Re: [femm] FEMM Vs. commercial code

[David Meeker <dmeeker@xxxxxxxx>]

Keith Gregory wrote:
> [...]
> That said it is not bullet proof and David 
> would be the first to admit that it does fall over occasionally in ways 
> that most commercial code would not. 
Ha! I certainly would be the first to admit that the program isn't bullet-proof!
>
As had been noted, it's the entire group of people that are associated with
the program that are a big strength.  That is, people (e.g. Keith) who thoughtfully
answer questions; people who report bugs, make suggestions, and ask questions
when the program is doing something unexpected.  Several people have also
made essential contributions to the program on the source code level, e.g.
Si Hang with his fast triangle search algorithm and Robin Cornelius by adding
in the Lua scripting capabilities.  What's sort of the coolest thing about
it is that the people who are associated with the program come from all over
the world and hail from a number of different backgrounds, providing everyone
with a unique global networking opportunity.

The biggest "FEMM vs Commercial Code" difference that hasn't really been
mentioned yet in this thread is that the complete source for FEMM (even including
the LaTeX source for the manual) is available for download.  In contrast,
there is little insight into how the commercial programs are doing what they
do, aside from the occasional article in Transactions on Magnetics.  Now,
the typical user might not care a whole lot about the source, but the source
could be potentially valuable in a number of ways:

-- For students.  During the creation of FEMM, I kept finding that there
were crucial implementation details and "gotchas" that have been omitted
in the standard treatments on the subject.  Having the source allows students
to lift up the hood and check out an attempt at stringing all of the pieces
together into a working tool, resolving these "hidden" fine points.  Now,
the methods in FEMM might not always be the best or most optimal way that
one could write things (see above comment about "bullet-proof" and let me
invoke Lubarsky's Law of Cybernetic Entomology one more time), but thinking
about how things could be done better might be a useful exercise in itself.

--  For "power users."  I know of several people who have been twisted enough
to modify FEMM to suit their needs for some unique problem, which is possible
because the source is available.  This level of flexibility just isn't available
with commercial solvers.

-- If there are bugs.  If the program is doing something that seems "weird,"
it's possible to look inside it and see exactly what the program is doing
and whether or not it is reasonable.

-- It enables the code to "grow" faster.  Having the source has enabled people
to contribute a number of extensions to the program that have expanded the
functionality of FEMM in some big ways (see above).

At any rate, if anyone is interested in the range of what's out there as
far as commercial codes, I've listed all of the ones that I know about at:
http://femm.berlios.de/links.htm
 Exploring those links will paint a pretty good picure of the state of the
art as far as commercial programs.

Dave.
--
David Meeker
dmeeker@xxxxxxxx
http://femm.berlios.de/dmeeker