cst 500thz

Dear users,
using transient solver in CST Microwave Studio I have the following problem:
I would like to calculate S-Parameters through a metallic periodic structure in the visible/near infrared. I use two ports in a waveguide to excite the signal and to detect the reflected/transmitted intensity. Unfortunately S-parameters change when I set the lower frequency in "frequency range" to 0THz. If the lower frequency is finite (i.e. 1 or 250THz) no change in S-Parameters is visible. On the other hand I have heard, that simulations run faster and more accurately when setting the lower frequency limit to 0 THz.
So my question is: Which results can I trust more?
Attached to this post you can find the simulated S-parameters for a lower frequency of 0Thz and 1THz respectively. The upper frequency was 500THz. Especially in the region of 300THz which is of special interest in my case, large differences in the S-parameters appear.
Thanks a lot for your support,
Markus.

mbroell
Are you using the same solver accuracy setting for both cases? You may try increasing the solver accuracy of the lower freq=1THz case by 20dB and see if the results begin to converge with the lower freq=0THz case.

Hi Dave,
sorry for the late reply, the simulation takes a while. Unfortunately the S-Parameters do not converge, even for a higher accuracy. In the case of 1THz I have a clear resonance behaviour at about 390THz. This resonance has disappeared if I take 0THz as lower frequency. Is there any physical meaning why results are different for different frequency ranges? Why is it favourable to take 0THz as lower limit?

[quote="Is there any physical meaning why results are different for different frequency ranges? Why is it favourable to take 0THz as lower limit?[/quote]
I would take the lower freq=0THz case as more accurate. Remember, T=1/df, where df = bandwidth of signal, so the bandwidth of the lower freq=0THz case is larger and at the same time the simulation will be shorter.
I do see a small, but distinct resonance, at around 340THz in the lower freq=0THz case, which is probably the same ripple that you see in the lower freq=1THz case. In my opinion, the smoothing you see in the lower freq=0THz case is because of the increased bandwidth of the excitation signal.
Since you increased the solver accuracy of the lower freq=1THz model and it is already well converged, one more thing to try is to run the lower freq=0THz case with an increased solver accuracy of 20dB and see if the resonances which get deeper and shift up in frequency a little. By increasing the simulation time you are the solver is forced to integrate the S-parameters using more data points and you may see an overal improved resolution in the frequency domain.