Cad requirements for good mesh

Prep (CAD and Meshing) Meshing
1

Dear all,

Since my last call at the forum, I have been more in touch with NS EM and got a little bit more experience. But not enough to simulate successful CAD which is not explained by the examples.

To make it clear in advance, I can simulate the structure one, but the results are wrong, so I have to investigate more time to figure out why. In the meantime, I’ll ask you about the important impacts of the CAD structure for the mesh.
For that, I add two examples of the same antenna. One with a volume body, the other only as surfaces.
Why did I do this? To figure out how high the impact of small surfaces is. Simulating the volume body (data two) didn’t work, so I transformed it into a Sim-Object as simple as I could (data one). The simulation works now, but it needs a long time. Why? I don’t know.

Back to my topic. In future, I can’t simulate only surfaces I have to add some material properties (not only PEC) so I need a volume. Also, a volume is more easy to create as CAD.
What should I look for, if I create a CAD taking into account the mesh(size). I can’t imagine, that I can completely avoid small areas.
What could I do if I get some surfaces which size are << than the wavelength and so also << the mesh size?

And addition to that, what happens if I have a huge bandwidth for example 20 GHz for simulation, so between the minimal wavelength and the maximal are a factor by 20, I think that has also an impact on the simulation time.

Please ask me if there are any questions.

OK. Is it possible to upload the step file or sat or stp? Or cub5? Otherwise I’ll upload a few pictures.

Data one

The surface body of the antenna

HF906_Antenna_Voltage_Strip_surface
HF906_Antenna_Voltage_Strip_surface1266×612 158 KB

Data two

The volume body of the antenna

HF906_Antenna_Voltage_Strip_volume
HF906_Antenna_Voltage_Strip_volume1266×612 70.4 KB

journal

The journal for the antenna by data one
reset

The dimension of the geometrie are in mm

##Multiply the size of parameter by 1000 to be in mm. CAD is in mm and lambda in m

#{f_max=18e9}

#{c_0=299792458}

#{lambda=(c_0/f_max)*1000}

import acis “E:/Michael/Nullspace_workspace/HF906_Antenna/Horn2019_HF906_eddit_with_out_coax_surface v1.sat” heal attributes_on separate_bodies

create surface rectangle width {2} height {3} zplane

rotate Surface 59 angle 90 about Y include_merged

move Surface 59 x -25 include_merged

split surface 59 direction curve 245

imprint all

merge all

nsem voltage source ‘port1’ pos surface 60 neg surface 61 impedance 50

nsem load material library ‘E:\Michael\Nullspace_workspace\Material\materials.h5’

nsem assign volume all material ‘pec’

nsem assign surface 17 58 material ‘RO3003’

set duplicate block elements off

block 1 add surface all

block all element quad9

surface 98 99 interval 1

mesh surface 98 99

surface 1 3 4 5 6 10 22 9 23 38 36 56 size {lambda/10}

mesh surface 1 3 4 5 6 10 22 9 23 38 36 56

surface not is_meshed size {lambda/5}

mesh surface not is_meshed

python script

Simulation script
Sim_HF906.txt (3.6 KB)

Material

Material Python File
materials_V2.txt (862 Bytes)

2

Michael,

Thanks for your questions. Let me dive into your files and get back to you.

3

Thanks for your detailed questions, Michael! It’s really helpful for the rest of our user community!

4

Michael, can you attach that SAT file from your journal? If it won’t post here, please send to info@nullspaceinc.com.

1 Like
5

Have you get the files via Mail?

6

We did receive them. Thanks for sending those. We’ll take a look.

7

Michael,

There appears to be multiple problems with the “Horn2019_HF906_eddit_with_out_coax_surface v1.sat” model. Please consider observations below.

[1] I observe multiple overlapped volumes. In general we recommend removing overlap volumes for they can cause confusion and confuse the merging/meshing operations.
[2] The line [nsem assign surface 17 58 material “RO3003”] is an invalid material assignment in NSEM. One cannot assign a material to a surface. It can only be assigned to volumes.
[3] There are intricate features at the bottom of the horn model that are likely not going to impact your engineering data. These features may also create meshing issues that are difficult to overcome. In situations like this it is usually better to start w/ a simpler model in which these troublesome features are either removed or altered to yield a better suited model for meshing. In general, features smaller than wavelength are likely not going to contribute significantly to the engineering data. Those too are candidates for simplification.
[4] In general, in performing a wide band simulation, one should mesh at the maximum frequency of the bandwidth. Yes, it may impact run time since the mesh at maximum frequency will produce a denser mesh.
[5] You need to specify a mesh scheme and mesh density. We recommend the pave mesher.
[6] For electrically larger models NSEM supports compression, but I think your model is likely better to perform without compression.

I would recommend trying to perform the simulation on a simpler version of the horn to gain a baseline understanding of the model. To that end I attempted to fix some of the volume overlaps and remove most of the intricate features at the bottom of the horn. See the journal file below. I hope it may help guide you on questions/issues above.

I hope this helps,
Mark

reset

#Multiply the size of parameter by 1000 to be in mm. CAD is in mm and lambda in m
#{f_max=18e9}
#{c_0=299792458}
#{lambda=(c_0/f_max)*1000}

import acis “Horn2019_HF906_eddit_with_out_coax_surface v1.sat” nofreesurfaces heal attributes_on separate_bodies

create surface rectangle width {2} height {3} zplane
rotate Surface 59 angle 90 about Y include_merged
move Surface 59 x -25 include_merged
split surface 59 direction curve 245

simplify the model

delete surface 41 45 46 49
delete volume 23 29 31 32 33 36 37 39 41 42 43 44 45 46
create surface curve 30 73 116 125

imprint all
merge all

nsem load material library “materials.h5”
nsem voltage source “port1” pos surface 60 neg surface 61 impedance 50
nsem assign volume all material “pec”

This is an invalide NSEM material assignment

##nsem assign surface 17 58 material “RO3003”

block 1 add surface all
block all element quad4
surface all scheme pave

#{mesh_max = lambda/4}
surface all scheme pave
mesh surface all

save cub5 “horn.cub5” overwrite journal

8

Dear Mark,

Thank you for your detailed and accurate answer. I’ll try out your tips and recommendations step by step. I’ll let you know if I get it; if not, I’ll ask again.
Thank you for your time and support!

Best regards
Michael

1 Like