vnaproperty.h

[wolfgang6444]

apparently the file vnaproperty.h is missing in the tar-ball:
n2pkvna.h:43:10: fatal error: vnaproperty.h: No such file or directory

Many thanks,

Wolfgang

[scott-guthridge]

Hi Wolfgang. Thank you for trying this software out. The vnaproperty.h file comes from a library in a different git repo: https://github.com/scott-guthridge/libvna) Have you installed that? You can build libvna from source, or under Releases, you'll find both RPM and Debian binary packages.

I haven't contributed any updates to n2pkvna for a while because I wasn't getting any complaints from anyone, so I figured no one was using it. If I get feedback, though, I will respond.

[wolfgang6444]

Hi Scott,
many thanks for the answer and your work. I very much like the hardware, but I do not like the fact that I have to run some closed source software under XP hence my interest in your solution. In the long run I would like to make the hardware stand-alone with a raspberry pi.
Now libvna and n2pkvna compile and install fine, but upon startup I get:

n2pkvna-gui
Can't locate Browser/Open.pm in @inc (you may need to install the Browser::Open module) (@inc contains: /usr/local/lib64/perl5 /usr/local/share/perl5 /usr/lib64/perl5/vendor_perl /usr/share/perl5/vendor_perl /usr/lib64/perl5 /usr/share/perl5) at /usr/local/bin/n2pkvna-gui line 21.
BEGIN failed--compilation aborted at /usr/local/bin/n2pkvna-gui line 21

I am on Rocky-Linux 8.10

Wolfgang

BTW: I had to install libyaml-devel for libvna; maybe it's worthwhile mentioning that in the README or automatically checking the dependency in the config

[scott-guthridge]

Thanks for pointing out missing dependencies. The GUI needs the perl-Browser-Open package. And I see that the n2pkvna build does currently need libyaml-devel, but I think I can rid of that [direct] dependency since n2pkvna doesn't compile against that library. Build of libvna still needs it, but its README-BUILD says that.

I don't have Rocky, but I started up a clean fedora39 in a container -- so not much is already installed -- and created a non-root user with sudo access. Here are the steps I followed:

# Install build dependencies
sudo dnf install libusbx-devel gcc libyaml-devel make

# Add these for build from git
sudo dnf install autoconf automake git libtool

# Install runtime dependencies
sudo dnf install fxload
sudo dnf install perl-Browser-Open perl-Cairo perl-Carp perl-Clone \
        perl-File-Temp perl-Glib perl-Gtk3 perl-open perl-Thread-Queue \ 
        perl-threads perl-YAML-LibYAML

# Install libvna (would need to build from source if not x86_64)
curl -LO https://github.com/scott-guthridge/libvna/releases/download/v0.4.0/libvna-devel-0.4.0-1.x86_64.rpm
curl -LO https://github.com/scott-guthridge/libvna/releases/download/v0.4.0/libvna-debuginfo-0.4.0-1.x86_64.rpm
curl -LO https://github.com/scott-guthridge/libvna/releases/download/v0.4.0/libvna-0.4.0-1.x86_64.rpm
sudo dnf install libvna*.rpm

# Clone n2pkvna source and enter
git clone https://github.com/scott-guthridge/n2pkvna.git
cd n2pkvna

# Bootstrap Autoconf and Configure
./bootstrap
./configure

# Build
make -j12

# Install
sudo make install

I will update the docs and try to get rid of n2pkvna's dependency on libyaml-devel. Thanks.

[scott-guthridge]

I pushed a fix to the "development" branch you can try. I will create a new release soon.

[wolfgang6444]

With your help I now can compile and install.
Upon startup I get a message telling that the device is missing:

n2pkvna-gui
n2pkvna: error opening n2pkvna device: No such file or directory
Perl exited with active threads:
1 running and unjoined
0 finished and unjoined
0 running and detached

I have created the group n2pkvna and added the current user. It does not work as root either.
Are you looking for a particular device in /dev ? Do I need a udev-rule?

[scott-guthridge]

You do need a udev rules file -- the make install step should have installed it. Look in /usr/local/lib/udev/rules.d/. You may need to tell you system to re-scan the rules to notice it:

udevadm control --reload-rules

To debug this further, run lsusb. You should see a line like this:

Bus 004 Device 008: ID 0547:100b Anchor Chips, Inc. N2PK VNA V5

Note the 100b (as opposed to 100a). This indicates that the firmware was successfully loaded.

I'll add that and a couple other details to the README, too.

[scott-guthridge]

Something that might simplify this would be building from RPM source instead of doing everything manually. This has the advantage of automatically checking for missing build-time dependencies and automatically installing all the run-time dependencies. It also does things like running ldconfig after installing a library and running udevadm control --reload-rules after installing UDEV rules, so you don't need to do those by hand. There are fewer steps overall. And it finally has the advantage of installing in the standard /usr/bin, /usr/lib, etc. directories and keeping track of the installed version. Here are the steps for building through RPM:

curl -LO https://github.com/scott-guthridge/libvna/releases/download/v0.4.0/libvna-0.4.0-1.src.rpm

sudo dnf install -y gcc libyaml-devel rpm-build
rpm install libvna-0.4.0-1.src.rpm
rpmbuild -bb rpmbuild/SPECS/libvna.spec
sudo dnf install -y rpmbuild/RPMS/*/*.rpm

sudo dnf install -y autoconf automake git libtool libusbx-devel
git clone https://github.com/scott-guthridge/n2pkvna.git
cd n2pkvna
./bootstrap
./configure
make rpm
sudo dnf install -y rpm/*/*.rpm

When I get a chance, I'll make n2pkvna provide a .src.rpm file like libvna does, then you won't even need to install autoconf or git!

[wolfgang6444]

the firmware seems to be my problem:

Bus 001 Device 008: ID 0547:100a Anchor Chips, Inc.

Do I need to upload a specific firmware? Is this different to what G8KBB is using with his XP-software?

BTW: Yes I could try the rpm, but as I will probably try to contribute, I need to be able to compile from source anyway.

[wolfgang6444]

I just realized, that I don't have /sbin/fxload

[wolfgang6444]

O.K. I had to install
sudo yum install fxload.

Now the gui is opening. I guess I need to read the doc now. Calibration is not obvious here ...

[wolfgang6444]

Apparently Rocky Linux does not read udev-rules from /usr/local/lib/udev/rules.d
I had to move 85-n2pkvna.rules to /etc/udev/rules.d/

[wolfgang6444]

O.K.,
the doc does not talk much about the GUI.
Could you maybe provide a short step-by-step instruction on how to perform the most simple measurement: S11 on a single port with SOL calibration as a kick-start?
I am using a T1-6T reflexion bridge at ADC1

many thanks for your patients,
Wolfgang

[scott-guthridge]

Calibration is not obvious here

If you go to Help / Manual, it should bring up the HTML manual. Go through the quick start guide and run through the checkoff list. Basically, you just go through the tabs in the app: First go to Setup and create a configuration for your particular hardware config. Likely, the defaults will be close if not exact. Then go to calibrate and create a calibration. Finally, go to Measure and do the measurement.

Give me feedback on what's confusing -- I tried to make it really easy. But there are a bunch of different setups for the N2PK VNA: simple reflection bridge on one port, Full S parameter test set, RF-IV, and any of those with the attenuator. And then I offered a bunch of different calibration options because libvna supports much more than simple SOLT. But if you just want the usual SOLT, select calibration type E12, and you should be set.

[scott-guthridge]

The scans are a little slow because libn2pkvna measures each point 8 times and averages. It's a pretty easy code change to have it measure just 4 (like the Windows version does by default), and that will double the speed. I plan to make it configurable with default probably at 4. Minimum possible is 2 measurements per point to measure the I and Q to get the phase, but the results are a lot more accurate with 4 because this subtracts out DC offset in the ADCs.

For prioritization purposes, let me know what missing features bother you the most: slow scans, missing Smith chart graphs, or something else.

[wolfgang6444]

Yes indeed, I did not realize the bulit in help and have just looked at the .pdf.
The help is excellent!

Just a few things I have run across in my first measurement:

1. I had to install gnuplot to get graphs.
2. For some reason my graphs just show crosses not lines (probably just a config for gnuplot?)
   3)trivial, but the instruction box asking me to start the actual measurement talks about an "open"-measurement, while I obviously have to connect the DUT.
   

As far as additional features are concerned:
-the obvious thing for a VNA is of course the SMITH-CHART
-markers would be fine
-some SCPI-interface to control the instrument remotely. (I know SCPI is an awful language, but I guess it still is the standard)

I still have to look at many of the details like RF/IV etc. - but I am very enthusiastic so far.

Really nice job!

I am not sure if we need things like modeling and the xtal-measurement. While I frequently use the xtal-measurement I feel this is too specific for a VNA-software. This is maybe better handled by a separate script controlling the VNA (e.g. via SCPI). Waht do you think?

You mentioned averaging. So far I am pleased with the performance. Have you considered the special sampling methods using phase-rotation (call it correlated double sampling) G8KBB has implemented to suppress coupling inside the instrument (increase isolation) and/or suppress harmonics.

[scott-guthridge]

• gnuplot dependency: OK I'll add that to both the README and the RPM/Debian dependencies
• crosses instead of lines: missed that because I have a ~/.gnuplot file containing "set grid" and "set style data lines". I can fix that.
• On the "Prepare VNA for open measurement": when you did your calibration, did you specify that it needed manual steps? It looks like it's prompting because of a manual step called "open".
• Smith chart: I almost got that done, but then I got myself into a rabbit hole trying to generalize it to handle points outside of the unit circle, i.e. DUT's with negative resistance.
• markers: what kind of markers do you think would be useful?
• SCPI: it's actually not that bad of an interface -- much better than GPIB pre-SCPI [have some instruments on my bench that talk the older style one and two character commands with no standardization at all]. It wouldn't be that hard to create a daemon on top of the CLI or integrated into the GUI to do that. This would be a good little project for someone to contribute.
• "correlated double sampling": I wonder if that's what the library is already doing -- could you send me a link describing that? The scan code measures with RF and LO in phase, then it steps LO through 45, 90, 135, 180, 225 and 270, adding the projections of each measurement onto the real and imaginary axes. So if there is a DC offset in the ADC, it gets subtracted out by the 180 degree opposite measurement.

I was also thinking about how you mentioned the calibration wasn't completely obvious. I think you have a point. It's probably putting too much emphasis on the error parameter model and not enough on the calibration method. Also, it currently doesn't support 'unknown' parameters, so can't do TRL and UXYZ (unknown through) methods. And it doesn't conveniently let you enter delays, capacitance and inductance polynomial fit and losses that some calibration kits specify.

[wolfgang6444]

-On the "Prepare VNA for open measurement": I need to double check
-To me Smith charts only make sense for S11 and S22 not S21 or S12. Measurements with |Sxx| > 1 are specific. For me issuing a warning and not showing these points would be fine, as long as they still show in the exported files.
-Yes you could think of many different kind of markers depending on the type of graph: manual, peak-search, tracking ...
To me one or two simple manual markers showing the related frequency and values would cover most usecases.
-O.K. I can take a look into the SCPI interface. My christmas holidays are over and I am usually quite busy, but that is not urgent ... ;-)
-I never had a look into G8KPP's software, but what you describe and what apparently is already implemented is also how I understood it:
Leave the LO-phase fixed - step the RF-phase by 2*pi/n - "rotate" the measured I and Q back numerically by mulitplying with exp(-2 * pi /n *index), -sum them all up and divided by n. That way you get rid of ADC-offsets, "self-mixing' in the mixer, and I think also some harmonics generated in the DUT, mixing with harmonics of the LO (I need to think further). You kind of rotate the phasor digitally all across the path RF-DDS up to and including the ADC.

[wolfgang6444]

-On the "Prepare VNA for open measurement": O.K. I named the b11 in the RB setup "open". That's why it showed "open" in measure. Everything fine here.
-The calibration is the key to VNA measurements. It is involved and cannot be made 'simple'. I think you have done a very good job in describing the error-terms and their correction ( I have not seen it being described that way before). Yes, a possibility to enter model-parameters for the standards would be important (at 60MHz it is probably not a "killer" if you don't have them but still). I don't know which form is most popular: I guess S-parameters or discrete (frequency dependent) components and a delay could be used.
Adding more advanced calibration techniques is always I good. I always wanted to take a closer look into TRL but haven't gone there yet.
-Also general deembedding e.g. with entered S-Parametes for the embedding system could be interesting.

-Many ideas, much work ... I know ...

BTW: I see the "attenuation"-button on some panes. I am not aware the N2PK-hardware can do that standalone, so I guess it would require extra hardware?

[scott-guthridge]

Latest code on the "development" branch should fix the plots to use lines instead of points and to show the grid. It also fixes the missing dependency on gnuplot.

Back on the "Prepare VNA for open measurement", in the majority of setups, there should be no "manual steps". Let me give examples of when those do and don't apply.

First, a case where you would use them: suppose that you have a manual switch (not a relay) that swaps the two VNA ports, and you suspect that switch is not perfect and want to include corrections for it in the calibration. Then you would set up two manual steps, e.g. (1) "Measure S11 and S21", (2) "Measure S12 and S22". These would prompt you to turn the switch to the right position before each measurement is made. Edit: if that switch is between the reflection bridge and DUT (as opposed to being on RF out behind two reflection bridges), then you have to use E12 error terms to correct for it.

Second, an example where you would NOT use any manual steps: you have a VNA with reflection bridge on port 1 and just the detector on port 2 that measures S11 and S21 only. Here, to measure all four S parameters, you do need to manually switch the two probes on the DUT. But these are NOT the manual steps the Setup is talking about. In this case, the number of manual steps should be set to zero. The measure screen will still prompt you to swap the ports, but this swap doesn't need to be included in the calibration because any connection non-repeatability you might have won't be consistent anyway, so you don't really want that in the calibration.

[wolfgang6444]

-Yes, the latest upload in the development branch connects the points in the graph.
-I got the point about manual steps in the setup: These are only required if the setup as seen from the VNA changes during the measurement sequence like with manual switches. Just swapping connectors does not (reproducably) change the setup - hence no manual steps.
-What is the function of the "use references" button in the setup?
-Coming back to the "correlated double sampling" or rather "correlated multi sampling": I think it is really important NOT to change the phase of the LO. If you change both the RF- and LO-phase by the same amount, nothing much changes for the DUT nor the mixers, nor the ADC. It is essentially just a shift in time by a couple of ns. This may reduce some phase- or amplitude differences in the DDS, but those get canceled out by pretty much any calibration anyway.
If you change the RF phase and leave the LO-phase fixed (or vice-versa), you essentially convert the ZERO-IF architecture into a very low IF architecture. Now the harmonics of the RF-signal no longer coincide with the LO-harmonics and thus mix down to some rotating ADC-values. Only for the fundamental the ADC-results rotate at exactly the same rate as the software introduced rotation of the RF-phase - the harmonics rotate the ADC at different rates and thus get canceled out if you properly combine the ADC-values. LO-feedthrough mixes to DC (as always) but now gets canceled out by the software rotation applied to the ADC-results prior to averaging. The same holds for (DC-) Offset of the ADC.

[scott-guthridge]

I added the Smith Chart and improved plot ranges and units getting preserved when changing what to plot.

What is the function of the "use references" button in the setup?

All the those buttons (use switch, use RF-IV, use references) do is add and remove choices in the table below. If you check use-references, then the a11, a12, a21, and a22 inputs are added. Look at the "Full S-Parameters With Reference Measurements 2x2" example on the Setup page of the manual for how reference inputs could be used. I'm not sure how realistic this example is because the errors introduced by the relays would have be less than the errors the reference inputs are trying to correct. You can do it without relays, though, for a one-port VNA setup. You just need a bi-directional reflection bridge.

[correlated double sampling] ... important NOT to change the phase of the LO

I think the windows software holds the LO phase constant and rotates the RF phase. This library holds the RF phase constant and rotates the LO phase. I chose to hold RF constant, thinking that if the DUT is a really high order filter with long transient, it's better to keep the stimulus steady while measuring. I think that which one you hold constants doesn't really matter, though, because it's only the relationship between them that we're measuring.

I made a lot of code changes -- see if you can break it. If not, I'll make a new release.

[wolfgang6444]

Sorry for the late reply, I was very busy but gave it a try today.
So far it seems to work fine. Smith-Chart is really nice. I tried it with an open cable and the result is as expected.
In the Smith-Chart I get I get individual points - no lines. Not sure if that is intended. Of course a straight line between points is hardly ever a correct interplotaion in a Smith-Chart, but it would at least give and indication of the frequency trend, which might otherwise be hard to see at first glance.
I agree, that it is better to rotate the LO-phase than the RF phase. Is the number of phase steps configurable, or is that hardcoded?
Are markers already included? If yes, how do I activate them?
Many thanks

Wolfgang,

[scott-guthridge]

I've seen a lot of examples of Smith charts with just points and others with connected points. It would be easy to change it to lines and points -- let me try it and see how that looks.

BTW, I didn't add labels on the Smith chart -- followed LibreVNA's example of no labels, which is clean-looking -- but could add them if needed. The Smith grid circles of constant resistance are at: 0, 0.2, 0.5, 1.0, 2.0, 5.0 and infinity. Circles of constant reactance are at: 0.0, ±0.2, ±0.5, ±1.0, ±2.0 and ±5.0. And if you expand the dimensions outside of the unit circle, then you'll see negative resistance circles of -0.5 and -0.2 to the left, -1 on the vertical line tangent to the right hand edge of the unit circle, and -5.0 and -2.0 to the right. Handling the negative resistance case wasn't that hard, as it turns out, though you have to understand the theory a bit to understand why they're where they are.

Number of phase steps is currently hard-coded at 8, but I'd like to make that configurable.

I haven't added markers. I'm not sure what I should mark, really -- especially on the Smith chart. Can you show some examples of what you'd like?