[FieldTrip] real-value spatial filter from FFT dipole localization?

[Björn Herrmann]

Dear fieldtrip users,

maybe someone can point me in the right direction. I try to obtain a spatial filter from localizing the sources of an auditory 3Hz peak in my FFT spectrum. I would like to apply the filter to my single trial real data, i.e. basically reducing my 204 gradiometer channels (neuromag system) to one or two channels in source space.
I was using DICS beamformer and cfg.realfilter = 'yes' but the source localization results (NAI) looked not too well compared to the nice auditory gradiometer topographies.
I then used ft_dipolefitting, however, I have trouble understanding the output. This is where I wondered whether someone could help me, in telling whether it is possible to obtain a real-value spatial filter from this:

source = 
     label: {204x1 cell}
       dip: [1x1 struct]
     Vdata: [204x408 double]
    Vmodel: [204x408 double]
      freq: 3
    dimord: 'chan_freq'
      grad: [1x1 struct]
       cfg: [1x1 struct]

source.dip =
        pos: [2x3 double]
        mom: [6x408 double]
        pot: [204x408 double]
         rv: [1x408 double]
        pow: 2.7948e+18
        csd: [6x6 double]
    fourier: [6x204 double]

I would appreciate helping comments, as I tested the different methods for some time now, but was unable to extract a reasonable real-value spatial filter for my auditory cortex activations.

Best regards,

Björn