Author 
Topic 

liping
7 Posts 
Posted  Oct 06 2023 : 13:38:33

Hello, I'm facing an issue related to automated 2D capacitance simulation. When I use automated methods to set the parameter d0.5, I've noticed that the solution runs faster in certain input scenarios. This behavior seems inconsistent with the documentation. Interestingly, when I directly run the menu with d0.5 through FasterCap, it indeed produces a faster solution.
My question is, when I use FasterCapObject.Run() to set d0.5, does FasterCap also potentially modify other parameters simultaneously? Because when I input a0.001 d0.5 for the dataz2.lst file, the iteration time becomes excessively long and the simulation cannot complete successfully. I've noticed that removing d0.5 allows the simulation to run normally. What's confusing to me is that when I input a0.001 d0.5 through FasterCap in the menu, it runs fine, which suggests that it's unlikely to be an issue with the input file. Moreover, this issue seems to only occur in certain cases, as it runs fine when the input file is dataz.lst. I would greatly appreciate any insights or assistance you can provide on this matter.
* 2D  dataz *
* data is included in a single file *
C cube_0.txt 2.5 0.0 0.0
C cube_1.txt 2.5 0.9539636313562916 0.24
C cube_2.txt 2.5 1.8704052341272623 0.24
C cube_3.txt 2.5 0.0 0.48
C cube_4.txt 2.5 1.745984802890689 0.48
* gnd plane below the signal lines C gnd_plane.txt 2.5 0.0 0.24
File cube_0.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.5662797250455648 0.06 S cube_2d 0.5662797250455648 0.06 0.5662797250455648 0.0 S cube_2d 0.5662797250455648 0.0 0.0 0.0 End
File cube_1.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.0920727372874171 0.06 S cube_2d 0.0920727372874171 0.06 0.0920727372874171 0.0 S cube_2d 0.0920727372874171 0.0 0.0 0.0 End
File cube_2.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.0770810568901302 0.06 S cube_2d 0.0770810568901302 0.06 0.0770810568901302 0.0 S cube_2d 0.0770810568901302 0.0 0.0 0.0 End
File cube_3.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.13 S cube_2d 0.0 0.13 1.3614262053193862 0.13 S cube_2d 1.3614262053193862 0.13 1.3614262053193862 0.0 S cube_2d 1.3614262053193862 0.0 0.0 0.0 End
File cube_4.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.13 S cube_2d 0.0 0.13 0.254015197109311 0.13 S cube_2d 0.254015197109311 0.13 0.254015197109311 0.0 S cube_2d 0.254015197109311 0.0 0.0 0.0 End
File gnd_plane.txt
0 1 unit segment * S gnd 0 0.0 2 0.0
End
* 2D  dataz2 *
* data is included in a single file *
C cube_0.txt 2.5 0.0 0.0
C cube_1.txt 2.5 0.9537777053676844 0.24
C cube_2.txt 2.5 0.0 0.24
C cube_3.txt 2.5 1.7211230623140632 0.24
C cube_4.txt 2.5 0.0 0.48
* gnd plane below the signal lines C gnd_plane.txt 2.5 0.0 0.24
File cube_0.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 1.2104390758085908 0.06 S cube_2d 1.2104390758085908 0.06 1.2104390758085908 0.0 S cube_2d 1.2104390758085908 0.0 0.0 0.0 End
File cube_1.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.0924445892646309 0.06 S cube_2d 0.0924445892646309 0.06 0.0924445892646309 0.0 S cube_2d 0.0924445892646309 0.0 0.0 0.0 End
File cube_2.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.0694995325917452 0.06 S cube_2d 0.0694995325917452 0.06 0.0694995325917452 0.0 S cube_2d 0.0694995325917452 0.0 0.0 0.0 End
File cube_3.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.0327350444904012 0.06 S cube_2d 0.0327350444904012 0.06 0.0327350444904012 0.0 S cube_2d 0.0327350444904012 0.0 0.0 0.0 End
File cube_4.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.13 S cube_2d 0.0 0.13 1.4986775766195883 0.13 S cube_2d 1.4986775766195883 0.13 1.4986775766195883 0.0 S cube_2d 1.4986775766195883 0.0 0.0 0.0 End
File gnd_plane.txt
0 1 unit segment * S gnd 0 0.0 2 0.0
End
The following is the report when run through automation, with the middle part omitted
Running FasterCap version 6.0.7 Copyright 2019 FastFieldSolvers S.R.L. ht*p://w*w.fastfieldsolvers.com, All Rights reserved Starting capacitance extraction with the following parameters: Input file: C:/Users/#38134;#27122;#20043;#21073;/Desktop/boxing/fastcap_solve/dataz2.lst Auto calculation with max error: 0.001 Remark: Auto option overrides all other Manual settings
2D Solver Engine invoked Solution scheme (g): Collocation, GMRES tolerance (t): 0.0005 Outofcore free memory to link memory condition (f): 5 Potential interaction coefficient to mesh refinement ratio (d): 0.5 Mesh curvature (mc): 3 Number of input panels to solver engine: 21
Iteration number #0 ***************************
*************************************** Refining the geometry.. Refinement completed Mesh refinement (m): 1e+032 *************************************** Computing the links.. Number of panels after refinement: 22 Number of links to be computed: 62 Done computing links *************************************** Precond Type(s) (p): Jacobi GMRES Iteration: 0 1 2 3 4 5 6 7 GMRES Iteration: 0 1 2 3 4 5 6 7 GMRES Iteration: 0 1 2 3 4 5 6 7 GMRES Iteration: 0 1 2 3 4 5 6 7 8 GMRES Iteration: 0 1 2 3 4 5 6 7 8 Capacitance matrix is: Dimension 5 x 5 g1_cube_2d 1.86845e010 6.72977e012 1.48021e011 2.8681e012 1.14486e010 g2_cube_2d 7.70814e012 5.47727e011 3.65003e012 2.33042e012 3.89788e011 g3_cube_2d 2.47472e011 1.20123e012 3.82928e011 8.23898e013 1.19584e011 g4_cube_2d 7.53201e012 5.94661e012 5.87511e013 3.35812e011 1.59169e011 g5_cube_2d 4.90338e011 1.96906e011 1.30791e011 6.72791e012 1.55671e010 Warning: capacitance matrix has a nonnegative offdiagonal element at row 1 col 4 Warning: capacitance matrix has a nonnegative offdiagonal element at row 2 col 3 Warning: capacitance matrix has a nonnegative offdiagonal element at row 4 col 1 Warning: capacitance matrix is not diagonally dominant due to row 3
Solve statistics: Number of input panels: 21 of which 21 conductors and 0 dielectric Number of input panels to solver engine: 21 Number of panels after refinement: 22 Number of potential estimates: 89 Number of links: 84 (uncompressed 484, compression ratio is 82.6%) Max recursion level: 4 Max Mesh relative refinement value: 1.63392 Iteration time: 0.013000s (0 days, 0 hours, 0 mins, 0 s) Iteration allocated memory: 16468 kilobytes
.......................................................
Iteration number #13 ***************************
*************************************** Increasing the geometric refinement.. Delta in refined panel and link count w.r.t. previous iteration less than 10% (Panels # 12032, Links # 129198404) Automatically increasing the refinement parameters Refinement completed Mesh refinement (m): 0.000746188 *************************************** Computing the links.. Number of panels after refinement: 22379 Number of links to be computed: 348923200 Estimated memory required for storing panel interaction links is: 5324 Mbytes Available free memory left is: 11737 Mbytes Estimated links memory is more than 20% of the available free memory. Going outofcore. Outofcore temporary files folder: 'C:\Users\#38134;#27122;#20043;#21073;\AppData\Local\Temp' Done computing links *************************************** Precond Type(s) (p): Jacobi GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 Capacitance matrix is: Dimension 5 x 5 g1_cube_2d 2.53204e010 3.62172e011 2.93502e011 3.24414e012 4.73204e011 g2_cube_2d 3.61954e011 7.7316e011 4.68352e014 7.77318e013 3.88815e011 g3_cube_2d 2.94876e011 6.35262e014 6.67483e011 2.16869e013 3.29524e011 g4_cube_2d 3.46781e012 8.01761e013 2.0838e013 4.3621e011 2.05264e011 g5_cube_2d 4.84927e011 3.90529e011 3.28927e011 2.04973e011 1.65168e010
Weighted Frobenius norm of the difference between capacitance (auto option): 0.00172351
Solve statistics: Number of input panels: 21 of which 21 conductors and 0 dielectric Number of input panels to solver engine: 21 Number of panels after refinement: 22379 Number of potential estimates: 312541289 Number of links: 348945579 (uncompressed 500819641, compression ratio is 30.3%) Max recursion level: 31 Max Mesh relative refinement value: 0.000746188 Iteration time: 3609.763916s (0 days, 1 hours, 0 mins, 9 s) Iteration allocated memory: 2406158 kilobytes
Iteration number #14 ***************************
*************************************** Increasing the geometric refinement.. Delta in refined panel and link count w.r.t. previous iteration less than 10% (Panels # 24064, Links # 516816624) Automatically increasing the refinement parameters Refinement completed Mesh refinement (m): 0.000373094 *************************************** Computing the links.. Number of panels after refinement: 44758 Number of links to be computed: 1395714960 Estimated memory required for storing panel interaction links is: 21296 Mbytes Available free memory left is: 11733 Mbytes Estimated links memory is more than 20% of the available free memory. Going outofcore. Outofcore temporary files folder: 'C:\Users\#38134;#27122;#20043;#21073;\AppData\Local\Temp' Done computing links *************************************** Precond Type(s) (p): Jacobi GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 GMRES Iteration: 0 1 2 Warning: user break request received, stopping execution, please wait..
Total allocated memory: 2421575 kilobytes Total time: 15657.608398s (0 days, 4 hours, 20 mins, 57 s)
Warning: Program execution stopped on user request!


liping
7 Posts 
Posted  Oct 06 2023 : 14:05:48

I seem to have found the issue. Can I ask, when using automation, if you don't include "d," is "ap" enabled by default, and if you add "d," is it defaultly disabled? 


Enrico
543 Posts 
Posted  Oct 06 2023 : 17:39:04

Hi Liping,
yes, you found the reason. In a nutshell, the defaults when running FasterCap via automation are necessarily different from the defaults you have when you use the GUI.
The point is, ap is used to tell FasterCap to use the automatic preconditioner selection. When launching FasterCap via automation, if this was the default, then you would need a switch to tell FasterCap NOT to use the automatic selection. The logic in the interface is that you select something if you want it, not that you select something that you don't want. This is the reason why in the GUI dialog you have a 'check' mark on ap. I realize this may seem confusing. Perhaps we should update the documentation explicitly stating this point. However I was also wondering how many people are really using the Automation interface; my main point is that this is actually pretty Windowsspecific. Possibly it could be a better solution to have an output parsable log. As you may know, you can launch FasterCap in pure text "shell" mode, so you could fully use it as a script command. Your opinion is welcome here, as obviously you are one of the users of the Automation interface.
Said that, regarding your last question, no: there is no change of parameters or meaning of the parameters when you specify 'd'. Simply, when using d0.5, you get much more links between panels, and this triggers a different preconditioner selection. The default Jacobi preconditioner is fine enough in your case when d is 1, so you don't appreciate a difference in speed; but when you go to d0.5 this becomes inefficient.
One final note, I'm not sure why you use a0.001 but most of the time this is kind of excessive. The expected accuracy of FasterCap is anyway around 1%. Actually if you really want to speed up your solutions, and your geometries are different but kind of similar expecially in the dimensions and dimension ratios, you may be better off by launching some automatic solves and then copy the parameters of the last minus one simulation and use those in manual mode. Last minus one because you are already at convergence at n1 iteration (the Frobenius norm tells you that between iteration n and iteration n1 the result has changed, in weigthed norm, less than the allotted value); manual mode because you avoid running all the n1 iterations and 'jump' directly to the last one.
Best, Enrico



liping
7 Posts 
Posted  Oct 06 2023 : 19:44:54

Thank you for your response, but I made some attempts earlier and still have some doubts. When I set only "a0.001" in the GUI dialog with all other settings at their defaults, the solving time becomes excessively long. However, when I change "d" from the default value of 1 to 0.5, it actually speeds up the process. This is entirely done through the fastercap interface and doesn't involve any automation. As for why I set "a0.001," it's because I noticed that when I set it to "a0.01," I get a warning after the last iteration , saying: "capacitance matrix has a nonnegative offdiagonal element at ......". Here is my input and output for reference * 2D  dataz 73.lst *
* data is included in a single file *
C cube_0.txt 2.5 1.4312336291624494 0.0
C cube_1.txt 2.5 0.8843663727106793 0.24
C cube_2.txt 2.5 0.0149953096784077 0.24
C cube_3.txt 2.5 1.1993942734652152 0.24
C cube_4.txt 2.5 1.3844605536900192 0.24
C cube_5.txt 2.5 0.1076211082903343 0.48
* gnd plane below the signal lines C gnd_plane.txt 2.5 0.0 0.24
File cube_0.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.5687663708375506 0.06 S cube_2d 0.5687663708375506 0.06 0.5687663708375506 0.0 S cube_2d 0.5687663708375506 0.0 0.0 0.0 End
File cube_1.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.2312672545786413 0.06 S cube_2d 0.2312672545786413 0.06 0.2312672545786413 0.0 S cube_2d 0.2312672545786413 0.0 0.0 0.0 End
File cube_2.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.0530056819859769 0.06 S cube_2d 0.0530056819859769 0.06 0.0530056819859769 0.0 S cube_2d 0.0530056819859769 0.0 0.0 0.0 End
File cube_3.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.1373056340489094 0.06 S cube_2d 0.1373056340489094 0.06 0.1373056340489094 0.0 S cube_2d 0.1373056340489094 0.0 0.0 0.0 End
File cube_4.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.06 S cube_2d 0.0 0.06 0.0655792964947632 0.06 S cube_2d 0.0655792964947632 0.06 0.0655792964947632 0.0 S cube_2d 0.0655792964947632 0.0 0.0 0.0 End
File cube_5.txt
0 cube 2d * S cube_2d 0.0 0.0 0.0 0.13 S cube_2d 0.0 0.13 1.8923788917096656 0.13 S cube_2d 1.8923788917096656 0.13 1.8923788917096656 0.0 S cube_2d 1.8923788917096656 0.0 0.0 0.0 End
File gnd_plane.txt
0 1 unit segment * S gnd 0 0.0 2 0.0
End
this is d0.5
Running FasterCap version 6.0.7 Copyright 2019 FastFieldSolvers S.R.L. ht*p://w*w.fastfieldsolvers.com, All Rights reserved Starting capacitance extraction with the following parameters: Input file: C:/Users/#38134;#27122;#20043;#21073;/Desktop/boxing/boxing/fastcap_solve/dataz 73.lst Auto calculation with max error: 0.001 Remark: Auto option overrides all other Manual settings
2D Solver Engine invoked Solution scheme (g): Collocation, GMRES tolerance (t): 0.0005 Outofcore free memory to link memory condition (f): 5 Potential interaction coefficient to mesh refinement ratio (d): 0.5 Mesh curvature (mc): 3 Number of input panels to solver engine: 25
Iteration number #0 ***************************
*************************************** Refining the geometry.. Refinement completed Mesh refinement (m): 1e+032 *************************************** Computing the links.. Number of panels after refinement: 26 Number of links to be computed: 80 Done computing links *************************************** Precond Type(s) (p): Jacobi GMRES Iteration: 0 1 2 3 4 5 6 7 8 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 GMRES Iteration: 0 1 2 3 4 5 6 7 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 GMRES Iteration: 0 1 2 3 4 5 6 7 8 GMRES Iteration: 0 1 2 3 4 5 6 7 8 Capacitance matrix is: Dimension 6 x 6 g1_cube_2d 8.02025e011 1.40321e012 2.56639e013 4.35743e012 2.15319e011 2.2334e011 g2_cube_2d 1.00075e011 9.3269e011 4.27875e012 1.74136e011 4.66139e013 1.05734e010 g3_cube_2d 3.96622e013 5.0887e012 3.26152e011 2.91966e013 5.24016e013 1.52407e011 g4_cube_2d 6.74516e012 2.29657e011 2.23913e012 8.79952e011 3.29277e011 2.99988e011 g5_cube_2d 2.13931e011 3.01123e013 8.00186e014 4.47225e011 6.87758e011 4.69428e012 g6_cube_2d 3.17019e012 3.40823e011 1.19022e011 1.11601e011 9.0407e012 1.8198e010 Warning: capacitance matrix has a nonnegative offdiagonal element at row 2 col 1 Warning: capacitance matrix has a nonnegative offdiagonal element at row 2 col 3 Warning: capacitance matrix has a nonnegative offdiagonal element at row 2 col 5 Warning: capacitance matrix has a nonnegative offdiagonal element at row 3 col 1 Warning: capacitance matrix has a nonnegative offdiagonal element at row 4 col 3 Warning: capacitance matrix is not diagonally dominant due to row 2 Warning: capacitance matrix is not diagonally dominant due to row 4 Warning: capacitance matrix is not diagonally dominant due to row 5
Solve statistics: Number of input panels: 25 of which 25 conductors and 0 dielectric Number of input panels to solver engine: 25 Number of panels after refinement: 26 Number of potential estimates: 120 Number of links: 106 (uncompressed 676, compression ratio is 84.3%) Max recursion level: 4 Max Mesh relative refinement value: 1.64631 Iteration time: 0.026000s (0 days, 0 hours, 0 mins, 0 s) Iteration allocated memory: 16475 kilobytes
……………………………………………………………………………………
Iteration number #14 ***************************
*************************************** Increasing the geometric refinement.. Refinement completed Mesh refinement (m): 0.00566714 *************************************** Computing the links.. Number of panels after refinement: 1975 Number of links to be computed: 3129744 Done computing links *************************************** Precond Type(s) (p): Twolevels, twolevels preconditioner dimension (ps): 1024 GMRES Iteration: 0 1 2 3 4 5 GMRES Iteration: 0 1 2 3 4 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 GMRES Iteration: 0 1 2 3 4 5 6 7 GMRES Iteration: 0 1 2 3 4 5 6 7 GMRES Iteration: 0 1 2 3 4 5 Capacitance matrix is: Dimension 6 x 6 g1_cube_2d 1.38469e010 2.84054e012 3.38167e013 8.90175e012 1.99676e011 3.12811e011 g2_cube_2d 2.82938e012 1.09138e010 8.16555e013 3.46024e011 1.05666e012 5.10358e011 g3_cube_2d 3.44747e013 8.27885e013 5.09718e011 3.39038e014 6.79974e015 3.2174e011 g4_cube_2d 8.90349e012 3.45975e011 3.39971e014 1.25137e010 5.02952e011 2.53544e011 g5_cube_2d 1.99751e011 1.05806e012 1.35654e014 5.02956e011 9.70013e011 2.42398e011 g6_cube_2d 3.13339e011 5.11059e011 3.2173e011 2.5353e011 2.41956e011 1.99866e010
Weighted Frobenius norm of the difference between capacitance (auto option): 0.000655675
Solve statistics: Number of input panels: 25 of which 25 conductors and 0 dielectric Number of input panels to solver engine: 25 Number of panels after refinement: 1975 Number of potential estimates: 2210376 Number of links: 3131719 (uncompressed 3900625, compression ratio is 19.7%) Max recursion level: 23 Max Mesh relative refinement value: 0.00566712 Iteration time: 0.658000s (0 days, 0 hours, 0 mins, 0 s) Iteration allocated memory: 312306 kilobytes
Total allocated memory: 312306 kilobytes Total time: 1.669000s (0 days, 0 hours, 0 mins, 1 s)
this is d1
Running FasterCap version 6.0.7 Copyright 2019 FastFieldSolvers S.R.L. ht*p://w*w.fastfieldsolvers.com, All Rights reserved Starting capacitance extraction with the following parameters: Input file: C:/Users/#38134;#27122;#20043;#21073;/Desktop/boxing/boxing/fastcap_solve/dataz 73.lst Auto calculation with max error: 0.001 Remark: Auto option overrides all other Manual settings
2D Solver Engine invoked Solution scheme (g): Collocation, GMRES tolerance (t): 0.0005 Outofcore free memory to link memory condition (f): 5 Potential interaction coefficient to mesh refinement ratio (d): 1 Mesh curvature (mc): 3 Number of input panels to solver engine: 25
Iteration number #0 ***************************
*************************************** Refining the geometry.. Refinement completed Mesh refinement (m): 1e+032 *************************************** Computing the links.. Number of panels after refinement: 26 Number of links to be computed: 80 Done computing links *************************************** Precond Type(s) (p): Jacobi GMRES Iteration: 0 1 2 3 4 5 6 7 8 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 GMRES Iteration: 0 1 2 3 4 5 6 7 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 GMRES Iteration: 0 1 2 3 4 5 6 7 8 GMRES Iteration: 0 1 2 3 4 5 6 7 8 Capacitance matrix is: Dimension 6 x 6 g1_cube_2d 8.02025e011 1.40321e012 2.56639e013 4.35743e012 2.15319e011 2.2334e011 g2_cube_2d 1.00075e011 9.3269e011 4.27875e012 1.74136e011 4.66139e013 1.05734e010 g3_cube_2d 3.96622e013 5.0887e012 3.26152e011 2.91966e013 5.24016e013 1.52407e011 g4_cube_2d 6.74516e012 2.29657e011 2.23913e012 8.79952e011 3.29277e011 2.99988e011 g5_cube_2d 2.13931e011 3.01123e013 8.00186e014 4.47225e011 6.87758e011 4.69428e012 g6_cube_2d 3.17019e012 3.40823e011 1.19022e011 1.11601e011 9.0407e012 1.8198e010 Warning: capacitance matrix has a nonnegative offdiagonal element at row 2 col 1 Warning: capacitance matrix has a nonnegative offdiagonal element at row 2 col 3 Warning: capacitance matrix has a nonnegative offdiagonal element at row 2 col 5 Warning: capacitance matrix has a nonnegative offdiagonal element at row 3 col 1 Warning: capacitance matrix has a nonnegative offdiagonal element at row 4 col 3 Warning: capacitance matrix is not diagonally dominant due to row 2 Warning: capacitance matrix is not diagonally dominant due to row 4 Warning: capacitance matrix is not diagonally dominant due to row 5
Solve statistics: Number of input panels: 25 of which 25 conductors and 0 dielectric Number of input panels to solver engine: 25 Number of panels after refinement: 26 Number of potential estimates: 114 Number of links: 106 (uncompressed 676, compression ratio is 84.3%) Max recursion level: 4 Max Mesh relative refinement value: 1.64631 Iteration time: 0.023000s (0 days, 0 hours, 0 mins, 0 s) Iteration allocated memory: 16475 kilobytes
…………………………………………..
Iteration number #20 ***************************
*************************************** Increasing the geometric refinement.. Refinement completed Mesh refinement (m): 0.00070839 *************************************** Computing the links.. Number of panels after refinement: 15809 Number of links to be computed: 77773854 Done computing links *************************************** Precond Type(s) (p): Twolevels, twolevels preconditioner dimension (ps): 1024 GMRES Iteration: 0 1 2 3 4 5 6 GMRES Iteration: 0 1 2 3 4 5 GMRES Iteration: 0 1 2 3 4 5 6 7 8 9 GMRES Iteration: 0 1 2 3 4 5 6 7 GMRES Iteration: 0 1 2 3 4 5 GMRES Iteration: 0 1 2 3 4 5 6 Capacitance matrix is: Dimension 6 x 6 g1_cube_2d 1.38515e010 2.8426e012 3.53255e013 8.90465e012 1.99848e011 3.13906e011 g2_cube_2d 2.85605e012 1.09198e010 8.39346e013 3.46503e011 1.05733e012 5.12006e011 g3_cube_2d 3.41624e013 8.2684e013 5.10072e011 3.40156e014 7.88542e015 3.21835e011 g4_cube_2d 8.89621e012 3.46544e011 2.58353e014 1.25322e010 5.04151e011 2.53052e011 g5_cube_2d 1.99945e011 1.0571e012 1.53708e014 5.04117e011 9.7151e011 2.42605e011 g6_cube_2d 3.13496e011 5.11183e011 3.22076e011 2.53579e011 2.42128e011 1.99814e010
Weighted Frobenius norm of the difference between capacitance (auto option): 0.0013294
Solve statistics: Number of input panels: 25 of which 25 conductors and 0 dielectric Number of input panels to solver engine: 25 Number of panels after refinement: 15809 Number of potential estimates: 50384934 Number of links: 77789663 (uncompressed 249924481, compression ratio is 68.9%) Max recursion level: 28 Max Mesh relative refinement value: 0.00070839 Iteration time: 16.836000s (0 days, 0 hours, 0 mins, 16 s) Iteration allocated memory: 1498021 kilobytes
Iteration number #21 ***************************
*************************************** Increasing the geometric refinement..
Warning: user break request received, stopping execution, please wait..
Total allocated memory: 3189 kilobytes Total time: 38.044998s (0 days, 0 hours, 0 mins, 38 s)
Warning: Program execution stopped on user request!



Enrico
543 Posts 
Posted  Oct 10 2023 : 18:21:45

quote: When I set only "a0.001" in the GUI dialog with all other settings at their defaults, the solving time becomes excessively long. However, when I change "d" from the default value of 1 to 0.5, it actually speeds up the process
This can be expected in case the default compression of couplings between the different conductors that you set with d1 is actually excessive. This is a parameter that is not very easy to set automatically; in particular, you should decrease d when the conductors, or part of them, are very close to each other. In this case, the process is faster as convergence is reached in a smaller number of iterations, as the matrix is in fact better conditioned. The cons of this approach is that of course the more links you have, the more memory you use and the more number of operations you need to perform. If this improves convergence, there is a breakeven point in the speedup: at a certain point, you have no compression any more and having a full matrix to invert will require order(N^2) operations, not order(N).
Best, Enrico



liping
7 Posts 
Posted  Oct 11 2023 : 18:48:15

I also suspected that this might be the case, but the documentation didn't mention that reducing the 'd' could potentially make the process faster, which left me somewhat puzzled. Sincere thanks for your continued patient responses all along. 



Topic 


