Re: [femm] FE magnetics commercial ware

["Steven Stretz" <stretzsj@xxxxxxxxxx>]

Thanks for the input. I'm sure I will use it as I take the next dive into
finite element packages. As I was reading your reply I thought of some of
the user interface issues with 3D that I had dealt with in the past. You
must admit that creating the 3D geometry is a challenge, in general (coils
are particularly fun). That's why it is important that the software support
some sort of "foreign", but standard, 3D file format for geometry input.
I've told every vendor so far that it is rediculous wasting time trying to
write a "3D CAD" front end when there are perfectly good package on the
market that do a much better job (Solidworks, Solid Edge, Cadkey, Ironcad,
on and on) and at a reasonable price. Some are listening.

3D has a long way to go, still, from pre and post processors to the solvers
themselves. Compute time is an issue that will take care of itself as it
always has (Last year at this time did you think you could have a 1.2GHz
processor on your desk?). The more basic issue is need. There are times
when you NEED 3D. There are may more times when it would be nice but
provides little improvement in accuracy or insight than what you can acheive
with 2D and an understanding of what you're analyzing. Try getting a handle
on 3D eddy current analysis or a reliable value for iron loss, low and high
frequency, how about 3D transient analysis. The Magsoft and Ansoft approach
that allows external circuits to be used with a 2D model and interface with
other modeling packages like Symplorer offer some powerful capabilities,
beyond 3D in many ways.

Sorry for rambling, thanks again.
----- Original Message -----
From: "David Meeker" <dmeeker@xxxxxxxxxxxxxxxxx>
To: <femm@xxxxxxxxxxx>
Sent: Thursday, January 04, 2001 2:38 PM
Subject: Re: [femm] FE magnetics commercial ware


> Well, my first run-in with Amperes was back in about 1996, where a guy
that I
> worked with was running an old beta version of Amperes that some friend at
IES
> had given him. It was slow, hard to use, and gave some pretty whacky
results.
> However, the program has come along quite a bit in the last 5 or 6 years.
At
> work, we have a copy of it, and I use it on occasion, mostly for "open
core"
> problems where a 3-D program is really necessary.
>
> Anyhow, I've checked it in the past by modeling some relatively
complicated
> axisymmetric geometries as 3D problems in Amperes, and then comparing to
the
> results from femm. I have generally been satisfied by the level of
accuracy in
> the recent versions of Amperes, although it's certainly not _impossible_
to give
> it a pathological geometry. The user interface is sort of unusual, but
not too
> hard to use--it took me about a day of messing around with it to
comfortably say
> that I knew how to run the program. On the down side, it can take a
_long_ time
> run, and it can requires a really high-end machine to run it effectively.
It is
> normal for it to make require multiple gigabytes free as swap space for
> relatively dense meshes. It can also take a long time to do the numerical
> integrations that are required to evaluate forces. And, of course, the
program
> is licenses for on the order of $10K US per year (which seems like a lot,
but
> it's not out of line for low sales volume scientific software).
>
> We also have multiple seats Ansoft's Flux 2D and Flux 3D. The program
seems
> accurate enough, but the user interface looks like something from the
early
> 1980's. I've done a few sample problems on it, but it was such a drag to
use
> that I didn't re-install it when I changed computers about 8 months ago.
> However, the version that I have is about 3 years old, so things might
have
> improved since then.
>
> Some people that I work pretty closely with use Opera (Vector Fields) and
Magnet
> (Infolytica). I've generally been pleased with the results of these two
> programs. Recently, I've been involved in designing some machines that
really
> test the envelope of Infolytica's 3D abilities. We ran into some snags,
but the
> Infolytica people have been particularly responsive in helping us out, for
which
> they should be commended.
>
> Also, I've compiled a list of links to all of the commercial low-frequency
> magnetics codes that I could find, some 24 different sites. It's been
hiding on
> the femm website for a while at:
http://members.aol.com/GMagnetics/links.htm
> If I've missed any, let me know.
>
> "Steven Stretz" <stretzsj@xxxxxxxxxx> wrote:
> > Dave, I don't know if you get enough praise and thanks for your work.
> > It would be, is, great to have an open source finite element program
> available.
> > It would also be nice if it were to develop like a "Netscape" and the
> magnetics
> > community could improve and expand it. But alas that is not an easy
task. I
> > only wish I could contribute more but my expertise in the theory and
math is
> > limited. Anyway thanks for your time, work, and creativity. Best of
luck in
> 2001.
>
> I've sort of been hoping to entice someone to help extend the program's
> capabilites in future versions. I've been getting bogged down with other
stuff
> lately, and I haven't been able to devote lots of time to working on femm.
The
> program's capabilities are roughly equal to that of a "low-end" commercial
> solver, but it still lacks a lot of the useful features of the "high-end"
> programs. Who knows--it might even be fun to take a stab at a 3D
program....
>
> Dave.
>
>
>