Dave:
I was very glad to hear that my codes can work
for FEMM program, and thank you for testing it for me. You are
welcome.
The problem is that the field near corners isn't very accurately interpolated from nodal values over triangles. The field at the corners can be singular (or nearly so), and the direction of the field can change abruptly as you go around a corner. I have been dealing with this by just sampling the flux density in the elements directly around the corner node, choosing the biggest flux density, and then setting the direction based on the "original" direction of the flux. This isn't really that satisfying a way to do things, but it makes things look ok in density plots... Another thing. These days, I wrote the fast point location routines that Si hang Wow--this is excellent. I was able to compile your code without any problems, and it seems to work great! For example, the time that it took to compute the torque on a model of a synchronous PM motor with about 42,000 node points was reduced by a factor of 8. It also works fine when the domain is non-convex and multiply connected. Thank you very much for taking the time to put this together--it definitely improves the program's performance. Dave Meeker -- http://members.aol.com/dcm3c Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service. |