From: Paul
Date: Sun, 13 Aug 2000 09:36:31 +0100
Subject: [TSSP] Some progress to report.
The laplace solver has been vastly improved and can now
be relied upon to give reasonable answers. Problems with
incomplete relaxation, mentioned in the last message have
been solved. Changes to tlap are:
* Extended relaxation time with improved detection of
convergence.
* Increased math precision was necessary.
* Reduced grid size from 4096 to 2048. Further reduction
to 1024 introduced strong aliasing errors into the distributed
capacitances.
* Introduced subset solving to greatly improve performance which
offsets the increased relaxation time.
* Major tidyup.
* Added support for toroids and primary.
The simulator program has also been extended to deal with
toroid capacitances. The distributed capacitance between
toroid and coil seems to have quite a large effect on
the resonances.
The first test run, using figures for the thor system
under construction by Marco Denicolai,
(http://www.saunalahti.fi/dncmrc/thor.htm)
give
Measured Simulated Error
f1 65.46 kHz 66.7 kHz +1.8%
f2 222.75 kHz 220.1 kHz -1.1%
f3 346.3 kHz 325 kHz -6.3%
which I'm quite pleased with. These results are from 32 coil
segments, so a quarter wave at f3 is spanned by only 6 segments,
which may account for the error at f3.
Plans:
I'll be able to release v0.2 next week after more tidying
and documentation.
Next step will be to collect measurements on around six
systems of varying sizes and compare them with simulations.
Regards All,
--
Paul Nicholson,
Manchester, UK.
--
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.