RF power combiner

Oct 22, 2025 Last reply: 8 months ago 11 Replies

While musing over a text about transmission line transformers, combiners, splitters and so on, I came up with a power combiner that promises a bandwidth spanning from some kHz well into the GHz domain, with good isolation between input ports, and that does _not_ have a null where its transmission lines are half a wavelength long.



I could find no other examples of this arrangement. The usual combiners are wire- or coax- wound transformers, with 3 decade bandwidths in the 100kHz-100MHz ballpark, or Wilkinson things that only begin to work above half a GHz.



This design should be able to work from a few kHz up to well into the GHz domain. Anyway, for what it's worth, here it is.



Jeroen Belleman



=========================



Version 4 SHEET 1 1044 680 WIRE -432 160 -480 160 WIRE -320 160 -352 160 WIRE -224 160 -240 160 WIRE -112 160 -144 160 WIRE -64 160 -112 160 WIRE 176 160 32 160 WIRE 448 160 176 160 WIRE 688 160 544 160 WIRE -64 192 -112 192 WIRE 112 192 32 192 WIRE 240 192 112 192 WIRE 448 192 400 192 WIRE 592 192 544 192 WIRE 112 240 112 192 WIRE 176 240 176 160 WIRE 688 288 688 160 WIRE 752 288 688 288 WIRE 800 288 752 288 WIRE 896 288 880 288 WIRE -48 368 -96 368 WIRE 112 368 112 320 WIRE 112 368 48 368 WIRE 240 368 400 192 WIRE 240 368 112 368 WIRE 400 368 240 192 WIRE 464 368 400 368 WIRE 592 368 560 368 WIRE -432 400 -480 400 WIRE -320 400 -352 400 WIRE -112 400 -160 400 WIRE -48 400 -112 400 WIRE 176 400 176 320 WIRE 176 400 48 400 WIRE 320 400 176 400 WIRE 464 400 320 400 WIRE 688 400 688 288 WIRE 688 400 560 400 WIRE 320 432 320 400 WIRE 320 544 320 512 FLAG -480 400 0 FLAG -112 192 0 FLAG -480 160 0 FLAG 896 288 0 FLAG -112 400 inb FLAG -112 160 ina FLAG -96 368 0 FLAG 592 368 0 FLAG 592 192 0 FLAG 320 544 0 FLAG 752 288 sum SYMBOL tline 496 176 R0 SYMATTR InstName T2 SYMATTR Value Td=2n Z0=50 SYMBOL tline 512 384 R180 SYMATTR InstName T3 SYMATTR Value Td=2n Z0=50 SYMBOL tline 0 384 R0 SYMATTR InstName T5 SYMATTR Value Td=2n Z0=50 SYMBOL tline -16 176 R0 SYMATTR InstName T6 SYMATTR Value Td=2n Z0=50 SYMBOL res -144 384 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 50 SYMBOL res -128 144 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R2 SYMATTR Value 50 SYMBOL voltage -224 400 R90 WINDOW 0 -32 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName V1 SYMATTR Value ac 0 SYMBOL voltage -224 160 R90 WINDOW 0 -32 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName V2 SYMATTR Value ac 2 SYMBOL res 160 224 R0 SYMATTR InstName R3 SYMATTR Value 50 SYMBOL res 128 336 R180 WINDOW 0 36 76 Left 2 WINDOW 3 36 40 Left 2 SYMATTR InstName R5 SYMATTR Value 50 SYMBOL res 896 272 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R6 SYMATTR Value 25 SYMBOL bv -336 160 R90 WINDOW 0 -32 56 VBottom 2 WINDOW 3 73 153 VTop 2 SYMATTR InstName B1 SYMATTR Value V=0*exp(-0.5*pow(time/sigma-4,2)) SYMBOL bv -336 400 R90 WINDOW 0 -32 56 VBottom 2 WINDOW 3 73 153 VTop 2 SYMATTR InstName B2 SYMATTR Value V=0.99*exp(-0.5*pow(time/sigma-4,2)) SYMBOL res 336 528 R180 WINDOW 0 36 76 Left 2 WINDOW 3 36 40 Left 2 SYMATTR InstName R4 SYMATTR Value 1G TEXT -304 584 Left 2 !.ac dec 100 100k 1G TEXT -584 584 Left 2 !.param sigma 500p TEXT -40 584 Left 2 !;tran 20n TEXT -240 56 Left 2 ;Guanella In-phase combiner - Jeroen Belleman, 20251021.


<snip>

That is one weird-ass combiner, for sure. Looks like it has to be made out of coax, on account of the switcheroo in the grounds.

I like it!

Cheers

Phil Hobbs

The way I drew it was a bit clumsy. It's neater to arrange it to have the center conductors cross over instead of the screens. It should work quite well though, if constructed correctly. I got

100kHz-9GHz (-3dB) out of a similar thing, with better than 40dB of isolation between the inputs over almost the full bandwidth.

It appears I'm 38 years late though. A guy names Edwards, working for Rockwell, came up with exactly the same thing in 1988! US patent 4774481.

Oh well.

Jeroen Belleman

Why not go the whole hog and wind it with semi-rigid coax - RG405 - and use SMA soldered coax sockets. Even RG405 is 2.2mm OD so you might not be able to get in enough turns to get down to 100KHz, but it would look serious.

A useful comparison is "Coaxial AC Bridges" by Kibble and Rayner,

1984, 107 pages.

Joe

The low frequency behavior relies on the unphysical LTspice T-line model, unfortunately. In spherical-cow land, the two ends are uncoupled, like a perfect transformer with a delay.

The combiner model works even better with 2 ps delay lines. ;)

Cheers

Phil Hobbs

Of course, I'm aware of that. A practical implementation would begin to work at 100kHz, give or take depending on the number and type of magnetic toroids you'd slip over the bits of coax.

The hybrid bridge from which I derived this worked from about

30kHz, and that took lots (lots!) of ferrite toroids.

Jeroen Belleman

I read that, but it's a long time ago. Maybe I should read it again.

Jeroen Belleman

It isn't. Kibble and Rayner weren't interested in very high frequency transformers wound with coax transmission lines but rather fully shielded transformers working at rather lower frequencies. And my copy has 203 pages, 196 pages of text and seven pages of index.

When I used a transmission line transformer back in 1978, I cited a 1968 paper, Matick R.E. "Transmission line pulse transformers"Proc. IEEE volume 56, pages 47-62.

Matick was pretty comprehensive. I'm surprised that Rockwell found anything to patent.

Heh. Spice in general did that. I recall having to explain this to engineers who designed only digital logic, and were assuming that one could send unrelated signals down each conductor of shielded twisted pair cables. Such engineers were called "Digits" then.

What eventually worked was to point out that each conductor had a magnetic field due to the current flowing down it, and that voltages would thus be induced in all nearby roughly parallel conductors.

Joe

I made a more realistic coax model, using three of the Tlines.

I have a diffusion problem, driving an ITO coating over a pretty hi-K dielectric film. I was thinking about using an LTline to roughly model that case, rather than doing calculus or anything nasty like that.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

You can wrap quite a lot of RG405 around an RM14 pot core. You might have to wrap the copper outer with plumber's Teflon tape to avoid shorted turns.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required