About the inserted power in the structure.

[psaizcor13]

First of all, I would like to congratulate @thliebig and all the collaborators on this project; it is really nice.

I wanted to ask a question regarding the power injected into the structure. I'm simulating a patch antenna. It works at two frequencies, so you will see two resonances.

Below are the frequency responses of the powers (incident, reflected, and total):

I'm injecting a Gaussian pulse between 1450 MHz and 1750 MHz. You can observe that the incident power is not a perfect Gaussian spectrum. I tried increasing the mesh density around the port and improving the mesh, without many changes. I changed also BC to PML_8. I think the structure is not perfectly matched to the port, although at the resonance frequencies I see 50 ohms.

In the end, the S11 is computed from the ratio of powers, so the result makes sense. However, I'm a little worried about numerical errors. Furthermore, you can observe a ripple in the S11 (although the level of energy decay was reached).

I would expect a power spectrum like this one (this is obtained with a 50 ohm trace):

My question is: Has anyone experienced this same behavior? Does anyone have advice to overcome this issue?

Thank you in advance.
Kind regards

[VolkerMuehlhaus]

I would expect a power spectrum like this one (this is obtained with a 50 ohm trace):

@psaizcor13
Please show your code where you define Gaussian pulse. It seems that you are using too narrow bandwidth, which explains the ripple at the edges of your simulation frequency range.

Your excitation frequency range can (and should) be wider than your antenna bandwidth, so that S11 can be calculated accurately across the entire band.

From what I understand, the incident power will not show the gaussian pulse, but the incident voltage will. Please check that:

u1 = np.abs(port.uf_inc)

[psaizcor13]

Hallo @VolkerMuehlhaus

My definition of the Gaussian pulse is this one (in the Python interface):
f0 = 1.6*1e9
fc = 0.15*1e9
FDTD.SetGaussExcite(f0, fc)

I tried to increase the bandwidth of the simulation. This time I used:
f0 = 1.6*1e9
fc = 0.3*1e9
FDTD.SetGaussExcite(f0, fc)

with similar results.

I don't like so much when the S11 has ripple and the values are higher than 0dB at the edges of the bandwidth.

I plotted also the absolute value of the voltages:
np.abs(port.uf_inc)

with similar results (the y-axis states Pin, but in this case is the Voltage):

Maybe is not enough and I need to increase more the bandwidth(?) For this high resonant structure the FDTD method may not be the most convenient, however I was expecting to get better results.

The beginning of the time domain signal has this shape:

which it is like a Gaussian pulse, although after the first samples there is an oscillating signal, due to the two big resonances of my S11. At the end my energy decay criteria is reached (I fixed -40dB). According to my experience with other EM solvers, this energy decay setting is enough, but with openEMS I often see the -50dB value.

Since I'm simulating a patch antenna with double resonance, could a degenerate mode be causing this unusual shape in the Gaussian pulse? My structure is a patch antenna with two trimmed edges to achieve dual resonance.

[VolkerMuehlhaus]

The beginning of the time domain signal has this shape:

That is the expected shape, a modulated Gaussian pulse. This is because you have limited bandwidth and no DC component.

The ripple at the edges of your analysis range looks like insufficient convergence to me, so make sure that simulation does not stop before due to number of iterations. Your second screenshot shows quite large residual voltage at the end of the plot !?
You can also try to set convergence limit to -50dB or lower, but even -40dB should look better than your screenshot (less ripple).