Re: [femm] Re: number of points in material defintion

["Robert Macy" <macy@xxxxxxxxxxxxxx>]

I took points from manufacturer's data curve. When femm plotted the data,
the curve did not follow the data points.

Yet, the original data was monotonic and "made sense". Please explain.

- Robert -

----- Original Message -----
From: mailto:David Meeker <dmeeker@xxxxxxxx
To: femm@xxxxxxxxxxxxxxx
Sent: Tuesday, February 04, 2003 6:47 AM
Subject: [femm] Re: number of points in material defintion


--- In femm@xxxxxxxxxxxxxxx, "Michael Salloker
<michael.salloker@xxxx>" <michael.salloker@xxxx> wrote:
> Hello,
>
> I wonder how many points un B-H-curve are necessary to get a good
> description of a ferromagnetic material.
>
> I found out, that increasing the number of points also increase the
> speed of the calculation, so its necessary to find a sufficient large
> number of data points.
>
> So imagine I have a material with a saturation of about 1T, i could
> do a spacing of 0.1 T, or 0.05T aso. In general I believe its better
> to make aquidistant points for B and for H, due the nonlinearity.
>
> Whats is the experts opnion?
>
> Thanks
>
> Michael

The program interpolates the entered points using cubic splines.
Cubic splines are twice continuously differentiable, so you don't need
an excessive number of points to get a smooth BH curve.

Generally, it is good to have more points around the "knee" of the BH
curve. If there are not enough points around the knee, the program
can't fit a monotonically increasing BH curve through the points. The
program then smooths the data points repeatedly until the fit becomes
monotonic--the cost for this smoothness is that the curve doesn't pass
exactly through the defined points.

If you push the "plot BH curve" button on the BH point entry dialog,
the program plots up the fitted curve, along with the data points. If
the program had to smooth, the fitted curve doesn't exactly pass
through the data points. You can add more points (typically around
the knee) to get things so that the program doesn't have to do any
smoothing--although the smoothed curve that doesn't exactly pass
through the defined points might be good enough for your purposes
without having to add in extra points.

Another consideration is that you'd typically want to make sure that
you extend out the BH curve far enough so that you aren't
extrapolating the BH curve at any of the points in your solution.

Typically, I try to make the points somewhat evenly spaced in sort of
a line integral sense. For example, if dB and dH are the differences
in B and H respectively between successive points on the BH curve, and
Bmax and Hmax are the last point on the curve, I try to make:
sqrt((dB/Bmax)^2 + (dH/Hmax)^2)
be sort of constant. However, this is just sort of a rule of thumb.

Typically, most of the points in the materials library have about 15
or so points.

Dave.
--
David Meeker
http://femm.berlios.de/dmeeker


Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.