Gaussian Pulse in Free Space

[zeroedus]

Hi:

There is a source radiating into a vacuum (free-space) and a detector at a certain distance from the source. If the signal transmitted is a Gaussian pulse modulated by a carrier freq w0 i.e. exp(-a(t-td)^2)*sin(w0t), will there be a frequency shift at the detector?
This is done in a FDTD simulator (Full-wave), and when I run the simulation and plot the FFT of the source and monitor field, I get a 10% frequency shift.

Thanks in Advance.

V

 

[flatulent]

I suspect your problem is caused in the simulation method. Try a longer time period to capture the signal. A gaussian pulse has infinite time duration in both positive and negative time.

Try a baseband pulse on wires and see what you get.

 

[irfan1]

I have to agree with flatulent. You need to set a higher stop time. Setting a higher stop time will increase your resolution in terms of frequency. If you still suspect something, you may put another monitor at a more distant point and chech whether there is a shift and if there is a shift, does this shift change with distance.

 

[zeroedus]

I am attaching the plot for 5 monitors. I have set the stop time to 2^10. I increased it to 2^12
Any suggestions.
Irfan: Did you face problems like this when you used FullWave?

Thanks,
V

 

[irfan1]

As far as I can remember I did not face such a problem.

 

[irfan1]

I did some similar calculations. I also noticed those shifts in my simulations. But not as significant as yours. Could you please try to do the same simulations by refining the mesh further? I mean by using smaller mesh steps and smaller time steps?

 

[zeroedus]

The pulse time parameter I chose was 0.4. Hence, to get a good resolution from the FFT either the time samples have to be increased or the grid and time spacing must be decreased.
The observed shift maybe because the FFT needs more time samples to resolve the field. I kept the pulse parameters same but decreased the grid and time step size to 1/10th the previous value. Now the shift is very small, but the simulation time has increased by a factor of 30. Hence, the tradeoff!

Thanks for the suggestions and help.

V

 

[Zottiri Kusuta]

As far as i can see from the plots you have attached, even the gaussian pulse you have excited (source) isn't symmetric. See the x-axis cross points. The left one is smaller than 1 but the right one is larger than 1. Am I wrong? The problem may be you make a mistake with the gaussian excitation, a small indexing problem?

 

[zeroedus]

As I mentioned before, for a smaller pulse time, the # of samples will have to be more. Hence for a slower varying pulse, there will be less distortion after the FFT.
This can be seen from the FFT of a Gaussian pulse with pulse time =2 as compared to the one shown above. (where the pulse time is 0.75).
Hence, the difference.

 

[irfan1]

it seems like a bug to me. I will report this issue to the RSOFT's support.

 

[zeroedus]

I tried to explain it to the Support. They say that it's a problem with the # of time samples I have taken. They say to increase it in order to get the "wanted" result.

So please let us know what Support tells you. If they accept it, then that'll be great!

Thanks,
V