How to optimize my S-parameters in the EM-Simulation Model

[DieBemme]

Hey everyone,

for a student project of mine, I am designing a hairpin filter, with a bandwith that can be controlled via a diode. The simulation in my schematic works wonderful and shows the expected behaviour.
It may seem kinda overloaded but the part I'm mainly struggling with is the layout (see further down).

Yet, after designing the layout (leaving proper space for lumped elements), naming the ports and defining the vias, the simulation result show a much worse "chebychev-kind" response. On one part the filter attenuation is way higher, and on the other hand I dont get such a clear passband, as one peak is more dominant than the rest... I hope you can give me some tips, on what I can optimize in my design to optimize my filter response.

This is the first time I am designing a microstrip filter and had a really great time playing with all the possibilities, but thought that you guys probably have way more experience and another viewpoint... Thx in advance!

And here the pics for the layout side of design:

 

[volker@muehlhaus]

No answer to your question, but a hint for getting accurate results: for such coupled line designs, you need to enable edge mesh in Momentum settings.

EM-based optimization of planar filters is a research topic on its own. One of the guys who published a lot on efficient methods is Dan Swanson.
Here is one of his papers, this one uses Sonnet EM but you might get the idea:

https://www.sonnetsoftware.com/support/downloads/techdocs/port-tuning-a-microstrip-folded-hairpin-filter-revc.pdf

--- Updated Yesterday at 4:29 PM ---

Two more comments, in addition the edge mesh:

1) Having the open ends - where you have maximum E field - over a ground plane cutout is a big difference to the microstrip models that assume solid ground.

2) For the pins/ports on the middle of the resonator arms, make sure you have defined a port with (small) finite width. Placing a "dot" pin there is invalid, because the port will extend along the entire edge.