Dumpfile timestep output

[wisnukusuma]

I was wondering why the output of my dumpfile time step was in order and sometimes not. what could have been affected from the setup, is the excitation? or the mesh grid?

[thliebig]

Yes this has to do with the timestep. If you mesh is (very) dense, the timestep gets very small. Then dumping every timestep is a complete waste of disk space. openEMS calculates a good sampling rate that is somewhat larger/better than Nyquist requires and that than means that only every 10th or 100th etc. timestep is required.

[wisnukusuma]

@thliebig related to the excitation, I was wondering what is the bandwidth of step function, I saw in the code that the init requires a f_max. As far as I know the step function has shae like squarewave but only in the rise edge, so it has bandwidth range, what is the bandwidth range for step function?

[biergaizi]

As far as I know the step function has shae like squarewave but only in the rise edge, so it has bandwidth range

The bandwidth of the step excitation is entirely defined by its rise time, and an ideal step function (similar to the one used in openEMS) goes up instantaneously. So to my best knowledge, it doesn't have a well-defined bandwidth.

In fact, it has been previously shown by Ted Yapo that using step excitation in openEMS can excite high-order non-TEM modes in microstrip transmission lines, causing invalid simulation results. He also spent great effort on creating a bandwidth-limited version of the step function using a custom excitation formula. See: A Note on Gaussian Steps in openEMS

Thus the step excitation must be used with extreme caution, and probably shouldn't be used at all for parameter extraction. BTW, it also seems to be a nice addition to implement a bandwidth-limited step excitation in the openEMS upstream code in the future, perhaps I can submit a patch.

[wisnukusuma]

nice explanation @biergaizi, thank you for your answer.