Click here to go to our main page on rat race couplers
Click here to go to our main page on Gysel couplers
Click here to go to our main page on combiners
New for November 2019. Updated for December 2020 with improved image quality, so you can actually read the data! And we added the full reference to Steven March's paper at the bottom of the page, like all good boys and girls do. Gallant would be proud!
In June 1968, the IEEE MTT published a short paper (a letter, actually), titled "A Wideband Stripline Hybrid Ring" by Steven March, of HRB Singer in State College, PA. His combiner is a brilliant modification (performance is much improved!) of the rat-race coupler that was (is) used to provide 180 degree outputs. By extension, the March coupler (and the rat-race) can provide in-phase outputs by shifting the input from the "delta" port to the "sum" port, like a Gysel combiner, something he did not mention in his article. If Steven March is out there somewhere, tell him we said "Good job!"
In order to appreciate the March combiner, we will compare it to a rat-race coupler. The rat race is shown below. It has a perimeter of 1.5 wavelengths, each segment is one quarter wave. The ring perimeter 'ZR" is ideally SQRT(2)*Z0, or 71 ohms in 50 ohm system. The ports are numbered clockwise, starting from the lower left.
We have never been able to figure out who invented the rat-race coupler. Perhaps it was Rizzo the Rat, shown below being electrocuted.
The March hybrid is shown below. The lower three sections have been replaced with a 3dB coupled-line coupler, with two of the ports shorted (pay attention to which ports, it matters). The perimeter (or diameter if you lay it out in a circle) is reduced 33%. ZR for the coupled line is ideally SQRT(2)*Z0; depending on design rules you may have trouble reaching this value, but life is analog and full of compromises. All of the RF inputs are DC grounded, which could be a good thing or a bad thing depending on your design. The ports are numbered clockwise, starting from the lower left. Port 1 is the "delta" port (180-degree coupler) and port 2 is the "sum" port (0 degree, or Gysel, coupler). The even and odd mode impedances are 170.7 and 29.29 ohms, which represent 3.01 dB coupling (perfectly even power split) in a SQRT(2)*50 ohm system, For more information in calculating coupler even and odd mode impedances, visit this Microwaves101 page on coupled-line couplers in Microwave Office or ADS.
The key to the March coupler is that shorted coupler. The reflection coefficient of a short circuit is on the left side of the Smith chart, so it provides 180 degree phase shift at all frequencies. This is sometimes called a phase inverter. It behaves much better than a 180 degree line that is used to shift phase in a rat-race. At F0, a 180 line provides 180 phase shift, but at F0/2 it is only 90 degrees long. That short circuit is 180 degrees until the cows come home!
First, let's look at port match (also agnostic to when the coupler is used for 180 or 0 degree power split). Below is the rat-race. We did not plot ports that would be the same response by symmetry. Looks like it provides 30% bandwidth (8.5 to 11.5 GHz) at 20 dB return loss.
Below is the March combiner. It is slightly improved, to 40% bandwidth for 20 dB return loss. But this is not the real measure its higher capabilities.
Not let's look at the isolation of each design. Port 3 is isolated from port 1, and port 4 is isolated from port 2. Below, the rat race combiner has a deep isolation null at center frequency, and provides 20 dB isolation over less than 40% bandwidth (~8 to 12 GHz).
The March hybrid provides 20 dB isolation at all frequencies, until it starts moding (when the substrate height supports an additional mode, a topic for another day). We weren't lying when we said the March coupler provides a major improvement!
OK, let's hook up some 180 combiners to the delta port (port 1). The rat-race has an issue with the "coupled" port (port 2) having a different response than the "thru" port (port 4). It is really a narrow band network, but designers often extent the bandwidth by over-coupling it to increase the range where the amplitudes are within a certain window (like within one dB).
Now comes the light bulb moment. Check out the March combiner. There is barely any difference between the two transmission coefficients. The response is flat, and has much more bandwidth. As Paul Teutel would say, it's so good it's ridiculous.
Now let's hook up 0 degree dividers (the Gysel) to the sum port (port 2). Below, the rat-race/Gysel shows a narrowband response, like it's 180 ring hybrid brother.
And here again, the March coupler shows superior, broadband response.
Last, let's compare the phase response of the four couplers. Below we compare the 180 degree hybrids. Here, the phase of the two output signals have been subtracted from each other. The Y-axis increment of 22.5 degrees is useful as it is 360 degrees divided by 16, of the minimum phase bit of a four-bit digital phase shifter. If we consider 11 degrees (360/32) as the figure of merit, the rat-race works over 40% bandwidth, and the March works over all frequencies.
Here is the phase response of the 0 degree hybrids. The relative bandwidths of the two designs are similar to the 180 case, and March wins again.
In conclusion, the March hybrid offers a huge improvement over the original rat-race/Gysel combiner. It is 33% smaller as well.
If you have a design that uses the March combiner, please contribute an image to this page!!
References
Steven March, "A Wideband Stripline Hybrid Ring" by Steven March, IEEE Transactions on Microwave Theory and Techniques, vol. 16, issue 6, pp. 361-361, June 1968.