transparent source fdtd
can anyone explain ...... the " all possible types of source implementations in FDTD "...
like ... Point source.. Dipole source..
anything related to source implementation will help me..
thanks in advance..
Wow that is a big question! 99% of the time, there are two main sources used. I describe them briefly below.
Total-Field/Scattered-Field
================
This is essentially a one-way source injected at one plane in your grid. It can be either a plane wave or something else. It propagates in only a single directly from where it was injected and does not propagate backwards. Waves scattered are able to propagate back through it so it is transparent in this sense. This is perhaps the most common source type.
Dipole Source
==========
The dipole source is used most often to simulate the feeding of antennas. It is also used to excite cylindrical waves.
What are you interested in modeling? Given that, I could perhaps help you better.
-Tip
Hi, thanks for the reply..
i would be greatful if you can really help me..
can i hav your mail id so that i can contact you in directly...
regards,
milli.
I'll send you my e-mail, but it would be useful to the group to have the discussion here.
What are you interested in modeling?
-Tip
Hi, may I join the discussion?
I feel perplexed when referring to 'dipole source' and 'point source'? Are they exactly the same thing except for the names?
Many thanks
A dipole source and a point source are the same thing when referring to FDTD. Keep in mind, however, that you can excite a dipole source with either an electric field or a magnetic field and each will produce a different configuration of fields and likely a different response from whatever you are modeling. The dipole source usually refers to the electric field point source type.
-Tip
Thanks a lot. let's say... if I want to add a 'dipole source' in FDTD in 2D and 3D simulation, I should use the statement in the FDTD loop like
F(i,j,k)=S(t) for hard source
or for soft source
F(i,j,k)=F(i,j,k)+S(t)
in a word, I should assign or add input wave to one particular field component at one single point and then use it to excite a dipole source.
Is that corrent? If so, does it work for both 2D and 3D simulation? Thanks. and may I have your email or other IM ID as to contact you directly?
That's it! In a 2D simulation you will be creating a cylindrical wave. In a 3D simulation you will be creating a donut shaped wave, so you will not get propagation in the direction parallel to the field you are exciting.
-Tip
Great!!! really appreciate your help...
one more question, in formula
F(i,j,k)=S(t) for hard source
or for soft source
F(i,j,k)=F(i,j,k)+S(t)
can S(t) be of any type? regardless of sinusoidal or gaussian modulated?
Yes, your function S(t) can be anything. The only restriction is that your time steps are sufficiently small to resolve S(t) so that it is "smooth and continuous." Abrupt jumps in that function can cause problems.
-Tip
respected rrumpf
except sinusoidal,gaussian function.
is it possible to apply rectangular,triangular function as source to FDTD.
can you clarify this?
I suspect that in the way you mean it, the answer is no. The S(t) function and at least its first derivative must be continuous. This rules out "perfect" rectangular and triangular functions. However, if your "round" the edges a little bit on your rectangular pulse or triangular pulse, your simulation should work. I would also suggest using somewhat smaller time steps so the "rounding" does not have to be as severe.
The problem with 'any function' is the same as 'any model'. FDTD discretizes both time and space
If your Yee cells are too big you can use things like average permittivity but cannot expect accurate
results.
Similarly if your function cannot properly be resolved with the FDTD time step you cannot
accurately model it. Roughly speaking, FDTD connects the discrete function values of your source
function with more or less straight lines. So a rectangular pulse will get some steep but not infinite
slope.
I have a doubt, when I do a simulation of plane wave propagation through vacuum, using TF/SF scheme, I find as distance of an observation point from the face of incident wave increases, the amplitude of wave keeps on increasing slowly. When I decrease my cell size further (lambda/30) this amplitude increase decreases a little, but still I get some amplitude increase particularly after around 20 lambdas of propagation distance.
Does anybody has some idea, what could be the possible cause of this error.
I have another doubt regarding the source implementation while studying the propagation of EM waves through plasma. In the TF/SF technique generally we use six faces to describe the incident plane wave. Do we need to change the conditions which we apply here at the boundaries between TF and SF region while doing plasma simulation. Is it possible to start plasma in the TF region just from the SF/TF boundary, or do we need to leave some cells.
I have not done any plasma simulations so I cannot be of any help there. As for the TF/SF problem, have you tried putting periodic boundary conditions all the way around your grid and just letting your code run for a long time? If you do this, the wave will just keep propagating and perhaps produce some neat standing wave patterns. The point will be to look of the field amplitude is stable over a long simulation. If it explodes or goes to zero relative quickly, there is a problem with your code and I would suspect first that it is your update coefficients. At least, this is usually my problem.
-Tip
when i did my simulations i never faced such problems like increasing amplitude..
in your case i think either of field profile or your update coefficients might be causing the problem..
check your field profile and at the same time whether the field points are being updated regularly or not..
i hope this might help you..
regards,
milli
1) as the others said check your coefficients
2) what increases in the amplitude do you see (increases by a factor of 1.0001 and by a factor of 10 will ususally have different reasons)
3) do you use normal incidence?
4) what type of wave do you add
5) how do you compute the wave you input at the boundary (analytic formulas are not necessarily best because of the numerical dispersion)
Unless you have a concrete reason for not doing so I would keep the material across the SF/TF interface constant (makes implementation and debugging much easier). I have seen papers where inhomogeneous SF/TF interfaces are considered but never tried one of these methods (I don't think there is a universal method.)
Finally, how do you model the plasma? Do you use any of the dispersive FDTD models?