ring coupler 1 to 4 dividing network

[yefj]

Hello , i need to split my input power into 4 then amplify then reunite them again using quadrature coupler.
I have built one as shown bellow ,as a unit its very good. but its geometic properties are probplematic because i need to supply my amplfiers in parralel.
Is there some good geometcial strtegy to construct 1 to 4 divider from these components?
Thanks.

--- Updated Nov 5, 2023 ---

UPDATE:
i tried to do the following but its not going that good.

 

[volker@muehlhaus]

Why do you want to use branchline coupler instead of wilkinson coupler?

Long ago I had designed a PA with two transistors and branchline coupler for input/output. There I just added some curves for better fit to the layout, so that input/output are along the x-axis.

For four outputs it could be routed like this:

But don't forget the phase offset between the outputs. Ports 2 and 4 in my example are in phase, port 3 is +90° relative to port2/4, port 5 is -90° relative to port 2/4.

 

[yefj]

Hello Wolker,i am following the idia shown in the diagram bellow.
I have build a splitter as shown below(port input is 3 and outputs are 2 4 5 6)
the s-param results are as shown bellow.The S5p file of the splitter is attached.
the problem is when i try to recombine the power back i get -40 dB insertion loss.
I try to connect the two halfs not in simetrical way ,still its not working.
why its not adding back?
even if you try schematics way using the attached S5p in ADS its not working.
Where did i go wrong?
Thanks.

ring_180_single_best_net52.s5p

drive.google.com

 

[volker@muehlhaus]

I'm quite sure you signals in the 4 paths are out of phase, and don't add up.
For each of the amplifier paths through the couplers, you should write down the total phase offset from the couplers.

I round your phase results a little bit and get these phases at the input of the amplifiers: 90°, 180°, 90°, 0°.
If use the same couplers on the output side, the phase offsets double: 180°, 360°, 180°, 0°
In other words, the signals from the 4 amplifiers cancel at the output.

You need to compensate the phase offsets by extra line length in some (not all!) paths, so that all paths are in phase at the combined output.

 

[FvM]

Without additional delay lines, relative output phases of the divider are 0°, two 90°, 180°. In post #3 schematic, outputs seem to be combined correctly.

But which circuit are you simulating? -40 dB sounds like almost perfect cancellaton, hard to achieve in this circuit. Or non-functional amplifier.

 

[yefj]

it seems that maybe there is a cancelation on the way back.
i dont know why maybe the combiner and divider are not suppsed to be identical?

--- Updated Nov 5, 2023 ---

UPDATE:
The answer is that in the mirror each sub coupler need its input legs to be switched(except the isolation)
I'll be happy to understand will for splitting and recombining on the recombining part we need to flip legs?
Thanks.

 

[volker@muehlhaus]

dont know why maybe the combiner and divider are not suppsed to be identical?

This was your choice - you use Branchline coupler which have 90° offset between the outputs. You finally need to do your homework and calculate the phase for each branch/path with cascaded couplers!!!!

--- Updated Nov 6, 2023 ---

In post #3 schematic, outputs seem to be combined correctly.

But if we put two of these coupler configurations back to back, so that 4 amplifiers can be inserted, they do not combine with correct phase. They actally cancel at the output.

I have attached such a configuration for my 1.9 GHz couplers, which work perfect fine as long as we don't combine them like this. You can see that output signal cancels "very well".

--- Updated Nov 6, 2023 ---

Appendix to my example above: when looking at all paths, we can see that power goes to the termination resistors in the output stage (ports 5 and 6).

 

[FvM]

Depends on what you consider back-to-back. Post #3 circuit is connected so that all pathes add in phase. Like sketched below.

 

[volker@muehlhaus]

Thank you FvM!

I'm not sure how to implement that in layout, and still wonder why a branchline coupler (with phase offset and resulting layout difficulties) was used here. Isn't a wilkinson divider solution much nicer here, where output signals are in phase?

--- Updated Nov 6, 2023 ---

Ok, I figured out a layout solution without crossover that works:

 

[biff44]

i am not a big fan of ring hybrid couplers. for one thing, they are very narrow band, and as such a minor error in physical length may cause a big error in power split and phase tracking.

it seems like branch line couplers are a little big easier to model, not sure why. In any event, if i was forced to use a ring type hybrid coupler, i would def use an electromagnetic field solver, and keep the substrate thickness small so i did not run into aspect ratio issues and multiple modes happening

 

[volker@muehlhaus]

i am not a big fan of ring hybrid couplers.

What we discussed here are normal branchline couplers implemented with curved lines. These are not ring hybrid couplers (rat race couplers)!