[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: ACMI bug fix



Original poster: "Paul Nicholson by way of Terry Fritz <twftesla-at-qwest-dot-net>" <paul-at-abelian.demon.co.uk>

Bart,

> Out of curiousity, what was the bug?

A mistake in the use of the radius when calculating M. This bug
would not show up with a constant-radius coil, ie a solenoid. 
I noticed it when running Dave's three trifilar strands - the 
resulting k between strands was well short of unity - so I went
hunting for the bug.  The mistake had crept in when I extended the
code to do arbitrary coil profiles.  The mistake stayed in because
I failed to test it properly with the data you and others have
previously provided - I just used solenoid models, not the flats (:

With the broken v0.6a, you might well have got different values of
M if one of your coils was a flat spiral, and you change the order
that they appear in the input file.  Bart - I think you noticed
something odd?

You'll find that there's an extra digit of precision in some of the
inductance outputs.  This is usually a *bad thing* since it's a
nonsense to report more digits than the accuracy of the calculation.
In this case, the extra digit helps to give smooth convergence when
you wrap a gradient descent optimisation around acmi - for seeking
geometries which max the L.  Just something I'm playing with but will
probably take out again. 
--
Paul Nicholson
--