On Tue, 14 Nov 2006, Huazhong Wang wrote:
i found harminv only works after some step functions, not all step
functions. Harminv runs after after-sources, after-sources+ T, and
after-time T. It stopes after other step functions, for examples:
at-end, at-time T.
this is very strange.
Do I understand correctly that you are trying to do (at-end (harminv
...))?
This will not work because the harminv function cannot just be called at
the end. (harminv ...) produces a step function that is called at every
time step (or every time step after-sources etc.) in order to keep track
of the fields to analyze, and then finishes the analysis at the end.
So, you can only use (harminv ...) inside of step functions like
after-time, after-sources, etcetera that call it over a range of times.
It won't work with at-end or at-time because then you are not giving it
any data to work with.
a gaussian source was used in my simulation and i need to do frequency
analysis at a point which is far away from the source. this is like one of
meep examples: modes of a ring resonator. that means a period of time is
needed for wave to propagate from source. which step function is the best
for my simualtion if i need an accurate harminv result? In the meep manual,
it is recommended the step function after-sources is sufficient. but i found
the harminv result is related with the step-function, i.e. which time step is
used for harminv.
For resonant mode examples there is normally no need to put your analysis
point far from the source; you want to put them both into the cavity.
What are you trying to do?
by the way, in my simulation, the center frequency of gaussian source is
0.35, but the harminv results gives a shifted frequency less than 0.35. why?
the structure in my simulation is a square-rod photonic crystal. does
photonic crystal shift frequency?
The frequency never "shifts" in a linear system. However, if you put a
pulse/source into a system, then the frequency response is generally
*filtered* by the properties of the structure. For example, in a photonic
crystal the frequencies inside the photonic band gap are filtered out.
In general, if you put in a gaussian pulse you don't get out a gaussian
pulse; what you get out depends on the geometry.
Second, it sounds like you are misunderstanding harminv. Harminv is
appropriate when you are looking for slowly (exponentially) decaying
resonant modes, corresponding to a Lorentzian line shape. It is not an
appropriate tool for picking out arbitrary spectral features of the
source, such as a Gaussian pulse.
If you try to apply harminv to data that are not exponentially decaying
modes, i.e. if you break the assumptions underlying harminv, then what you
get out is unpredictable.
If you want to analyze an arbitrary spectrum, the best thing to do is to
just use the Fourier transform, e.g. as computed by Meep's dft-flux
feature.
Steven
_______________________________________________
meep-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
