Discussion Forum

You appear to be misunderstanding the terminology. "Far-field" does not refer to some location merely moderately far from the antenna. It is for computing the angular distribution of the radiated fields in the limit as R goes to infinity, but scaling them to a particular R (to yield finite fields). (You can learn more about this if you search this forum or refer to the help system). The domain you select for computing the far field pattern is not intended to be, and certainly doesn't need to be, in the "far-field region" itself. You don't have to mesh the "far field" region either, since any physical locations (feld points, if you will) for the far field values are not part of the finite-element solution at all! Rather, this is all done in post-processing, via the Stratton Chu aperture field (AF) integration method, applied to FE-computed near fields. The surface(s) you choose for the FE-computed fields to be used in the integrand for applying AF integration should be on surface(s), aka, "aperture(s)", that radiate. The aperture of a horn is one such simple example. Another example would be the surface of an enclosed computational region containing that horn. In principle, using different enclosing (or partially enclosing) surfaces, if they include most of the dominant contributing radiating fields, will give nearly the same result. If you include only the aperture of a radiating horn you will miss some of the diffraction effects around the edges/flange of the horn aperture. (For a dramatic example of that effect, model an open ended waveguide including just the guide aperture vs an end-enclosing aperture, and compare). Finally, if you want to compute fields that are not close enough to the antenna that you can easily generate an FE solution directly there, and yet not so far that you can approximate them well by scaling a far-field result, then you should consider employing the radiating near-field method, which I describe in https://www.comsol.com/community/exchange/672/ . I hope this helps. Finally, for a very old example (not making use of any newer Comsol features) of computing field patterns from a pyramidal horn, see https://www.comsol.com/community/exchange/192/ . Added: You might also want to take a look at example files provided by Comsol in the Application Library, in which far-fields of some simple antennas (e.g., dipole & patch, if I recall correctly) are computed.

Thank you for the answer, you say "It is for computing the angular distribution of the radiated fields in the limit as R goes to infinity, but scaling them to a particular R (to yield finite fields)". What is the particular R at which the electric far field (emw.normEfar) is calculated then ? We know that in the far field domain the fields scale as 1/R. When I add radiation pattern node in the polar plot group to comsol and make a polar plot of electric far field ( emw.normEfar ) , the result will depend on the value of R at which the plotted far field is calculated. To make a comparison with experiment where I measure radiation pattern at a particular distance R_experimental from the antenna I need to know the R at which comsol far fields are shown in order to rescale them ( with a factor R/R_experimental ) and compare comsol simulation to measurements. I know that R doesn't change the angular distribution in far field limit, but it changes the magnitude of fields, which is proportional to what I measure experimentally.

R = 1.0 m.

Is it always R=1.0 m in any case ? I ask becuase I have to test some antennas with different dimensions. If for some of those antennas the far field limit is beyond 1 m having the far field calculated at R=1.0 m would result in an error since it would not be in far field region. Is there a command to modify R in order to tell COMSOL where to compute far field with Stratton Chu integration?

I suspect that you don't fully appreciate the meaning of the term "far field." The R=1m value simply corresponds to imposing an appropriate scaling factor, as if (whether or not) the field falls off proportional to 1/R, and as if (whether or not) the pattern is independent of distance. If you were to change that reference value R, all it would do is compute is the same values, but scaled by the ratio of the old R to the new R. There would be no new information content! The far field calculation doesn't compute field details that are unique to the near field, whether the true near field region of your antenna extends beyond R=1m or not. If you need to compute details in the reactive near field region, use the FE mesh directly and compute the fields in that region (which may be computationally intensive). If you need to compute fields in the radiating near field region, you can use the method I described in my 1st reply to you. And if you need to compute fields in the far field region, but you want to deal with actual field values (not just the pattern) at some finite R, then use the far-field computation tool that Comsol provides and scale your results in accordance with the "actual" distance to your field point. I hope that helps.