package but they are probably rated to about 1W and probably a bit more before they actually get damaged. So U1 can probably dump something in the ballpark of 1W into R1 if there is a bad overload. But much more than this and something is going to release some magic smoke somewhere...

[altaic]

Dave, I really liked the front-end breakdown and explanation of the block diagram!

As people mentioned in the thread for the previous video, it'd be great to have a fundamentals friday explaining how SAs work and what each stage of the front-end does to the input signal to do the business. Actually, a video series on analog front-ends in general would be awesome-- you've already done one on multimeters; 'scopes and SAs to come?

[AF6LJ]

By the way Dave;
The common way to pick off signals for sampling is to use a directional coupler.
Their PCB implimentation of a directional coupler is tried and true. They look like 10DB couplers. IF I had to put a number on it.
Good job on the teardown.
One of your very best RF teardowns yet...


 
 

[Someone]

Daylight = 500THz

Terahertz

[AF6LJ]

Daylight = 500THz

Terahertz

Yup

[tszaboo]

I really liked the added effort for this RF equipment. It was too often: "Look RF voodoo and strange looking antennas!". OK, not really, but almost. This was very good.

[bitseeker]

Best teardown, Dave. The extra time investment clearly shows and the level of detail is appreciated.

[richnormand]

Really appreciated this tear-down and analysis with the diagrams and the board walkthrough of the various stages.
Cheers and thanks Dave.

PS: considering the comments on the last video comparing the two SP and the fact you did not have it for long I wonder what are the reactions from Siglent on the level of design detail given as well as just opening it?

[tautech]

Really appreciated this tear-down and analysis with the diagrams and the board walkthrough of the various stages.
Cheers and thanks Dave.

PS: considering the comments on the last video comparing the two SP and the fact you did not have it for long I wonder what are the reactions from Siglent on the level of design detail given as well as just opening it?

Dave has a reputation for teardowns so any manufacturer shouldn't be at all surprised that he did.

Don't want a teardown and exposure of your design to the world ..... then don't send it to Dave, simple.

[nctnico]

From what I have seen from the teardown there is nothing Siglent should be ashamed off. However the hardware engineers at Siglent should go after their co-workers in the software department with pitch forks more often to make them do a good job as well 

[richnormand]

From what I have seen from the teardown there is nothing Siglent should be ashamed off.

Fully agree. At this price point this is excellent work.

[EEVblog]

PS: considering the comments on the last video comparing the two SP and the fact you did not have it for long I wonder what are the reactions from Siglent on the level of design detail given as well as just opening it?

I just got an email from Siglent, they loved the teardown.

[bitseeker]

That excellent, Dave. Now, how can we use that knowledge to work for us?

[AF6LJ]

PS: considering the comments on the last video comparing the two SP and the fact you did not have it for long I wonder what are the reactions from Siglent on the level of design detail given as well as just opening it?

I just got an email from Siglent, they loved the teardown.

Good Deal 

[coppice]

PS: considering the comments on the last video comparing the two SP and the fact you did not have it for long I wonder what are the reactions from Siglent on the level of design detail given as well as just opening it?

I just got an email from Siglent, they loved the teardown.

They might have liked it less if you'd done a parallel teardown of the 3.2GHz model, and found the hardware was identical.

[tautech]

PS: considering the comments on the last video comparing the two SP and the fact you did not have it for long I wonder what are the reactions from Siglent on the level of design detail given as well as just opening it?

I just got an email from Siglent, they loved the teardown.

They might have liked it less if you'd done a parallel teardown of the 3.2GHz model, and found the hardware was identical.

We already know it is.

[coppice]

PS: considering the comments on the last video comparing the two SP and the fact you did not have it for long I wonder what are the reactions from Siglent on the level of design detail given as well as just opening it?

I just got an email from Siglent, they loved the teardown.

They might have liked it less if you'd done a parallel teardown of the 3.2GHz model, and found the hardware was identical.

We already know it is.

Where is the information which confirms this? All I have seen it that the feature packs are enabled by a key.

[Smokey]

Paging Shahriar... Shahriar... Please report to the EEVBlog for comments on RF gear....
https://www.eevblog.com/forum/profile/?u=8240


I wonder if Dave violated their video blogger to video blogger agreement where Dave won't do RF gear and Shahriar won't take apart stuff people send him in the mail.

[tautech]

PS: considering the comments on the last video comparing the two SP and the fact you did not have it for long I wonder what are the reactions from Siglent on the level of design detail given as well as just opening it?

I just got an email from Siglent, they loved the teardown.

They might have liked it less if you'd done a parallel teardown of the 3.2GHz model, and found the hardware was identical.

We already know it is.

Where is the information which confirms this? All I have seen it that the feature packs are enabled by a key.

Not officially, but read this thread and make up your own mind:
https://www.eevblog.com/forum/testgear/hack-of-sigllent-spectrum-analyzer-ssa3021x/


BTW already linked in reply #7

[EEVblog]

I wonder if Dave violated their video blogger to video blogger agreement where Dave won't do RF gear

This is only my 3rd spectrum analyser teardown 

[rf-loop]

PS: considering the comments on the last video comparing the two SP and the fact you did not have it for long I wonder what are the reactions from Siglent on the level of design detail given as well as just opening it?

I just got an email from Siglent, they loved the teardown.

They might have liked it less if you'd done a parallel teardown of the 3.2GHz model, and found the hardware was identical.

We already know it is.

Where is the information which confirms this? All I have seen it that the feature packs are enabled by a key.

Is it better to read this EEVblog forum bit more deeply.

Independent of this and without deep knowledge how different manufacturers optimise they production and models.
If look Dave's photographs from boards it do not need very deep knowledge to know that if make these, example qty 100 and run statistics with several test parameters you find that every single unit is different, more or less. Analg RF world is - fun. (Btw, in images I can not fid signs about laser or other method individually trimmed filters or other things etc...)
In finally you may find (I do not say Siglent find) that some units are wise to sell as 1.5G models, some may be ok for 2G and some may be ok for 3G. It depends how we set our parameters limit values. Then configure (digitally) these units to different models. If I'm manufacturer I do this.  Also Analog Devices do even with simple chips labelling them to 40MHz or 100MHz.... and so on..

[nowlan]

Re: RF sections;
All those hititte chips made me wonder if it was a reference design, or outsourced.

[rf-loop]

Re: RF sections;
All those hititte chips made me wonder if it was a reference design, or outsourced.

Really? Do you really believe that company like Siglent do not have enough well educated and experienced RF engineers for design this kind of simply front end where is nothing special. Siglent is also ODM so if you want something like this you can buy this design from Siglent.
Just simple cheap front end without any kind of special "High-End" or "State of art" level things. Why it need outsource design with high risk. All these kind of things need tightly keep inside house. Also they have experince with they older SA models 1010 and 3030. If development team do  years work and this is result I do not wonder at all this result, perhaps I hope result is better... but, this is much more than crap, so public "raise the cup and give a round of applause"  to Siglent SSA product department development team. In this price result is not bad.

[classical]

Thank you for this great teardown! The Image with block diagram overlay was really well done and extremely helpful. Threre is a lot of work in it. I do appreciate it.

[coppice]

Where is the information which confirms this? All I have seen it that the feature packs are enabled by a key.

Is it better to read this EEVblog forum bit more deeply.

Independent of this and without deep knowledge how different manufacturers optimise they production and models.
If look Dave's photographs from boards it do not need very deep knowledge to know that if make these, example qty 100 and run statistics with several test parameters you find that every single unit is different, more or less. Analg RF world is - fun. (Btw, in images I can not fid signs about laser or other method individually trimmed filters or other things etc...)
In finally you may find (I do not say Siglent find) that some units are wise to sell as 1.5G models, some may be ok for 2G and some may be ok for 3G. It depends how we set our parameters limit values. Then configure (digitally) these units to different models. If I'm manufacturer I do this.  Also Analog Devices do even with simple chips labelling them to 40MHz or 100MHz.... and so on..

So, what you are trying to say is you have no idea, right? It might be the same hardware. It might not. It almost certainly fairly similar, but maybe they use a higher spec version of some key parts.

[xrunner]

I'd probably consider it if I hadn't already bought a DSA815-TG last year. Something new and better always comes out, that's the way it goes. But for my hobby uses the DSA-815 is perfectly fine.

[bitwelder]

This teardown was really good, thesignalpath-good
My vote:  O0 O0 O0 O0 (4 Shariars)

[nctnico]

So, what you are trying to say is you have no idea, right? It might be the same hardware. It might not. It almost certainly fairly similar, but maybe they use a higher spec version of some key parts.

I don't believe they leave producing the 2 or 3GHz model to chance. With modern production processes you can build every unit to precise specifications. If you have different versions you'll also need different microstrip filters so you'll also need two different board designs. Taking the chance with lower spec components isn't going to cut it either. For example: there are many RF switches on the board. If some of them operate to >3GHz and some not the chance of creating a 3GHz capable SA is nearly 0. Also if demand is greater for the 3GHz model you will be left with a whole bunch of boards which you can't sell. So either all units are the same and work up to 3.x GHz or they have specific designs for the 2.xGHz and 3.xGHz models.

[albert22]

 
VERY good video.
Thanks

[TurboTom]

There are quite a few amazing details about the Siglent SSA3021X and the Rigol DSA815 that they have in common. The design of the RF path right to the first mixer is almost identical (a few components are different but function-wise it's exactly the same). Quite funny is the use of the first mixer "in reverse", i.e. RF is applied at the "IF" output of the mixers (HMC488 in case of the Siglent, probably HMC213 in the Rigol -- the surface is lasered blank here) while the mixing product is taken from the RF input. Since the mixers are passive devices (double-balanced diode bridges), this is possible and permits input frequencies (thoretically) down to DC. Both LO and RF "inputs" of the mixers are transformer-coupled and hence have a lower frequency limit. What makes me wonder though if the existing hardware of the 2.1G siglent spectrum analyzer is really capable of 3.2G is the fact that the mixer's IF range is actually specified to 2.5G only. This may be just a limit imposed by the way Hittite has tested this device but it gives some reason for doubts. The HMC213 that I assume to be used in the DSA815 is specified to 1.5G IF so this matches the operational range of this analyzer.

Well, that's just food for thought.

Another interesting detail is that Siglent used a one-wire unique ID ROM chip (U48, DS2401) on their application processor. Could it be possible that Siglent got batches of ID ranges that refer to a certain frequency version of the analyzer? Or maybe have odd IDs on 2.1G machines and even on 3.2G or vice versa? Or is it just there to be able to identify every single analyzer in the field and prevent it from being tampered with? I would really like to know if the machine still runs without that chip...

I guess the Siglent SA will become a new "hacker's paradise"... 

Oh - just noticed: In the lower right corner of the control board, next to the buzzer, there's an unpopulated footprint of a four-pin header labled "uart". I guess that's the place too look first... Maybe it's easier to get into that thing than expected!

Cheers,
Thomas

[nctnico]

Still what makes me wonder is that many Hittite chips used in this SA are declared obsolete.

[tautech]

Still what makes me wonder is that many Hittite chips used in this SA are declared obsolete.

Really? How?

As rf-loop has alluded to, development will have taken some years and in that time ANY IC's production status can change for whatever reason. Part obsolescence can even result from a change to ROHS compliance or better spec'ed chip in the same package but who knows.
Have you never encountered this (of course you f***ing have) or are you just simply trolling the new Siglent SA threads? 

[Smokey]

How long did it take to put this video together.  This one was definitely not "off the cuff".  Lots of prep it looks like.
Did you get some help with this one?

[tautech]

How long did it take to put this video together.  This one was definitely not "off the cuff".  Lots of prep it looks like.

Do the maths, this is when Dave first announced he had it:
https://www.eevblog.com/forum/testgear/siglent-ssa3000x-spectrum-analyzers/msg962513/#msg962513

Did you get some help with this one?

If you study the following posts linked above you'll see some forum members gave advice.

[PartialDischarge]

Still what makes me wonder is that many Hittite chips used in this SA are declared obsolete.

Have you never encountered this (of course you f***ing have) or are you just simply trolling the new Siglent SA threads? 

I dont think hes trolling but its nctnico and his crusade against rigol and siglent products.

[EEVblog]

How long did it take to put this video together.  This one was definitely not "off the cuff".  Lots of prep it looks like.

This one took about two days by the time I dicked around with all sorts of little stuff. Started as usual on Tuesday and wanted to release Tuesday, but had to do some family stuff and I also decided to do the new overlay thing during it. Wasn't finished and uploaded until Wed night.
Almost all my videos I shoot and edit in a day, so you could say roughly twice as long as a normal teardown.
"prep" isn't the right word, there was still essentially zero "prep", I pressed record and started yapping. Then I decided I'd try to use the photos I had and do all the teardown commentary during the edit, rather than in-camera mic work as I normally almost always do.
So the only "prep" was figuring out the blocks and drawing them all, looking up datasheets etc. Once I had all that it was my usual press record and start yapping technique. There was no script or planning what I was going to say, as with all my videos, so IMO, still "off-the-cuff".

Did you get some help with this one?

Not really. But I was talking about it on the forum and posted an early 1st guess block overlay before I wet to bed and realised I had a lot of it wrong, as was confirmed by several others. So I took another shot at it the next day going through it more carefully.

[EEVblog]

I don't believe they leave producing the 2 or 3GHz model to chance. With modern production processes you can build every unit to precise specifications. If you have different versions you'll also need different microstrip filters so you'll also need two different board designs. Taking the chance with lower spec components isn't going to cut it either. For example: there are many RF switches on the board. If some of them operate to >3GHz and some not the chance of creating a 3GHz capable SA is nearly 0. Also if demand is greater for the 3GHz model you will be left with a whole bunch of boards which you can't sell. So either all units are the same and work up to 3.x GHz or they have specific designs for the 2.xGHz and 3.xGHz models.

Yep. I'd bet my bottom dollar all units are fully speced to 3.2GHz.

[EEVblog]

Really? Do you really believe that company like Siglent do not have enough well educated and experienced RF engineers for design this kind of simply front end where is nothing special.

It makes we wonder if they have poached a designer or two from one of the biggies?, or maybe even Rigol? As I have heard that Rigol have been losing people...

Not that I'm suggesting Siglent couldn't have done this with their existing people, but to me it just seems like it had a bit more RF spit'n'polish then you'd expect from their first shot at a spectrum analyser.

[tautech]

Really? Do you really believe that company like Siglent do not have enough well educated and experienced RF engineers for design this kind of simply front end where is nothing special.

It makes we wonder if they have poached a designer or two from one of the biggies?, or maybe even Rigol? As I have heard that Rigol have been losing people...

Not that I'm suggesting Siglent couldn't have done this with their existing people, but to me it just seems like it had a bit more RF spit'n'polish then you'd expect from their first shot at a spectrum analyser.

And again we remind readers this is not Siglents first "shot" at a spectrum analyser although they haven't had one to offer in their lineup for the last year or two.
https://www.eevblog.com/forum/testgear/siglent-ssa3000x-spectrum-analyzers/msg964143/#msg964143

[Brumby]

Yes, Dave.  A good job indeed.

I found this to be not only good as a tear-down, but informative as to the nature of the beast and actual implementation of various functional blocks.

I know it's extra work, but I think it's something I would like to see more of.

[rf-loop]

, but to me it just seems like it had a bit more RF spit'n'polish then you'd expect from their first shot at a spectrum analyser.

What "first shot"?

SSA1010 and SSA3030 are predecessors for SSA3000X

[EEVblog]

, but to me it just seems like it had a bit more RF spit'n'polish then you'd expect from their first shot at a spectrum analyser.

What "first shot"?
SSA1010 and SSA3030 are predecessors for SSA3000X

Yep, I stand corrected.
Did those things ever really sell?
The SSA1010 shows zero hits on Youtube.
The SSA3030 gets two hits on one channel.
Zero hits for reviews anywhere

[coppice]

, but to me it just seems like it had a bit more RF spit'n'polish then you'd expect from their first shot at a spectrum analyser.

What "first shot"?
SSA1010 and SSA3030 are predecessors for SSA3000X

Yep, I stand corrected.
Did those things ever really sell?
The SSA1010 shows zero hits on Youtube.
The SSA3030 gets two hits on one channel.
Zero hits for reviews anywhere

Maybe they never exported them, but they've been easy to find in China. See https://world.taobao.com/search/search.htm?_ksTS=1466844976484_55&spm=a21bp.7806943.20151106.1&search_type=0&_input_charset=utf-8&navigator=all&json=on&q=SSA1010&cna=L%2Fc7DLx%2FFQcCAXvL8GZogKJ6&callback=__jsonp_cb&abtest=_AB-LR517-LR854-LR895-PR517-PR854-PV895_2462 and https://world.taobao.com/search/search.htm?_ksTS=1466844934601_41&spm=a21bp.7806943.20151106.1&_input_charset=utf-8&navigator=all&json=on&q=SSA3030&cna=L%2Fc7DLx%2FFQcCAXvL8GZogKJ6&callback=__jsonp_cb&abtest=_AB-LR517-LR854-LR895-PR517-PR854-PV895_2462

[tautech]

, but to me it just seems like it had a bit more RF spit'n'polish then you'd expect from their first shot at a spectrum analyser.

What "first shot"?
SSA1010 and SSA3030 are predecessors for SSA3000X

Yep, I stand corrected.
Did those things ever really sell?
The SSA1010 shows zero hits on Youtube.
The SSA3030 gets two hits on one channel.
Zero hits for reviews anywhere

Maybe they never exported them, but they've been easy to find in China.

They did, a 2013 quote I've just looked up had them retailing at over $5k for the SSA3030.
Needless to say I supplied my customer with something else as he didn't need 3GHz.

[rf-loop]

Did those things ever really sell?
The SSA1010 shows zero hits on Youtube.

There is whole world also outside of youtube.
Also someone selling used SSA3030 http://maxanalyzer.com/products.php?product=Siglent-SSA3030-9KHz-to-3-GHz-Spectrum-Analyzer  (btw, look this equipment rear panel, RS232, LAN, GPIB, VGA, USB, )

eBay have still new ones (and expensive),

Also it is good to realize that there is big domestic markets inside mainland China.  There is lot of things what no one sell in western markets - some company may think that it is perhaps even wise to stay out from western markets.  Western markets do not make all happy. Try find example ZLG oscilloscopes from western markets.
Living there do not even need know that there is some periphery world outside China.

[Garan]

Great job on the teardown Dave! I liked how you overlaid the block diagram on the circuit board photo.

A few comments:
[7:00] The unloaded IC, U15, to the left of the Spartan 6 FPGA is probably a for second bank of DDR memory that isn't shared with the host processor. The additional memory is to provide data buffering for the DSP operations in the Spartan 6. This would enable a number of high performance features: real-time spectrum analyzer operation, digital modulation signal analysis, etc.

[23:17] The input switch U1 switches the RF input to the 50 ohm dummy load when there is an overpower condition. The unloaded diodes D31-D34 could be the detector or they could be limiters with an adjustable threshold. Perhaps these diodes were not loaded because they degraded input IP3 and/or caused a rolloff in the frequency response.

The following switch U2 has a trace on pin 1. It is likely for a calibration signal as G0ZHU mentioned. By providing a reference signal, the signal path amplitude drift vs temperature can be compensated to some degree.

The preamp used non-reflective SPDT switches to increase isolation and terminate any stray leakage into the thru path, which feeds back into the input of the preamp, causing oscillation.

[33:07] Your block diagram and photos provided enough information to deduce the frequency plan without doing any measurements.

[44:35] The ADC uses bandpass sampling to sample the IF. The Nyquist criterion requires that the signal bandwidth has to be less than 20 MHz. It doesn't mean the IF frequency has to be less than 20 MHz. With a bandpass filter around the IF signal, aliasing can be used to downconvert the signal in the digital domain.
70 MHz is the 3rd IF. Since the ADC samples at 40MSPS, 70 MHz will alias to 10 MHz, which in in the center of the first Nyquist zone.
810 MHz is the 2nd IF as you already discovered. This allows determining the 2nd and 3rd LO frequencies:
3rd LO = 880 MHz (810 MHz + 70 MHz)
2nd LO = 3520 MHz (880 MHz x 4)
The 1st IF is thus 4330 MHz. (810 + 3520 MHz)
1st LO range at mixer port: 4330 MHz to 6430 MHz (to 7530 MHz for 3.2 GHz model)

The 2.1 GHz and 3.2 GHz model share the same RF design as one would not use such a high 1st IF frequency and use a frequency doubler in the 1st LO path if the design were optimized only for 2.1 GHz.

[15:47] The TG offset synthesizer IC is a HMC835 (33-4100 MHz) according to the part markings, not a HMC832. It is set to 4.33 GHz, the same frequency as the 1st IF.

[tautech]

Welcome to the forum Garan.

Sounds like you've had a good study of them, from personal experience maybe ? 

[Garan]

Welcome to the forum Garan.

Sounds like you've had a good study of them, from personal experience maybe ? 

Why, yes....I have lots of experience in spectrum analyzer design and teardowns.