AnsweredAssumed Answered

Edge mesh setting in Momentum

Question asked by anhchu_iis on Sep 29, 2020
Latest reply on Oct 5, 2020 by anhchu_iis

Like • Show 0 Likes0
Comment • 4

Dear all,

We are em simulating a slab inductor at 263GHz using Momentum and got confusing results from different settings for the edge mesh width. The inductor is a 7.5x45um aluminium trace with a thickness of 3um. At 263GHz, the skin depth is ~160nm. We created two edge ports p1 and n1 for the simulation (see part 1 of the figure). In the simulation setting (see part 2 of the figure), we fixed the mesh density and the TL Mesh and changed the edge mesh setting. For automatic setting, we got edge mesh with a width of 669nm and resistance of 0.672 Ohm. Being aware that this edge mesh may be too thick (thicker than the skin depth), we tried different value for the edge mesh width, from 1nm (very fine) to 100nm. You could see the change of the resistance and the current density. Noticeably, for finer edge mesh, the current does not flow on the two side walls of the trace, but rather inside. The resistance has a peak value for 10nm edge mesh width. Overall, the Q value of our slab inductor varies from about 25 to 55.

We are quite confused about what edge mesh width we should pick for the realistic loss of the trace. Could you provide some hints on this?

Thank you a lot.

Anh

Attachments

edge_mesh_MoM.JPG403.3 KB

No one else has this question

Outcomes

volker_muehlhaus Oct 1, 2020 11:46 AM

Hi Anh,

it doesn't make sense to manually force edge mesh to values much smaller than skin depth. From my experience, automatic edge mesh width works well. Current density also looks trustworth for automatic skin depth.

I have some doubts about using TL mesh here, with very small mesh in the xy-plane but not in z direction. I would expect a more "balanced" mesh density to work better.

Just or reference, I used another trusted MoM solver which gives 681 mOhm for your line. That is very close to your Momentum result with default edge mesh.

Best regards

Volker

Actions

anhchu_iis Oct 5, 2020 1:52 AM

Dear Volker,

Thank you a lot for your reply. I always think that the optimum edge mesh width should be close to the skin depth so that the edge singularity effect could be properly approximated. The automatic value of 669nm is however too thick to me. The current density is confusing since it is much higher on the two sides of the trace than on the top and bottom sides. Could you comment on that?

I used the TL mesh to maintain the accuracy without having a large mesh as in the case of using cells per wavelength setting alone. Here I actually follow the guidelines from the book “Microwave Circuit Modeling Using Electromagnetic Field Simulation” by Daniel G. Swanson and Wolfgang J. R. Hoefer that using 4-6 subsections across the conductor width could reach an error of about 1% (page 101). The coarse mesh in the z-direction is a good point. However, I think it is from the way Momentum models the thick conductor so I actually don’t know if we could add more elements. Could you provide here a hint?

I did some more simulations using automatic edge mesh without the TL mesh and the results are in the attached image. The edge mesh width decreases and the resistance actually increases with finer mesh. At 2000 cells per wavelength (CpW) setting, the current density approximates that in my previous simulation with 20nm edge mesh. Should I trust the finer mesh setting here?

One more thing is the resistance of the trace is much higher in microwave mode than in RF mode (see parts 2 and 3 of the image). I could guess the additional radiation loss but the trace is about 1/7 of the wavelength (wavelength in the substrate is about 330um). Do you think the radiation effect is that lossy?

Thank you very much.

Best,

Anh

Actions

volker_muehlhaus @ anhchu_iis on Oct 5, 2020 2:18 AM

Hi Anh,

I understand your point, but don't know if edge mesh width = skin depth is a good choice, because I don't know what possibly smart correction or special MoM basis function is built into that cell.

What Momentum solves is a hollow conductor, with sheets of a given surface impedance. That surface impedance of current sheets is calculated from conductivity and geometry, which is not trivial when the conductor has such an aspect ratio.

Regarding TL mesh, I have great respect for Dan Swansons work, but that advice is for flat conductors. As we see here, transmission line mesh does not add extra lines on the side walls. But from a physical understanding, skin effect will happen on all sides. I have no answer what is best in this situation.

Regarding extra loss in microwave model, I think that is radiation - or in more general, some higher mode effect. I have seen the same thing for RFIC inductors at high frequency, and could confirm that it was radiation in my case, confirmed by another independent 3D FDTD solver. You can look at the radiation pattern in Momentum and see if that looks physically meaningful.

I would suggest that you contact Keysight support, and hopefully the Momentum team in Gent can provide some authorative answer how to handle this in the most reliable way.

Best regards

Volker

Actions

anhchu_iis @ volker_muehlhaus on Oct 5, 2020 4:36 AM

Thank you a lot Volker for your honest answers. I will contact Keysight support as you suggested.

Best,

Anh

Actions

4 Replies

volker_muehlhaus Oct 1, 2020 11:46 AM
Mark Correct
Correct Answer
Hi Anh,

it doesn't make sense to manually force edge mesh to values much smaller than skin depth. From my experience, automatic edge mesh width works well. Current density also looks trustworth for automatic skin depth.

I have some doubts about using TL mesh here, with very small mesh in the xy-plane but not in z direction. I would expect a more "balanced" mesh density to work better.

Just or reference, I used another trusted MoM solver which gives 681 mOhm for your line. That is very close to your Momentum result with default edge mesh.

Best regards
Volker
Like • Show 0 Likes0
Actions
anhchu_iis Oct 5, 2020 1:52 AM
Mark Correct
Correct Answer
Dear Volker,

Thank you a lot for your reply. I always think that the optimum edge mesh width should be close to the skin depth so that the edge singularity effect could be properly approximated. The automatic value of 669nm is however too thick to me. The current density is confusing since it is much higher on the two sides of the trace than on the top and bottom sides. Could you comment on that?

I used the TL mesh to maintain the accuracy without having a large mesh as in the case of using cells per wavelength setting alone. Here I actually follow the guidelines from the book “Microwave Circuit Modeling Using Electromagnetic Field Simulation” by Daniel G. Swanson and Wolfgang J. R. Hoefer that using 4-6 subsections across the conductor width could reach an error of about 1% (page 101). The coarse mesh in the z-direction is a good point. However, I think it is from the way Momentum models the thick conductor so I actually don’t know if we could add more elements. Could you provide here a hint?

I did some more simulations using automatic edge mesh without the TL mesh and the results are in the attached image. The edge mesh width decreases and the resistance actually increases with finer mesh. At 2000 cells per wavelength (CpW) setting, the current density approximates that in my previous simulation with 20nm edge mesh. Should I trust the finer mesh setting here?

One more thing is the resistance of the trace is much higher in microwave mode than in RF mode (see parts 2 and 3 of the image). I could guess the additional radiation loss but the trace is about 1/7 of the wavelength (wavelength in the substrate is about 330um). Do you think the radiation effect is that lossy?
Thank you very much.
Best,
Anh
Like • Show 0 Likes0
Actions
- volker_muehlhaus @ anhchu_iis on Oct 5, 2020 2:18 AM
  Mark Correct
  Correct Answer
  Hi Anh,
  
  I understand your point, but don't know if edge mesh width = skin depth is a good choice, because I don't know what possibly smart correction or special MoM basis function is built into that cell.
  
  What Momentum solves is a hollow conductor, with sheets of a given surface impedance. That surface impedance of current sheets is calculated from conductivity and geometry, which is not trivial when the conductor has such an aspect ratio.
  
  Regarding TL mesh, I have great respect for Dan Swansons work, but that advice is for flat conductors. As we see here, transmission line mesh does not add extra lines on the side walls. But from a physical understanding, skin effect will happen on all sides. I have no answer what is best in this situation.
  
  Regarding extra loss in microwave model, I think that is radiation - or in more general, some higher mode effect. I have seen the same thing for RFIC inductors at high frequency, and could confirm that it was radiation in my case, confirmed by another independent 3D FDTD solver. You can look at the radiation pattern in Momentum and see if that looks physically meaningful.
  
  I would suggest that you contact Keysight support, and hopefully the Momentum team in Gent can provide some authorative answer how to handle this in the most reliable way.
  
  Best regards
  Volker
  Like • Show 0 Likes0
  Actions
  - anhchu_iis @ volker_muehlhaus on Oct 5, 2020 4:36 AM
    Mark Correct
    Correct Answer
    Thank you a lot Volker for your honest answers. I will contact Keysight support as you suggested.
    
    Best,
    Anh
    Like • Show 0 Likes0
    Actions

Edge mesh setting in Momentum

Attachments

Outcomes

Related Content

Recommended Content