Re: plane waves in free space

[Kevin Thomas <kjt@cray.com>]

Brian,

You don't mention the magnitude of the residual field that you
are seeing.  If the magnitude is small, then this is to be expected.
Often this effect is called "numeric noise", because it is an
artifact of the limited precision of the computer representation
of floating point numbers.  Both the mantissa (number of bits of
numeric precision) and the exponent (number of bits to represent
the range of possible values) affect the actual "noise floor".
With IEEE single precision floating point, one can expect noise
around 1e-6 v/cell or lower.  The values should fluxuate
randomly with time around this level, and thus appear to change
in a non-physical manner.

Note that v/cell is quite different from v/m if
your cell size is small compared to a meter.  If the cell size is
1 mm, then the noise magnitude will appear 1000 times greater
when displayed per meter.

Likewise, observed magnetic field values are smaller than the
electric fields values because they are scaled by the intrinsic
impedance of free space.

However, magnitude of the noise should always be 10000 to 100000 times
smaller than the peak value of the plane wave.  This
ratio (usually expressed in dB) is sometimes called the "dynamic range"
of the simulation.

ABCs of course play a part in dynamic range, too, so the use of
PML is recommended.

-- 
Kevin Thomas    kjt@cray.com   tel 1-651-605-9072
http://lc.cray.com/~kjt/        or 1-800-284-2729 x6059072

"Brian T. Schwartz" wrote:
> 
> Hi folks,
> 
> I'm still trying to find the s-parameters for a structure.
> 
> But at this point, however, I am simply trying to propagate a plane wave
> pulse across a region of free space to eventually find s21 = t = 1, and
> s11 = r = 0.
> 
> I figure I should see the plane wave in my xy-probe, and after the pulse
> is done, I should see nothing in the contour plot.  But this does not
> happen. The pulse comes and goes, but there is residual field that
> continues seemingly forever.
> 
> Any ideas on what I'm missing?
> 
> 1.  I'm using PML's.
> 
> 2.  My source is a plane wave, not part of the model geometry.  So there's
> no hard source issues.
> 
> 3.  I see the source waveform in View Source Waveforms", and my simulation
> runs far after when the pulse is over.
> 
> Brian.