Siglent SSA3000X and SSA3000X-Plus Spectrum Analyzers

[Pinkus]

FYI for future readers (currently the threads are close together but that will change):
I did here: https://www.eevblog.com/forum/testgear/comparison-rigol-dsa815-tg-with-siglent-ssa3021x-spectrum-analyzer/msg965435/#msg965435
a brief comparison between the Rigol DS815-TG and the Siglent SAA3021X.

[EEVblog]

Here is the 10MHz phase noise comparison at 10KHz offset with a -110dBc source.
No contest, the Siglent wins hands down.

[EEVblog]

45min comparison with the Rigol 815 rendering now.
All I've got time for now, want to do a teardown tomorrow before it needs to go back.

[rf-loop]

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.

After @Pinkus previous answer I take next small amount different and simple approach to Q
This also tell something about  levels measurement. It need know that always if we are near noise level, where ever it is with our settings, level measuremet is unreliable due to signal + noise is what then measure and if noise is near signal all we know that 1+1 is 2   and 1+-1 is 0 but if signal level is enough below signal 1+0.01 is 1.01 and 1+-0.01 is 0.99. Independent of brand, math realities are  same.

In these images signal source level is also quite reliable but if need real accuracy this need other methods. But source is HP8644B with optional better level accuracy. (also I check that using two calibration level 30dB external attenuator with -70dBm out result was same as direct -130dBm.



With these settings in image it can measure accurately 10dB over ref level (if someone wonder these yellow A traces.)
There is many basic fundamental things nicely explained in this application note (chaper levels). http://cp.literature.agilent.com/litweb/pdf/5952-0292.pdf

Traces C and  D with bit more narrow RBW  for get more distance to floor.
Note: Trace C Detector accidentally forget to Normal instead of P-Peak (but in this case result is same so no error, just quickly checked it)






NOTE: Att 50dB! (if you wonder noise level)

(sweep time is nice)

For clarify. Next image only changed is that i release  trace C from freeze (also changed to P-PK but no effect here).  A, B and D stay freezed)
After then I set turn Attenuator from 50 to 0dB.  Signal is exactly same -40dBm.
As can see noise level drops as expected and measured level stay close same so also attenuator do not have big error between 0dB position  and 50dB position.

Noise level do not drop 50dB (of course)
But signal level drop and this is what need happend.

[cio74]

Thanks, much appreciated.

[ted572]

Here is the 10MHz phase noise comparison at 10KHz offset with a -110dBc source.
No contest, the Siglent wins hands down. 

The Signet data provided looks flawed to me.  Please see comments here ->  https://www.eevblog.com/forum/testgear/comparison-rigol-dsa815-tg-with-siglent-ssa3021x-spectrum-analyzer/msg966028/#msg966028

[Pinkus]

In Daves video at around 38:30 it shows a unit with firmware 100.01.02.07.06.
I am wondering where this FW 07.06 comes from, as the latest official available firmware is 07.05?

Interestingly his unit has a serial # just approx. 30 larger than mine which was calibrated 2 weeks earlier. That makes an average of 3 finished units per day. So it seems they do not expect to flood the market at this moment.

[markone]

Here is the 10MHz phase noise comparison at 10KHz offset with a -110dBc source.
No contest, the Siglent wins hands down.

Personally i would consider this peculiar performance gap togheter with larger frequency coverage clear winner points against Rigol DS815, bigger screen is a nice addition.

I am eagerly waiting to see your tear down video, with the hope to see a decent internal build quality, because hammered metal shield, rust and solder balls would be clear deal breakers.

[tautech]

In Daves video at around 38:30 it shows a unit with firmware 100.01.02.07.06.
I am wondering where this FW 07.06 comes from, as the latest official available firmware is 07.05?

I'm guessing it's a beta FW, we do occasionally get pre-release beta FW versions sent to us.
If it is we may have to wait a little for the next official FW release as the couple of bugs Dave spied might get fixed in the 7.xx. 

[rf-loop]

There is now so much talking about noise floor and phase noise things including misunderstandings and mixing things etc.

Close carrier phase noise is better what I expect with specifications. Specifications tell only Phase Noise for 10, 100 and 1000kHz offset and only for 1GHz.

Here 2 images.
First one is 10MHz carrier in center and 5MHz LSB and USB. Here used 1kHz RBW
Second image is also 10MHz carrier in center and then 10kHz LSB and USB.
Both images have 3:1 RBW/VBW

10MHz signal come from HP Z3801  owen (quite good phase noise also in near carrier where it beats normal cheap Rb's just completely) and holdoff (GPS antenna shutted off).

In first image there is 4 traces.
Trace B is for reference without signal (and then freezed) with all same settings.
Trace A Detector is Positive Peak
Trace C Detector is Normal
Trace D Detector is VideoAverage
No any trace averaging in use so that no one can develop paranoid thoughts what is hidden using average.


Second image  also have 4 traces.
Trace A detector is Pos Peak


Trace C is 20 times Trace Average from Detector: Pos Peak
Trace B detector is VideoAverage
Trace D is 20 times Trace Average from Detector: VideoAverage

Sidenote: When marker function is Noise marker do not use it for signal/peaks  level measurements. It is NOT for this!
Simplified: It force marker source trace input from VideoAverage detector output. (this is ok (only) for purposes what it is designed as its name tell). But it is not direct value from VideoAverage detector output. It show "noise corrected" 1Hz normalized value.
I have not ispected what correction values they exatly use. (this factor depends many internal details in filter shapes and detector details etc. In some individual case it looks like this value was 2.5dB.  (but my quess is that always in SSA3kX it is between 2.1 and 2.8dB (?))
So if you look VideoAverage and RBW is 1kHz... not -30dB from this but -27.5dB for Hz (noise).

[EEVblog]

SPOILER!
NOTE: Not finished yet, I need to go home and get some sleep 
Click for full res pic.
The 3.2GHz bit is obviously the preamp which can be switched in or out.

Siglent RF Section Block Diagram (NOT FINISHED) by Dave Jones, on Flickr

[1design]

The signal flow is a bit wrong on the above picture, my take at it is as follows:

• The first compartment are only switches: protection, calibration, 20dB attenuator
• Second compartment: LNA
• Marked 1st mixer: tracking generator mixer, LO enters from the top filter array and VCO then exists to the right into the LNA
• Next mixer to the right is the first downconverter and it get its reference from the main LO, the signal exists upwards towards the 810MHz SAW IF filters
• Mixer marked MCL downcoverts the IF to DC by mixing with LO/4 and outputs the signal upwards into the "RF switch"  section to U86

BR

[G0HZU]

Yes, there's lots wrong with the labelling at the moment.

If you send me an unmarked version I'll mark it up with some more detail with what I think is correct. I think the main (1st LO) VCO is the Zcomm part in the metal can that you have marked as VCO.

To get the full LO1 range this has to be doubled and then the doubled output has to pass through one of three interdigital BPFs before it reaches the first mixer. So you have the yellow arrow direction backwards through the three LO1 BPFs.

The section you have labelled MAIN  VCO LO1 looks like the VCO for LO2 to me. The yellow signal path arrows through the first IF along the bottom of the image looks to be backwards to me as well.

[G0HZU]

There also appears to be an optional/switchable (narrow?) BPF at the final IF to the left of U86 but these SMD parts aren't fitted. This option at the final IF may be to minimise the BW fed to the ADC when the full IF bandwidth isn't required.

If the (locked out?) control for this is found in the firmware and unlocked then you could fit a BPF with BW of your own choice and select it from the UI.

[G0HZU]

Looking at the microstrip dimensions (and the pale colour of the PCB dielectric material) the top layer PCB material is likely to be Rogers 4003C or 4350 at 0.020" thick.

If so, you could guesstimate the frequency of the first IF by measuring the resonator length in the first IF filter. It's going to be somewhere around 4GHz. Probably 4.4GHz? But that's just a guess based on a guess at the PCB material and a guess at the frequency plan. Looking at the resonator spacing, the filter is going to be quite wide. Probably 500MHz wide?
Each one of the three LO1 BPFs needs to be about 1GHz wide to cover the total 3GHz tuning range for LO1 and the resonator spacing in these filters looks to be about right for this bandwidth.

[EEVblog]

Yes, Z-COMM unit is the 1st LO I think.
I must engage brain next time. I'll start again.
Original photo is here:
https://www.flickr.com/photos/eevblog/27749729501/

[1design]

The amplifier on the 1st IF is a hmc716, so the frequency is between 3.1 and 3.9GHz.

[nctnico]

It is also nice to see they have used an elliptic / Cauer low pass microstrip filter (bottom left corner). It is the first time I have seen one in the wild.

[G0HZU]

The amplifier on the 1st IF is a hmc716, so the frequency is between 3.1 and 3.9GHz.

It might be down as low as 3.9GHz but that would mean that the IF rejection spec for the input LPF would be quite tough to meet.

Also, LO2 is clearly a low side LO wrt IF1 and if the IF2 is 810MHz then this would mean that LO2 would sit inside the analyser's main tuning range of 0 to 3200MHz. Even with IF1 at 3.9GHz this would happen.

But it's late and I can't think so clearly right now...  off to bed...

[EEVblog]

Anyone know what a 955C SOT23-6 is?
buggered if I can find it.
Obviously some RF switch

[1design]

Those are all SPDT RF switches. I couldn't find the manufacturer.

BR

[EEVblog]

This is a better effort:

SiglentRF-Final by Dave Jones, on Flickr

[bson]

Do we know yet what the real-time FFT span is?

[radar_macgyver]

They did well to choose the HMC835 for the second LO PLL. It has a low phase noise on its own and really shines when combined with an external VCO with a higher Q. I've used it in conjunction with Zcomm's coaxial resonant oscillators to make some compact, low phase noise synthesizers. I haven't played with the HMC703, but it looks like it has the same frac-N core as the 835.

I'm also hoping that Siglent will implement some dynamic parameter switching on these PLLs to tune up the close-in phase noise in certain modes. Agilent does this on the MXA and PXAs, which offer a low phase noise mode that trades off sweep speed for lower close-in noise.

Can anyone read the part number on U83/U84 (bottom right of picture)?

[EEVblog]

Can anyone read the part number on U83/U84 (bottom right of picture)?

HMC860, HMC976, linear regs