HFSS vs. CST

Guys,
Anyone can comment on that? MWS - time-domain solver - for fine speed very wide bandwidth - that's clear.
But if I need my antenna be simulated 15-17GHz and use the PML, PML is no good for wide freq. band for many reasons. Besides, to be good at the lower freqs the PML must be wastedly afar. This is where I appreciare HFSS alot. With MWS the PML (transient solver) PML are embodied with big penalty in that case. To tackle that problem I had to write very long macro with MWS. And the answer to the question - Up to what freq is the PML in MWS good is not obvious. Frienda mine did a Ph.D on PML and found somewhere around 6-10 layers to suffice for a real decent bandwidth and reflection losses.
regards,
cheng

HFSS uses PML if you want very good results and you have the same problems. Otherwise you can use ABC 2° order but with some approximations. Further if you use HFSS for radiation problems you need of a lot of memory and time and it is easy to exceed your system capability.

Regarding Cheng's question concerning PML at low frequency, I think if you need to place it very far away in MWStudio, you would need to do the same thing in HFSS. With this, you can't say HFSS is better.

I think the performance of PML should be stated more carefully. To say that 6-10 layer is alway good is misleading. Other conditions such as the distance to the discontinuity or the kind of radiator are also important and should be combined together.

One of the short-comings of CST MWS is that you can not directly define and analyze driven problems in which at least one port is not located on a plane parallel to one of XY, YZ, or ZX planes. An example is a U-type rotary joint or an alpha-degree waveguide or microstrip bend, in which alpha is not equal to K times 90.
wave-maniac

To my knowledge, HFSS has similar problem. How do you draw a coaxial line which makes 30degrees with the z-axis, and define the port?

Loucy, you are missing the point here.
My point was that I DON'T CARE about low frequency if I need 15-17Gigs and I can optimize my PML (as many layers as I will) for ANY particular band. With MWS one has no choice but to resort to the full bandwidtht (for obvious reasons) and what I say is that you can't really trust the lower freq. band but the one you optimize your PML for. Why do I care about bogus performance between 1-14Gis if I care about 15-17Gigs? That was my point. If I need my lower frequency in HFSS I will optimize for thaty freq. range too.
cheers,
cheng
In MWS number of layers is set to like 4. But for real decent simiul. 6 seems preety good choice - even more.

Are you kidding loucy? You can define a port on any outermost planar face of a body. Simply click on it when defining boundary conditions, and set it as a port. If you still insist that's not true, please u/l the 3D structure to show you that.
wave-maniac

In MWS PML are placed at minimum distance in part of the wavelength at the upper frequency chosen (1/8 is default); therefore if you chose 0-15 GHz or 14-15 GHz, the minimum distance is the same. Obviously, in the first case you have lower accuracy in the lower frequency band. Anyway it is possible change the default.
Time domain solver are very speed with broadband problems, of course. But, it is to be considered that FEM codes are very inefficient with radiation problems; often it is not possible to begin the similation!

Hi, Cheng
If you don't care about 1-14GHz, how will the performance of PML at 1GHz affect your result at 15GHz, in time domain simulation? Maybe dispersion might have some effect, but can we neglect that?
Also, can we limit the spectral content of the excitation to near 15-17GHz? I have used MWStudio for only a few times, and I don't know the answer.
Regards,
Loucy

To Wave-Maniac,
I think defining a tilted coaxial port is a bit more complicated than picking a single planar surface. You have to do the cap, the end of inner and outer conductor, and also the calibration line. Have you ever seen HFSS meshed/modeled it secessfully?
Regards,
Loucy

Dear loucy;
I need a 3D or 2D sketch. The words can not define the picture clearly.
I'll be looking forward to receiving the geometry.
Even the existance of a case, doesn't mean that HFSS can not define any relatively tilted port. I've analyzed/designed many passive structures with tilted ports. CST MWS can not define even a single one.
One of the problems with PBA, at least the one used in MWS, is that any refinement in a point, creates extra refinement in subspaces from that point towards all the 6 planes defining the outermost box. A true CFDTD has no such problem, i.e. the refinement is limited to that point.
HFSS mesh generator is cool, though not as powerful as that of ANSYS. I wish the mesh could be imported from outside HFSS.
Regards
wave-maniac

The master/slave boundaries do not need to be parallel to the xz and yz axes. We did a comparison of of the CST and HFSS using a probe fed cavity backed patch. A triangular lattice can be modeled in HFSS with 3 sets of master/slave pairs. To my knowledge, which is limited when it comes to CST, you can't set up more than 2 pairs of periodic boundaries.
One thing that probably unfairly bias's me toward HFSS is it is running on a UNIX box, which handles the big stuff much better. I also frees up my desktop processor! :D

Hi Loucy,
The performances of PML at low frequencies are not relevant for the response in the higher frequency band; some influences are possible only for very non linear media.
It is possible to limit the excitation between 15-17 GHz; in this case we have a longer time similuation due to the inverse relation between the frequency band and time duration of the signal excitation. Anyway it is possible to reduce this effect by applying windowing to the signal or using autoregressive filters.
Regards
yuyu

Hi wave-maniac,
probably in the next version of CST, 5, it will be possible to have the refiniment limited to the point without extra cells in subspaces toward the all directions.
regards
yuyu

After reading the above, I have to agree that HFSS is better than MWStudio in the area of tilted port, and I think defining a tilted coaxial port is possible.

Hi ALL !
I caclulate mostly oversized, overmoded passive WG structure.
For that problem CST is totally useless !
MWS (even last v4.1) requires very long time for simulation and the accuracy is so low ....

That is a surprise. Oversize or Over-mode shouldn't be a problem for the method. It might not have de-embedded correctly.

I would add that the geometry modelling in HFSS is not so good, it lacks accuracy. For example, I take the intersection of a sphere and a cylinder, paste back the sphere, check for overlapping, OK. Then I do a rotation on the intersection. HFSS would report finding overlap.
This problem sometimes cause error in meshing or unreasonable high number of elements.

in my design of antenna, the measured results are agree well to the design result by Hfss.