FreeLists / si-list / [SI-LIST] question concerning socket for SDRAM module - mechanical problem?

["Jan Vercammen" <jan.vercammen1@xxxxxxxx>]

Hello si-list,

I have a problem with the dynamic range of a FDTD code (with LC to be
specific).
However, the problem is general in nature and other users of FDTD solvers
could
have had similar problems.  I have posted the same questions on the
LC-users list.

I think I have exhausted all alternatives in the hope of improving the
problem. The
dynamic range of the FDTD code seems to be of the order of 1E-7 to 1E-8 or
about
140 to 160dB. I say dynamic range, but the problem seems to focus on a
trailing
DC-offset of a pulse.

I am using a gaussian pulse and lanch it on a parallel plate system
in the TEM mode, other modes are not excited. I am using magnetic
wall boundaries to enforce the TEM mode of this 2D problem. The pulse
is excited in a hard manner, but is switched of (truncated) after it
finishes.

The problem I face is a DC offset (plus some noise) trailing the gaussian
pulse,
it does not reach zero, but rather sticks to the range 1E-7 to 1E-8.
Depending on the
sigma choosen it could be positive or negative. A larger sigma is better.
For
example a 6 sigma gives the pulse a dynamic range of exp(-36)=2.31e-16,
which is huge. The source pulse is defined in the range -6 to +6 sigma,
after which it
is removed from the simulation.


This is what I tried:
-1- a larger sigma is better,sigma 4.5 to 5 is a good choice, above that
there is not much benifit
-2- I have tried the 1st derivative of a gaussian pulse for the source as
there is no DC component in it.
      However, I get the same reults.
-3- I have tried a smaller time step, but no result. Actually it results in
a slightly larger DC offset!!!
-4- I have tried a finer mesh (1:2), but no improvement.
-5- tried a slower rise-time. No result either.
-6- I do not see any problem with the ABCs (Mur 1st order), there is some
reflection but they do not
produce DC offsets.


I am suspecting either (1) dispersion or (2) the finite range of floating
point numbers or (3) possible the
fact that a DC offset is a solution of Maxwell's equation. So for (3) I
assume that the pulse consists of
a small step + a gaussian.

I can do something about (1), I increased the rise time or use finer cels,
but this was without result. So is
(2) a possible cause?? Or is it (3). I have read an article about the
latter, where people used resistors
to get rid of DC offsets in antenna simulations (using FDTD).

Actually, does anyone know the floating point format used for the Linux LC
version? Floating, double or
long double? Has anyone experience with this kind of problem? Can I run
long double floating
point on a Linux platform? Has anyone had similar problems with other FDTD
codes?

I am stuck here  and it seems that I do need a big dynamic range.


Any hint or help is appreciated.

Kind regards,

Jan Vercammen





Kind regards,

Jan Vercammen
Agfa-Gevaert NV
Mortsel, Belgium



------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field

or to administer your membership from a web page, go to:
http://www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field

List archives are viewable at:     
                http://www.freelists.org/archives/si-list
or at our remote archives:
                http://groups.yahoo.com/group/si-list/messages 
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu