I am simulating simple waveguide model using FDTD Method for S Parameter extraction. But, I am unable to simulate it successfully. Output waveform always goes unbounded....and input waveform also does not start properly. It starts with ripples. Is there any solution for this???? If it is necessary, i can upload the code also......
I have taken waveguide having cutoff at 3 GHz. I calculated the dimensions of waveguide using the cutoff frequency. I modeled it.
I use dt=dt=1/(c*((1/cell_x)^2+(1/cell_y)^2+(1/cell_x)^2)^0.5);
I sample geometry such that 100 samples can be accommodated in one wavelength................
I use PEC BC condition like this...
ex(1: xsize,1: ysize,=0;
ey(1: xsize,1: ysize,=0;
ez(1: xsize,1: ysize,=0;%if wave propagation is in z direction......this apply for all four faces of waveguide......
For excitation, ez(xsize-2,ysize-2,round(ksize/2))=ez(xsize-2,ysize-2,round(ksize/2))...
+ exp(-(t-0.5e-12).^2/1e-25)*5.*exp(i*5e15*t); %I give excitation in between structure
For two faces on direction of wave propagation, I define ABC boundary.........
I try to extract time domain signal at middle axis of waveguide geometry......but, and separate the incident and reflected signal........but, not able to do it since output waveform is going infinite......
Can you please tell me what could be the problem???? if you need, i will upload the code also......
You are using the maximal stable time step dt_max (actually the stability is only guaranteed for dt < dt_max). If you are using floats then this can be a problem
I sample geometry such that 100 samples can be accommodated in one wavelength................
Since you are using a Gaussian pulse for the excitation I am not quite sure which wavelength you mean here
I use PEC BC condition like this...
ex(1: xsize,1: ysize,=0;
ey(1: xsize,1: ysize,=0;
ez(1: xsize,1: ysize,=0;%if wave propagation is in z direction......this apply for all four faces of waveguide......
For excitation, ez(xsize-2,ysize-2,round(ksize/2))=ez(xsize-2,ysize-2,round(ksize/2))...
+ exp(-(t-0.5e-12).^2/1e-25)*5.*exp(i*5e15*t); %I give excitation in between structure
I try to extract time domain signal at middle axis of waveguide geometry......but, and separate the incident and reflected signal........but, not able to do it since output waveform is going infinite......
Can you please tell me what could be the problem???? if you need, i will upload the code also......
Hi
dt plays the Major Role in stability, reducing it from that magic value increases the stability while increasing simulation duration. Thus at first step reduce dt by 10% and more and more until it get stable.
Hi
dt plays the Major Role in stability, reducing it from that magic value increases the stability while increasing simulation duration. Thus at first step reduce dt by 10% and more and more until it get stable.
if dt satisfies CFL then the simulation is stable. If dt > CFL limit then except for exceptional cases the simulation will get unstable. But once dt < CFL limit you cannot 'increase' stability. Also the more you decrease dt the less accurate your simulation gets (e.g. numerical dispersion increases).
Another question is what materials (e.g. metamaterials) are used and how they are modelled.
As you said, i will try with dt<dt_max(10%dt_max)......
I am taking 100 samples in one wavelength...wavelength considered here is the minimum wavelength of operation......
For truncation at two faces, I am using Mur's First order ABC.......
In excitation, please remove that exp(i*5e15*t) term......i am using cos term in place of it......I apologies for my mistake in writing......
And about material, I am using vaccum having Er=1 Mu=1 for waveguide cavity.........
Is all these thing ok...........????????????
I am also not able to understand how can I get scattered field which is coherent with Input Waveform??? I read Allen Teflove's Book....but still have the confusion.....can you please give some ideas????
This may be the reason for your problems. I occasionally used 2nd order Mur which does have
stability problems. I think I read somewhere that 1st order Mur has less problems but I am not
sure.
I think you should dump the signals at several locations, especially close to the boundary and check where the signals start growing. If it is at the boundary you should consider implementing
other ABCs (or check for bugs)
And about material, I am using vaccum having Er=1 Mu=1 for waveguide cavity.........
I am also not able to understand how can I get scattered field which is coherent with Input Waveform??? I read Allen Teflove's Book....but still have the confusion.....can you please give some ideas????
I would first try to deal with the instability. But the simplest method is to simply wait until the input has sufficiently decayed and then use the remaining bit as the reflected/scattered signal. This works if your input pulse is relative short and you measure far enough from the scatterer