| T O P I C R E V I E W |
| ksyew |
Posted - Jan 18 2008 : 10:53:34
Hi,
Regarding this command line option "-l {n|auto}" which is about the partitioning level of the multipole algorithm. May I ask
1) what is multipole algorithm?
2) what is the default number on n?
3) If the n number is higher, will the inductance accuracy better?
4) Is there any relevancy between this level partitioning and the segmenattion of the ground plane?
Thank you very much. Nice day |
| 1 L A T E S T R E P L I E S (Newest First) |
| Enrico |
Posted - Feb 01 2008 : 13:35:34
1. Simply put, the multipole algorithm is an algorithm used to perform a fast multiplication of a large dense matrix by a vector.
The topic anyway is quite complex, so I suggest you to read the articles about the fast algorithms behind FastCap and FastHenry.
You can find them following the 'links' page on w*w.fastfieldsolvers.com, then clicking on 'MIT CAD group'.
You can look under 'Research->Publications' or 'People->Alumni'. I suggest you to start with K. Nabor's thesis "Fast Three-Dimensional Capacitance Calculation" that, even if related to capacitance, gives you the background of the algorithms then employed in FastHenry as well.
2. Default is 'auto', so the software will decide when to stop; there is no default fixed n
3. Yes, but the relationship is not direct; n is related to the accuracy of the multipole algorithm, which in its turn is used as part of a matrix inversion algorithm. So if, for instance, your input discretization is insufficient, there is no way to get better results by increasing n.
4. Indirect again, as per 3. Anyway, if this is your question, increasing n won't increase the segmentation of the gnd plane
Enrico
|
|
|
