| T O P I C R E V I E W |
| firat |
Posted - Dec 02 2015 : 21:09:17
Is there a way to compute inductance matrices with port terminals connected to different metallic islands? I wasn't successful with modifying simple example given in FastHenry manual.
Firo |
| 5 L A T E S T R E P L I E S (Newest First) |
| Enrico |
Posted - Dec 23 2015 : 17:08:53
I'm not 100% sure I follow your reasoning, but not knowing how your structure is, and what is your ultimate goal, it is difficult to provide further suggestions. I may recommend however to read Matt Kamon's PhD thesis "Fast Parasitic Extraction and Simulation of Three-Dimensional Interconnect via Quasistatic Analysis" available for free on the M.I.T. web site w*w.rle.mit.edu/cpg/ (just replace the * with w, this is the anti-spam filter changing the characters). This paper in particular deals with the quasistatic EM field parasitic extraction problem (i.e. not only magnetic or electrical, but the combination of the two, while still under the quasi-static assumption).
Regarding instead your direct question, yes with FastCap of course you can do whatever you want, as you have full access to the source code. There is a debug mode (to be set with a #define statement in the source) that enables outputting the panel charges.
If instead you want a ready-to-use solution, FasterCap integrates directly the option, as described in the embedded online help:
quote:
'Dump charge densities in output file (-c)' dumps a set of FasterCap mesh geometries containing charge densities information. One set of files is output, in the same directory of the input file, for each conductor specified. These files are in FastModel compatible format, so they can be loaded into FastModel for charge density visualization. Please be aware however that for high levels of refinement the output files can be huge, and can also take some time to be stored on the mass memory device.
The main files are named after the specified input file, merged with the conductor name, with the extension '.lst'. All the other input files are referenced inside the main files.
Ciao
Enrico
|
| firat |
Posted - Dec 21 2015 : 17:47:23
Hi Enrico,
Thank you for your detailed response. A collection of metallic islands with terminals of each port defined on different islands is an electrostatic system(given that the size of the system is small compared to the wavelength of interest). By electrostatic I mean that once we take the excitation frequency to DC, magnetic fields will vanish and we'll be left only with non-zero electric fields. This suggests that FastHenry is not the proper tool to use since it does magnetostatic analysis.
What I would like to do is then to extract parasitic inductances associated with an electrostatic system. To do this I can first solve the system with FastCap to get the charge distributions. Then as the next order correction to the electrostatic approximation the continuity equation will give me div(J) assuming time-varying charges. Then I can solve for J assuming further an irrotational J.
Hence my question is now: is there a way(maybe using FreeCAD or so) to import/export meshes together with the solution data so that I can post-process the data (solve a Poisson equation to get J for example)? I have already access to Ansys tools(HFSS,Q3D etc.) but couldn't export meshes and simulation data somehow.
Ciao
quote:
Originally posted by Enrico
I'm afraid you cannot. Let's discuss a bit the topic, and start from Maxwell's equations (for sinusoidal steady state). The Ampere-Maxwell law says that
Curl(H) = J + jweE
Where H is the magnetic field, j is the imaginary constant, w is the angular frequency, e the permittivity, E is the electrical field, J is the current density.
The quasi-static assumption is to consider frequencies of interest small enough such that the term jweE (the displacement current) can be neglected.
Therefore, you cannot account for any gap in the circuits, where the actual current (as movement of electrons) is 'replaced' by an electric field. The simplest and classical case is a capacitor along a closed path.
In mathematical terms, if you take the divergence of the above equation, you get
div(J) = 0
(as div(curl(A)) = 0 for any vector A). This is exactly current conservation.
Seen in another way (theoretical guys will excuse me), as the quasistatic assumption means that all interactions happen instantaneously (neglecting the finite speed of light), for small gaps you can just short the nodes together as I already mentioned, or otherwise you need an actual segment in between to have a closed circuit (path) for the current to flow between the port terminals.
Best Regards,
Enrico
Firo |
| Enrico |
Posted - Dec 10 2015 : 17:57:21
I'm afraid you cannot. Let's discuss a bit the topic, and start from Maxwell's equations (for sinusoidal steady state). The Ampere-Maxwell law says that
Curl(H) = J + jweE
Where H is the magnetic field, j is the imaginary constant, w is the angular frequency, e the permittivity, E is the electrical field, J is the current density.
The quasi-static assumption is to consider frequencies of interest small enough such that the term jweE (the displacement current) can be neglected.
Therefore, you cannot account for any gap in the circuits, where the actual current (as movement of electrons) is 'replaced' by an electric field. The simplest and classical case is a capacitor along a closed path.
In mathematical terms, if you take the divergence of the above equation, you get
div(J) = 0
(as div(curl(A)) = 0 for any vector A). This is exactly current conservation.
Seen in another way (theoretical guys will excuse me), as the quasistatic assumption means that all interactions happen instantaneously (neglecting the finite speed of light), for small gaps you can just short the nodes together as I already mentioned, or otherwise you need an actual segment in between to have a closed circuit (path) for the current to flow between the port terminals.
Best Regards,
Enrico
|
| firat |
Posted - Dec 07 2015 : 17:32:48
I mean connecting the two sides of the port to electrically disconnected islands. Even in the quasi-static regime one can excite the islands with an AC voltage source between them and induce magnetic fields and currents. I wonder if I can extract inductances associated with this description with FastHenry.
Firo |
| Enrico |
Posted - Dec 04 2015 : 18:31:28
Do you mean connecting the two sides of the port to electrically disconnected islands? In this case, no: FastHenry does not model field radiation, as it works in quasi-static regime, so not having any current loop will prevent you to have a solution.
If instead you are looking for a way to connect different electrical parts together without using a segment in between, you can use the .equiv statement to logically short two nodes (but be careful as the current will 'jump' from one node to the other; so this is an approximation that can be used only in some cases, e.g. very short distance, where the actual current path between the two nodes can be neglected)
Best Regards,
Enrico
|
|
|
