[FieldTrip] Power normalization using mtmfft,

[Wonkyung Choi]

To fieldtrippers,

Hello, I have a question about the raw power using 'mtmfft'. When I put 60
seconds data vs. 120 seconds data with the same sampling rate for
simulation, the latter has roughly 1/2 of raw power compared to the formal.

When I looked into *ft_freqanalysis>ft_specest_mtmfft*

dum = fft(ft_preproc_padding(bsxfun(@times,dat,tap(itap,:)), padtype, 0,
postpad),[], 2);
dum = dum(:,freqboi);
% phase-shift according to above angles
if timedelay ~= 0
dum = dum .* exp(-1i*angletransform);
end
dum = dum .* sqrt(2 ./ endnsample);

it normalizes by sqrt(2/N) where N is the number of samples including
padding.

(1) why is it normalizing using square root function but not 2/N?
(2) why does the raw power still differ when the input differs in length in
the time domain (e.g. 60 s vs. 120 s)?

Please advise. Thank you so much.

Best regards,
Woni
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20240205/4fc3f76c/attachment.htm>