Hack of Sigllent spectrum analyzer ssa3021X?

[Lunasix]

Hi,
I have done step 6. After that, the display shows a flat line (with noise), around -60dB up to 3.2GHz, instead of a line reaching -50dB at this value before (not sure of exact values). Test with a 950MHz signal shows the right level.
Info system shows same options as for SSA, but all valid for a limited time.
Before and after step 6, telnet is never valid at startup, I must connect to web page, send DEBTTT SCPI command and connect to port 23 with root and ding1234. Step 3 has been done, file is modified, but it doesn't work (or just after modification and first reboot, but after some time, it no longer work).
Some time later, I've done step 8 (skip step 7). And after that, system info shows many new options, all permanently valid (except one, don't remember what option), and the display shows no longer a flat line but a line slowly climbing, as before cross flash. And then I've done step 7 (4 lines deleted, product type 2 to type 5).
Tested VNA mode, and of course Smith chart indicates cable is missing.
Tomorrow, I will open the Siglent, solder the 2 connectors and add the cable. I've found SMA connectors with center pin smaller, will fit in the printed board.
I think the most important is done, with success !

[Sly]

Hello,
I tried to measure the ESR of a 10MHz quarz crystal with my upgraded SSA3021.
Measuring |S11|=1.13dB with the Siglent in SA mode whicht indicates an ESR of around 14ohm.
Measuring |S11|=5.93dB with the Siglent in VNA mode (SOL calibrated, 7001 points) gives an ESR of 98 ohm.
Showing the Smith chart with the Siglent in VNA mode (SOL calibrated, 7001 points) gives a semi circle chart slightly outside the unity circle with an ESR of 99.75 ohm.
Showint the Smith chart with a Nano-VNA gives an almost perfect circle with an ESR of around 11ohm.

So in SA mode I get a realistic value for the ESR which sort of corresponds with the Nano-VNA.
In VNA mode the ESR is far from realistic around 100 ohm and the Smith chart is partly outside the unity circle.
To test if the high Q is part of the problem I tested with a resistor of 1k2 parallel to the crystal with about the same result.
To test if the high Q is screwing up the input circuitry and setpoints I tested with a 10dB attenuator between the VNA and crystal with about the same results.
Applying a short, an open, a load and a resistor around 12 ohm gives me totally correct values on the horizontal axis of the Smith chart.

The display indicates Cor (Calibrated) in the top left corner.

Can someone point me out if this is a PEBKAC problem or is there something missing/wrong with the updated SSA3021 in VNA mode?

Thanks in advance.

[RoV]

I have seen results like that, but it was caused by a calibration frequency range very different from the measurement range. E.g. cal 0.1-3200 MHz on 1601 points, followed by a measurement range 9.95-10.05 MHz on the same number of points.
Does the Siglent show "CAL" on the top left?
In those circumstances the VNA tries to interpolate the calibration points, but it seems not to work properly with xtals. Don't know if it is caused by a poor interpolation algorithm, or if it is normal for any VNA.

[S13]

I was thinking a calibration issue as well.

Also, looking at a detailed explanation of w2aew on measuring crystals with a vna, im not sure if you can just read the ESR from the smith chart? Take a look at this video, particularly at the 9:50 mark where he measures the motional resistance of a crystal:



On his smith chart it would read as 57.1ohms, while his calculation for motional resistance is 7.152ohms
Now i know motional resistance and ESR are not the same, but his example isnt far off from what you are seeing on your Siglent VNA. So i wonder which of your VNAs is correct?

[Sly]

The display indicates Cor which means calibrated according to the manual.

[Sly]

He measures S11 there while the back of the crystal is not grounded but connected to the other input meaning he measures Rm + 50 ohm.

[S13]

A yes 7.1 + 50 = 57.1 makes sense.

[RoV]

Ok, found! I was wrong with my initial assumption of bad calibration interpolation.
The issue is caused by the very high Q of the crystal, that requires a slower sweep respect to the analyzer default. Alternatively, a lower IF BW could be used, but this feature, although present, is "grayed" on the SVA1032X VNA menu.
Fortunately, sweep time is available and works as desired, causing a delay between frequency setting and measurement readout.
In the images I show the masurement of an 8 MHz crystal with 1601 points on 10 kHz around the series resonance frequency: with the default sweep time of ~1 s, with 5 s, with 10 s. The last setting is really necessary to get stabilization. Obviously a shorter time would suffice with less points.
Interestingly, interpolation seems to work rather well, because the instrument was calibrated on the full frequency span.

Another point: the default -5 dBm power can be too high for small crystals, causing them to warm, drift and increase the ESR. In this measurement I used -15 dBm.

Probably the nanoVNA performed better because it is slower...

[S13]

Interesting!
Im still new learning VNAs, so is it fair to assume this is something to look out for when doing measurements on any high Q circuit?

Should you always be comparing results with higher sweep times to see if the measurement remains consistent, for any type of measurement? Like, is that common practice?
Or is this more of a siglent thingy?

[RoV]

Interesting!
Im still new learning VNAs, so is it fair to assume this is something to look out for when doing measurements on any high Q circuit?

Should you always be comparing results with higher sweep times to see if the measurement remains consistent, for any type of measurement? Like, is that common practice?
Or is this more of a siglent thingy?

Normally not necessary. Personally, I have found only 2 scenarios requiring to slow down a VNA:
- very high Q
- high group delay
The first with crystals, the second when measuring reflections from the end of a long coaxial line, or reflections from a target at a distance from an antenna (requiring time domain analysis).

[S13]

I can confirm your solution.
So what does changing the sweep speed do actually? Does it change the IF bandwidth? I cant tell, because like you said the sva1000 greyes out the menu option, and it always shows as 10kHz regardless of what sweep speed i select.

My measurement of a 10MHz crystal:

[RoV]

I can confirm your solution.
So what does changing the sweep speed do actually? Does it change the IF bandwidth? I cant tell, because like you said the sva1000 greyes out the menu option, and it always shows as 10kHz regardless of what sweep speed i select.

It doesn't change the IF bandwidth. It introduces a larger (than default) delay between each new frequency setting and the start of the relevant measurement period. Devices with high Q respond slowly to excitation changes and require additional time.

[S13]

ok thank you for clearing that up! 

[Sly]

Thank you very much!

I can confirm indeed that increasing the sweep time makes all issues dissapear here too.
Was wrestling in my head with the fixed bandwitdth (i.c.w. the hight Q) but never thought about the sweep time.
This gives me back my trust in the instrument!

TL;DR
Measuring the S11 AND S21 of high-Q circuits like crystals and resonators in VNA mode might require setting the sweeptime manually (long enough) to get a useful result.
In SA mode this is not necessary because reducing the RBW also fixes this.


Thanks again 

[tautech]

Not trying to be petty but this VNA mode stuff should really be in the SVA1000X thread......
https://www.eevblog.com/forum/testgear/siglent-sva1015x-1-5ghz-spectrum-vector-network-analyzer-(coming)/

Would you mind quoting your own posts and pasting them into a new reply there please.

One further comment about slowing sweeps to get a more accurate result....posts above I see are just using a small portion of the 10001 points available.
If you were to set points to max and leave Sweep as Auto, could you get the same result ?

[Sly]

Thanks, didnt take it as petty at all.

I initially posted this issue here because I naively thought it could very well be related to modding the SA to the VNA model.
Of course now it definitely turned out to be a PEBKAC issue this is not the logical place.

With regards to the sweep speed:
I initially did increase the number of points to 10001 but that didnt work, only increased the time calibrating which I had done over and over again.

Now I’m not certain what you suggest me to do now. Shall I make one post in the SVA thread wrapping it al up? And removing my posts here?

[tautech]

Thanks, didnt take it as petty at all.

I initially posted this issue here because I naively thought it could very well be related to modding the SA to the VNA model.
Of course now it definitely turned out to be a PEBKAC issue this is not the logical place.

With regards to the sweep speed:
I initially did increase the number of points to 10001 but that didnt work, only increased the time calibrating which I had done over and over again.

Reducing the Span some to only cover the frequencies of interest can help immensely with sweep speed then Cal on the narrower Span.

Now I’m not certain what you suggest me to do now. Shall I make one post in the SVA thread wrapping it al up? And removing my posts here?

I'll leave that to you, it's an interesting discussion especially for those finding their way with VNA use but it really should be in the SVA1000X thread as no factory SSA3k analyzer (except SSA3000X-R) has a standard VNA mode.

[RoV]

...
One further comment about slowing sweeps to get a more accurate result....posts above I see are just using a small portion of the 10001 points available.
If you were to set points to max and leave Sweep as Auto, could you get the same result ?

This would alleviate a bit the problem, reducing the frequency step. But not solve it, because the new frequency would produce a measurable effect on the crystal with a delay longer than the inter-step time.

[mojoe]

Do these instructions still work for the latest SSA3021X-Plus?

https://www.eevblog.com/forum/testgear/hack-of-sigllent-spectrum-analyzer-ssa3021x/msg5746107/#msg5746107

The link he provided is dead. I think I found the same script here: https://www.online-python.com/PX3v0J8nbs

Is a firmware downgrade required for the latest shipping model?