R
RobertMacy
Guest
On Sun, 01 Mar 2015 11:37:55 -0700, Tim Williams
<tiwill@seventransistorlabs.com> wrote:
No analytical output, only calculated results given a geometry, current,
frequency.
However, femm has LUA script. Using LUA script you can set up a 'sequence'
of analyses, just like you manually putting in values, drawings, etc. That
way, you can 'step through' changes in geometry and frequency. I often use
the script to step through frequency and for each result addend to an
output text file until completely through the spectrum. Then using the
text file as an input use FREE octave, Matlab clone, to plot and curve fit
to the accuracy desired. At least doing that sequence I have a bit more
trust in the formulas.
Also, for physical position, where two objects are moved around in space,
there is a 'trick' to maintaining relative accuracy WITHOUT increasing the
mesh to the point we're talking about a day long set of calculations, but
more like 20 minutes or less. For example, moving an object and changing
the mesh yields around 0.05% accuracy, BUT moving the object but NOT
changing the mesh can yield down into the 1ppm accuracy. but again that's
'relative' accuracy. but makes for much easier curve fitting.
<tiwill@seventransistorlabs.com> wrote:
"RobertMacy" <robert.a.macy@gmail.com> wrote in message
newsp.xutcqseu2cx0wh@ajm...
Still say, get a copy of femm 4.2 and model two cores in axisymmetric
mode. Both to get good estimates and to gain understanding of just how
insidious magnetic fields can be!
You can even 'recreate' Tim's formula.
But it doesn't have analytical output, does it?
You could do a plot of given geometry (some fixed ratio of width and
height) vs. distance and get a series of values, then curve fit a similar
equation. Which of course has the same problems as trying to draw up the
approximate equation from first principles, but with a known error bar
and
no need for hefty E&M equations. On the upside, even if you end up with
linear interpolations between data sets, you're doing pretty good.
Tim
No analytical output, only calculated results given a geometry, current,
frequency.
However, femm has LUA script. Using LUA script you can set up a 'sequence'
of analyses, just like you manually putting in values, drawings, etc. That
way, you can 'step through' changes in geometry and frequency. I often use
the script to step through frequency and for each result addend to an
output text file until completely through the spectrum. Then using the
text file as an input use FREE octave, Matlab clone, to plot and curve fit
to the accuracy desired. At least doing that sequence I have a bit more
trust in the formulas.
Also, for physical position, where two objects are moved around in space,
there is a 'trick' to maintaining relative accuracy WITHOUT increasing the
mesh to the point we're talking about a day long set of calculations, but
more like 20 minutes or less. For example, moving an object and changing
the mesh yields around 0.05% accuracy, BUT moving the object but NOT
changing the mesh can yield down into the 1ppm accuracy. but again that's
'relative' accuracy. but makes for much easier curve fitting.