Siglent SSA3000X and SSA3000X-Plus Spectrum Analyzers

[PartialDischarge]

Phase noise for a 1GHz 0dBm carrier from my Rubidium + Marconi 2030 for a 1MHz offset on both the Rigol and Siglent.
Significantly different PLL responses.

Very strange picture on Siglent.
Why does Siglent show no warning that input power 0dBm is higher than Ref -40dBm, Rigol shows UNCAL warning.
Why trace peak on Siglent is much much lower (-65dBm) than actual input power (0dBm) ?

Also why does the Siglent have a attenuation of 10dB set?

[Pinkus]

That's not what the UNCAL warning is for on the Rigol, it has to do with sweep time.
The Siglent will eventually show an "ADC overload" warning message and beep continually if the input is overloaded.

Exactly. In this setting everything below 100Hz RBW will show the UNCAL on the Rigol; in other cases in might appear earlier or later, probably depending on the sweep time.
However, Dave if showing such comparisons you should use the same settings on both machines. I recognized at least the 10dB  attenuator on the Siglent (0dB on the Rigol) (as the poster before me already posted).

The lower signal on the Siglent appears to be a bug. I recognized that with lower RBWs (probably only with narrow band signals - this needs some more time for checking) the signal is not being displayed correctly anymore (EDIT: not a bug, system was in video average mode and there were not enough data on the screen; with a smaller span (=more data) it is OK). See attached screenshots showing RBW of 10KHz, 3KHz, 1KHz, 300Hz, 100Hz, 30Hz). (Edit: in PeakPeak mode it is correct, thus Daves screenshot would probably look correct it he would change Detect mode from Video Average to PeakPeak).
It is always the same 10 Mhz signal, no difference in settings but RBW. Signal level on the screen falls from >0dB to -75dB.

[rf-loop]

Why does Siglent show no warning that input power 0dBm is higher than Ref -40dBm, Rigol shows UNCAL warning.

That's not what the UNCAL warning is for on the Rigol, it has to do with sweep time.
The Siglent will eventually show an "ADC overload" warning message and beep continually if the input is overloaded.

Can you explain what those images are supposed to show please, cos it's making no sense to me to set the ref to -40dBm and feed in an 0dBm signal.

We can set reference level lower than signal peak level. But, there need be careful also and know what do so that result in sideband is still valid. I will later show some test images what may explain this better but this may take bit time due to queued work and hobby situation.
Basic fundamentals of spectrum analyzers may also help.

[hendorog]

Why does Siglent show no warning that input power 0dBm is higher than Ref -40dBm, Rigol shows UNCAL warning.

That's not what the UNCAL warning is for on the Rigol, it has to do with sweep time.
The Siglent will eventually show an "ADC overload" warning message and beep continually if the input is overloaded.

Can you explain what those images are supposed to show please, cos it's making no sense to me to set the ref to -40dBm and feed in an 0dBm signal.

We can set reference level lower than signal peak level. But, there need be careful also and know what do so that result in sideband is still valid. I will later show some test images what may explain this better but this may take bit time due to queued work and hobby situation.
Basic fundamentals of spectrum analyzers may also help.

Interesting - so you are overdriving it for a portion of the sweep so that you have enough sensitivity left to see the sideband?

The SA I use has a phase noise mode for doing that and the dBc conversion. Possibly because it does not show anything like a sensible trace when overdriven to that level. I didn't realise other types permit it.

Not only is the 10dB attenuator set, the Siglent also has 10x averaging turned on.

[rf-loop]

Why trace peak on Siglent is much much lower (-65dBm) than actual input power (0dBm) ?

Because in this case Siglent use Detector mode what is ok for measure noise (average), not ok for measure signal peaks. Independent of this, there is still 0dBm signal what, due to phase noise, is rising sideband (noise) level what is (in principle) ok - exept that I suspect bit this result....

Also need note that 100kHz from carrier or 1M from carrier are in same ballpark. But if we look more low signals I suspect that  Rigol meet its noise floor level very soon.  But totally different game start if we go near to carrier. There is just - game over.
Many times I need look near carrier center, say example inside 1kHz from carrier center.

Siglent need repair and improve some things in Markers and specialy what are related to dBc/Hz and or dBm/Hz
It is borderline that I can name it as bug. But least it need improve if not even repair.
(Do not trust dBm/Hz readings exept if you measure just noise and nothing else. Specially if use delta cursors so that there is just noise and signal peak. For more trusted result you need change Detector. But now happend strange thing. It reset dBm/Hz and show agen just dBm. If at this time need use it there can set one trace for math. Trick is: do math for this trace sot that output is input (add zero). After then take this trace for Marker source. Now select Noise marker. (it now set VideoAverage detector  for this

Other thing what Siglent need think agen is frequency settings. If I do, Peak marker - move it to start freq. After I want change span I want keep this start freq but Span is always calculated related to CF and I can not select that  "keep start freq" (or stop freq or center). It is permanently fixed to "keep center freq". This is not nice for sideband working. Of course it can do also now but but need push lot of buttons after change span. Also proportinally moving Markers when change span need perhaps some more option what user want set. Keep abs positions or change proportinally after user change span.

[rf-loop]

Why does Siglent show no warning that input power 0dBm is higher than Ref -40dBm, Rigol shows UNCAL warning.

That's not what the UNCAL warning is for on the Rigol, it has to do with sweep time.
The Siglent will eventually show an "ADC overload" warning message and beep continually if the input is overloaded.

Can you explain what those images are supposed to show please, cos it's making no sense to me to set the ref to -40dBm and feed in an 0dBm signal.

We can set reference level lower than signal peak level. But, there need be careful also and know what do so that result in sideband is still valid. I will later show some test images what may explain this better but this may take bit time due to queued work and hobby situation.
Basic fundamentals of spectrum analyzers may also help.

Interesting - so you are overdriving it for a portion of the sweep so that you have enough sensitivity left to see the sideband?

The SA I use has a phase noise mode for doing that and the dBc conversion. Possibly because it does not show anything like a sensible trace when overdriven to that level. I didn't realise other types permit it.

Not only is the 10dB attenuator set, the Siglent also has 10x averaging turned on.

Specifications
"Displayed Average Noise Level (DANL)....sample detector, trace average >50"

As I have previously told sample det with trace average (enough sweeps) give very similar result as give VideoAverage. (When average mode is LogPower)
This can also use for measure sideband noise levels in case of phase noise. But it need note that only Peak detector give also carrier level reliable for measure its level. It come very important if user use example delta markers for measuring and there is easy acciddentally walk and drop to trap.

[nctnico]

Why trace peak on Siglent is much much lower (-65dBm) than actual input power (0dBm) ?

Because in this case Siglent use Detector mode what is ok for measure noise (average), not ok for measure signal peaks. Independent of this, there is still 0dBm signal what, due to phase noise, is rising sideband (noise) level what is (in principle) ok - exept that I suspect bit this result....

Can you show a 10MHz signal with a 1MHz span at 30Hz RBW to verify Pinkus used the wrong settings which caused the wrong results?

[tautech]

Why trace peak on Siglent is much much lower (-65dBm) than actual input power (0dBm) ?

Because in this case Siglent use Detector mode what is ok for measure noise (average), not ok for measure signal peaks. Independent of this, there is still 0dBm signal what, due to phase noise, is rising sideband (noise) level what is (in principle) ok - exept that I suspect bit this result....

Can you show a 10MHz signal with a 1MHz span at 30Hz RBW to verify Pinkus used the wrong settings which caused the wrong results?

See edited post
https://www.eevblog.com/forum/testgear/siglent-ssa3000x-spectrum-analyzers/msg964577/#msg964577

[Pinkus]

Can you show a 10MHz signal with a 1MHz span at 30Hz RBW to verify Pinkus used the wrong settings which caused the wrong results?

I edited my post above. The problem was VideoAverage mode on "ON" (I tried to use the same basic settings Dave was using). The same problem applies to Daves shot. As of this there is not enough data on the screen for the average, thus not enough data. With VideoAverage "Off" or a smaller SPAN (=more data) it is fine.
Edit: I added 4 pictures to this post:
1st: 100Hz in Average mode (displayed signal is already a bit too low)
2nd: 30Hz in Average mode (now even more dropped level)
3rd: 30Hz in Average mode but with a Span of 25Khz instead of 200Khz (everything OK as now more data available for averaging)
4th: 100Hz in PeakPeak mode and 200 Khz Span (everything OK even with a wider Span as now peaks are displayed)

So it is obvious that the problem is a result of not enough data on the screen for video averaging. I should have noted this before.
Dave: can you please produce your shot again in Trace: Average; Detector: PeakPeak mode?
The Rigol DSA815 does have the same problem when RBW is small enough, it has the the same wrong display as the Siglent; when Detector is set to Peak and Trace to Avg everything is OK here too.

[rf-loop]

Why trace peak on Siglent is much much lower (-65dBm) than actual input power (0dBm) ?

Because in this case Siglent use Detector mode what is ok for measure noise (average), not ok for measure signal peaks. Independent of this, there is still 0dBm signal what, due to phase noise, is rising sideband (noise) level what is (in principle) ok - exept that I suspect bit this result....

Can you show a 10MHz signal with a 1MHz span at 30Hz RBW to verify Pinkus used the wrong settings which caused the wrong results?

Imho,  This is not that Pinkus selected wrong settings.

Is this peak noise? No it is not, it is signal carrier peak. But he use Noise cursor function what set wrong Detector for measure nothing other than Noise in this named  case..

But yes I will show it (also it is explained indirectly earlier in this thread when I tell why I have used one trace for Math (for intermediate step for jump over Siglent forced detector selection)

[Pinkus]

OK Guys,
I will have the Siglent and the Rigol here for another 2 hours.
I already took the opportunity and made some checks and comparisons (with screenshots) which I will post later (maybe tomorrow).

Now: if you have specific needs and/or wishes where I should compare those two analyzers, please post them here. I will try to accomplish your wishes.
Time is running......

[rf-loop]

Why trace peak on Siglent is much much lower (-65dBm) than actual input power (0dBm) ?

Because in this case Siglent use Detector mode what is ok for measure noise (average), not ok for measure signal peaks. Independent of this, there is still 0dBm signal what, due to phase noise, is rising sideband (noise) level what is (in principle) ok - exept that I suspect bit this result....

Can you show a 10MHz signal with a 1MHz span at 30Hz RBW to verify Pinkus used the wrong settings which caused the wrong results?

Here some images (2nd is what you ask).
There is used different Detectors.
PosPeak (pink), Normal (Yellow), VideoAverage (Green) this last one is good to noise but very careful with other things, as described shortly also in Usermanual.



All images is made so that carrier peak is 0dBm. (also in displayed level range exept last "extreme")
Signal come from HP/Agilent 8644B also checked with Z3801 DOCXO (in free run mode)  and with HP8642B with its own option reference. I do not use directly example Rb for reference because they are so poor in short time stability.

Last image: This is NOT what you get when you buy SSA3000X!  This is out from SSA3021X and SSA3032X features and specifications, totally.
In last image I have set ref to -10dBm because noise level drop partially out from display. 10dB over ref so that noise level rise enough for looking. Carrier peak is still  measured right by Marker and it is there still ok 10dB over reference level (but not more if want level accuracy (in this particular combination of settings)

Also it looks like Siglent frequency reference is really good.  It looks more like if want competite higher class SA's there need bit improve PLL or soething.

----
@Pinkus. If you still have available Rigol also, please look 1kHz span 100Hz RBW and example 4k span with 0dB signal and using normal pos peak and video average detectors and so that ref is also 0dBm. (perhaps you need also use some 2-5 times  trace averaging for more steady image?)
Because, there is real drop out game.

In images, note that Marker 1 is only what measure narrow carrier "peak" level right. (Marker 2 just show how wrong it can be with Detector what is not good for used purpose. User need know what he is doing.


1GHz   Span 1M, RBW30Hz


10MHz   Span 1M, RBW30Hz


10MHz   Span 200kHz, RBW30Hz



Do not dream about this!

[egonotto]

Hello,

i want ask what the real difference in noisefloor (DANL) is.

As in Daves picture "SiglentSSA3000-1.JPG" it seem to be almost 20dB.

I have trouble to believe that.
Do he use comparable adjustment (perhaps detector)?

Best Regards
egonotto

[Pinkus]

@Pinkus. If you still have available Rigol also, please look 1kHz span 100Hz RBW and example 4k span with 0dB signal and using normal pos peak and video average detectors and so that ref is also 0dBm. (perhaps you need also use some 2-5 times  trace averaging for more steady image?)
Because, there is real drop out game.

OK - done. Almost in the last minute...the Rigol is gone now - no more comparison possible.
Please find attached the shots you asked for.
All show the same 10 Mhz signal with (almost) 0dB level. All shots with 20x video average (detector: peak-peak).
Attenuator had to be 20dB at Siglent (Rigol 10dB), otherwise the signal was to strong. All other settings are identical.

Screenshots:
Rigol 4 Khz SPAN, 100 Hz RBW
Siglent 4 Khz SPAN, 100 Hz RBW

Rigol 1 Khz SPAN, 100 Hz RBW
Siglent 1 Khz SPAN, 100 Hz RBW

Rigol 1 Khz SPAN, 30 Hz RBW
Siglent 1 Khz SPAN, 30 Hz RBW

Siglent 1 Khz SPAN, 10 Hz RBW (10 Hz not available at Rigol)

[nctnico]

Still I'm not sure whether it is correct the signal shows so low on the SSA3000 when using the normal detector. I just tried with my Advantest SA and with RBW=1kHz and VBW=10Hz a 0dBm signal stays at 0dBm. Could it be Siglent compresses multiple points into a single point by taking the average of several measurements instead of the maximum measured value?

[rf-loop]

Still I'm not sure whether it is correct the signal shows so low on the SSA3000 when using the normal detector. I just tried with my Advantest SA and with RBW=1kHz and VBW=10Hz a 0dBm signal stays at 0dBm. Could it be Siglent compresses multiple points into a single point by taking the average of several measurements instead of the maximum measured value?

I do not exatly understand what test or what image you have in mind with this comment.

Perhaps this clarify it better. There is marker now for every trace and every trace have different detector.

Signal 10MHz, 0dBm, Span 4M, Center 10M, RBW 1kHz, VBW 10Hz.

Detectors:
A Normal e as in some previous images this "fat" yellow trace where even point draw positive and odd point negative and repeating over
B PPeak
C Npeak
D VideoAverage

Detector mode Sample not used (it moves between B and C as can see in my some previous images)

Traces just all new in every sweep (no trace averaging)

Display line is adjusted exactly where is VideoAverage level when signal is off



As can see PPeak and Normal result is same.


Note: Ref is -10dB but signal is 0dB. Top is over image but it measure it still ok, (this is not problem in this case)

[nctnico]

Yes, but what happens if to decrease the RBW to 30Hz with a 1MHz span? IMHO normal detector mode should give a level around 0dBm.

[rf-loop]

Yes, but what happens if to decrease the RBW to 30Hz with a 1MHz span? IMHO normal detector mode should give a level around 0dBm.

Yes it give also 0dBm, exactly. Images later.

[Andy2]

The Siglent certainly appears to put the Rigol in the shade re phase noise. But (I'm sorry, I can't resist this as I've had an early Scotch & Dry...) My little Signal Hound SA44b does even better:

[rf-loop]

Yes, but what happens if to decrease the RBW to 30Hz with a 1MHz span? IMHO normal detector mode should give a level around 0dBm.

Here previously missed images.
As can see, both detectors give same result.
Also there can see VidAve trace is same independent of VBW as expected.

It is possible that in some my Siglent images made confusion becaause no information abput what marker is for what trace. Also it need know that trace display priority is D, C, B, A.  No it is not accident it is just clever. First you use default trace A. After then you want add some setting... take trace B. It write over A, then you think you want still keep these  but take something other visible and you add C  and now this most new write over these old...  This is good as default.  But I hope they later add feature where user can call what ever trace over others just pressing trace selector and this trace pop up  over others and so on..  Just like in oscilloscopes.

[rf-loop]

The Siglent certainly appears to put the Rigol in the shade re phase noise. But (I'm sorry, I can't resist this as I've had an early Scotch & Dry...) My little Signal Hound SA44b does even better:

Yes.

There can also see that SH filter shape factor is different.
And, it take nearly 7 times more time for sweep 4kHz with RBW100Hz
How it is with 10Hz RBW.

SH near carrier phase noise looks exellent!

[Andy2]

Moving in to 1KHz span and 10 Hz RBW begins to show the Hound's very close in noise:

[rf-loop]

Moving in to 1KHz span and 10 Hz RBW begins to show the Hound's very close in noise:

NOTE: In Siglent image top level (Ref) is -20dB. Signal top is there 0dBm but unvisible.
(Siglent display setting max is 10dB/div so max visible range is 100dB)



Signal 0dBm 10MHz, AM modulation 1% with 80Hz

Green trace is average what give quite ok result for also average noise level.
Yellow trace is Positive peak detector Positive peak hold. "Max hold".
Pink trace is Positive peak detector Negative peak hol. "Min Hold"

Just explanation if someone is confused with Atten and Ref level settings... Siglent can not turn for 120dB display range. So visible window can in some situations  move down if there is some interesting under -100dBc.
Why 21dB Attenuator instead of Default 20dB. Just with 10Hz RBW if use default -20dB Attenuator and Ref is 0dBm and Signal is 0dBm it goes too close to some internal limiting factors. (With 30Hz RBW there is more room before hit limitst, and with 100Hz RBW agen more.) If situation is very very bad with levels then there come also visible and audible warning of course but then we are really in true ADC clipping) There is not this situation with 10Hz RBW, 0dBm signal 0dBm Ref 20dB atten. But there strat minimal level instability in measurements and something like 0.2-0.4dB random variations. This is why I set 1dB more Attenuator from 20dB to 21dB so signal peak level measurement is rock solid stabile and accurate.

Shortly: do not worry my settings. They are exactly ok for this case and do not give any kind of too bad or too good result..

Signal is there just ok over top of visible window with these settings. Also of course automatic tracking cursor follow this peak.

Why there is this modulation. Just for imagine what is situation if look same with example Rigol. These AM sidebands just perhaps barely detectable. (But not at all if minimum RBW is 100Hz, but I have heard someone have hacked it to 10Hz RBW)

Disclaimer:
Last two images: No one can promise if these are available and true in SSA3kX. World is full of many kind of experimental things out from specifications....
Specifications for SSA3000X what Siglent sell and promise at this time is: Minimum available RBW is 10Hz. Period!



Next is just for show  that previous image signal level is ok when it exeeds display reference level.
Only change is ref level.

[cio74]

Apologise if I have missed it, have someone injected or/and sweep a signal very close to the noise floor level, say -80dBm, is it properly displayed/measured?

Thanks.

[Pinkus]

Apologise if I have missed it, have someone injected or/and sweep a signal very close to the noise floor level, say -80dBm, is it properly displayed/measured?

Thanks.

Here you go: 0dB signal with external attenunator switched to
0dB
-20dB
-40dB
-60dB

and then
-70dB
-80dB

and then
-80dB with PreAmp switched on.
-80dB with PreAmp switched on and also internal attenuator of 40dB.
Nothing to complain I would say.