Default mesh settings

I have a question about EM simulation accuracy. I'm working with a very simple layout to learn the Momentum portion of ADS. The layout is just a ground plane with a transmission line on top. With default settings of 20 cells/wavelength and transmission line mesh unchecked my S11 or S22 show a high inductive component of the impedance. All other calculators show just pure resistive impedance. If I increase the c/w to a high number or enable transmission line mesh with a number close to 10, the reactive portion of the impedance goes almost to zero. Changing these settings, however, increases simulation time noticeably. I suspect that once I import a real board with many layers and complex geometries, the simulation will take days to complete. My question is if I can't get accurate results for a transmission line simulation with default settings,will all other simulation be inaccurate? Am I missing something? I believe I have all ports set correctly. My simulation frequency range is 1-2GHz.

Hello,

this sounds wrong. Can you show some more details and results? Or perhaps upload an example project to show your issue?

Best regards

Volker

Hi Volker. I created a new workspace just to work on my question. For some reason the reactive portion is lower now but it's still there. Both the schematic microstrip simulation and linecalc show zero or almost zero reactance. In EM simulation the results depend on how the ground plane is configured. If the polygon is extended left and right beyond the ports, for example, you get one result. And if the left and right edges of the polygon line up with the ports, the result is different. I'm attaching the workspace. Thanks.

附图/附件

Practice_wrk.zip

Hi dmitry,

I see what you mean.We might have two effects here.

One effect is port calibration: port calibration is switched off (not possible) if your drawn ground plane extends outside the ground pin location.

The second effect is indeed mesh density: in your testcase, the ground plane mesh is incredibly coarse. For accurate calculation of the transmission lines, we need to get the correct currents and charge on the mesh cells. If you think about the capacitance (charge) on the coarse mesh cells, it is easy to understand that this mesh can't provide an accurate capacitance from signal conductor to the ground plane.

It is important to understand that this groundplane-mesh-size issue happens because you have an explicit drawn, meshed ground plane. For infinite ground as a boundary condition, there is no such coarse mesh issue.

Not sure about slot planes, they might also behave perfect in that aspect. Maybe check yourself or ask ADS support.

If you really need explicit drawn ground planes: You might have seen that you can set mesh density - relative or absolute - for each layer individually. I would try refining the ground plane mesh to some reasonable absolute value, so that the line has reasonably dense ground mesh cells to couple to.

Best regards

Volker

This is some good stuff for me to learn, Volker. Thanks a bunch. EM simulation is now quick and agrees with the schematic S-parameter simulation. Can you answer another question a little off topic. What is the best format to use for layout import in ADS if a board was created by another tool?

You're welcome :-)

For layout data transfer between tools, I often use DXF. That's robust vector data and althought the format wasn't originaly designed for PCB, it usually works well. Gerber RS274X is only my second choice, because the drill data transfer can be nasty.

申明：网友回复良莠不齐，仅供参考。如需专业帮助，请学习易迪拓培训专家讲授的ADS视频培训课程。