Round to Square (Waveguides) ?

[Darktrax]

Often there is a round waveguide, such as from a circular feed horn.
It needs to join to square kit, like Polarizers, or Ortho-Mode Tranducers.

I already know from trying it out that the square can be set right up against the round, seemingly with no significant loss.
My question is, is it better to have the whole area of the circular waveguide fit inside the square, or should the whole square up to the corners fit inside the circle?

The one I tried at 8.5 GHz had a little of the corners of the square cut off by the circle, and a little of the circle cut off by the sides of square.
It works - but it looks wrong. It would be nice to know what is the right way to do this.

 

[jiripolivka]

Check the old book "Harvey: MIcrowave engineering". Transitions between various waveguides depend MAINLY on the modes of propagation in eachm one. Only knowing them you can figure out how best to make the transition.

A circular feed to a rectangular waveguide like you can see in satellite antennas can sometimes be joined with a gradual transition sometimes even without it if your signal bandwidth is narrow and power low. The best such transitions use electro-formed structure gradually changing the shape from circle to rectangle. They are expensive, too.

 

[Darktrax]

Thanks jiripolivka

Exactly because of expense, and simply how hard it is to find any transitions, I simply bolted the flanges together. They fitted, but looking through from each end, I could see the corners of the square cut off, and from the other side, four small segments of the circle slightly blocked by the sides of the square.

This did not seem to degrade the performance in any way. The joint was in a critical place, at the output of a circular horn feed into a polarizer, before a transmit-reject filter and low noise amplifier. In transmit mode (100W) using the other polarization, there did not seem to be a problem.

When the whole thing is "just working", and the assembly is very high up and takes a long time to align and verify, there is limited value, (or even choice!) in questioning the circle-square join.

I am as surprised as anybody that it works as well as it does. There must be thousands of these out there. Either I am missing something, or it is the best kept secret in the microwave industry? Can it be true that everybody simply bolts these together, and they all "just know" it that just does not matter?

 

[jiripolivka]

Thanks jiripolivka

Exactly because of expense, and simply how hard it is to find any transitions, I simply bolted the flanges together. They fitted, but looking through from each end, I could see the corners of the square cut off, and from the other side, four small segments of the circle slightly blocked by the sides of the square.

This did not seem to degrade the performance in any way. The joint was in a critical place, at the output of a circular horn feed into a polarizer, before a transmit-reject filter and low noise amplifier. In transmit mode (100W) using the other polarization, there did not seem to be a problem.

When the whole thing is "just working", and the assembly is very high up and takes a long time to align and verify, there is limited value, (or even choice!) in questioning the circle-square join.

I am as surprised as anybody that it works as well as it does. There must be thousands of these out there. Either I am missing something, or it is the best kept secret in the microwave industry? Can it be true that everybody simply bolts these together, and they all "just know" it that just does not matter?

The real industry does operate often in the way you detected. Some cases should be analyzed exactly starting from Maxwell's equations, some can work by trial and error. Keep your finding as your know-how. Make sure there is no leakage and no poor SWR in-band. Be proud of your achievement!

 

[Bob60]

China`s companies made cheap staff. US company ask 15K backs Chinas significantly less. But you can made, for instance, own septum adapter with two ports and isolation up to -50dB. Good staff, is it .

 

[davenn]

looks like most people have forgotten the ways of us ol' microwavers did for a rect. to round transistion ;-)
shape a wooden (preferable to metal) dowel to the size of your rectangular WG, and taper it down.

warm the copper circular WG and gently hammer the tapered dowel into the end of the round WG
keep going till you produce the appropriate sized rectangular opening on the WG.
Mount a retangular WG flange ... DONE

a VERY old, tried and true method !
cheers
Dave

 

[Bob60]

Progress, progress how much soldiers do we need to change damage bulb.
It looks a lot.

 

[davenn]

Progress, progress how much soldiers do we need to change damage bulb.
It looks a lot.

BOB
not sure what that had to do with round to rectangular WG ???

D

 

[Bob60]

Never mind, simply thinking happens in many languages. Probably at engl sound funny.
As for waveguides it depends and up to you. Circular, elliptic, polarization has some benefits. Otho mode as well, simply I`m quite confident. It`s my job. I`m playing with these staff. Spent a lot of money and I was close to made own design but finally found much more elegant solution.

 

[Darktrax]

looks like most people have forgotten the ways of us ol' microwavers did for a rect. to round transistion ;-)
shape a wooden (preferable to metal) dowel to the size of your rectangular WG, and taper it down.

warm the copper circular WG and gently hammer the tapered dowel into the end of the round WG
keep going till you produce the appropriate sized rectangular opening on the WG.
Mount a retangular WG flange ... DONE

a VERY old, tried and true method !

The 22mm (or near equivalent 7/8") copper pipe, persuaded onto a rectangular-to square mandrel made from a broom handle, then soft-hammered and/or rolled makes a smooth transition, over more than a wavelength, and reflects very little.

Better performance can be had from a square transition, bolted hard up onto the circle, with a couple of matching steps getting smaller. Its an expensive thing that is overall, a bit longer.

Good performance can be had by ignoring reflection matching optimizations, and simply setting a circle next to a square, spaced about 1/4 wavelength apart. Then have a circular cone try and make it into the square, and a square-base pyramid cone try and fit through the circle. Let them intersect. It looks wrong in there. There are clearly edges that could be shaved away, but this can be made as split-block and turned.

Finally - when I compared the result with what you get when you simply put the circle right up against the square, let the corner of the square get clipped out, let the sides of the circle get knocked off, and actually build it, saving the expense of transition, it was so good that I do not feel these "transitions" are needed in a practical situation. I may be wrong. There might be situations where it makes a difference, but not for me this time.
See the casual, non-optimized, transition that I not not use in the end. shows S21 and S11


P.S.
For a (sort of) comparison, I include a picture of predicted performance of a square waveguide simply butted up to a feed-horn, square to round, and simply forget about the cut-off corners. This was intended to propagate both components of circular polarization. Ignore the resonances from other features in the assembly, and look at the S11 match. It could maybe be made more beautiful, but to be honest, with a match around -40dB over that bandwidth, is it really worth knocking ourselves out to add expensive transitions?


P.P.S
Apologies that I am beating this thread to death, but I happen to have a picture of a really classy transition. It uses a 4-section "homogeneous" matching transformer to go between WR-430 and WC-504 .
Note especially that after going to all that trouble, finally, in the end, even this beauty still clips off bits of the circle end, although, to be fair, it does land better on the rectangle end . I do not have performance details, but even if they were exceptional, I know I would not bother to use one. It looks expensive!!

 

[Bob60]

OMT ( ortomode, two polarization) is little bit complicated.

 

[Darktrax]

OMT ( ortomode, two polarization) is little bit complicated.

It can be if you let it, but it need not be.
You can use a linear polarization signal (say vertical) at the same time as another linear polarization (say horizontal), as independent signals that have nothing to do with each other, provided you keep the polarizations at right angles to each other. This kind can use a pyramid shape feed horn.

Clearly, to get such a combination down a waveguide where each wants to have the same ability to get through it, you make the waveguide (equal opportunity) square. Square can support higher modes, so you keep the transition short. Square is what gets a OMT (Ortho Mode Transducer) on the end of it. Good versions can isolate between signals better than -30dB and more.

Circular polarization is using both linear polarizations together to carry the same signal , half-power to each, but with the phases messed with in a special way, such that one is delayed by 90° quarter wavelength. The result sums to make a polarization that not longer stays fixed, but rotates at the signal frequency. This stuff needs circular waveguides and feed horns, which, like square, also supports higher modes, and, like square, you don't give it much chance to get those going unless you deliberately intend to use such. Rotation direction depends on which one is delayed. As with linear, you can get better than -30dB crosstalk isolation between RHCP and LHCP simultaneously.

This is the difference between a OMT and a polarizer. You can use a OMT + a delay section to put together a single circular polarized result, but it is not the same as using a polarizer, where all the energy from RHCP ends up in one port, and all the LHCP ends up in the other, such as with a septum or other polarizer.

OMT extracts only the linear components, which is why each is 3dB down per port when receiving circular.
With polarizers, it is so much more convenient to have all RHCP in one port, and all LHCP in the other.

All the above means that at some point, we are joining round to square. For me, the main surprise was how hard it is to find a OMT to purchase as a separate item, instead of it being part of a bigger kit. Also, how often round is just stuck straight up against square somewheres - and how little it matters when this is done.

 

[Bob60]

Yes I see. I told that junction from circular to rectangle to design more or less easy against to OMT.