Refresh my memory, did you reply to me regarding the equations used
to calculate mutual inductance in that solnoid3 program?
That's a good example of knowing a program's limitations and there
ain't no way of knowing without knowing the internals (or is it
infernals)
Anyway, I never did find that Neumann double line integral, but one
day I *will* get Grover.
=======================================
Mike,
I vaguely remember an argument about Solnoid3 which almost extended to
fisticuffs. It began because some nitpicker had got a different answer to
what he had expected with a 2-turn coil. All the usual warriors joined in.
You will forgive the exaggerations.
As everybody knows a coil has length, diameter and nunber of turns. The
program was novel and ambitious in that it covers from one turn, via a
stretched-out helix, to a straight length of wire. The problem arose because
of how I had defined length. It becomes complicated at 1, 2 and 3 turns when
one is obliged to take into account both wire diameter and spacing between
turns or winding pitch. Bear in mind that the program uses an explicit
function like inductance L = F( Number of turns, Wire diameter, Winding
pitch, Length of coil former) which contains one variable too many for
comfort. But we cannot remove pitch because it affects length just as much
as the number of turns and wire diameter.
To simplify matters the number of turns was restricted to integers. But one
cannot avoid the fact that coil length changes in steps whenever the number
of turns changes. So the question arises between which two points along the
coil should the length be defined or measured. The existence of winding
pitch gets in the way.
I forget the details, and I'm not taking the trouble to sort through the
source code, but I defined length as being directly related to N-1 turns
which fits in very nicely with one turn, and with the all important maths.
Of course, beyond a very few turns, within the degree of accuracy required,
it doesn't matter whether its N or N-1. But there are substantial errors
with only 1 or two turns. How can winding pitch (required for wire
insulation and cooling) be accounted for?
It think the program reduces the length of 1 turn to be the same as wire
diameter.
The arguments soon veered, of course, away from how accurate was the program
but on to how many turns there were on a given length. It all depends on how
one thinks about it or which side of the coil it is viewed from.
This is where we came in - an error being caused because the user did not
understand how the program works. I solved the problem by amending the
program operating notes. But half of the participants in the argument were
still convinced there was only one number of turns and that was N and to
hell with the maths.
I also recollect taking the trouble to see what Grover had done about it.
Obviously he must have met with the problem. But he deliberately ignores
it. Winding pitch disappears from length measurements when there are only a
few turns.
He preferred to attach less importance to it than the newsgroup did. He was
probably right.
Incidentally, I didn't use Grover to produce the program. I used Nagaoka
with some modifications for most of it.
