Ex,y,z component far-fields from CalcNF2FF

[_neil_]

Can i access the x,y,z, component E-(far)fields directly after the CalcNF2FF routine has run?

In the calculation of the radar cross-section (tutorial RCS_Sphere.m), the far field Prad is very nicely calculated from the CalcNF2FF routine and accessible from the returned structure. However, i'd like to access the x,y,z E-field far field components directly in the code. Can i do this directly?

I see there are *.h5 files created by the CalcNF2FF and i guess perhaps the E-fields components are contained within. However, is there a method to get the far-field components out directly, without having to access the *h5 files?

many thanks,
Neil

[thorsten]

I'm pretty sure that all this data is included in the nf2ff data returned by CalcNF2FF as this is what is plotted as a pattern for example.

regards
Thorsten

[_neil_]

Thank you for your help.

Yes, i see i can get far-fields directly from CalcNF2FF in the frequency domain. However, I need far-fields in the time domain. I managed to get the time domain far-fields by making an inverse FFT on the frequency domain far-fields. The resulting time domain signals look correct (probably Nyquist sampled ~lambda/2), but the sampling on this was far sparser than on the local E-fields from the probe, these being sampled at the lambda/20 criterion for a reliable functioning FDTD calculation.

So the question is, can i get time domain far-fields directly from CalcNF2FF at the same sampling (ie lambda/20) as that for the probe time domain (near-field) E-fields?

many thanks,
Neil

[thorsten]

Hi,

I'm a but confused about your request. lambda/20 is a length (e.g. in meter) and thus has little to do with time-domain??

Furthermore the nf2ff (at least the one in openEMS) cannot handle time-domain input and/or output...

regards
Thorsten

[_neil_]

Thank you for getting back to this question and apologies for the confusion.

Isnt the (near-field) sampling (time resolution) in the time domain returned by RunOpenEMS determined by something like (lambda/20)/speed of light? Being lambda/20, that's effecively 10 times better than Nyquist, as Nyquist sampling is lambda/2.) Is this not the case, or are there other parameters determining the time domain time resolution?

If you take the Fourier transform of the (complex) frequency domain information returned by nf2ff, you get something which looks very much like a far-field time domain response, sampled at something like Nyquist. This seems logical, as time and frequency domains are related through the FT. Simulations on a few targets seem to confirm this. Would you agree with that?

I took the approach of taking the FT of the frequency domain information, as i was wanting a far-field time domain response, but actually was wanting a better sampled response, more like the near-field returned by RunOpenEMS. Might there be another approach to obtaining a finely sampled far-field time domain response?

many thanks,
Neil

[thorsten]

I'm not I understand everything you were asking, but the parts that I think I understand I do not agree with

The near-fields are all recorded in frequency domain, because (afaik) nf2ff does not make any sense in time-domain?

In short, far field patterns are calculated for the given frequencies and this is what nf2ff returns...

regards
Thorsten

[_neil_]

grateful all the same for your effort in trying to understand the problem.

The question still remains then, what's the best (most efficient & quickest) way to determine the far-field electric fields in the time domain?

many thanks,
Neil

[rml]

I suspect there is a little confusion here.
If, for example, you want to investigate the effect of electromagnetic fields on human tissue at a distance of several wavelengths from an antenna (hence already in the "far-field" region), you need the normal near-field data.
But please describe what you really want to do, in case I am wrong.

[_neil_]

that's right, so how do i get the far-field components Ex, Ey, Ez in the time domain? Thank you for any help, Neil

[rml]

Maybe I'm thinking too simply, but couldn't this approach work?
You perform one simulation with the antenna and the human body model and store the fields in the area of interest near the antenna.
In a second simulation run, you set all the tissue properties of the human body model to air, but don't change anything else, especially not the mesh.
Isn't the difference then your intended result?
Surely you should use PML as the BC.
Using a plane wave instead of the antenna should work the same way.