Menu

Discussion Forum

New Discussion
Filtern

Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Antenna far field postprocessing

Posted 24.05.2011, 10:17 GMT-4 RF & Microwave Engineering, Results & Visualization Version 4.1 22 Replies

Edgar Kaiser
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Dear All,

I am modelling VHF/UHF directive antennas in 3D and I am specifically interested in the far field patterns of the antennas.
I notice that the far field postprocessing is extremely time consuming. My impression is that it takes much longer than actually the integration of the model.
The results look reasonable. But any change to the visualization again may take an hour to compute whereas the model without the far field node may take 30 minutes.

Any hints, anybody?

Cheers
Edgar
22 Replies Last Post 29.04.2013, 02:10 GMT-4
Edgar Kaiser
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.05.2011, 06:08 GMT-4
Hi All,

I want to give an update on recent findings:

As long as only 1D plots and polar plots are used, the far field post processing is reasonably quick. This is mostly sufficient for far field investigations. You want to know the azimuthal and elevation polar plots. And you want to have impedance plots which can be done in 1D too.

If you want to see the near field it gets worse. Here I find slices in 3D most useful. And even if you don't plot the far field data, but the E-field for example, it gets unacceptably slow again.

So, it looks like I neeed to switch off the far field calculation node if I want to see anything in 3D. This is actually not a very nice situation.

Actually it looks like a bug, because even for 1D azimuth and elevation polar plots the system needs to integrate the full far field. Can we hope this will be improved in 4.2?

Btw, I am running the latest 4.1 version 185.

Cheers
Edgar
Hi All, I want to give an update on recent findings: As long as only 1D plots and polar plots are used, the far field post processing is reasonably quick. This is mostly sufficient for far field investigations. You want to know the azimuthal and elevation polar plots. And you want to have impedance plots which can be done in 1D too. If you want to see the near field it gets worse. Here I find slices in 3D most useful. And even if you don't plot the far field data, but the E-field for example, it gets unacceptably slow again. So, it looks like I neeed to switch off the far field calculation node if I want to see anything in 3D. This is actually not a very nice situation. Actually it looks like a bug, because even for 1D azimuth and elevation polar plots the system needs to integrate the full far field. Can we hope this will be improved in 4.2? Btw, I am running the latest 4.1 version 185. Cheers Edgar
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.05.2011, 07:05 GMT-4
Hi Edgar

Do you happen to have a simple model to upload, to try to load into v4.2 ? Just to see if I can udnerstand you better ;)

--
Good luck
Ivar
Hi Edgar Do you happen to have a simple model to upload, to try to load into v4.2 ? Just to see if I can udnerstand you better ;) -- Good luck Ivar
Edgar Kaiser
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.05.2011, 08:36 GMT-4
Hi Ivar,

first I must correct myself. If no farfield data is plotted in 3D it works fine.

Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes.
I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing.
Looking forward to your results in 4.2
Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field.

Thanks
Edgar

Hi Ivar, first I must correct myself. If no farfield data is plotted in 3D it works fine. Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes. I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing. Looking forward to your results in 4.2 Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field. Thanks Edgar

Attachments:

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26.05.2011, 03:48 GMT-4
Hi Edgar

on my WS with v4.2 it was not that much better: with FF 17 seconds to solve (against your 106) but then about 90 sec for the FF calculation

I were using 12 of my 24 multi-threaded CPUs (2xX5680 3.33GHz Win7-64), as the model is small <= 1Gb, I suppose it remained mostly in the CPU cash

With my powerful WS I find it acceptable, but your model is simple.

We have noticed too, when working on complex geometries (>1000 objects) the "refresh time can come up in minutes, so sometimes you must take your time and not be too nervous with the mouse

--
Have fun Comsoling
Ivar
Hi Edgar on my WS with v4.2 it was not that much better: with FF 17 seconds to solve (against your 106) but then about 90 sec for the FF calculation I were using 12 of my 24 multi-threaded CPUs (2xX5680 3.33GHz Win7-64), as the model is small 1000 objects) the "refresh time can come up in minutes, so sometimes you must take your time and not be too nervous with the mouse -- Have fun Comsoling Ivar
Edgar Kaiser
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26.05.2011, 04:27 GMT-4
Hi Ivar,

thanks a lot for taking your time and checking.

I am just wondering, why it is like that. Is a far field calculation consuming that much more calculation effort than solving the FE model?
Maybe in 3D it is like that. The whole boundary must be integrated for each direction. It might be worthwhile to consider something like a resolution parameter in the farfield node that allows a coarser resolution of the far field. Or is it possible to implicitely control that by the meshing on the far field boundary?

Thanks again
Edgar
Hi Ivar, thanks a lot for taking your time and checking. I am just wondering, why it is like that. Is a far field calculation consuming that much more calculation effort than solving the FE model? Maybe in 3D it is like that. The whole boundary must be integrated for each direction. It might be worthwhile to consider something like a resolution parameter in the farfield node that allows a coarser resolution of the far field. Or is it possible to implicitely control that by the meshing on the far field boundary? Thanks again Edgar
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26.05.2011, 07:42 GMT-4
Hi Edgar

Unfortunately I cannot answer, I do not know.
But I remember from the time I made my own near-field/far field DOE (Diffractive Optical Elements) calculations, it took a lot of time, and at that time (a couple of decencies ago) we restricted ourselves to 2D

When I find a moment I would like to try some diffractive optics with COMSOL ;)

--
Good luck
Ivar
Hi Edgar Unfortunately I cannot answer, I do not know. But I remember from the time I made my own near-field/far field DOE (Diffractive Optical Elements) calculations, it took a lot of time, and at that time (a couple of decencies ago) we restricted ourselves to 2D When I find a moment I would like to try some diffractive optics with COMSOL ;) -- Good luck Ivar
David Smith
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.11.2011, 06:45 GMT-5
Hi,

Thanks for some interesting posts.

I downloaded the UHF/VF antenna model and wanted to know if you used a swept mesh for the PML domain?

I'm interested in far-field radiation patterns and impedance and was wondering if PML would still be effective if a standard free-quad mesh was used?


Thanks,
David
Hi, Thanks for some interesting posts. I downloaded the UHF/VF antenna model and wanted to know if you used a swept mesh for the PML domain? I'm interested in far-field radiation patterns and impedance and was wondering if PML would still be effective if a standard free-quad mesh was used? Thanks, David
Edgar Kaiser
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.11.2011, 07:20 GMT-5
Hi,

I used a standard free mesh without any refinement.

Cheers
Edgar
Hi, I used a standard free mesh without any refinement. Cheers Edgar
David Smith
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.11.2011, 08:44 GMT-5
Hi Edgar,

Thanks for your reply.

I will try the same approach. I previously haven't used a PML domain only the far-field - so I'm interested to see if there are any differences.

What did you base the radius of PML on? I was thinking r1=wavelength/2 (inner radius of the PML) and r2=r1*1.5 (outer radius of the PML).

Thanks again,
David
Hi Edgar, Thanks for your reply. I will try the same approach. I previously haven't used a PML domain only the far-field - so I'm interested to see if there are any differences. What did you base the radius of PML on? I was thinking r1=wavelength/2 (inner radius of the PML) and r2=r1*1.5 (outer radius of the PML). Thanks again, David
Edgar Kaiser
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.11.2011, 10:25 GMT-5
David,

the PML is important because otherwise reflections will result in an unrealistic far field. I would make it a little bigger, at least a wavelength for the inner sphere and 1.5 to 2 wavelengths for the outer sphere.
You will see if the wave is nicely damped in the PML or you need to make it bigger (or can make it smaller).

Cheers
Edgar
David, the PML is important because otherwise reflections will result in an unrealistic far field. I would make it a little bigger, at least a wavelength for the inner sphere and 1.5 to 2 wavelengths for the outer sphere. You will see if the wave is nicely damped in the PML or you need to make it bigger (or can make it smaller). Cheers Edgar
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.11.2011, 10:37 GMT-5
Hi

just a comment, I have noticed that PML works best with layered (sweep) mesh. This is "easily" made if your external shell is cut up i.e. if its spheric, or circular , use a "layer" in your geometry, also for cubes. Then mesh the interior, and finish the mesh with a (or several for cubics sides) sweep of the outer PMl layer.

this makes nice "onion layer mesh for the PML, but is far more tricky to achieve if you do not use the new " geometry - layer" feature in V4

--
Good luck
Ivar
Hi just a comment, I have noticed that PML works best with layered (sweep) mesh. This is "easily" made if your external shell is cut up i.e. if its spheric, or circular , use a "layer" in your geometry, also for cubes. Then mesh the interior, and finish the mesh with a (or several for cubics sides) sweep of the outer PMl layer. this makes nice "onion layer mesh for the PML, but is far more tricky to achieve if you do not use the new " geometry - layer" feature in V4 -- Good luck Ivar
David Smith
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.11.2011, 10:43 GMT-5
Thanks Edgar for your comments on the sizing.

Ivar, do you know if a spherical shell can be cut into layers in Comsol (I don't have the option of import CAD 3D geometries)?

Thanks,
David
Thanks Edgar for your comments on the sizing. Ivar, do you know if a spherical shell can be cut into layers in Comsol (I don't have the option of import CAD 3D geometries)? Thanks, David
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.11.2011, 11:36 GMT-5
Hi

I usually make my "air" bloks or sphères in COMSOL, while I import often my models from my CAD (SW in my case) and the "layer" feature is part of the COMSOL geometry items

--
Good luck
Ivar
Hi I usually make my "air" bloks or sphères in COMSOL, while I import often my models from my CAD (SW in my case) and the "layer" feature is part of the COMSOL geometry items -- Good luck Ivar
David Smith
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24.11.2011, 11:55 GMT-5
Hi Ivar,

Thanks very much - the 'layer' option is very useful.

I will try a PML with and without swept mesh.

Best Regards,
David

Hi Ivar, Thanks very much - the 'layer' option is very useful. I will try a PML with and without swept mesh. Best Regards, David
Glenston Miranda
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23.12.2011, 02:26 GMT-5

Hi Ivar,

first I must correct myself. If no farfield data is plotted in 3D it works fine.

Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes.
I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing.
Looking forward to your results in 4.2
Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field.

Thanks
Edgar




[QUOTE] Hi Ivar, first I must correct myself. If no farfield data is plotted in 3D it works fine. Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes. I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing. Looking forward to your results in 4.2 Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field. Thanks Edgar [/QUOTE]
Glenston Miranda
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23.12.2011, 02:26 GMT-5

Hi Ivar,

first I must correct myself. If no farfield data is plotted in 3D it works fine.

Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes.
I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing.
Looking forward to your results in 4.2
Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field.

Thanks
Edgar




[QUOTE] Hi Ivar, first I must correct myself. If no farfield data is plotted in 3D it works fine. Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes. I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing. Looking forward to your results in 4.2 Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field. Thanks Edgar [/QUOTE]
Glenston Miranda
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23.12.2011, 02:27 GMT-5

Hi Ivar,

first I must correct myself. If no farfield data is plotted in 3D it works fine.

Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes.
I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing.
Looking forward to your results in 4.2
Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field.

Thanks
Edgar


Hi Edgar,

Just look at the last reply that i have posted
There was a problem with my machine.

Let me know how to delete a reply.
Cheers!!

Regards
Glenston


[QUOTE] Hi Ivar, first I must correct myself. If no farfield data is plotted in 3D it works fine. Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes. I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing. Looking forward to your results in 4.2 Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field. Thanks Edgar [/QUOTE] Hi Edgar, Just look at the last reply that i have posted There was a problem with my machine. Let me know how to delete a reply. Cheers!! Regards Glenston
Glenston Miranda
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23.12.2011, 02:48 GMT-5

Hi Ivar,

first I must correct myself. If no farfield data is plotted in 3D it works fine.

Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes.
I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing.
Looking forward to your results in 4.2
Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field.

Thanks
Edgar



Hi Edgar,

I know very little about FEM Theory. I can only guess why far field takes so much time.
Comsol uses Stratton Chu formula to calculate its far field. In other words it uses the near field to calculate the far field. The equation is present in the Comsol documentation in far filed calculation theory. The integration involved is strenuous as it has to find E & H and it has to do a surface integration . This could be time consuming.

Far field in itself is an approximation in theory. An alternative to find the 3D far field radiation pattern is by creating a parameterized surface and use that as the data set. Resolution can be varied by changing the value in the settings window. Here it need not find the values of E and H for every nodal element and then integrate which could be very very time consuming. I've done the 3D radiation pattern for a resolution of 100. I've used your model to demonstrate it.

Let me know if you have questions!!
Cheers!!

Regards
Glenston
[QUOTE] Hi Ivar, first I must correct myself. If no farfield data is plotted in 3D it works fine. Here is a simple model of a dipole in 3D. On my machine the solution took some 2 minutes and the farfield processing took about 10-15 minutes. I am aware that the model has some imperfections, but all should only affect the solution and not the time for the farfield processing. Looking forward to your results in 4.2 Btw, COMSOL support attributes it to the 3D rendering. Actually I am using software rendering because of graphics card issues, but the rendering is good enough for all other kinds of 3D plots that do not involve far field. Thanks Edgar [/QUOTE] Hi Edgar, I know very little about FEM Theory. I can only guess why far field takes so much time. Comsol uses Stratton Chu formula to calculate its far field. In other words it uses the near field to calculate the far field. The equation is present in the Comsol documentation in far filed calculation theory. The integration involved is strenuous as it has to find E & H and it has to do a surface integration . This could be time consuming. Far field in itself is an approximation in theory. An alternative to find the 3D far field radiation pattern is by creating a parameterized surface and use that as the data set. Resolution can be varied by changing the value in the settings window. Here it need not find the values of E and H for every nodal element and then integrate which could be very very time consuming. I've done the 3D radiation pattern for a resolution of 100. I've used your model to demonstrate it. Let me know if you have questions!! Cheers!! Regards Glenston

Attachments:

Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26.04.2013, 16:41 GMT-4

Hi

I usually make my "air" bloks or sphères in COMSOL, while I import often my models from my CAD (SW in my case) and the "layer" feature is part of the COMSOL geometry items

--
Good luck
Ivar


Hi Ivar,

I am coming back to this old thread because I am desperately seeking for the geometry layer feature you mentioned. I don't find it in 4.3a.

Cheers
Edgar

--
Edgar J. Kaiser
www.emphys.com
[QUOTE] Hi I usually make my "air" bloks or sphères in COMSOL, while I import often my models from my CAD (SW in my case) and the "layer" feature is part of the COMSOL geometry items -- Good luck Ivar [/QUOTE] Hi Ivar, I am coming back to this old thread because I am desperately seeking for the geometry layer feature you mentioned. I don't find it in 4.3a. Cheers Edgar -- Edgar J. Kaiser http://www.emphys.com
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 27.04.2013, 06:52 GMT-4
Hi

is it this you are looking for ?
(open well up all the tabs in the geometry and check if you do not need to turn on to "show" some of the advanced features, I have turned all on ;)

I'm not sure the FF looks OK though

--
Good luck
Ivar
Hi is it this you are looking for ? (open well up all the tabs in the geometry and check if you do not need to turn on to "show" some of the advanced features, I have turned all on ;) I'm not sure the FF looks OK though -- Good luck Ivar

Attachments:

Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 27.04.2013, 17:08 GMT-4

Hi Ivar,

it still took me a while until I figured out how simple it is. Sometimes I am searching for the glasses sitting on my nose.

Thank you
Edgar

--
Edgar J. Kaiser
www.emphys.com
Hi Ivar, it still took me a while until I figured out how simple it is. Sometimes I am searching for the glasses sitting on my nose. Thank you Edgar -- Edgar J. Kaiser http://www.emphys.com
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 29.04.2013, 02:10 GMT-4
Hi

Well COMSOL is becoming as complex as Multi-Physics itself ;)

Now I was a bit puzzled by the Far Field pattern, but I didn dive into the model really, I just remade it quickly as I would have mounted i in 4.3at, from the one above in 4.1.

I must admit that I'm stuck in structural and static FSI these days for my clients ;)

--
Good luck
Ivar
Hi Well COMSOL is becoming as complex as Multi-Physics itself ;) Now I was a bit puzzled by the Far Field pattern, but I didn dive into the model really, I just remade it quickly as I would have mounted i in 4.3at, from the one above in 4.1. I must admit that I'm stuck in structural and static FSI these days for my clients ;) -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.

Suggested Content