Siglent SVA1015X and SVA1032X 1.5, 3.2GHz Spectrum & Vector Network Analyzers

[Kibabalu]

I have a question (possible bug I found?) about the EMI software package.

I am taking EMI measurements through a 10dB attenuator. Based on the manual I need to set the "Ref Offset" value in the Amplitude menu to match the external attenuator.

When in normal spectrum analyzer setting the "Ref Offset" value does shift the amplitude of the plot by the correct amount. If I enter 10dB the entire curve shifts up by 10dB on the scale.

However, when in the EMI mode setting the "Ref Offset" value shifts everything, including the scale and the limit mask. This means that all the readings and the limit mask for checking EMI emissions levels is incorrect if there is any external attenuator that is corrected via the "Ref Offset" value.

Am I misunderstanding something? The behavior seems to be correct for just spectrum analyzer mode. I looked through all the options specific to the EMI mode and I do not see any alternative way to set a ref offset value. The limit masks can be shifted up or down, but that still means that all the amplitude values are incorrect with an external attenuator.

We discussed this issue in

https://www.eevblog.com/forum/testgear/using-siglent-ssa3021x-(no-hacking-here)/msg3455418/#msg3455418

It's clearly a bug that makes this machine inapplicably for serious EMI work.

[tubularnut]

The behaviour is just the same - when you press calibrate, calibrate, 1-port the frequency range always resets to 100khz to 3.2Ghz.

I have a SVA1032X, born SVA1032X, with SW 3.2.2.4.0 20201220-1 and in this unit the freq does not change

Works ok on mine too, upgraded SSA3021X, same software versions.

[eloso]

Wow Thanks ! - I guess it was kind of obvious but yes, plugging in a mouse really makes a difference - recommended ! 

There are still unresolved issues around VNA calibration and saving of calibration.

One has been referred to already by Ghislain

Am I correct in assuming that in such case, after the user calibration (OSLT) has been successfully performed, the status in the upper left of the screen would have to change to Cor (=calibrated)?
Now it shows --- regardless of what I do.

According to the user manual on page 76:

3.5.6 Calibration
Set calibration related items. The calibration status is displayed in the upper left corner of the screen.
The calibration status and display are as follows:
No calibration data --- (displayed in gray)
Calibrated Cor (displayed in blue)
Correction Off Off (displayed in gray)
Need to re-calibrate C? (displayed in blue)

I experience the same - the unit seems to be calibrated (I know this becuase I am calibrating with a known dodgy load and it accounts for it perfectly). But it is disconcerting to see that the status is ---

I did find that if I do 1-port Cal followed by just an "Open" Cal the status changes to "Cor" and the 1-port cal is not lost. So there is another good work around.

It begs the question:  What is the point of having a menu option just for "Open" calibration and another for "Short" Calibration when there is none for "Load" Calibration ? I mean if you want to do a calibration you either select "1-port" or "Through" which include all elements.

OK, next problem. When you Save the calibration to a file, switch the instrument off, switch on again, it has lost the calibration. Aha - but you saved it to a file ? Load the file and, yes you guessed it, you have still not got your calibration back. The only answer is to repeat the calibration.

Now this is not a huge problem because I think you probably need to calibrate at every use anyway - especially if you've had the instrument off and come back for another session. But you are supposed to be able to do this and it makes me queasy that if some of these small items don't work, what is there that also doesn't work and that might give you really dodgy results when you don't realise it?

After having read everything I can find, I am now starting to think that this might all be down to the unit being an "upgraded" SSA. Perhaps the firmware is trying to deal with a use case that it wasn't written for - like maybe there isn't a proper factory VNA calibration file and it is getting into a twist as it tries to combine a non-existent file with a user cal file.

I did have a look at the logs and thought I may have the answer but there are log entries to say successful factory vector calibration applied, so it probably isn't that after all. But perhaps something similar that I haven't thought of ?

Regards

Eloso

[hendorog]

Wow Thanks ! - I guess it was kind of obvious but yes, plugging in a mouse really makes a difference - recommended ! 

There are still unresolved issues around VNA calibration and saving of calibration.

One has been referred to already by Ghislain

Am I correct in assuming that in such case, after the user calibration (OSLT) has been successfully performed, the status in the upper left of the screen would have to change to Cor (=calibrated)?
Now it shows --- regardless of what I do.

According to the user manual on page 76:

3.5.6 Calibration
Set calibration related items. The calibration status is displayed in the upper left corner of the screen.
The calibration status and display are as follows:
No calibration data --- (displayed in gray)
Calibrated Cor (displayed in blue)
Correction Off Off (displayed in gray)
Need to re-calibrate C? (displayed in blue)

I experience the same - the unit seems to be calibrated (I know this becuase I am calibrating with a known dodgy load and it accounts for it perfectly). But it is disconcerting to see that the status is ---

I did find that if I do 1-port Cal followed by just an "Open" Cal the status changes to "Cor" and the 1-port cal is not lost. So there is another good work around.

It begs the question:  What is the point of having a menu option just for "Open" calibration and another for "Short" Calibration when there is none for "Load" Calibration ? I mean if you want to do a calibration you either select "1-port" or "Through" which include all elements.

OK, next problem. When you Save the calibration to a file, switch the instrument off, switch on again, it has lost the calibration. Aha - but you saved it to a file ? Load the file and, yes you guessed it, you have still not got your calibration back. The only answer is to repeat the calibration.

Now this is not a huge problem because I think you probably need to calibrate at every use anyway - especially if you've had the instrument off and come back for another session. But you are supposed to be able to do this and it makes me queasy that if some of these small items don't work, what is there that also doesn't work and that might give you really dodgy results when you don't realise it?

After having read everything I can find, I am now starting to think that this might all be down to the unit being an "upgraded" SSA. Perhaps the firmware is trying to deal with a use case that it wasn't written for - like maybe there isn't a proper factory VNA calibration file and it is getting into a twist as it tries to combine a non-existent file with a user cal file.

I did have a look at the logs and thought I may have the answer but there are log entries to say successful factory vector calibration applied, so it probably isn't that after all. But perhaps something similar that I haven't thought of ?

Regards

Eloso

Quick answer to part of it, this might clear up a bit of confusion:

The Open and Short calibrations are not part of the 1 Port cal. Instead they are for quick, approximate cals where you only use one standard.
The data is being normalised to a standard - almost the same thing as if you 'Save Data' to Memory and use the Data/Mem math function.

A quick test suggests that this Open Cal is not exactly the same thing as the Data/Mem math function as there is a -21ps delay coming from somewhere.  I haven't researched it further.
There is no directivity correction of course.

A load calibration on its own would be bit different - one point is that a perfect load has no reflection and so can't be used to normalise phase or magnitude.

I did come up with a quick and dirty, poor mans calibration at one stage. From memory it was like this:

* Sweep the open port/open ended test cable - use that as the open standard measurement
* Alter the open sweep data by 180 degrees - use that as the short standard measurement.
* Sweep a real load.
* Calculate the correction in the normal way.

I'm sure there are better ways to do it, that was just something I dreamed up.

[hendorog]

Also - check you are on the latest firmware?

I am not able to replicate the Save/Load cal issue either.

[tautech]

Also - check you are on the latest firmware?

I am not able to replicate the Save/Load cal issue either.

The trick is to change the Save type to a Cal file type so it gets stored internally and is accessible to load again.

[eloso]

Thanks everyone.

Yes I am on latest firmware. 3.2.2.4.0.r2 is displayed on the System Screen.

And sorry but no, the answer is not to change the file type to CAL. I have already done this as a matter of course.

Interestingly, loading a saved CAL file does make a difference - ever so slight difference to the display. But it is a long way from being what it should be. I am using a very good 50 ohm load and a fairly dodgy one, in order to give me a reliable baseline and test subject.

I have heard two folk say now that this behaviour is not apparent on their instruments.  I am wondering if there is an issue relating to it being an upgraded machine rather than a native born one.  It would be really helpful if folk can state if their machine was purchased as an SVA1032X or was transformed into one.


I had no problem with the upgrade and Linux has been part of my career for a long time. I have checked several times and looked at several different (but more or less the same of course) descriptions of the upgrade process so am as confident as it is possible to be that the upgrade went according to plan and should be just like everyone else's.  I'm experienced enough also to know that this may not be the case of course and I may have made a mistake. 

Cheers


Eloso

p.s. The three issues in a nutshell are

1. The resetting of the stimulus to 100khz-3.2Ghz when pressing 1-port, calibrate, calibrate.
2. The stubborn refusal for the cal status indicator to show anything other than "---" after the 1-port calibration routine and
3. The incorrect calibration status after loading a previously saved CAL file.   

Just to clarify, there is no error message, nor indication that the loaded CAL failed.  It is just that the instrument doesn't go back to the proper calibrated state that it was in when the CAL data was saved. This is evidenced by the trace of a 50 ohm load  that is lossy above 2Ghz but that is used for calibration. After calibration the load of course looks perfectly flat.  Switch off, switch on, load the cal file and it reverts to its true picture of poor return loss.

[hendorog]

Thanks everyone.

Yes I am on latest firmware. 3.2.2.4.0.r2 is displayed on the System Screen.

And sorry but no, the answer is not to change the file type to CAL. I have already done this as a matter of course.

Interestingly, loading a saved CAL file does make a difference - ever so slight difference to the display. But it is a long way from being what it should be. I am using a very good 50 ohm load and a fairly dodgy one, in order to give me a reliable baseline and test subject.

I have heard two folk say now that this behaviour is not apparent on their instruments.  I am wondering if there is an issue relating to it being an upgraded machine rather than a native born one.  It would be really helpful if folk can state if their machine was purchased as an SVA1032X or was transformed into one.


I had no problem with the upgrade and Linux has been part of my career for a long time. I have checked several times and looked at several different (but more or less the same of course) descriptions of the upgrade process so am as confident as it is possible to be that the upgrade went according to plan and should be just like everyone else's.  I'm experienced enough also to know that this may not be the case of course and I may have made a mistake. 

Cheers


Eloso

Ok, I think I've found a/the problem:

There must be a calibration done first before loading the cal has any effect.

I tested by doing an 'open cal'  with a load on the port (because it was quicker as it only requires a single standard).
(Note that I used a Load here instead of an Open on purpose, I wanted the calibration to be obviously different to what is normally shown)

After the cal was completed, I applied it and then removed the load from the port.
The trace moves up to 30-40dB, and is obviously different to what an open port would normally show (which is a flat line at 0dB)

Then I saved the cal, reset the machine and loaded the cal again.
Observed that nothing changed, which is a bug.

Then I did an open cal with the port open, and applied that cal.
Now I loaded the saved cal again, and this time it shows the expected trace from the saved cal.

[scotty7]

Hi everyone,

I bought an SVA1032X as my first VNA. I have a concern about it though. After a fresh calibration using the "enhanced response" cal, looking into the instrument and measuring the S11 it doesn't seem to be very well matched. Any chance anyone else can confirm this with their unit? The VNAs I've used at work don't have this kind of response so I'm curious as to why mine looks like this.

First picture is what I'm concerned about, second picture is the test setup.

Thanks!

[hendorog]

Looks wrong to me, but I haven't used it myself. Make sure you are on the latest FW as there have been improvements with the calcs.

I am not in a position to replicate at the moment - but just wondering if you can post the same picture using just an OSL cal?
Also the measurement of the same load used in the cal - for both a 1 Port cal and an Enhanced Response cal.

That should show if the Enhanced response is doing anything at all, or just not improving the port 2 match.

[scotty7]

There's no improvement over the OSL cal. I tried that as well.

[hendorog]

There's no improvement over the OSL cal. I tried that as well.

Ok thanks I will report it as a bug.

[ExaLab]

Hi everyone,

I bought an SVA1032X as my first VNA. I have a concern about it though. After a fresh calibration using the "enhanced response" cal, looking into the instrument and measuring the S11 it doesn't seem to be very well matched. Any chance anyone else can confirm this with their unit? The VNAs I've used at work don't have this kind of response so I'm curious as to why mine looks like this.

I am not a sva1032x owner but the graph seems credible. It represents the s11 of port 2 (which is usually not very efficient in terms of vswr) seen through a transmission line less than a meter long.

[hendorog]

Hi everyone,

I bought an SVA1032X as my first VNA. I have a concern about it though. After a fresh calibration using the "enhanced response" cal, looking into the instrument and measuring the S11 it doesn't seem to be very well matched. Any chance anyone else can confirm this with their unit? The VNAs I've used at work don't have this kind of response so I'm curious as to why mine looks like this.

I am not a sva1032x owner but the graph seems credible. It represents the s11 of port 2 (which is usually not very efficient in terms of vswr) seen through a transmission line less than a meter long.

Trawled through some old documents to refresh the memory, and I think that it depends upon which version of "Enhanced Response" is implemented.

The version in this document was the one I had in my head. I've commented on it before, and (assuming I'm understanding it correctly) then it does correct for load match. (e22):
https://www.researchgate.net/publication/263306917_1_GHz_Automatic_2-Port_Vector_Network_Analyzer_Using_Common_Laboratory_Instruments

However, it is most likely you are right and Siglent have implemented the more simplistic HP version. That doesn't correct for Load match and is just an OSL cal plus Thru normalisation:
https://ena.support.keysight.com/e5071c/manuals/webhelp/eng/index.htm#measurement/calibration/basic_calibrations/enhanced_response_calibration.htm

[eloso]

The lumpiness looks like it repeats around every 166Mhz -  consistent with the length of the cable at 1/4 wavelength. Just a guesstimate ....

Eloso

[nike75]

I found some bugs with my crossflashed SVA1032X - firmware 3.2.2.4.0

1.Very often but not every time: after powering on, having previously selected vector analyzer mode with applied user calibration, the device starts without applying the user calibration.
Even more - correction button is "On" state and 'Cor' is showed. When I press the correction button (On/Off), it works in opposite - when corrections are off, actually user calibrations are applyed, and when corrections are on, calibration is not applied.
This problem remains so either until a new calibration is performed or until the mode of the instrument is changed and returned to the vector analyzer.

2. Very often when making a 1-port calibration the process does not finish - the device does not complete some measurement(open/short/load) and it is necessary to select the calibration process again. This happens more often when working with more points - for example 751. Is it possible that this problem is due to the fact that the instrument is cross-flashed and does not have a factory calibration of the vector analyzer?

[TK]

Does it happen only on crossflashed units or all SVA units?  Mine has similar symptoms but it is crossflashed. We will need confirmation from an original SVA unit

[eloso]

I have experienced this and a variety of quirks on my cross-flashed unit.  I do use the instrument mostly below 500Mhz and I often use my genuine Deepelec NanoVNA in preference to the Siglent because it just kind of works better and provides sufficient precision.  Yhat's a bit sad really from the Siglent point of view but let's face it, I am getting benefits from Siglent cross flashing that are essentially free so no complaints there at all - in fact just gratitude for all the work that has been done to help me upgrade the instrument.

I have a lot of experience in software development and I do have a sneaky suspicion that the software has been written to include code that expects a factory calibration file to exist and that causes some unexpected nehaviour if it doesn't.  I would love to try out someone elses factory calibration file, or even convert one of my own user cal files to become the factory calibration if it turns out they are in the same format.  It matters not a jot if the factory cal isn't the right one for that machine, what would be important is that it exists in the right place to be picked up by the code that otherwise misbehaves if it unexpectedly doesn't find it.

Is there a dump of the file system available anywhere of an original genuine SVA or can some kind person make one available? Or perhaps just a factory cal file or a dump of the cal directories to allow me to make my own investigation ? Or just more info about the factory cal file format/naming ?

Thanks !

Eloso

[tv84]

Indeed, it looks like the "original" SVA1000X has 2 more files than an original SSA3000X+.

I attach them below.

Try them to see if it solves the problem.

[Bicurico]

Great finding!

 

Hopefully this closes the cross-grade.

Would be great if these two files are user updateable usinig some selectable auto-calibration function.

Regards,
Vitor

[eloso]

My cross-flashed SVA1032X does indeed already have these two files in the firmdata0/cali directory.  File sizes are identical. sha256 checksums are different to those of the files you attached. My own version are datestamped to coincide with the last calibration that I carried out earlier today.

Nevertheless, I made full backups of everything as usual and swapped the files for those that you have attached.

This doesn't appear to have made any difference to the quirky behaviour I have documented earlier.

It was a nice thought but perhaps we are barking up the wrong tree. Unless there are factory calibration files elsewhere. Rather than comparing file lists from an SSA with a genuine SVA can we compare filelists from a genuine SVA with my upgraded version ?  I can do this if someone can provide a full treelist of files from a genuine SVA.

Cheers

Eloso

[tv84]

It was a nice thought but perhaps we are barking up the wrong tree. Unless there are factory calibration files elsewhere. Rather than comparing file lists from an SSA with a genuine SVA can we compare filelists from a genuine SVA with my upgraded version ?  I can do this if someone can provide a full treelist of files from a genuine SVA.

The thought of missing calib files was good. The thought of something missing elsewhere doesn't make much sense. All other files probably are installed by a FW package.

Do you want the full calib directory of a SVA1032X to replace yours?

[eloso]

Yes, I am sure you are right.

In the early days incidentally I checked my debug log and there were many entries pertaining to factory calibration data file load  failing. Curiously this changed at some point and for the last month I regularly get messages to say factory calibration file load was a success.  There is even a huge string of exclamation marks after it (did the author think that success was such a rare event that exclamation marks would be needed ?    )

I don't think there is much mileage in me trying to take this any further. Wait and see if anyone else reports anomalous behaviour in VNA mode. At the end of the day it seems to work fine and gives the kind of results I would expect when using the little tricks and workarounds.

Thanks for stepping in.

cheers

Eloso

[nike75]

My crossflashed device did not have factory cali files for vna. Now i copied other device files (from tv84) and things look an idea better.

Issue 1 (post#641) with the state of the user correction after powerOn remains. 
It will soon become clear whether issue 2 with random failed user calibrations will disappear.

Thanks tv84!

[antenna]

Anyone have pictures of the SVA1032x apart?  I just noticed the VNA port on mine is slightly loose and I need to know how to disassemble it to get them tightened. Normally, I'd just tear into it, but I got as far as cutting the warranty seal and realized its a bit too expensive to tear apart before seeing if anyone has advice on the topic.  Anyone do this before?