Lange Coupler HELP ! Please come in if you had experience with Lange coupler design!

Hello all,

First of all, I am using HFSS to simulate this structure.

Substrate: 99.5% Alumina
Trace: Copper

I need help and advise on how I can solve this problem. I designed a Lange coupler that looks like the one below. The response looks fine when I did not add the extra 50 ohms microstrip lines on the 4 ports. However, when I change the length of the microstrip line to a very long length, The response of the coupler looks completely different going from equal coupled to undercoupled. I am expecting the coupling to be the same.

Also, the coupled port and direct port only has a phase difference of about 85 degrees. From my simulation results, I figured that as the separation between the fingers gets smaller, the phase difference is a lot closer to 90 degrees. The problem with small separation is as the separation gets smaller, the response gets way overcoupled.

I don't understand why this is happening. Please comment and discuss. This is an urgent project.

Thank you!

You are seeing interactions from your 90 degree bends and effectively have some tuning stubs hanging on your output ports, which means your Lange is not yet optimally tuned. Remove as much of the 50 ohm lines as you can (effectively de-embed the coupler), then tweak the spacing, widths and lengths until you reach the absolute minimum insertion losses and a phase difference very near 90 degrees. This should be near the optimal design. When you add 50 ohm lines, the performance should not move at all... that's when you know you've reached the optimum point (minimum internal reflections).

I had exactly this problem while designing a wilkinson power divider a few months ago... changing the lengths of my 50 ohm input lines just trashed the performance. After optimizing the divider all by itself, the 50 ohm in/out lines had zero adverse effects on the combiner's performance.

Hi,

Thank you for your comment. May I ask what simulator did you use for your wilkinson project?

Thank you!

I primarily use ADS since I do both system-level and device-level simulations, and like the ability to combine them together into hierarchical designs. All of the wilkinson sims were done with Momentum, and a fair bit of the optimization was done using parametrized layouts and look-alike EM Sim components. This allowed me to let the optimizer run iterations overnight and get me close to the optimal design, then I manually tuned and sim'ed to reach the final design.