Time reduction in HFSS simulation by splitting the model
I'd like to reduce simulation time of a suspended stripline fed- circular waveguide travelling wave antenna in HFSS by splitting the structure in two halves. Since one of the two halves is geometrically constant (the radiating section, I don't have to change anything during the optimization), I supposed I can simulate it once and thus connect the S matrices of the two halves using an Ansoft Designer - HFSS dynamic link, doing the optimization just for the matrix containing the feed network and considering the radiating section such as a constant 'load'. As proof of my idea, I tried with a simple circular waveguide. I simulated the whole structure in HFSS, then I simulated the halved structure. By the union of the halved structure's S matrices, I wanted to reproduce exactly the same behavior of the total waveguide simulated in HFSS. The result is: IT DOESN'T WORK... Why, I ask. Do u think I negleted some particular or something? I even tried defining a multimode (3 modes) port in HFSS for the halved model port used as connection with the other half, but I can't manage to obtain the exact behavior of the total HFSS model. Here are some explicative pictures:
[img][/img]
[img][/img]
I really hope someone can help me. I need to perform lots of simulations for my project and if I can manage to use this 'trick', I think this can save a lot of time! And time's money, as they say!
Many thanks in advance,
Ivan
PS- The graphs plot the S11 for the HFSS total model and the Designer re-joint model...
how do you join 2 S matrices together?
Shouldn't you be multiplying the ABCD matrix?
Also, when you join 2 matrices, the effect simulated is one of the two:
The two objects are either in series or in parallel. Especially considering the fact that you have a one port network.
How did you divide the waveguide? Did you simulate half the length?
do you consider the boundary condition at the symmetric plane?
The elements are in series for sure.... Two non ideal transmission lines in cascade. The cascaded S matrices provides a generalized S matrices, but not through a simple matrix product! So I expect the S matrix of a cascaded network, just like using ABCD matrices product. I just avoided the passage through ABCD transformation. The waveguide has been divided in extactly one half of the originary. The waveguide was divided by reducing the waveg. length of half of the originary. Just this. In both cases I have two-port networks, the former for the total structure and the latter for the half waveguide, to be connected with another half S matrix.
The pictire you attached is quite unclear. What are you trying to simulate, hollow circlular waveguide ?
You must use proper symmetry plane or define polarization lines for each mode in the port (the same way for each port).
Multiplacation of scattering S-matrixes is a solid, straightforward method which works for sure.
It's better to upload your test project.
It's just a test of dividing structures in more sub-structures and connect the sub-structures' matrices in Designer to check wether the results are the same of the original design. I need to demonstrate this 'cause I want to apply the principle to the design of a array travelling wave antennas I'm working at. I just thought to use circular waveguides as a simple test. I need to reduce simulation time in my real design cause it's a very large and high structure. Anyway, I'll upload the files as soon as possible.
Regards.
I.
Added after 1 minutes:
As test structure a used a teflon-filled circular waveguide
So let me get this:
You reduce the simulation time by:
1 - Reducing the length of the waveguide by half.
2 - reducing the wavelength of the system (by increasing the frequency)
3 - You acquire the S matrix for both halves and you multiply them together.
Dear Pelada,
I use the same solution freq. for both the total and half waveguide. The second step would be to join (not just simply multiplicate) the two S matrices of the two halves, aiiming at reproducing the exact behavior of the initial complete waveguide. I'm just trying to validate this trick in order to apply it on a more complicated project I'm working at. In this way I could let a part of the structure to be fixed (i.e. the radiating termination of my travelling wave antenna) and optimize and/or parametrize the remaining part . As someone said il previous replies, I mean as for the use of simmetry boundaries, my intent is not to use them. I already used simmetry bound. and I know what they are and how to use them. I'm very practical with HFSS, boundaries in HFSS, etc. My question now is quite different. I'd like to validate the method of dividing blocks for a simple waveguide, divided in two halves of the original, but I expect this technique to be working for non- simmetrical structures as well (if it works for symmetrical ones). So I'm not mainly interest in simmetries, 'cause the cutting plane in my real project doesn't permit to define any simmetry.
Maybe I haven't been so clear in explaining my idea/peoblem. Anyway, I hope the concept is more clear now.
Thanx
I.
Ivan, it would be helpful if you uploaded your hollow circular waveguide design up on here. Then i can help you!