Couplers and/or Combiners for Amplifier Modules

There are a number of ways to combine the output of the modules for more power, and they all have their advantages and disadvantages. The important design parameters are:

Here are the 3 popular methods, using 3 db couplers/combiners:

Type Advantages Disadvantages
Wilkinson
  • simple to make
  • inexpensive
  • requires the use of a high-power 100 Ohm resistor, isolated from ground
  • Isolation is poor
Sage Wireline
  • compact in size
  • Expensive ($20 per foot, minimum order quantity of 10 feet)
  • requires the use of a 50 Ohm high power resistor (10w minimum) for the isolation port
  • Difficult to work with (fragile wires, Kapton insulation

Note: I could not get this stuff to work well at 23cm. Could be I just cut it to the wrong size, but I got poor port tracking and isolation with my prototypes.
Branchline Hybrids
  • inexpensive
  • easy to make
  • requires the use of a 50 Ohm high power resistor (10w minimum) for the isolation port

I like the branchline hybrid coupler myself, and prototyped several on .031 FR4 and .030 R4350B. The ones described below are good for up to 75w combined output on .031 FR4. As I refine the design on R4350B (a lower loss microwave substrate that will take more power), I'll post that data here as well.

You can find useful information on the design of these couplers on the Microwaves101 web site here:

This first attempt was modeled with AppCad, a free program available on the Agilent Technologies web site. 

design center = 1280 MHz
35 ohm = 97x1272 (1217+55)
50 ohm = 55x1350 (1253+97)

(because I aligned the adjacent microstrip traces center-to-center, I corrected the lengths from AppCad by the width of the adjacent branch lines)

The resulting measurements indicate good isolation and port tracking. The difference in port coupling factors needed some work, though, as they were .4 db apart. Not bad, but I thought they should match up better.
So, I applied a design correction using the Microwaves101 unequal split calculator.

design center = 1280 MHz
34.74 ohm = 98.5 x 1274.5 (1216+58.5)
48.3 ohm = 58.5 x 1347.5 (1249+98.5)
50 ohm leaders = 55 x 187

(this adjusts v6 to bring the main and aux coupling factors closer together. -.3 was used in the correction formula)

The results were better, but clearly I needed more correction. And what's with the isolation curve moving down from the design center? More on that in a bit...
design center = 1280 MHz
33.3 ohm = 104 x 1279 (1212+67)
44.5 ohm = 67 x 1344 (1240+104)
50 ohm leaders = 55 x 187

(additional v6 adjustment using -1 as correction in formula to bring the main and aux coupling factors closer together)

OK, nice port tracking, and the coupling matches within a tenth of a db. But look at the isolation, it moved down even more from the design center???
After pondering this a bit, I finally figured it out; as I widened the transmission lines to adjust coupling ratios, my connection to the 4 ports became less precise (this was caused by my placement of the 50 ohm lead-ins). Essentially, this error caused the horizontal lines to become long, and the vertical ones to become short. Duh!

Once I corrected for this, all was well, and the device behaved as predicted, with excellent performance over the entire 23cm band.

Clicking the picture below will provide artwork to scale in .pdf format.