SIGLENT SSA3032X input related spurious

[RFDUK]

Our second faulty SSA3032X returned under warranty to Germany today.

Now 9 weeks from initial purchase, but we like this instrument, so waiting on a replacement ... third time lucky? Hope so.

[Performa01]

Our second faulty SSA3032X returned under warranty to Germany today.

Now 9 weeks from initial purchase, but we like this instrument, so waiting on a replacement ... third time lucky? Hope so.

Fingers crossed!

In a situation like this, at least by now, a good distributor should test any new instrument for this specific flaw before sending it to you (or anyone else for that matter). Have you negotiated something like this?

[RFDUK]

Our second faulty SSA3032X returned under warranty to Germany today.

Now 9 weeks from initial purchase, but we like this instrument, so waiting on a replacement ... third time lucky? Hope so.

Fingers crossed!

In a situation like this, at least by now, a good distributor should test any new instrument for this specific flaw before sending it to you (or anyone else for that matter). Have you negotiated something like this?

I'm working on the assumption that Batronix forwarded the 2nd instrument to me on the basis the 1st was an isolated failure. The 1st replacement unit I assume was not checked in Europe, it looked sealed from the factory.

When I received a second instrument with the same fault, I suggested to Batronix that Siglent had a problem in production. That was on Dec 28th.

What I've asked for on the 28th Dec ... Batronix to establish a dialogue with Siglent to confirm they know of the issue and give us confidence they can deliver an instrument in specification in a reasonable timeframe? They agreed to do this.

I'm optimistic if Batronix live up to their customer service claims we are now on a fast track to progress. I'll let you know.

[rf-loop]

Here is some things about  60MHz below carrier spur.

Here is case when carier moves slowly up.
If slowly move carrier frequency up this spur move down until it hop up 2.5MHz. After jump up it again continue down until it have moved 1.25MHz  and again hop up (if carrier continue moving up).
It moves like zig-zag.  Moves slowly down same amount what carrier moves up. Until it reach limit where it need hop up.

These movements are roughly (not measured enough accurately) 1.25MHz down following carrier but opposite direction, and  then 2.5MHz hop and again start moving down following carrier but opposite direction.

I have not opened my SSA for look what happen if take bottom and top aluminium block out and try do some GND contackt-seal trick for reduce possible leak between blocks.  Area between IF3 and 810MHz saw filter looks "terrible" if think high grade isolation but I have not opened it for look aluminium wall between these blocks. Also between these blocks GND is cutted for air gap but....  is it possible also that this aluminium block do some "fun" things and then every SA is different. I dso not like idea that there is just plain aluminium surface "contacting" (or not) with PCB GND surface "randomly" here and there and mostly just near screws (if they are tightened right. But how can tighten chinese soft screws... 
In professional equipments there is mostly also good flexible rf-seal in this kind of places between aluminium blocks and PCB.

Attached images may clarify how this spur frequency acts relative to carrier frequency.
Note that for this 10MHz span 60MHz image I have rised 120MHz carrier level so that spur level rise for better visibility.
Other image show same carrier but now with -30dBm mixer level and there can see spur is under -70dBc.

[Orange]

Here is some things about  60MHz below carrier spur.

Here is case when carier moves slowly up.
If slowly move carrier frequency up this spur move down until it hop up 2.5MHz. After jump up it again continue down until it have moved 1.25MHz  and again hop up (if carrier continue moving up).
It moves like zig-zag.  Moves slowly down same amount what carrier moves up. Until it reach limit where it need hop up.

These movements are roughly (not measured enough accurately) 1.25MHz down following carrier but opposite direction, and  then 2.5MHz hop and again start moving down following carrier but opposite direction.

I have not opened my SSA for look what happen if take bottom and top aluminium block out and try do some GND contackt-seal trick for reduce possible leak between blocks.  Area between IF3 and 810MHz saw filter looks "terrible" if think high grade isolation but I have not opened it for look aluminium wall between these blocks. Also between these blocks GND is cutted for air gap but....  is it possible also that this aluminium block do some "fun" things and then every SA is different. I dso not like idea that there is just plain aluminium surface "contacting" (or not) with PCB GND surface "randomly" here and there and mostly just near screws (if they are tightened right. But how can tighten chinese soft screws... 
In professional equipments there is mostly also good flexible rf-seal in this kind of places between aluminium blocks and PCB.

Attached images may clarify how this spur frequency acts relative to carrier frequency.
Note that for this 10MHz span 60MHz image I have rised 120MHz carrier level so that spur level rise for better visibility.
Other image show same carrier but now with -30dBm mixer level and there can see spur is under -70dBc.

I think the frequency hopping is due to the fact that the analyzer does the analysis based on an FFT. If FFT is involved the LO probably is set at fixed intervals.
You can try and deselect FFT, and see if the frequency hopping still occurs...

[G0HZU]

I really don't think anyone should take their analyser apart yet but if anyone is brave enough to take their analyser apart then please be VERY careful not to flex the big PCB because you could easily cause cracks inside the many ceramic SMD caps on the main board. Even a tiny amount of flex/stress could damage lots of capacitors. Not good!

We don't know what is causing the variability in rejection. It could be they changed something on the PCB or it could be a metalwork/isolation issue. It could just be bad luck with how the stopbands interact when two SAW filters are connected in series like this. Probably best to wait until RFDUK gets the third analyser to try out.

The zigzag nature of the spurious as described by TurboTom in post #13 and by rf-loop is exactly what I'd expect to see because the analyser is in FFT mode for all the screenshots I have seen so far.

[rf-loop]

I think the frequency hopping is due to the fact that the analyzer does the analysis based on an FFT. If FFT is involved the LO probably is set at fixed intervals.
You can try and deselect FFT, and see if the frequency hopping still occurs...

I did not wonder at all this hopping. Naturally and of course I know it. But I tried demonstrate this to some other peoples who are not familiar with this equipment but interested how example this spur act. Also tiny bit more information for purpose if some people want think what are possible sources for this mixing product (spur). (here is also peoples who do not have this equipment but still want think this case with very deep knowledge about this kind of equipments )

[rf-loop]

I really don't think anyone should take their analyser apart yet but if anyone is brave enough to take their analyser apart then please be VERY careful not to flex the big PCB because you could easily cause cracks inside the many ceramic SMD caps on the main board. Even a tiny amount of flex/stress could damage lots of capacitors. Not good!

We don't know what is causing the variability in rejection. It could be they changed something on the PCB or it could be a metalwork/isolation issue. It could just be bad luck with how the stopbands interact when two SAW filters are connected in series like this. Probably best to wait until RFDUK gets the third analyser to try out.

The zigzag nature of the spurious as described by TurboTom in post #13 and by rf-loop is exactly what I'd expect to see because the analyser is in FFT mode for all the screenshots I have seen so far.

Yes, this warning is extremely important. We who are familiar with this kind of things know how to handle these, even more fragile things,  but... there is lot of peoples who do not know....

Idea for this tiny test was just because previously I understand you think this (after TurboTom msg) and I want support your thinking  so that now you know, without this equipment in your hand, that it is as you think.

[Performa01]

I think the frequency hopping is due to the fact that the analyzer does the analysis based on an FFT. If FFT is involved the LO probably is set at fixed intervals.

I think this is why rf-loop added the text “Because this is frequency hopping sweep spectrum analyzer…” to his 2nd image


-----------------------------------------------------------------------
Btw. I have asked my friend for some more tests and it’s now confirmed that the spur is related to the input signal in a linear way. He has tested this for mixer levels of -40dBm - -20dBm.

So it actually would be some sort of image response and I think the concerns voiced by rf-loop, regarding poor contact (or no contact at all) between aluminum and the PCB traces sound very plausible to me. Trying to contact aluminum usually is a true nightmare…

[RFDUK]

I think the frequency hopping is due to the fact that the analyzer does the analysis based on an FFT. If FFT is involved the LO probably is set at fixed intervals.

I think this is why rf-loop added the text “Because this is frequency hopping sweep spectrum analyzer…” to his 2nd image



-----------------------------------------------------------------------
Btw. I have asked my friend for some more tests and it’s now confirmed that the spur is related to the input signal in a linear way. He has tested this for mixer levels of -40dBm - -20dBm.

So it actually would be some sort of image response and I think the concerns voiced by rf-loop, regarding poor contact (or no contact at all) between aluminum and the PCB traces sound very plausible to me. Trying to contact aluminum usually is a true nightmare…

Yes, very much agree with both additional points ... Btw ...

The unit to unit range of performance on this spur is typically -70dBC for an older instrument, -57dBC on my first and then -50dBC on my second. Poor aluminium screen contact to the PCB is a strong favourite for this wide range.

[TurboTom]

I consider the cover issue only valid if Siglent changed from the machined parts to completely die-cast configuration without any post-processing. CNC machining is so accurate nowadays that even if they only machine the surface mating with the PCB, shielding performance should be good enough that there wouldn't be the inaccuracies present that we are observing on the recent instruments. If Siglent doesn't employ any post-processing of the cast components, okay. But since machining of the covers only requires any dated three-axis-mill, I don't think they will risk the problems induced by using bare, die-cast shields. Anyway, I may be proven wrong - which would be a relly stupid move on behalf of the manufacturer...  . I would be curious to see a recent SSA being torn down for ananlysis...though I wouldn't recommend anyone to do this.

Cheers,
Thomas

[RFDUK]

I consider the cover issue only valid if Siglent changed from the machined parts to completely die-cast configuration without any post-processing. CNC machining is so accurate nowadays that even if they only machine the surface mating with the PCB, shielding performance should be good enough that there wouldn't be the inaccuracies present that we are observing on the recent instruments. If Siglent doesn't employ any post-processing of the cast components, okay. But since machining of the covers only requires any dated three-axis-mill, I don't think they will risk the problems induced by using bare, die-cast shields. Anyway, I may be proven wrong - which would be a relly stupid move on behalf of the manufacturer...  . I would be curious to see a recent SSA being torn down for ananlysis...though I wouldn't recommend anyone to do this.

Cheers,
Thomas

That got me thinking on the aluminium contact areas to the PCB in the region of the 3rd mixer and input SAW filters. Here are a couple of picts from Dave's teardown.

Can anyone explain the patchy layer that only partially covers the screen cover contact areas on the SAW filter side of the PCB?

Another variabilty could be the solder resist integrity. PCB manufacturer 'keep out' settings sometimes deliver not as intended resist dimensions. Excessive resist dimensions could perhaps get under the aluminium contact area?

Are the groundplane void slots bridged by the aluminium screen wall? There is a witness line running into one of the slots from the cover.

[nctnico]

I strongly doubt changes in casting or milling will have a large effect on how the PCB (doesn't) make contact with the case. After all the aluminium will be covered with an insulating oxide skin anyway. A change in components is much more likely.

[RFDUK]

Interesting observation. The milled housings would benefit from an Alocrom coating or similar.

[G0HZU]

I would expect that the aluminium would be coated after machining but it should be a conductive coating not insulating. I seriously doubt they would use a coating that insulated the aluminium. At work we have been making RF PCBs like this for decades and have tried various types of screening. All our machined aluminium screens are coated and the coating is conductive. It's hard to tell they are coated with a casual glance.

How well the screens work (or don't work) depends a lot on the type of RF circuits they are covering so it isn't always obvious how much grief the screens can cause if something is lacking. There are also lots of things that can cause variation on screening performance when making basic screens like this. Often its possible to examine a dozen identical screens and spot flaws or differences with the naked eye. Sometimes the edge deburring gives variable results and other times there can be subtle issues with the milled faces. Other times, contaminants can get on the PCB or the metalwork and these can be hard to spot unless a magnifier is used. Other times there simply aren't enough screws or the PCB surface itself isn't as flat as hoped. Sometimes, RF absorber is needed to be stuck inside the lid of the milled pockets. For our high performance products we now use a squashy EMC 'filler' that sits in a groove in the machined faces. This works a lot better than trying to make a reliable and repeatable seal using just flat machined faces against a PCB.

Whatever the cause for the poor image rejection, I think it is affecting all of the analysers that I've seen in this thread to some degree because I doubt that the designer would have wanted 'only' -70dBc rejection here. We often use cascaded SAW IF filters at work but we generally put active or passive parts in between them (to help prevent passband and stopband issues) and try and have plenty of screws holding the metalwork together at vital locations like this. Even -70dBc image rejection here is not that good despite the low cost of this analyser.

[nctnico]

True but all those extra measures mean extra cost for a mass produced low cost spectrum analyser. If it where my design I'd try very hard to not make it depend on mechanical stuff being done exactly right. Also 15 dB difference between the specification and actual circuit seems like a lot to me for a problem with shielding. I think pictures of the inside of a newer model would shine a light on what may be causing the problem.

[G0HZU]

Yes, it could be they changed the SAW filters or maybe changed something about the PCB or the metalwork or maybe the design needs some RF absorber. Usually, if there is a metalwork issue then pressing or squeezing the lid (or tweaking with the screws) will change the level of the unwanted spurious signal a bit.

[TurboTom]

Yeah, maybe Siglent just fired the guy with the Popeye-like forearms who used to tighten the screws on the previous machines for eating too much spinach and replaced him with a tiny woman... 

Whatsoever, I've got to agree with nctnico that only a shielding problem without any additional change in hardware would not be very likely to have so much effect to change the image frequency rejection by 20dB on some machines. But I'm quite certain that here on the forum, we will find out the culprit for the problems sooner or later.

Cheers,
Thomas

[G0HZU]

Whatsoever, I've got to agree with nctnico that only a shielding problem without any additional change in hardware would not be very likely to have so much effect to change the image frequency rejection by 20dB on some machines.

Doesn't that depend on what makes up the final (image) signal you see at the input to the third mixer? I don't know what your background/experience is but usually a metalwork issue can cause summing/cancelling effects depending on the phase and amplitude of whatever signal paths can conduct or leak to the input of the third mixer. If two signal paths arrive here at the same amplitude (-60dBc?) but unknown phase, then you could get >30dB cancellation or you might get a 6dB summing effect depending on phase. So >36dB variation in such a case? Tweaking a screw slightly can make a big difference in this scenario if it spoils a cancellation effect for example..

I can't tell for certain of the metalwork is OK. Where I work we would have to look at the drawings and get our hands on the hardware and do some tests to be as certain as you. Even then, there could be problems with yield when tested across (say) 100 units. But you seem confident the metalwork is OK? It might help if you gave something about your background and experience wrt RF converter design and testing?

[rf-loop]

It looks like many peoples have more like audio experience than rf.

But then, just like G0HZU told there really can be many reasons including this aluminium block. Also I do not want release any part from list of suspects.

If we get true random problem distribution and we find older and new units randomly have some units what have clearly more bad this -60MHzc spurious then situation is different. Now we have only very small amount of data. How many observations we have! With this data we can not even tell sure that all older units are "good" and more new units "bad". Just this we can not do exept if we jump to world of  trumpths instead of truths.
Only we know that  some devices have been detected with quite high input related spurs, out of specifications. And these these very few observations have been with these more new manufactured units. But how many older units we have real test data. Only a few.

But also we have at this time no older unit what show this bad. This counter is still zero.
Also we do not know if there is good "new"  units. 

We can now suspect that perhaps there is something changed and perhaps more new units have this high spur. But at same time we do not know if there is also somewhere some thing better if Siglent have changed some things. Some times these kind of things  go sdo that make one thing better other think may slip more bad.
Also it looks like same time when this individual named spur is rised also these units have rised noise floor!!  Perhaps need also take it into thinking. 

But even with this small amount of observations I'm very interested to see what are internal all changes between older and newer units!

Of course when equipment is clearly out of specifications thisa must not acvcept. But inside specifications we need accept all variations. What ever variations.

We do not even know at this time what signals and where produce this spur. So we can not say if it is more or less like signals shielding problem between parts of circuit or some components characteristic problem if they have changed some component type or just manufacturer.
But still commonly can say that shielding is much more complex and difficult to predict than just digital-audio-LF peoples think mostly.
Many times it need lot of experimental work. Some times it still is like magic because it is so complex that it is more easy to think magic than math. The less you know the easier things seem to be.



Also some times it is good to reset and take ground under foots and also reset the scale of thought.
Lets look somehow comparable (mostly more bad) Keysight N9320B RF Spectrum Analyzer.
If we look input related spurious. (I can ask if I'm a little nasty, did Keysight designer want this high level of spurius and why)

Keysight N9320B Spurious response specification:
Input related spurious < -60 dBc | -30 dBm signal at input mixer.


About Siglent. We need more observations. We need explanation what is this. Is it some component (also Al block and PCB is component) change to bad direction and how this is solved back to ok (or better)  or if these single units have just been failed units. If they are just defect units then explanation how these bad units can jump over factory quality control to factory out gate. (I do not accept transport damage explanation with this kind of failure. Nearly sure this failure have been there before it exit factory)

[Orange]

It looks like many peoples have more like audio experience than rf.

So instead of starting to insult 'peoples' Have you as a Siglent dealer, asked Siglent for a reaction on this issue ?

[rf-loop]

It looks like many peoples have more like audio experience than rf.

So instead of starting to insult 'peoples' Have you as a Siglent dealer, asked Siglent for a reaction on this issue ?

I talk here technical things what are my hobby and profession and what are interesting for me independent of if I some times sell some equipment or not. I'm much more interested about technical things than any kind of business.  Not talking "sales"man.
Also this is good place to talk things what also are sure read by Siglent.  Siglent RF designer engineer can answer here if he/she like.

[TurboTom]

I openly admit that I'm more experienced with digital, lower frequency and high power stuff than with RF/microwave circuitry but I tend to have quite a good "gut feeling" and I'm curious and brave (stupid??  ) enough to tear down my SSA for science's sake.

So please have a look at the attached photos. First of all, I found the RF PCB in my SSA to be warped round about 1mm in the centre -- probably this has no detrimental effect when it's sandwiched between the shielding plates. The fingerprints on the PCB don't originate from me but must have been there when the instrument left siglent's production line -- I handled the PCB only by the edges.
By the way, the lands of the shields completely cover the "isolation slots" as shown.

I removed the PCB, apparently it's exactly the same configuration as in Dave's teardown. And now I put some standoffs in the places where the long screws bolt the RF assembly to the chassis in order to install the RF PCB completely witout any shielding.

So please have a look at the screenshots... (full span is without input signal, just for comparison and to visualize the effect of the missing shields). I guess this rules out the (suggested lack of) shielding as the cause for the 60MHz spur one for all times.

Cheers,
Thomas

[rf-loop]

I openly admit that I'm more experienced with digital, lower frequency and high power stuff than with RF/microwave circuitry but I tend to have quite a good "gut feeling" and I'm curious and brave (stupid??  ) enough to tear down my SSA for science's sake.

So please have a look at the attached photos. First of all, I found the RF PCB in my SSA to be warped round about 1mm in the centre -- probably this has no detrimental effect when it's sandwiched between the shielding plates. The fingerprints on the PCB don't originate from me but must have been there when the instrument left siglent's production line -- I handled the PCB only by the edges.
By the way, the lands of the shields completely cover the "isolation slots" as shown.

I removed the PCB, apparently it's exactly the same configuration as in Dave's teardown. And now I put some standoffs in the places where the long screws bolt the RF assembly to the chassis in order to install the RF PCB completely witout any shielding.

So please have a look at the screenshots... (full span is without input signal, just for comparison and to visualize the effect of the missing shields). I guess this rules out the (suggested lack of) shielding as the cause for the 60MHz spur one for all times.

Cheers,
Thomas

Lot of thanks about this. So, now need in the first instance think other reasons.

What they have changed  (or is it still possible these bad examples have been somehow defected units  -- what feels very unlikely because RFDUC get one bad unit after one bad)




It was also my plan to soon tear down my SSA and do around same things after 1-3 days when I have first cleaned my work bench which had just been prepared allready for some SDS1202X-E repair-reworks.




EDIT/ADD:

As previously suspected. Main reason IS shielding/Aluminium block.
Some metal work supplier have done some tiny material surface changes in aluminium shielding block. This may cause this problem in some individual units.
Siglent will repair all units delivered back to Siglent what suffer  this issue.

It is (IMHO) also possible that there is quite simple solution for DIY repair if user have enough knowledge and skills for handling this kind of very sensitive and fragile things inside this kind of equipment,  but in mainstream: Siglent have done this manufacturing error, Siglent is responsible.

[G0HZU]

So please have a look at the screenshots... (full span is without input signal, just for comparison and to visualize the effect of the missing shields). I guess this rules out the (suggested lack of) shielding as the cause for the 60MHz spur one for all times.

Thanks for doing this. Forgive me for appearing a bit stubborn here but your published tests are interesting but not conclusive (yet)... 

By taking off the lid completely you remove one potential leakage path (via the metalwork) and replace it with an open air radiated path. So its difficult (for me at least) to draw any definite conclusions yet.

Based on this single test, what would you have concluded if the spur appeared lower or higher than before?  I'm not sure I could conclude very much in either case. I'm still being cautious here because I've been down this route many times in the past with screens.

You may have already tried this but a better test would be to fit the lid with the screws only partly tight and then wiggle and press it to see if the spur moves as the lid is pressed and the screws tightened. If it doesn't budge then I guess the metalwork can be ruled out with a fair degree of confidence. But obviously you have to be careful here if the metalwork doesn't have a locating dowel. You could easily short out and damage components if you try this.