sRate/3 PAC limitation

Hi and Thanks for your mafnificent program!
I have a question regarding the warning message and the limitation for the maximum frequency admisible to perform the PAC calculation, coming from the file process_pac.m :


I am used to work with frequencies up to sRate/2 (Nyquist Theorem) in Spectograms or Scalograms, but this is my first approach to PAC. I checked the bibliography mentioned in a similar Topic on this forum but I could not find anything regarding this limitation:
https://neuroimage.usc.edu/brainstorm/Tutorials/TutPac
https://neuroimage.usc.edu/brainstorm/Tutorials/Resting

The fact is, I have several EEG data at 500Hz and I need to analyze their PAC up to 250Hz.
At the moment, I've made some temporally modifications to the code in order to accommodate this requirement:
(1) process_pac.m (modifying sRate/3 -> sRate/2 in lines 259, 261 and 265)
% Check the nested frequencies
if (OPTIONS.BandNested(2) > sRate/2)
% Warning
strMsg = sprintf('Higher nesting frequency is too high (%d Hz) compared with sampling frequency (%d Hz): Limiting to %d Hz', round(OPTIONS.BandNested(2)), round(sRate), round(sRate/2));
disp([10 'process_pac> ' strMsg]);
bst_report('Warning', 'process_pac', [], strMsg);
% Fix higher frequencyy
OPTIONS.BandNested(2) = sRate/2;
end

(2) bst_pac.m (modifying sRate/3 -> sRate/2 in line 103
% Definitions
fmin = min(bandNesting);
fmax = sRate/2;

I compared the resaults with those obteined by using the original code after a resampling to 1000Hz and the differences are quite low.

If it is theoretically correct I would prefere not to resampling the original data. I hope you can tell me where to find the cause of the sRate/3 limitation and how it affects when using sRate/2.

Tanks for your time!

if the maximum fA is set to half the sampling rate, that makes only 2 time samples per fast cycle to measure possible coupling with a slower oscillation, which I believe methodological studies of PAC have found to be too little. Also, please note that plain over-sampling data originally sampled at 500Hz does not add new signal information, hence the similarity in the analysis outcomes.