Experimenting with waveguides using the LiteVNA

[joeqsmith]

Bent up the two ends and tacked (solder) it together.  I used some foil tape to seal it for now.  Yellow showing S21 with calibration, Red with an open.   Time to drill some more holes.

[coppercone2]

you can cut a v groove into it and bend it and then seal that with solder to get a right angle without the bending radius but you need a 90 degree endmill, I see click spring do that, but he uses fairly thick brass

[joeqsmith]

With it being a mm thick it may prove difficult to cut such a V.  I did try bending it tighter and eventually used a peen hammer.  Still no signs of cracking.  Compared with the PCB and fillet, I'm not sure we are any worse off.   

Interesting paper "Mismatch Caused by Waveguide Tolerances, Corner Radii, and Flange Misalignment"
https://library.nrao.edu/public/memos/edtn/EDTN_215.pdf

I tried to find Brady's paper he references from 1969 but no luck yet. 

Finished up the assembly and had to resort to propane  to preheat the brass to solder the seams.   No problem really. 

[coppercone2]

For milling thin material I think I would take a thick slab of metal and cut a slot in it that the cutter can fit through then clamp it down tight over the brass sheet so its held down very well and mill with a vacuum. It might work with a router

You need to level everything etc, probobly pain in the ass.

[joeqsmith]

May be a pain in the ass but for me the question is would it be worth it?

It would be interesting to compare a set of high quality transitions and waveguide with what I have shown to get some idea how much it would improve things.   Do you have any feel for this?   Do you own a Lite VNA and have some good WR90 bits that you could try and measure?  If you have a VNA better than the Lite and could run them on that, that would also be very helpful as a comparison.   

I've shown the noise floor of the Lite compared with my old Agilent.    If I run an SOLT with my LiteVNA torquing each connection, then attach three different thru adapters you can see the error band is pretty wide as we move beyond 8GHz. 

[joeqsmith]

Made some progress with it today.

[coppercone2]

I might have a section of brass WR90 I don't know what dimension it is. The one I screwed up trying to electropolish.

 I would need the coaxial adapter though.

I think I read some numbers somewhere that its 1-5% difference with the corner stuff, or maybe it was a book. I think the more rounder it is, the more propagation you get in the circular waveguide mode, which I think is more broad band, so you get like a less band pass feature from the wave guide.

[joeqsmith]

Would be great to have a sanity check.  Here's another paper that talks about the effects of the corner radii. 

https://iopscience.iop.org/article/10.1088/1681-7575/abd371

I look from time to time for used bits.  For $130 you can own this nice transition to use for your own testing.  Surprised to see the damage to the inside.   

https://www.ebay.com/itm/294867475982?_trkparms=amclksrc%3DITM%26aid%3D111001%26algo%3DREC.SEED%26ao%3D1%26asc%3D20180816085401%26meid%3De581e7ae342548db84848ee010d10aa7%26pid%3D100970%26rk%3D1%26rkt%3D10%26sd%3D294867475982%26itm%3D294867475982%26pmt%3D0%26noa%3D1%26pg%3D2380057%26brand%3DUnbranded&_trksid=p2380057.c100970.m5481&_trkparms=pageci%3Ae044fd31-7e59-11ed-b983-dafb5026ee95%7Cparentrq%3A22324c151850aa7105adb607fffebce2%7Ciid%3A1

[joeqsmith]

Here's another paper that discusses the corner radii.  Fig. 4 showing effects of radius vs return loss.  Mine is bent with 1-2mm inside. 

https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2016RS006027

[joeqsmith]

Note these production bits have an even more dramatic radius.   

Nice free haul.  Would like to have that adjustable short. 

[coppercone2]

dear lord, maybe for 35$

that is the wave guide that was being dragged behind plane it fell off of when it got stuck in a cargo net that ripped. You gotta de-rate the damaged hardware you know?? I love how there is a much nicer condition one for the same price

edit: remembered end launchers never had favorable prices

I don't have money to get into wave guide right now. I got too much other stuff. I need to unfuck the microwave equipment too like replace fans etc so I can stand being around it.

[coppercone2]

I don't think I will have a happy chirstmas and new year if I do this now. Maybe q1 2023

I was feeling great about it Q2 2022 but then I repaired too many things and i still don't want to look at the VNA etc.


when you get too heavy into broken HP

[joeqsmith]

If that were in my stash, I would gladly sell it to you for $35 and include free shipping.   No returns.   

[joeqsmith]

Here is my $20 worth of hobby brass and two decent SMA connectors running on both my old Agilent and the LiteVNA (both calibrated).   Note the Lite's frequency shift and dip in the passband.  Still for a $120 VNA that fits in your shirt pocket,  the fact it can show anything remotely close to the PNA impresses me.   I'm easily amused. 

With Dislord's latest changes to the Lite's firmware, I no longer have to be careful about the setup.  It just works as expected.   

[joeqsmith]

This is my friend Shrimpshowromma's LiteVNA that I have abused.  There are no TVSs on the front end.  The IF gain has been bumped.  Bypass has changed along with some other values. 

I used the included cables and cal standards supplied.  Note that I have modified the included short and open to add an end cap to allow me to use a wrench with both.  I used the load that I sorted (same part I used to cal the other Lite).   

Looking at S21 with a thru, the error window is improved and when looking at the filter, we no longer see the droop in the passband.  This VNA originally had higher noise than mine.  I'm surprised that it's still as good as it is. 

[joeqsmith]

Shown is a PCB version of the filter next to the hobby brass one.  Note the distance from coax transition to post on the PCB is no where near what is called for.   Oddly there is little difference between the two filters. 

The Lite was set to 1kHz IF.  I set the PNA to 1k and 150Hz to show the difference.   Both VNA's were then caled for 7.8 to 8.6G to zoom into the filter. 

[joeqsmith]

Looking at the LiteVNA groups.io,  there is a thread where people were experimenting to 15GHz before limiting it to 9.3G.    Shown is my last attempt of a making a bandpass filter.   No tuning this time.   

https://groups.io/g/liteVNA/topic/first_tests_above_6_3ghz/87993520?p=Created%2C%2C%2C20%2C2%2C20%2C0%3A%3Arecentpostdate%2Fsticky%2C%2C%2C20%2C2%2C20%2C87993520%2Cprevid%252525252525253D1643508180797191034%2Cnextid%252525252525253D1640273623127907015&next=1

[joeqsmith]

Of course to test that last waveguide filter with the Lite requires solving a few problems.  Even if Dislord were to extend the harmonic mode beyond 9.3GHz, the coupling and noise are poor.  Attached showing noise floor of the low cost VNAs along with my PNA.  The LibreVNA data was provided (JAN).   From 9 to 9.3G, the Lite is getting much worse and I am guessing the felt this was really the point where it was no longer useful.   

The other problem with that last waveguide is you can see how close the posts are.  It's center frequency is outside of the frequency range of the Lite.   10gGhz is data that someone posted on link I previously provided looking at extended frequencies.   All of the waveguides I have shown roughly follow the WR90 standard which has a cutoff of 6.557GHz.  Because of the limitations of the LiteVNA, we have been testing below the bandwidth ratio (8.2GHz). 

From Microwave Journal Frequency Matters:

Most rectangular waveguides are specified with a 2:1 width to height ratio to achieve a maximum bandwidth ratio of 2:1, i.e., the ratio between the maximum frequency and minimum cutoff frequency. This way, maximum power is carried by the waveguide before microwave breakdown, dielectric breakdown or multipaction can occur. This differs from a circular waveguide since it can only propagate a maximum bandwidth ratio of 1.3601:1, which is the maximum single mode frequency to minimum cutoff frequency. The recommended frequencies of operation for a rectangular waveguide are 30 percent higher than the cutoff frequency and 5 percent below the cutoff frequency of the next higher mode.

It's a tradeoff and because this thread is just for the fun of it we can experiment as we like.  I don't think you will find waveguides made from PCB in the wild nor people trying to use a $120 VNA to measure them.

[joeqsmith]

To get around this hurdle, we will need to use the LiteVNA in a region where it has better performance.  We can then upconvert that signal to something more useful for experimenting with our waveguides.  We can then downconvert that signal to something the LiteVNA can measure.   For S21, we can do this directly or for S11, we can add a coupler.   Of course the software would need to support this.   In 2003, I made such a system to extend my first VNA beyond 3GHz.   Results were poor at best as my 20 year older younger self wasn't very smart.  Not much has changed.     

Because the software I wrote for the LiteVNA is basically this this same software it has always supported adding mixers.  A few of you may have seen where I demo'ed this with the original NanoVNA.  I think I was somewhere in the GHz range with that.

[joeqsmith]

Over the last few days, I have been working on a simple converter for the LiteVNA.  I plan to use my old YIG for the local oscillator and the rest is just splitters, mixers, amplifiers and filters...

To date, Copper Mountain Tech, Keysight, Rohde Schwarz and Anritsu haven't knocked on my door offering to donate any equipment to my channel and because my home hobby budget prevents me from buying any high end equipment, we are pretty limited on how to test it. 

Thanks to members of this group,  I was able to extend the frequency range of my old PNA to 9GHz. 
https://www.eevblog.com/forum/testgear/agilent-e8357a/
The standards I have are not up to the task to measure down to the hairs on my butt cheeks but for these experiments, I think we can ignore that.  Thanks to  rf-messkopf, I was able to at least characterize them to 6GHz.  As mentioned, with the extender we can only look at S21 and are limited to normalization.   

Attached showing our homemade waveguide filter running on the LiteVNA up to it's 9.3GHz limit (Brown) using the provided standards and ideal model.  Red showing our PNA with my home made characterized standards and best load I have available.    Black is the same LiteVNA with my extender, sweeping all the way to 10GHz.   

The LiteVNA is pretty decent to 4GHz.  The YIG is limited to a bit over 8Ghz, allowing us to measure over 12GHz with this setup.     

[joeqsmith]

Being lazy,   I have been using Dennis G. Sweeney WA4LPR's Waveguide Filter Synthesis Program Version 1.3 to pick values for the filters shown.

Attached are the parameters I entered for the third filter.   Rather than buy any tubing to get closer to the recommended values, I used what I had on-hand.  The BW came out very close to 150MHz.   Center is off a bit which could very well be my eyeballing the extender's frequency (too lazy to hook up the counter).   The bat ears may be from the far off tubing sizes.  The first two filters include tuning stubs which were left off.

So far, the cost of the two antennas with coax, three filters and polarizer was maybe $60.  $22 in brass bits and the rest connectors.   LiteVNA was $120.  Add $150 for a better load, home made cables and adapters.   Many of the bits for the extender had come from eBay many years ago at reasonable prices.  Not sure what it would cost today.   

Hopefully I've shown you that you can play with this technology even on a budget.   Merry Christmas.

[joeqsmith]

I would like to test the waveguides closer toward 11GHz but the noise of the LiteVNA increases.  To work around this, I plan to mix with a higher frequency and use the lower sideband.  The software already supports it.  However this was the first time I ever tried ti use it in 20 some years and discovered a software bug when creating the Touchstone files (freq must increase).  Onto version 3.14.

Another problem is I am really limited on bits I can use as a sanity check at these higher frequencies.  For example, shown are three different 3dB attenuators.  One is a Mini-Circuits VAT3 rated to 6GHz.  One is from a really low cost Chinese set.  The last was a pull from some test equipment I had scraped.   These were swept from 8GHz to 11.5GHz.     

I have some attenuators from Midwest Microwave rated speced to 18GHz.   The second picture shows stacking 10dB attenuators while sweeping from 11 to 11.9.  These are all from the same lot and we can see the phase change as we increase the length.

[joeqsmith]

It seems that many of the training kits that were used at the universities/colleges used a fixed frequency oscillator.  This kit for example uses 10.425 GHz.  Mine uses 10.514 GHz and uses the WR90 standard. 

https://tecnoedu.com/Download/56-200.pdf

Using the Lite to sweep from 8-12GHz,  with two back-to-back transitions and comparing that with adding a 0dB attenuator between the two causes a fair change to S21.   In the region where they operate the kit, its stable within 1dB. 

[joeqsmith]

Looking at the slotted line with the probe removed compared with the attenuator set open.  Both parts appear to use the same slotted waveguide sub component.   

Not as stable as I would have expected. 

[joeqsmith]

Inserting the pin into the slotted line then moving it full left vs 18mm. 

Recording S21 while moving the slide from right to left.

Finally, leaving the slide fixed and normalizing the data and letting it run for several minutes, we can see the setup is fairly stable.