Testing return loss of 24 GHz antennas and transitions


Neil Smith G4DBN
 

I'm just embarking on machining some coax-input feedhorns and waveguide transitions to use with these Wavelab units on 24 GHz, but I don't have anything I can use to check the return loss of the transitions or horns. I can make up a 20 dB cross-coupler in waveguide, but while I can sort-of use that with my spectrum analyser as detector, assuming my best 50 ohm load looks vaguely resistive at that frequency, I can't think how to check a coax-fed feedhorn.

Without access to a 24GHz-capable VNA and the right connectors, obviously.

It is just a question of modelling it to get close and then trimming for maximum smoke on TX and use cold sky versus moon/sun noise to verify that it is working right?      Is there some magic bit of testgear I can make or obtain to get the thing tuned spot on and remove any uncertainty?

I'm thinking of using a 2.92 mm connector with a hermetic seal feedthru and a solid silver probe, because why not. They seem to be cheaper than 3.5mm or high-end SMAs. That much Sterling silver is about 10p.

It would be SO much easier with a waveguide launch PA/LNA and w/g relay, wouldn't it...

Neil G4DBN


Mark GM4ISM
 

Neil,

Amongst the things in my junkbox  are some nice Flann broadwall WG20 directional couplers and once in the WG domain I can measure return loss to a reasonable accuracy (Directivity  is 35dBmin)

Even with my best commercial coax transitions, the measurement uncertainly becomes  interesting.

 I have a 26GHz 50 ohm calibrated load  which helps  but without a proper coaxial S Parameter test set and Scalar  analyser,  I struggle.

The only mechanism I have left but not yet explored at 24GHz, is wobbulation.

A non-directional coupler  (signal sampler) and detector with a resolution of 0.1dB is all that needed. Absolute accuracy is unimportant

 You either change the sample point along the transmission line ( difficult in coax) or sweep over a suitable frequency range

 The resultant  amplitude vs frequency ( or distance) response will  vary,   FFT analysis will give you DTF / TDR*  information / antenna return loss information. You can do 'qualitative'  FFT anaylsis in your head :)  You wont be able to  determine accurate DTF  info this way but you can optimise antennas.  The technique does not lend itself well to very short transmission lines

 To calibrate this, ideally  you need a good know 50 Ohm load. With that I can probably help by 'matching' my good reference to a secondary load with a similar response.

 I believe I could achieve a secondary and traceable measurement system with a directivity of at least 25dB.  Not great but adequate.

On the subject of silver components, if you need silver, i have plenty and can cast into large ish  rough shapes suitable for your machining  skills, (Delft Clay casting method)

 I have about 300g of silver 'in stock' for casting, so there is scope for experimentation.

 There are few component that would benefit greatly from being made of silver  but if you find a need, give me a call.

 I can cast gold too of course but that is really not very cost effective. I don't have much surplus 18k + gold either :)

Good luck..  This is a field that few amateurs master with reasonable accuracy but it is possible  at reasonable cost, with effort, even at 24GHz.

Mark GM4ISM

* Note that the terms DTF (Distance to fault) and TDR (Time domain Reflectometry) are often interchangeable but there are are differences ,  associated with the way the data is measured. Measurement  parameters are  important  for these techniques and it is easy to  get things wrong and 'contaminate' results with errors  that are not obvious.

On 19/03/2021 20:09, Neil Smith G4DBN wrote:
I'm just embarking on machining some coax-input feedhorns and waveguide transitions to use with these Wavelab units on 24 GHz, but I don't have anything I can use to check the return loss of the transitions or horns. I can make up a 20 dB cross-coupler in waveguide, but while I can sort-of use that with my spectrum analyser as detector, assuming my best 50 ohm load looks vaguely resistive at that frequency, I can't think how to check a coax-fed feedhorn.

Without access to a 24GHz-capable VNA and the right connectors, obviously.

It is just a question of modelling it to get close and then trimming for maximum smoke on TX and use cold sky versus moon/sun noise to verify that it is working right?      Is there some magic bit of testgear I can make or obtain to get the thing tuned spot on and remove any uncertainty?

I'm thinking of using a 2.92 mm connector with a hermetic seal feedthru and a solid silver probe, because why not. They seem to be cheaper than 3.5mm or high-end SMAs. That much Sterling silver is about 10p.

It would be SO much easier with a waveguide launch PA/LNA and w/g relay, wouldn't it...

Neil G4DBN




--
This email has been checked for viruses by AVG.
https://www.avg.com


Neil Smith G4DBN
 

I have a 20 to 26 GHz YIG oscillator that makes a stonking great big signal and can sweep across a wide range. I made a sampler with two sections of copper semirigid with the sides of the sheath machined away, rejoined and soldered. That had enough coupling to get a decent signal level into my specan.  Just wondering if I could make a better version of that with a suitable terminating load, or perhaps machine up a coupler body and use some of those spiffy threaded SMAs I got from Paul G8AQA to make up an teensy air dielectric  directional coupler. With a few screws, I could probably get reasonable directivity. I have a couple of SMA loads that are rated to 26.5 GHz, although I think one of them is a 3.5.  I keep it for Sunday Best.

I found a nice SMA 26.5 GHz directional coupler on ebay but it is $$$$$$$$$.  I'm not looking at instrument-quality here, so maybe I'll dig out the twinned coax coupler and see if it has any directivity with the decent load on it.

Easy enough to machine up a decent waveguide coupler, but I see there are a few for sale as well if I'm feeling lazy.

The thought on using silver for the probe is that it's a lot nicer to machine than copper, although I don't know if I'd get away with a thermal shrink-fit or if I'd have to use soft solder. It would be fun to machine up some 134 or 241 GHz feedhorns from a cast silver master, although once I get my explosives precursors and poisons licence, the first job after a bit of anodising will be electroforming some corrugated SHF horns on an aluminium machined mandrels, then dissolving the aluminium away.

The other thought I had was whether a slotted line sampler is feasible in WR42. Slitting a 0.5mm slot into the broad face of a bit of waveguide after thinning the top, then making a carriage and slider arrangement would be a bit of fun to make.  Just need a suitable detector. My HP33330B runs out of puff at around 20 GHz. Directional coupler seems a lot simpler.

I'm trying to work out the downside of just tuning the transition or horn probe for maximum transmitted signal, then verifying sun/moon noise matches expectations

Neil G4DBN

On 19/03/2021 21:34, Mark GM4ISM via groups.io wrote:
Neil,

Amongst the things in my junkbox  are some nice Flann broadwall WG20 directional couplers and once in the WG domain I can measure return loss to a reasonable accuracy (Directivity  is 35dBmin)

Even with my best commercial coax transitions, the measurement uncertainly becomes  interesting.

 I have a 26GHz 50 ohm calibrated load  which helps  but without a proper coaxial S Parameter test set and Scalar  analyser,  I struggle.

The only mechanism I have left but not yet explored at 24GHz, is wobbulation.

A non-directional coupler  (signal sampler) and detector with a resolution of 0.1dB is all that needed. Absolute accuracy is unimportant

 You either change the sample point along the transmission line ( difficult in coax) or sweep over a suitable frequency range

 The resultant  amplitude vs frequency ( or distance) response will  vary,   FFT analysis will give you DTF / TDR*  information / antenna return loss information. You can do 'qualitative'  FFT anaylsis in your head :)  You wont be able to  determine accurate DTF  info this way but you can optimise antennas.  The technique does not lend itself well to very short transmission lines

 To calibrate this, ideally  you need a good know 50 Ohm load. With that I can probably help by 'matching' my good reference to a secondary load with a similar response.

 I believe I could achieve a secondary and traceable measurement system with a directivity of at least 25dB.  Not great but adequate.

On the subject of silver components, if you need silver, i have plenty and can cast into large ish  rough shapes suitable for your machining  skills, (Delft Clay casting method)

 I have about 300g of silver 'in stock' for casting, so there is scope for experimentation.

 There are few component that would benefit greatly from being made of silver  but if you find a need, give me a call.

 I can cast gold too of course but that is really not very cost effective. I don't have much surplus 18k + gold either :)

Good luck..  This is a field that few amateurs master with reasonable accuracy but it is possible  at reasonable cost, with effort, even at 24GHz.

Mark GM4ISM

* Note that the terms DTF (Distance to fault) and TDR (Time domain Reflectometry) are often interchangeable but there are are differences ,  associated with the way the data is measured. Measurement  parameters are  important  for these techniques and it is easy to  get things wrong and 'contaminate' results with errors  that are not obvious.





On 19/03/2021 20:09, Neil Smith G4DBN wrote:
I'm just embarking on machining some coax-input feedhorns and waveguide transitions to use with these Wavelab units on 24 GHz, but I don't have anything I can use to check the return loss of the transitions or horns. I can make up a 20 dB cross-coupler in waveguide, but while I can sort-of use that with my spectrum analyser as detector, assuming my best 50 ohm load looks vaguely resistive at that frequency, I can't think how to check a coax-fed feedhorn.

Without access to a 24GHz-capable VNA and the right connectors, obviously.

It is just a question of modelling it to get close and then trimming for maximum smoke on TX and use cold sky versus moon/sun noise to verify that it is working right?      Is there some magic bit of testgear I can make or obtain to get the thing tuned spot on and remove any uncertainty?

I'm thinking of using a 2.92 mm connector with a hermetic seal feedthru and a solid silver probe, because why not. They seem to be cheaper than 3.5mm or high-end SMAs. That much Sterling silver is about 10p.

It would be SO much easier with a waveguide launch PA/LNA and w/g relay, wouldn't it...

Neil G4DBN






Alan Beard
 

Hi Neil,

This is why I was looking at broadband waveguide to coax transitions
last year as I wanted a "no tune" design, a 23cm feed horn for my 2m dish.
Preferably with enough bandwidth to be useful at 13cm.
Also, I don't care which face the coax connector is on.

I ended up with this DUBUS design, which I built.

Also, for 3.4GHz I have a Cross-bar transition (circ 1950). But nowhere can
I find any design details ie. how broadband is it?

Alan VK2ZIW


On Fri, 19 Mar 2021 20:09:18 +0000, Neil Smith G4DBN wrote
I'm just embarking on machining some coax-input feedhorns and
waveguide transitions to use with these Wavelab units on 24 GHz, but
I don't have anything I can use to check the return loss of the
transitions or horns. I can make up a 20 dB cross-coupler in
waveguide, but while I can sort-of use that with my spectrum
analyser as detector, assuming my best 50 ohm load looks vaguely
resistive at that frequency, I can't think how to check a coax-fed feedhorn.

Without access to a 24GHz-capable VNA and the right connectors, obviously.

It is just a question of modelling it to get close and then trimming
for maximum smoke on TX and use cold sky versus moon/sun noise to
verify that it is working right?      Is there some magic bit of
testgear I can make or obtain to get the thing tuned spot on and
remove any uncertainty?

I'm thinking of using a 2.92 mm connector with a hermetic seal
feedthru and a solid silver probe, because why not. They seem to be
cheaper than
3.5mm or high-end SMAs. That much Sterling silver is about 10p.

It would be SO much easier with a waveguide launch PA/LNA and w/g
relay, wouldn't it...

Neil G4DBN


---------------------------------------------------
Alan VK2ZIW
Before the Big Bang, God, Sela.
OpenWebMail 2.53, nothing in the cloud.


Neil Smith G4DBN
 

I can certainly measure return loss up to 10 GHz in coax with a reasonable degree of precision, using some good SMA directioinal couplers or waveguide cross-couplers, and on 3.4, my NanoVNAv2plus4 and PocketVNA work very well and seem to agree with each other and with measurements using some HP directional couplers. I've made waveguide couplers up to 47 GHz, it's just that coax at 24 G is stretching me a bit.

It's simple enough to run a simulation in OpenEMS to find out the best balance of position, length, diameter and profile of the probe to get the desired return loss and bandwidth. The bandwidth does seem to come out right in the real thing, although defining the exact end of the PTFE insulation where it is trimmed does lead to a bit of uncertainty about the centre frequency.

I try to avoid tuning screws for aesthetic reasons, but that does lead to a lot of filing of the probes. I might relent, perhaps using sapphire rods mounted on screws, or threaded screws made from Rexolite, so there is no issue with contact corrosion. On some of the aluminium 3.4 GHz Super-VE4MA horns I made with welded backplates, I added a threaded tuning rod also made from aluminium.  That seemed a reasonable solution to prevent corrosion.

Using a fairly fat probe certainly helps to increase the bandwidth of the transition, at the cost of a lower peak return loss, but I guess anything over 17 dB or so is good enough?

Neil G4DBN

On 20/03/2021 04:49, Alan Beard via groups.io wrote:
Hi Neil,

This is why I was looking at broadband waveguide to coax transitions
last year as I wanted a "no tune" design, a 23cm feed horn for my 2m dish.
Preferably with enough bandwidth to be useful at 13cm.
Also, I don't care which face the coax connector is on.

I ended up with this DUBUS design, which I built.

Also, for 3.4GHz I have a Cross-bar transition (circ 1950). But nowhere can
I find any design details ie. how broadband is it?

Alan VK2ZIW


On Fri, 19 Mar 2021 20:09:18 +0000, Neil Smith G4DBN wrote
I'm just embarking on machining some coax-input feedhorns and
waveguide transitions to use with these Wavelab units on 24 GHz, but
I don't have anything I can use to check the return loss of the
transitions or horns. I can make up a 20 dB cross-coupler in
waveguide, but while I can sort-of use that with my spectrum
analyser as detector, assuming my best 50 ohm load looks vaguely
resistive at that frequency, I can't think how to check a coax-fed feedhorn.

Without access to a 24GHz-capable VNA and the right connectors, obviously.

It is just a question of modelling it to get close and then trimming
for maximum smoke on TX and use cold sky versus moon/sun noise to
verify that it is working right?      Is there some magic bit of
testgear I can make or obtain to get the thing tuned spot on and
remove any uncertainty?

I'm thinking of using a 2.92 mm connector with a hermetic seal
feedthru and a solid silver probe, because why not. They seem to be
cheaper than
3.5mm or high-end SMAs. That much Sterling silver is about 10p.

It would be SO much easier with a waveguide launch PA/LNA and w/g
relay, wouldn't it...

Neil G4DBN


Paul G8AQA
 

Hi Neil,

I have a quantity of 2mm tuning screws salvaged from 23GHz equipment. They have a locking nut and square section lockwasher. Let me know if you want any.

73
Paul G8AQA

On 20/03/2021 11:03, Neil Smith G4DBN wrote:
I can certainly measure return loss up to 10 GHz in coax with a reasonable degree of precision, using some good SMA directioinal couplers or waveguide cross-couplers, and on 3.4, my NanoVNAv2plus4 and PocketVNA work very well and seem to agree with each other and with measurements using some HP directional couplers. I've made waveguide couplers up to 47 GHz, it's just that coax at 24 G is stretching me a bit.

It's simple enough to run a simulation in OpenEMS to find out the best balance of position, length, diameter and profile of the probe to get the desired return loss and bandwidth. The bandwidth does seem to come out right in the real thing, although defining the exact end of the PTFE insulation where it is trimmed does lead to a bit of uncertainty about the centre frequency.

I try to avoid tuning screws for aesthetic reasons, but that does lead to a lot of filing of the probes. I might relent, perhaps using sapphire rods mounted on screws, or threaded screws made from Rexolite, so there is no issue with contact corrosion. On some of the aluminium 3.4 GHz Super-VE4MA horns I made with welded backplates, I added a threaded tuning rod also made from aluminium.  That seemed a reasonable solution to prevent corrosion.

Using a fairly fat probe certainly helps to increase the bandwidth of the transition, at the cost of a lower peak return loss, but I guess anything over 17 dB or so is good enough?

Neil G4DBN


On 20/03/2021 04:49, Alan Beard via groups.io wrote:
Hi Neil,

This is why I was looking at broadband waveguide to coax transitions
last year as I wanted a "no tune" design, a 23cm feed horn for my 2m dish.
Preferably with enough bandwidth to be useful at 13cm.
Also, I don't care which face the coax connector is on.

I ended up with this DUBUS design, which I built.

Also, for 3.4GHz I have a Cross-bar transition (circ 1950). But nowhere can
I find any design details ie. how broadband is it?

Alan VK2ZIW


On Fri, 19 Mar 2021 20:09:18 +0000, Neil Smith G4DBN wrote
I'm just embarking on machining some coax-input feedhorns and
waveguide transitions to use with these Wavelab units on 24 GHz, but
I don't have anything I can use to check the return loss of the
transitions or horns. I can make up a 20 dB cross-coupler in
waveguide, but while I can sort-of use that with my spectrum
analyser as detector, assuming my best 50 ohm load looks vaguely
resistive at that frequency, I can't think how to check a coax-fed feedhorn.

Without access to a 24GHz-capable VNA and the right connectors, obviously.

It is just a question of modelling it to get close and then trimming
for maximum smoke on TX and use cold sky versus moon/sun noise to
verify that it is working right?      Is there some magic bit of
testgear I can make or obtain to get the thing tuned spot on and
remove any uncertainty?

I'm thinking of using a 2.92 mm connector with a hermetic seal
feedthru and a solid silver probe, because why not. They seem to be
cheaper than
3.5mm or high-end SMAs. That much Sterling silver is about 10p.

It would be SO much easier with a waveguide launch PA/LNA and w/g
relay, wouldn't it...

Neil G4DBN










Virus-free. www.avg.com


Paul G8AQA
 

Neil,

I have 2 couplers from Macom. -17.5dB. One port terminated. 3.5mm connectors. No idea of the spec as it was probably military radar.  It could well be useful.  You are welcome to borrow one to play with. Maybe there is someone who could characterise it for us.

73
Paul G8AQA

On 20/03/2021 11:03, Neil Smith G4DBN wrote:
I can certainly measure return loss up to 10 GHz in coax with a reasonable degree of precision, using some good SMA directioinal couplers or waveguide cross-couplers, and on 3.4, my NanoVNAv2plus4 and PocketVNA work very well and seem to agree with each other and with measurements using some HP directional couplers. I've made waveguide couplers up to 47 GHz, it's just that coax at 24 G is stretching me a bit.

It's simple enough to run a simulation in OpenEMS to find out the best balance of position, length, diameter and profile of the probe to get the desired return loss and bandwidth. The bandwidth does seem to come out right in the real thing, although defining the exact end of the PTFE insulation where it is trimmed does lead to a bit of uncertainty about the centre frequency.

I try to avoid tuning screws for aesthetic reasons, but that does lead to a lot of filing of the probes. I might relent, perhaps using sapphire rods mounted on screws, or threaded screws made from Rexolite, so there is no issue with contact corrosion. On some of the aluminium 3.4 GHz Super-VE4MA horns I made with welded backplates, I added a threaded tuning rod also made from aluminium.  That seemed a reasonable solution to prevent corrosion.

Using a fairly fat probe certainly helps to increase the bandwidth of the transition, at the cost of a lower peak return loss, but I guess anything over 17 dB or so is good enough?

Neil G4DBN


On 20/03/2021 04:49, Alan Beard via groups.io wrote:
Hi Neil,

This is why I was looking at broadband waveguide to coax transitions
last year as I wanted a "no tune" design, a 23cm feed horn for my 2m dish.
Preferably with enough bandwidth to be useful at 13cm.
Also, I don't care which face the coax connector is on.

I ended up with this DUBUS design, which I built.

Also, for 3.4GHz I have a Cross-bar transition (circ 1950). But nowhere can
I find any design details ie. how broadband is it?

Alan VK2ZIW


On Fri, 19 Mar 2021 20:09:18 +0000, Neil Smith G4DBN wrote
I'm just embarking on machining some coax-input feedhorns and
waveguide transitions to use with these Wavelab units on 24 GHz, but
I don't have anything I can use to check the return loss of the
transitions or horns. I can make up a 20 dB cross-coupler in
waveguide, but while I can sort-of use that with my spectrum
analyser as detector, assuming my best 50 ohm load looks vaguely
resistive at that frequency, I can't think how to check a coax-fed feedhorn.

Without access to a 24GHz-capable VNA and the right connectors, obviously.

It is just a question of modelling it to get close and then trimming
for maximum smoke on TX and use cold sky versus moon/sun noise to
verify that it is working right?      Is there some magic bit of
testgear I can make or obtain to get the thing tuned spot on and
remove any uncertainty?

I'm thinking of using a 2.92 mm connector with a hermetic seal
feedthru and a solid silver probe, because why not. They seem to be
cheaper than
3.5mm or high-end SMAs. That much Sterling silver is about 10p.

It would be SO much easier with a waveguide launch PA/LNA and w/g
relay, wouldn't it...

Neil G4DBN










Virus-free. www.avg.com


Neil Smith G4DBN
 

Ta, I had a look in my box of tuning bits and found some silver plated tuning screws with collars and locknuts. Can't recall where from but they are in sealed plastic pockets and look rather nice.  Extremely fine thread on them, looks like they are intended for tuning cavity resonators.  Bit of a waste to use them for this application. I also found a packet of screws with sapphire rod ends, ditto. Also found a few very small silver plated screws of unknown thread though. They could be 10BA, a bit over 1.6 mm diameter and look to be about 72tpi. I need to get some thread wires and a micrometer on them to be sure. They might be perfect for the 24 GHz bits I need immediately, but I'll bear in mind the offer.

Neil G4DBN

On 20/03/2021 12:00, Paul G8AQA via groups.io wrote:
Hi Neil,

I have a quantity of 2mm tuning screws salvaged from 23GHz equipment. They have a locking nut and square section lockwasher. Let me know if you want any.

73
Paul G8AQA

On 20/03/2021 11:03, Neil Smith G4DBN wrote:
I can certainly measure return loss up to 10 GHz in coax with a reasonable degree of precision, using some good SMA directioinal couplers or waveguide cross-couplers, and on 3.4, my NanoVNAv2plus4 and PocketVNA work very well and seem to agree with each other and with measurements using some HP directional couplers. I've made waveguide couplers up to 47 GHz, it's just that coax at 24 G is stretching me a bit.

It's simple enough to run a simulation in OpenEMS to find out the best balance of position, length, diameter and profile of the probe to get the desired return loss and bandwidth. The bandwidth does seem to come out right in the real thing, although defining the exact end of the PTFE insulation where it is trimmed does lead to a bit of uncertainty about the centre frequency.

I try to avoid tuning screws for aesthetic reasons, but that does lead to a lot of filing of the probes. I might relent, perhaps using sapphire rods mounted on screws, or threaded screws made from Rexolite, so there is no issue with contact corrosion. On some of the aluminium 3.4 GHz Super-VE4MA horns I made with welded backplates, I added a threaded tuning rod also made from aluminium.  That seemed a reasonable solution to prevent corrosion.

Using a fairly fat probe certainly helps to increase the bandwidth of the transition, at the cost of a lower peak return loss, but I guess anything over 17 dB or so is good enough?

Neil G4DBN


On 20/03/2021 04:49, Alan Beard via groups.io wrote:
Hi Neil,

This is why I was looking at broadband waveguide to coax transitions
last year as I wanted a "no tune" design, a 23cm feed horn for my 2m dish.
Preferably with enough bandwidth to be useful at 13cm.
Also, I don't care which face the coax connector is on.

I ended up with this DUBUS design, which I built.

Also, for 3.4GHz I have a Cross-bar transition (circ 1950). But nowhere can
I find any design details ie. how broadband is it?

Alan VK2ZIW


On Fri, 19 Mar 2021 20:09:18 +0000, Neil Smith G4DBN wrote
I'm just embarking on machining some coax-input feedhorns and
waveguide transitions to use with these Wavelab units on 24 GHz, but
I don't have anything I can use to check the return loss of the
transitions or horns. I can make up a 20 dB cross-coupler in
waveguide, but while I can sort-of use that with my spectrum
analyser as detector, assuming my best 50 ohm load looks vaguely
resistive at that frequency, I can't think how to check a coax-fed feedhorn.

Without access to a 24GHz-capable VNA and the right connectors, obviously.

It is just a question of modelling it to get close and then trimming
for maximum smoke on TX and use cold sky versus moon/sun noise to
verify that it is working right?      Is there some magic bit of
testgear I can make or obtain to get the thing tuned spot on and
remove any uncertainty?

I'm thinking of using a 2.92 mm connector with a hermetic seal
feedthru and a solid silver probe, because why not. They seem to be
cheaper than
3.5mm or high-end SMAs. That much Sterling silver is about 10p.

It would be SO much easier with a waveguide launch PA/LNA and w/g
relay, wouldn't it...

Neil G4DBN