SIGLENT SSA3032X input related spurious

[rf-loop]

What about a re-calibration? There is no way anyone can change the RF shields and still say the cal is good.

You  know or just believe?
I have repaired some of these SSA's  RF units and know something how they act. Circuit design is good and it is not like crap hobbyist build hassle what drift wildly even  if you jump on the floor. This construction is rock solid.

I do not know! I only believe and predict based my limited knowledge and some limited experience, it do not need new factory calibration table after this repair if it is done right. I keep this until some proofment with data that I'm wrong in this named case.

Of course, who ever do works inside SSA need know how to handle things in this SA. It is not at all everyman work, even when it looks quite simple. There is so many traps what may lead damage or drift or latent damage what will later become active failure (example EOS/ESD latent damages). Also board is mechanically very fragile. Also some filters on the board need care and all need keep clean. But handled right it do not drift with these parameters what are in calibration table. This machine is only up to 3GHz so it is not very difficult and critical.
Of course aluminium block dimensions must not change out from its dimensional tolerances.
If I change some mixer or example front end attenuator it need new calibration data table.

Totally other case is if equipment have valid official NIST traceable calibration certificate.
When the calibration laboratory set the seals and they are broken the calibration certificate is not valid anymore. It needs new cal certificate. (even if nothing is changed but still needs a new valid certificate.) This is done by making measurements in an approved calibration laboratory but typically not adjustments (exept if asked for example improve accuracy and paid or if some value have drifted over specs and if it can adjust. If not, then repair. If can not. Then to recycle or use without certificate for purposes what do not need cert added with warning label about situation)

So, if SA have fresh valid cal cert. It is perhaps wise to wait for this repair until the cal cert needs to be renewed. This avoids the extra cost since usually the official cal cert is not cheap.

[nctnico]

What about a re-calibration? There is no way anyone can change the RF shields and still say the cal is good.

You  know or just believe?
I have repaired some of these SSA's  RF units and know something how they act. Circuit design is good and it is not like crap hobbyist build hassle what drift wildly even  if you jump on the floor. This construction is rock solid.

I do not know! I only believe and predict based my limited knowledge and some limited experience, it do not need new factory calibration table after this repair if it is done right. I keep this until some proofment with data that I'm wrong in this named case.

Of course, who ever do works inside SSA need know how to handle things in this SA. It is not at all everyman work, even when it looks quite simple. There is so many traps what may lead damage or drift or latent damage what will later become active failure (example EOS/ESD latent damages). Also board is mechanically very fragile. Also some filters on the board need care and all need keep clean. But handled right it do not drift with these parameters what are in calibration table. This machine is only up to 3GHz so it is not very difficult and critical.
Of course aluminium block dimensions must not change out from its dimensional tolerances.
If I change some mixer or example front end attenuator it need new calibration data table.

TheSteve is right. After a repair you can't say the cal is still good without testing. The number of IFs in your statement above also implies that.

[rf-loop]

What about a re-calibration? There is no way anyone can change the RF shields and still say the cal is good.

You  know or just believe?
I have repaired some of these SSA's  RF units and know something how they act. Circuit design is good and it is not like crap hobbyist build hassle what drift wildly even  if you jump on the floor. This construction is rock solid.

I do not know! I only believe and predict based my limited knowledge and some limited experience, it do not need new factory calibration table after this repair if it is done right. I keep this until some proofment with data that I'm wrong in this named case.

Of course, who ever do works inside SSA need know how to handle things in this SA. It is not at all everyman work, even when it looks quite simple. There is so many traps what may lead damage or drift or latent damage what will later become active failure (example EOS/ESD latent damages). Also board is mechanically very fragile. Also some filters on the board need care and all need keep clean. But handled right it do not drift with these parameters what are in calibration table. This machine is only up to 3GHz so it is not very difficult and critical.
Of course aluminium block dimensions must not change out from its dimensional tolerances.
If I change some mixer or example front end attenuator it need new calibration data table.

TheSteve is right. After a repair you can't say the cal is still good without testing. The number of IFs in your statement above also implies that.

Of course normal lab practice.
Example what I have done is that before anything, some (quite lot of) important measurements and cross checks and results write down. After finished, check that same measurements give same result or changes are enough small. It is natural that all peoples do not have this possibilty to do enough reliable measurements. So, typically this is not like DIY work for Mr Everyman.

Situation is very different if equipment have valid calibration certificate. After works done this calibration certificate is waste paper. Even if nothing have changed but if seals are broken certificate is just toilet paper. Period.
Only way for this case is get new cal certificate from official calibration lab. If chek do not meet specs can not write certificate and then need first repair/run "factory calibration" what corrects changes.
But it is like that after this named repair it still is ok without need of full service calibration what produce new calibration parameters to SA internal cal table. But if it need valid NIST traceable cal certificate then it need this calibration check procedure for new certificate.  Calibration and calibration check for producing valid calibration certificate are not at all same. Some times it is hard ti find which one peoples are talking about.
Totally different case if any critical components are changed in repair. It can easy imagine of look example SA internal cal table what have quite lot of data. It can say that mostly in these kind of repairs need produce new calibration data doing full "factory calibration". Example I do not have things what is needed for this.

[rfspezi]

My SSA3021X replacement just arrived and i immediately checked if the input related spurs got better.
The replacement unit gives me -65,15 dB.
Hmm.... ok, so it's tight within the specified <-65 dBc - however Siglent leaves me with the bad feeling that the value will worthen again over time  

[GDK_ATL]

Did they return the same unit or is it a new one? I thought I read where units without the problem are showing the spur down around -72 dBc or so.

[rfspezi]

It‘s another unit with production date January 2018.

[Safar]

Hi,

One else SSA:

System info:
SW1: 1.2.8.5a
SW2: 20170512-1
SW3: 000000D1
HW: 07.03.00
Calibration: 2017-04-22

I don't have clean source so I fed it from SDG2042x (Surprisingly works at 110MHz  )

Slightly not in spec....  ~61..62dB

But I bought SSA from some seller on Ali, that even does not exist anymore (but just for 1400 USD shipped!), so I think have no any chance for change it on this case.  May be I'll check screw later.

I attach AVG 10 times also.

Edit: Add 1280MHz and 3180MHz screens. Almost the same result (62..63 dB). M2 marker just for noise floor. All markers link to AVG20 trace

[rfspezi]

I fear that the units within spec and produced lately will degrade over the next view months to end up beeing out of spec just as the units produced some time agoe?
As untreated aluminum will build up an isolating oxide layer, i expect shielding performance to degrade?
I think we should perform a spurious test regularely before warranty period ends.

[Safar]

I fear that the units within spec and produced lately will degrade over the next view months to end up beeing out of spec just as the units produced some time agoe?

May be but for now first production units better than new one.

And I saw rf-loop message that deleted now, but I answer about spur near 110 MHz on my trace - this is SDG product only. Very possible garbage from some digital part. Very unstable frequency. And it passes to 2 outputs even if it disabled. Need to find.

[rf-loop]

There is also available one solution. It can learn from Keysight who have solved this same problem.
I have never seen lot of Keysight owners who complain this. Or is it because they pay more than four times this price.
Solution can read here, page 7. Beginning of this page.
I will recommend this solution also for Siglent.

[rfspezi]

 

[Safar]

I will recommend this solution also for Siglent.

Yes, yes... I think that spec of next Siglent SSA will be at Keysight spec (Input related spurious < -60 dBc) 

[G0HZU]

If I assume that the Keysight spec is a generic spurious specification that includes harmonic spurious terms then this figure will be dominated by the 2HI spec of about +30dBm. This isn't a bad 2HI spec for a budget analyser and it's better than some very expensive R&S analysers at low frequencies. Generally, the 2HI spec will be poor at the low end of the frequency range (on a budget analyser) for various reasons and I'd expect the Siglent and the Keysight analyser to both be about the same here in terms of spurious free range for 2HI terms.

So be careful of what you are laughing at here

By contrast, the rejection performance (or lack of it) at about a 60MHz offset on the Siglent analyser is not good even at -70dBc. I still maintain that something isn't right. I did suspect the metalwork straight away but it might be only part of the problem. If I had one of these analysers I'd take it apart and try and improve the performance here.

[Performa01]

By contrast, the rejection performance (or lack of it) at about a 60MHz offset on the Siglent analyser is not good even at -70dBc. I still maintain that something isn't right. I did suspect the metalwork straight away but it might be only part of the problem. If I had one of these analysers I'd take it apart and try and improve the performance here.

You have voiced your concerns about cascading high order filters directly without any isolation between them quite a while back and this has also been my immediate thought when I saw the corresponding picture of the PCB. Interestingly, no one else has picked up on that topic so far, so I would like to recall this once again.

Even though my experience is with much lower frequencies only, I can safely state that stop band performance will suffer significantly and unwanted resonances will build up as soon as we connect resonant structures – be it matching networks or filters – with quality factors Q > 0.7 involved directly. These structures are designed for a specific characteristic impedance and will only work properly if there is a corresponding broadband termination on their inputs & outputs. The output of the 1st filter doesn’t provide a proper wideband termination for the input of the cascaded 2nd filter and the same is true for the input of the 2nd filter with regard to the output of the 1st one. Termination will be not good for the pass band and atrocious in the stop band.

Consequently, I’ve used bandpass diplexers whenever feasible – which is of course only feasible for low Q networks and we cannot mirror a high order filter shape with a simple LC structure. For high selective bandpass filters, I would consider a 6db attenuator or an active buffer, but as you’ve suggested, a simple series resistor might already help a lot in damping unwanted resonances.

[rf-loop]

By contrast, the rejection performance (or lack of it) at about a 60MHz offset on the Siglent analyser is not good even at -70dBc. I still maintain that something isn't right. I did suspect the metalwork straight away but it might be only part of the problem. If I had one of these analysers I'd take it apart and try and improve the performance here.

You have voiced your concerns about cascading high order filters directly without any isolation between them quite a while back and this has also been my immediate thought when I saw the corresponding picture of the PCB. Interestingly, no one else has picked up on that topic so far, so I would like to recall this once again.

Even though my experience is with much lower frequencies only, I can safely state that stop band performance will suffer significantly and unwanted resonances will build up as soon as we connect resonant structures – be it matching networks or filters – with quality factors Q > 0.7 involved directly. These structures are designed for a specific characteristic impedance and will only work properly if there is a corresponding broadband termination on their inputs & outputs. The output of the 1st filter doesn’t provide a proper wideband termination for the input of the cascaded 2nd filter and the same is true for the input of the 2nd filter with regard to the output of the 1st one. Termination will be not good for the pass band and atrocious in the stop band.

Consequently, I’ve used bandpass diplexers whenever feasible – which is of course only feasible for low Q networks and we cannot mirror a high order filter shape with a simple LC structure. For high selective bandpass filters, I would consider a 6db attenuator or an active buffer, but as you’ve suggested, a simple series resistor might already help a lot in damping unwanted resonances.

I an very interested about this (and other things also there inside)
But is it so that this can split to two separate roads.

How to make/mod its hardware better what it is designed to be (this is interesting)

How to return/repair units what have this problem to meet they original and designed performance or lack of  performace... how want think about. (this is - must)

[TurboTom]

If I recapitulate my experiments with my SSA3021X with the shields removed / loose and compare the observed figures with the machines that had reportedly been corrected by Siglent, I can only whole-heartedly agree with "G0HZU" that inaccurate metalwork of the shields can only be half the truth. Unfortunately, unless we'll get some high-res teardown photos of a recent production machine's RF PCB, we're only "poking around in the fog". My guess is that Siglent (had to) change(d) the circuitry/components that led to the problem. The workaroud may be improved RF damping by installing shields with RF/microwave absorbers (ferrite blocks/plates) in certain compartments.

Nobody who owns one of the affected models brave enough to tear it down and take photos of the RF PCB? The "warranty void sticker" can be removed without damage using wax paper, some warm air and patience. And with some care and common sense, there's no risk of damaging anything. I guess many members here on the forum would be grateful for that.

Cheers,
Thomas

[rf-loop]

I can only whole-heartedly agree with "G0HZU" that inaccurate metalwork of the shields can only be half the truth.

Cheers,
Thomas

But as far as I know reason is not inaccurate metal work.
(if you mean dimensions)

What I have understooded problem is material. Metal work supplier have changed/added some kind of different surface oxidation process. But as long as Siglent do not make details/explanation officially public we stay in some "fog".

Where we can see result of repaired SSA. Evidence they are repaired, where? Sorry but I doubt a bit.

But also we can of course speculate if they have changed some component, example "compatible" other manufacturer same componnet.
I well remember what they did some day with SDG1000 series awg. Compatible output amplifier was not really compatible in this special case. All looks like just compatible and pin compatible (other manufacturer "looks like compatible") but deep inside data sheet there was some small imperceptible thing...and result was disaster in this individual case.. The problem was only noticeable in some situations, but it was bad when it did.

When I have time I will teardown my SSA and make some experimentals with this aluminium block. perhaps I can somehow mimic different surface layer between PCB and Al..  perhaps not useful but it do not cost anything and then I can also strenghten my SSA input ESD protection. (btw, this is what they have also perhaps changed in Siglent for later production what was perhaps not in production some time ago. How much this affect performance in front end - I do not know before I have suitable components and then done it)

[RFDUK]

When I have time I will teardown my SSA and make some experimentals with this aluminium block. perhaps I can somehow mimic different surface layer between PCB and Al..  perhaps not useful but it do not cost anything and then I can also strenghten my SSA input ESD protection. (btw, this is what they have also perhaps changed in Siglent for later production what was perhaps not in production some time ago. How much this affect performance in front end - I do not know before I have suitable components and then done it)

Has there been an ESD weakness reported in the past that resulted in SSA hardware changes?

The 4dB noise floor difference between old and new instruments may be explained by a small fixed attenuator at the input socket introduced to improve ESD ruggedness.

My January 2018 instrument spur performance of -71dBC has remained consistent for 4 weeks now. It's kept in a fairly consistent 20 degree C environment though.

The previous comments on the -65dBC input related spurious specification and comparisons to more expensive instruments are very relevant when judging what is reasonable.
Better than -60dBC spec is reasonable for higher level inputs, but -65dBC is not too great for lower level spurs that result from poor SA design or implementation that could be avoided at quite low unit cost. This is present at any input frequency and at any input level.
-50dBC to -60dBC is a SA toy, -70dBC is looking 'reasonable'.

Better than -80dBC would be a reasonable design target without adding a great deal of unit cost.

An old 'economy' general purpose SA here of the 1990s, an HP 8591E measures >-85dBC on the final mixer image spurious.

I wonder what a low cost Rigol SA performance is on the final mixer image spurious?

[rf-loop]

The 4dB noise floor difference between old and new instruments may be explained by a small fixed attenuator at the input socket introduced to improve ESD ruggedness.

What attenuator and where?

There is not any extra attenuator inside SSA after input connector and before ESD - over voltage sensitive things. Not yesterday, not today or tomorrow as far as I know.

Input is very ESD and EOS sensitive, much more sensitive than older SA constructions where is typically mechanical input attenuator before anything. As example in HP859xE  what I have repaired/serviced many in older times. But, if user manually force attenuator to 0 (turning knob alone can not so it do not happend accidentally) and then if he do more and go to menu and turn DC block off (and this case user really need know what he is doing specially if he do noth, 0 atten and DC block off.  Then there is direct DC pathway and with this combination it can kill easy.

In Siglent, naturally there can not turn DC block off at all. (DC is forbidden to 1st GaAs switch.)

But it really need handle with care.
If bad happend, it is still quite easy to repair, so no need put to recycling of electronic waste.

[RFDUK]

A fixed 3dB input attenuator would be very good for improved ESD ruggedness on a SA input if the sensitivity degradation is acceptable.

ESD event has rise time of <1nS, so DC block not relevant here?

Very difficult to improve ESD ruggedness of sensitive GHz instrumentation.

[TurboTom]

Regarding ESD sensitivity of the SSA input, I guess it all depends on the particular RF switch MMIC that Siglent used in their design. AFAIK nobody so far was able to identify it (case marking 955C). When I received my unit, I spent some time (unsuccessful unfortunately) trying to identify it. Since it seems that several manufacturers offer such switches with improved ESD tolerance, maybe the situation isn't as troublesome as we may think. I don't want to be the guinea pig though and rather use to treat the instrument's input quite carefully.

Cheers,
Thomas

[rf-loop]

A fixed 3dB input attenuator would be very good for improved ESD ruggedness on a SA input if the sensitivity degradation is acceptable.

ESD event has rise time of <1nS, so DC block not relevant here?

Very difficult to improve ESD ruggedness of sensitive GHz instrumentation.

DC block and and ESD protection, they all are different things.
I did not at all mean anything about DC block related to ESD!
Only told it about  HP and then tell that in Siglent DC block can not turn off. (of course can not because it is not implemented but also reason why it can not implement)



Up to some GHz there is still some things what can do for some level esd protection.

If there is well over 500V (up to ^ and perhaps potentially well over 1A (short time) peak current available... what is not at all unusual... 3dB attenuator can not do much (but of course tiny bit more than nothing) - simple physics.

If use external attenuator, as example I recommend many times, and then take it away only when really need. Still I want note that do not think and believe or trust this is ESD/EOS protection. Still need be careful.

[rf-loop]

Since it seems that several manufacturers offer such switches with improved ESD tolerance, ....

Cheers,
Thomas

Afaik it is not reinforced and from well known manufacturer.

(exept if it is short time ago changed)

[Safar]

Why not use some external coaxial limiter for ESD protection?

[TurboTom]

Slightly OT: Please find attached a quick comparison of the Siglent and the Rigol DSA815 V02.04 input circuitry. In case of the Siglent, I (maybe erroneously) assumed that the MMIC RF switches must feature some internal ESD protection since the footprints for the external suppressor diodes (D31~D34) are left unpopulated.
It's quite a surprise how much effort Rigol put into the design of their recent DS815 frontend circuitry. Clearly visible, they placed there some RLC circuitry and four(!) suppressor devices all in parallel. This input should be rather rugged.

Cheers,
Thomas