| T O P I C R E V I E W |
| yan |
Posted - Sep 27 2023 : 10:11:04
Dear Enrico:
I recently found a paper "Fast Parameter Extraction of General Interconnects Using Geometry Independent Measured Equation of Invariance", In this paper, it compared with FastCap, there is a scene is groud place(Fig.11 A right-angel bend over a ground plane)#12290;
for "the ground place", how will fastercap deal with? |
| 6 L A T E S T R E P L I E S (Newest First) |
| yan |
Posted - Oct 15 2023 : 05:25:31
Dear Enrico:
1. Thank you for your patient explanation. I understand the physical significance of the ground plane in 3D as you described. However, my confusion stems from the fact that many papers, even when addressing 3D problems, deliberately define a ground plane. I'm curious as to why they do this. Is it necessary to specify a ground plane when simulating a system on a chip or capacitance extraction of general three-dimension VLSI interconnect??
2. I'm trying to reproduce the results from the literature to validate my assumptions. I attempted to replicate the results from the paper "Fast Parameters Extraction of General Three-Dimension Interconnects Using Geometry Independent Measured Equation of Invariance", specifically "Fig5 Single plate on ground plane". Below is the file I defined and the output results.
sample.lst
* conductor | dielectric | offset
* name | constant | in space
*
C cube_10_5_1.qui 3.9000000 0.0 0.0 0.0
C cube_substrate.qui 3.9000000 2.5 -5.0 -2 +
C cube_substrate.qui 3.9000000 2.5 5.0 -2 +
C cube_substrate.qui 3.9000000 -2.5 5.0 -2 +
C cube_substrate.qui 3.9000000 -2.5 -5.0 -2 +
cube_10_5_1.qui
T 3 0.000000000 0.000000000 0.000000000 5.000000000 10.000000000 0.000000000 5.000000000 0.000000000 0.000000000
T 3 0.000000000 10.000000000 0.000000000 5.000000000 10.000000000 0.000000000 0.000000000 0.000000000 0.000000000
T 3 0.000000000 0.000000000 0.000000000 0.000000000 10.000000000 1.000000000 0.000000000 10.000000000 0.000000000
T 3 0.000000000 0.000000000 1.000000000 0.000000000 10.000000000 1.000000000 0.000000000 0.000000000 0.000000000
T 3 5.000000000 0.000000000 0.000000000 5.000000000 0.000000000 1.000000000 0.000000000 0.000000000 0.000000000
T 3 0.000000000 0.000000000 0.000000000 5.000000000 0.000000000 1.000000000 0.000000000 0.000000000 1.000000000
T 3 0.000000000 0.000000000 1.000000000 5.000000000 0.000000000 1.000000000 5.000000000 10.000000000 1.000000000
T 3 0.000000000 10.000000000 1.000000000 0.000000000 0.000000000 1.000000000 5.000000000 10.000000000 1.000000000
T 3 5.000000000 0.000000000 0.000000000 5.000000000 10.000000000 0.000000000 5.000000000 10.000000000 1.000000000
T 3 5.000000000 0.000000000 1.000000000 5.000000000 0.000000000 0.000000000 5.000000000 10.000000000 1.000000000
T 3 5.000000000 10.000000000 0.000000000 0.000000000 10.000000000 0.000000000 5.000000000 10.000000000 1.000000000
T 3 0.000000000 10.000000000 0.000000000 0.000000000 10.000000000 1.000000000 5.000000000 10.000000000 1.000000000
cube_substrate.qui
T 3 0.000000000 0.000000000 0.000000000 5.000000000 10.000000000 0.000000000 5.000000000 0.000000000 0.000000000
T 3 0.000000000 10.000000000 0.000000000 5.000000000 10.000000000 0.000000000 0.000000000 0.000000000 0.000000000
Capacitance matrix is:
Dimension 2 x 2
g1_3 2.50321e-009 -1.98481e-009
g2_3 -1.98435e-009 3.87537e-009
The value of C11 is similar to the results in Table 4 of the paper. Is my simulation result accurate?
Best,
Yan |
| Enrico |
Posted - Oct 13 2023 : 17:02:29
Hi Yan,
there is a fundamental difference in 2D and 3D electromagnetic simulations. You may find a formally correct explanation in any good physics text book, so I'll try to give you here a more "intuitive" clarification.
In 3D you mimic as much as possible the 'reality'. In this case, there no concept of 'ground plane' per se, as no conductor is 'at zero potential'. You should know that the value of a potential is defined only as a difference, and not in absolute value - as there is an arbitrary constant involved there. What we are interested then is the potential difference, that causes a charge accumulation on the set of conductors. No conductor is actually a 'ground'. You give this role to this or that conductor, according to what you are trying to achieve. Note however that it is often useful to assume that a conductor is at zero potential. This is usually 'solved' by assuming that the potential at infinity is zero, but this is an arbitrary assumption (it is actually natural, as potential in 3D goes as 1/r where r is the distance. So if the constant is zero, when r -> infinity the potential is zero). You could assume that the potential at infinity is 10, and you would be equally right. Said that, you may now assume that your ground is at zero volts, meaning that it has the same potential as the potential at infinity: there will be no potential difference between this conductor and infinity. Still, you are free to define the potentials on your conductors - you just fixed a reference value.
In 2D things work a bit differently. In this case, to use a visual image, you can imagine your conductors as having a third dimension, that is uniform and extending... to infinity. So you cannot assume that you have zero potential at infinity. This I think is clear from your question itself. Slightly more formally, you again have an arbitrary constant but now in 2D your potential goes as log(r) and this does NOT go to zero when r -> infinity. So to remove the constant you must instead impose the condition that the net charge of the system is zero; BUT this prevents you from being able to arbitrarily specify the potential on all your conductors. So you can only specify the potential between each of the conductors and a special reference conductor i.e. the ground one.
So coming to the conclusion, in 3D you do not need a ground plane per se - you need to represent all the conductors that are meaningful for your simulation. You may read also my paper "The Maxwell Capacitance Matrix" for some further insights, you can find it here ht*ps://w*w.fastfieldsolvers.com/Papers/The_Maxwell_Capacitance_Matrix_WP110301_R02.pdf (just change the anti-spam characters with the obvious replacements).
Best,
Enrico
|
| yan |
Posted - Oct 11 2023 : 04:00:22
Dear Enrico:
I apologize, but as a beginner, I revisited the documentation and came across 'cube_over_ground_plane.lst' in the samples. From there, I gathered that a 'ground plane' refers to the setup of a conductive plane. However, I still have a question. What are the practical use cases for a ground plane, and when is it necessary to implement one? In 2D simulations, a conductive plane is designated as the ground plane because it's not possible to specify an infinitely distant ground. However, in 3D simulations, when should one consider incorporating a ground plane? For instance, in your DRAM cell example, is it essential to include a ground plane? |
| Enrico |
Posted - Oct 10 2023 : 18:14:08
Hi Yan,
sorry but these are quite basic questions. I suggest you refer to the FasterCap embedded online help as well as to the sample files in the Samples directory.
Also the documentation on FastCap can help you, as your question are equivalent for this tool.
Best,
Enrico
|
| yan |
Posted - Oct 07 2023 : 03:36:18
Dear Enrico:
Thanks for your reply, I still have some questions. Can you help me answer them?
1. How to define a "ground place", use 'C' or 'D'?
2. In what scenarios does the ground place need to be used?
Best,
Yan |
| Enrico |
Posted - Oct 04 2023 : 12:46:50
Dear Yan,
surely you can simulate this structure. You can check the sample file capacitor.lst in the Sampes directory and work from there to create the geometry fitting the bend instead of a parallelepiped.
Best,
Enrico
|
|
|
