Strange behavior from rectangular waveguide

How to use openEMS. Discussion on examples, tutorials etc

Moderator: thorsten

Post Reply
PaulUK
Posts: 81
Joined: Wed 28 Oct 2015, 14:23

Strange behavior from rectangular waveguide

Post by PaulUK » Wed 18 Nov 2020, 14:24

I will post some images comparing a length of WG20 (WR42) simulated with both CST and openEMS. There are a number of discrepancies which I will list. I don't know if the issues are caused by incorrect use of openEMS or if it's a consequence of numerical noise or limitations of the software. I'm hoping that either Thorsten or someone else can shed some light on these issues.

List of issues:

1) When simulating the waveguide in CST FDTD solver (or CST F solver), the reflection S11 is below -90dB, openEMS gives values below -60dB but depending on the specified frequency range or z boundary PLM settings, it can vary. Is this a numerical noise limitation or something else? It could be an issue when designing waveguide filters or diplexers that require high isolation/rejection in the stop band.

2) When the openEMS waveguide model is changed from a normal waveguide structure using PEC with a finite wall thickness to just a waveguide defined only by the x and y boundaries being PEC, the S11 reflection now gets worse and is between -50 and -60 dB. The result should be identical to when the waveguide is defined by a rectangular channel of PEC. I don't know why there is a difference.

3) When taking the "realistic" rectangular waveguide model and replacing the PEC metal definition with a material definition with a given conductivity (kappa), the S11 is terrible regardless what conductivity I use that is close to any metal from brass, copper to silver. Also there is what seems like an oscillation on the S21 when you zoom into the plot. Is this a case of using the material definition function incorrectly?

I will post the two script files followed by some pictures, in this and the following thread.
Rect_Rect_WaveguideadapterWG20.m
Waveguide model with finite wall thickness.
(6.26 KiB) Downloaded 172 times
Rect_Rect_WaveguideadapterWG20_PEC.m
Waveguide model with PEC boundaries.
(6.19 KiB) Downloaded 167 times
CST Model
CST WG20 Model.png
CST model
CST WG20 Model.png (28.5 KiB) Viewed 4163 times

PaulUK
Posts: 81
Joined: Wed 28 Oct 2015, 14:23

Re: Strange behavior from rectangular waveguide

Post by PaulUK » Wed 18 Nov 2020, 14:27

Continued images...

CST Results
CST WG20 SParameters.png
CST Results
CST WG20 SParameters.png (36.25 KiB) Viewed 4162 times
openEMS Model
3DWG20.png
openEMS model
3DWG20.png (392.81 KiB) Viewed 4162 times
openEMS Results for finite wall thickness PEC.
PML_12 S11 S21.png
openEMS Results for finite wall thickness PEC
PML_12 S11 S21.png (44.48 KiB) Viewed 4162 times

PaulUK
Posts: 81
Joined: Wed 28 Oct 2015, 14:23

Re: Strange behavior from rectangular waveguide

Post by PaulUK » Wed 18 Nov 2020, 14:32

openEMS with PEC boundaries in x and y directions, vacuum inside.
PEC BC as WG Walls in x and y PML_8 S11_S21.png
openEMS with PEC boundaries in x and y directions
PEC BC as WG Walls in x and y PML_8 S11_S21.png (38.42 KiB) Viewed 4162 times
openEMS results when finite thickness model PEC is replaced with a material with conductivity "kappa".
Silver PML_8 S11_S21.png
PEC metal replaced with a material and finite kappa
Silver PML_8 S11_S21.png (35.63 KiB) Viewed 4162 times
End of post.

I hope that someone can shed some light on these peculiar observations. I hope that I have provided enough information, the scripts came from Gerald who is also a forum member. Thank you.

wettersat
Posts: 9
Joined: Fri 05 Jun 2020, 12:25
Contact:

Re: Strange behavior from rectangular waveguide

Post by wettersat » Sat 05 Dec 2020, 17:49

Hmmm. Looks like nobody has an idea or a tip.
Gerald, OE2IGL
http://wettersat.bplaced.net

PaulUK
Posts: 81
Joined: Wed 28 Oct 2015, 14:23

Re: Strange behavior from rectangular waveguide

Post by PaulUK » Tue 16 Feb 2021, 15:25

Has no one else seen this problem?

thorsten
Posts: 1438
Joined: Mon 27 Jun 2011, 12:26

Re: Strange behavior from rectangular waveguide

Post by thorsten » Thu 18 Feb 2021, 19:11

I assume you have PML at the start and end? Can you try to make it PEC inside the PML and only finite kappa inside the "normal" domain?
Maybe the PML does not behave well with high conductivity.

As for -50dB vs -100dB: the mode matching is based on the analytical modes and therefore never matches 100% the numerical results. If openEMS would use a numerical mode solver with the given mesh it might me able to get much lower, but I'm not sure if this is worth the effort. In a real device you would never get below -50dB or would even care to as this is already "nothing"...

regards
Thorsten

PaulUK
Posts: 81
Joined: Wed 28 Oct 2015, 14:23

Re: Strange behavior from rectangular waveguide

Post by PaulUK » Tue 23 Feb 2021, 15:42

Yes, PML has been used at each end. I tried different PML lengths and still no solution. I have not tried using PEC and a finite Kappa for the waveguide yet. I can give this a try. The PML has been fine in the past when using planar transmission lines.

I don't agree about the -50 dB level in real devices, I've had to design filters and diplexers that required the level in the stop-band to be greater than attenuated by 80 or 90 dB. These are usually transmit reject filters operating at high power and therefore it was important not to let the power back into the port with an LNA or LNB.

If it is the case of the way the modes are calculated, I think it would be a nice feature to have to allow numerical calculation of the port modes.

Post Reply