It doesn't make any difference, except that there are some excitations that don't have some frequencies in them. The "Gaussian" excitation has a "Gaussian" spectrum, and the result is divided by that shape to get the normalised results.

The "Gaussian" has the best compromise between the length of the pulse required and the frequencies that it contains - if the edges of the excitation are square then it contains high frequencies which require a finer mesh, and there may be nulls in the frequency content where you can't get an answer.

One advantage of using your own excitation is when you know what it will be in practice, and it is a pulse of some kind. Then you can get the exact time response directly. I think you can do this by post-processing anyway, even with a "Gaussian".

It isn't really a Gaussian. A Gaussian has no beginning and no end in time, and has no upper frequency limit. CST uses a shape similar to a Gaussian.

various excitations can result in Gibbs artifacts when filtered out or truncated. The same effect may be produced in the antenna pattern, which may complicate its analysis. Gaussian pulse has a very smooth shaping, and is equivalent to its own Fourier transform. Analysis with Gaussian pulse is less sensitive to truncation and does not produce that much artifacts or side lobes.

Another benefit is that Gaussian pulse is a solution of many wave-guiding problems, and is a pulse which you can actually find in many narrow-band systems: it is also a field distribution at the output of the dielectric wave-guides (optical fiber). So it makes the simulation somewhat closer to reality. But you can try other shapes, particularly because sides of the Gaussian window spectrum should be scaled before plotting the band response, i.e. dynamic range is worse there.

Hello,

Gaussian excitation covers wide range of frequencies and also works on pulse with frequency so it is better but all other excitation dose not change the results.

File is attached.

Thanks,