Voltage source mesh and plan wave polarisation

[Michael]

Dear all,

My next question is about the EM-Sources.
(Don’t worry, the antenna issue has not been forgotten, but it still needs some time.)

1. My professor and I are interested in what happens if the voltage source is meshed with more than one cell at the corner between the ‘pos’ and ‘neg’ cells.
   To my knowledge, the excitation has to be very precise meshed, because in this region the highest amount of field changing is there. In the online tutorial for the patch antenna, excited by a coax cable with the voltage source/gap as excitation, the voltage source has the same size as the coax pin and is so several times bigger than lambda. Also the mesh.
   So what happens at NSEM if the source is meshed by more than two cells in total?
   By trying it out I got some results (see picture), but how correct are there? For S11 and Z_in it doesn’t look realistic. The rest is similar to the original meshing interval.

**Result intervall = 4 cells**

**Result intervall = 1 cell**

**JOU**

reset

reset aprepro

Parameter, all definitions in mil

#{sub_h=31}

#{sub_w=787}

#{patch_w = 187}

#{coax_l = 100}

#{coax_r = 20}

#{pin_r = 6}

#{vs_w = 2*pin_r}

#{pin_l = sub_h+coax_l-vs_w}

#{feed_offset = 0.68}

#{mesh_size = (0.3/171000/25.41000)/10}

#{freq = 16e9}

#{co = 3e8}

#{lambda = co/freq}

brick x {sub_h} y {sub_w} z {sub_w}

volume 1 rename “substrate”

move Volume substrate x {-sub_h/2} include_merged

create surface rectangle width {patch_w} xplane

volume 2 rename “patch_vol”

create surface rectangle width {sub_w} xplane

volume 3 rename “gnd”

move Surface in volume gnd x {-sub_h} include_merged

create Prism height {pin_l} sides 6 radius {pin_r}

volume 4 rename “pin”

rotate Volume 4 angle 90 about Y include_merged

move Volume 4 x {-pin_l/2} z {-patch_w/2+feed_offset*patch_w/2} include_merged

create Prism height {coax_l} sides 6 radius {coax_r}

rotate Volume 5 angle 90 about Y include_merged

move Volume 5 x {-sub_h-coax_l/2} z {-patch_w/2+feed_offset*patch_w/2} include_merged

volume 5 rename “coax”

create surface rectangle width {vs_w} zplane

volume 6 rename “feed”

move Surface 25 x {-sub_h-coax_l+vs_w/2} z {-patch_w/2+feed_offset*patch_w/2} include_merged

split surface 25 direction curve 60

project surface 10 onto surface 18 imprint

subtract volume coax from volume gnd keep_tool

subtract volume pin from volume substrate keep_tool

subtract volume pin from volume coax keep_tool

subtract volume feed from volume coax keep_tool

imprint volume all

merge volume all

block 1 add surface all

nsem load material library “E:\Michael\Nullspace_workspace\Material\materials.h5”

nsem assign volume substrate material ‘RO3003’

nsem assign volume coax material ‘Teflon’

nsem assign volume patch_vol material ‘PEC’

nsem assign volume gnd material ‘PEC’

nsem assign volume pin material ‘PEC’

nsem assign Surface 19 20 21 22 23 24 28 49 52 material ‘PEC’

nsem voltage source ‘port1’ pos surface 27 neg surface 26 impedance 50

surface all scheme pave

curve 60 61 58 interval 4

curve 60 61 58 scheme equal

surface in volume feed size

mesh surface in volume feed

surface not is_meshed in volume coax size {coax_r/2}

mesh surface not is_meshed in volume coax

surface in volume patch_vol size {mesh_size/3}

mesh surface in volume patch_vol

surface not is_meshed size {mesh_size}

mesh surface not is_meshed

nsem print status

**short Log file/ sim overview**

system : DENSE

fill tolerance : 0.0001

solve tolerance : 0.0001

      element type = QUAD_4

gqCache beg

  sizeof(P) (Bytes) : 8.0e+00

  sizeof(Vector3<P>) (Bytes) : 2.4e+01

  sizeof(Jacobian<P>) (Bytes) : 5.6e+01

  gqCache (GB) : 9.9e-03

Nn : 797

Nb : 2968

Ne : 809

Nsb : 10

Ned : 1604

Nbe : 4

Nrhs : 1

process memory (GB) : -9.536743e-07 : -9.536743e-07

Sim time: 5.25min
10_GHz
RAM: GB

2. The plane wave excitation: To define the incidence angle of the plane wave in response to the origin I choose theta (t) and phi (p). So far so clear, but how I can determine the polarisation of the incoming plane wave?
   I understand that for the result I can choose the observation angle and get the different polarisation pp, pt, tp, tt. And that the excitation is automatically computes two linear polarisations, shifted by 90 °.
   So for example for a plane wave by t = 90° and p= 0° the polarisations are co and cross-polarised (or linear polarisation in horizontal (co) and vertical (cross) polarisation). But also for that, how do I know concerning Paraview results which polarization is which regarding the given theta and phi angle?
   In Para view the sheds called 0_0_Jp_Abs for positive current density and 0_1_Abs for negative current density. (0_2_, 0_3_,…,0_10_,0_11_, depends on the amount of excitation position).
   For some significant points and well-known geometries like a sphere, you can try to imagine how the polarization is by take a view of the surface current, but not for more complex chosen incidence angles and geometries. Or I’m wrong and it’s an easy thing but only my knowledge is not high enough, so please, could you explain how I get the polarization by looking at the results?
   I add an example from Parawiev with the chosen angle of incidence of the plane wave.

**surface current @ Paraview**

Incidence angle choosing by theta = 45° and phi = 90°

positive current density
Camera view @ Paraview Adjust Camera - Camera Paremeters - 0, 0.707, 0.707

positive current density by choose the second polarisation
( angle in respect to the origin unknown)
Camera view @ Paraview Adjust Camera - Camera Paremeters - 0, 0.707, 0.707

first plot by alignment of the coordinate axes x,y,z

**JOU**

reset

reset aprepro

#{f =10e9}

#{c=3e8}

#{lambda=c/f}

#{r=4*lambda}

create sphere radius {r}

nsem assign volume 1 material ‘PEC’

imprint all

merge all

set duplicate block elements off

block 1 add surface all

block all element type quad9

surface 1 size {lambda/5}

mesh surface 1

**short Log file/ sim overview**

system : DENSE

fill tolerance : 0.0001

solve tolerance : 0.0001

      element type = QUAD_9

gqCache beg

  sizeof(P) (Bytes) : 8.0e+00

  sizeof(Vector3<P>) (Bytes) : 2.4e+01

  sizeof(Jacobian<P>) (Bytes) : 5.6e+01

  gqCache (GB) : 1.5e-01

Nn : 22206

Nb : 44408

Ne : 5551

Nsb : 1

Ned : 11102

Nbe : 12

Nrhs : 2

process memory (GB) : -9.536743e-07 : -9.536743e-07

simulation time: 6.26 min
10_GHz
using peak RAM: 59.42293167114258 GB

[masha.petrova]

@Michael Thanks for another great question.

The positive/negative surface intersection needs to be a single edge at the intersection of no more than two pavers. If the voltage source is applied to an edge connected to more than two pavers the result will be wrong.

Does that help?

[Michael]

Thank you for your answer!

Yes, it helps. Two the second point, how is the polarisation of a plane wave in request to the origin defined? Is there a dependency between the polarisation and the phi and theta coordinates?

[daniel.faircloth]

Hi Michael,

When you request a plane wave angle, you actually get two excitations, Etheta and Ephi. The result in the far field dataset is the full four-polarization scattering matrix. Here is a clip from the rain drop example to help explain:

If you’re doing a monostatic plane wave case, the resulting far field or RCS data array will be 4D as shown above. The four polarizations for the plane wave are listed in the paragraph in the order they are stored in the data arrays. So in the plot example above, we’re plotting theta-theta polarization followed by phi-phi.

For Paraview, the data extraction is a little tricky. Since the 2023R3 release, we found an issue with the documentation. Apologies if that has led to some of the confusion.

Since Paraview is a separate software package, we have to reference the data in a method that Paraview can understand. In this case, we have to refer to the frequencies, theta, and phi angles by their index in the data arrays. As an example, I modified the simple_pec_sphere example discussed in the Getting Started section of the user guide. You can think of it this way, you have two incident polarizations, Nfq frequencies, Nth theta angles, and Nph phi angles. For the modified example, I left a single frequency but modified the incidence angles to have 2 angles in theta and 2 in phi. So there are a total of 4 incidence angles, which leads to 8 total excitations.

In Paraview, the first index is the frequency index, so it’s just index 0 in my example since there is only 1 frequency. See the screenshot from Paraview below. The second index lists the excitations in the order in which they are processed in the solver. So you can compute the index like:
index = iTh * Nph * 2 + iPh*2 + iPol

In my example, if you want to plot the surface currents at theta = 90, phi = 90, and the phi-polarization, this corresponds to
index = 0 * 2 * 2 + 1*2 + 1 = 3

Hope this helps. Please let me know if you have any further questions.

[Michael]

Thank you for your explanation!

It helps

[daniel.faircloth]

Michael, I realized after I replied that I didn’t properly explain. Please see my edited comment above.

[Michael]

Still thanks, that helps realy for understanding how it works in NSEM.

Two things could help in future,
to add the information that all indices start with 0 (if not mentioned)
(Not everyone is so familiar with python and how the indices are)

"model.set_plane_wave_grid([iTh = 0, iTh = 1], [iPh = 0, iPh = 1]) " (Also for Paraview where the first frequency point is zero
i.e. Range 3 to 20GHz → 0_0_Jp => f = 3GHz
It was not so obvious to me.

and second if possible, at the paraview method to fill up the first three positions with 0 so that you have 000_1_JP, 001_1_JP …
With the filter at properties, you can sort it of course but it’s not so good readable as the example below.

However, this is not necessary. The description of how the indices are compiled was very helpful!

[daniel.faircloth]

Thank you for the suggestions. As we continue to improve our products, these kind of suggestions are really helpful.

[masha.petrova]

@Michael

mation that all indices start with 0 (if not mentioned)
(Not everyone is so familiar with python and how the indices are)

"model.set_plane_wave_grid([iTh = 0, iTh = 1], [iPh = 0, iPh = 1]) " (Also for Paraview where the first frequency point is zero

Thank you, Michael! Really appreciate your questions and comments here.

All of your suggestions will help us to continue to improve our software and deliver great simulation tools for engineers.

Thanks for helping this whole Nullspace user community.