Re: [trouble]2D fdtd : wave shape distorted after propagatio
I am doing a project on FDTD simulation. I have used Taflove book as the reference. The code is implemented in Fortran90.
I am simulating TMz mode and exciting the Ez with a gaussian pulse at the center of the whole 2D space.
As Ez propagates in the plane, the gaussian wave is distorted.
This is what I did:
the 2D space is (500,500).
Ez(250,250) is excited.
The boundaries are PEC(Ez=0).
I observed Ez(100,100). And this is what I get. Why do I get a negative peak? Can you explain me the reason for this?
I suppose the above information is sufficient to understand my problem, but if you need any more specifics let me know.
Added after 1 hours 39 minutes:
I was trying the 2D code which is written in MATLAB and is available with the book. The result I obtained is similar. Can someone tell me why do we get such a result, instead of a gaussian pulse?
hi
u must check the medium permetevety and conductivety the wave propagate in free space or lossy dielectric or lossless dilectric what?
best regard
Added after 43 seconds:
hi
u must check the medium permetevety and conductivety the wave propagate in free space or lossy dielectric or lossless dilectric what?
best regard
as aaea3 said already, material parameters matter.
Also how do you add the source?
hard: Ez = signal(n)?
soft: Ez += signal(n)?
The signal you add at a point is not the signal you measure after the update equations.
(The flelds around your source point affect the value).
Finally, pulses become distorted by numerical dispersion. (Velocity depends on angle and
the factor lambda/cellsize which differs for different frequencies)
since it is PEC there are reflections, which will propagate in the simulation domain and add up to the pulse.
unfortunately it's not due to the material nor to the coherent interaction with the reflection from the PEC boundary, since you'll get pulse dispersion even in vacuum (eps_r=1).
It's a purely numerical artifact due to the discretization, that's why you should keep the time iterations of your simulation runs to a minimum.
Check Taflove for more details..
thank you.
i understand that it is due to discretization. But why is the phenomenon so obvious when i take a gaussian pulse, but is not visible when i consider a sinusoidal pulse? is this because of the higher frequencies in the gaussian pulse?
yes, it's because any pulse is made up by a range of harmonics, whereas an ideal sine wave contains only one frequency.
The odd pulse broadening you see is due to the fact that every harmonic in the pulse is delayed by a different amount.
In general, the slower the variation in time of something, the narrower its spectrum, and vice versa (t and f, or x and k for that matter, are said to be conjugated by the Fourier transform)