edge coupled bandpass filter

I'm designing Microstrip BPF now and I?ve got the same problem as in doc. h**p://www.ansoft.com/hfworkshop04/2004_HFSS_Users_Workshop_Bernard_Schmanski.pdf Does anybody know what kind of ?seeded mesh? was used in this example at site 16? There is very good distribution of tetrahedra: a lot of small ones near the irregularity and bigger far from irregularity. When I try to make something as at slide 16, my results anyway remain like at slide 9. If simply decrease length of the edge of tetrahedra, immediately is increased number of tetrahedra. I?ve tried to insulate some dummy (virtual) objects, but anyway simulating results stay far away from measured one. As in understand, main problem is to achieve right distribution of tetrahedra.

Hi an.t,

It's difficult from your slide to see what exactely you're trying to accomplish. In term of 'meshing', it depends on the complexity of the design. Typically, you'll have more mesh at points of high current density (i.e. Via, bends ...) while other areas of the design will be broken down in fewer meshes.

Hi, Boomerang!

My goal is to design interdigital BPF. After some attempts I understood, that HFSS is unable to perform right meshing by itself (at least in devices with strong edge-coupling effect). Very good example is h**p://www.ansoft.com/hfworkshop04/2004_HFSS_Users_Workshop_Bernard_Schmanski.pdf, please see 2 picture from this document below. Main idea of this paper is to insulate dummy (virtual) objects with seeded mesh. My question is how Bernard Schmanski achieve such good mesh distribution (at second picture)? I agree with you that higher density of smaller tetrahedra should be in points of high current density (i.e. Via, bends ...), AND in case of edge-coupling devices IN SLOTS between microstrips.
If do simply tetrahedra length limitation between coupled lines in filter (by inserting virtual object), results become better, but anyway I can not achieve such beautiful picture as Bernard Schmanski. In my project I have to draw a lot of small dummy object and set ?seeded mesh? with small length. As a result of work I get 150k-200k tetrahedra and very long processing time. My filter become similar as shown at slide 13 of Bernard Schmanski doc.

I got the same problem as you. Mesh distribution is very important for HFSS calculation. The simulation results are different from what 2.5D tools simulates(Center Frequency is low in HFSS). Since smaller tetrahedra is accepted, simulation time is very long.

I was somewhat surprised to see the comments in this thread. I have been modeling interdigital BPF filters in s and c bands using the default meshing. The results have been reasonalby close when we look at the measured values. At least for the performance requirements of my particular application. That may not be your case.

Seeding or manually modifying the mesh is certainly practical as is the addition of virtural objects. Mesh density can certainly be increased in high field areas using those features. I would tend to validate the simulator performance vs hardware rather than contrasting 3D and 2.5D tools.

Are you seeing corruption of your s-parameters? What gives you misgiviings about the results?

You can try to change the Solution Frequency. Sometimes this procedure is useful.

an.t shalom,

My name is Ital and I work for the Ansoft representative in Israel.
I'll try to explain what you've read in the document, and then the solution will be obvious. The HFSS adaptive mesh algorithm concentrates in places that have high energy. If for some reason S12 is very low in the initial mesh (below -30 dB), it means that all of the energy is conectrated near the ports, and the mesh refignment is done near the ports (which in your case means that S12 is still low).
So what you should do is set lambda refignment to 0.1 or 0.05 and set the adaptive frequency to the pass band (where S12 is not -70 dB). The initial mesh will be good enough to let the energy through the model and the adaptive mesh will kick in and refine the solution as expected.

Regards,
Itai