[Eeglablist] pop_newcrossf() timefreq winsize padratio question

Arnaud Delorme arno at ucsd.edu
Tue Dec 15 18:19:29 PST 2009


Dear Kate,


> We are using pop_newcrossf() in eeglab7_1_7_18b and
> trying to figure out winsize issues.
> Looking at timefreq.m,
> g.winsize = max(pow2(nextpow2(frame)-3),4);
> g.nfreqs= g.winsize/2*g.padratio+1;
> It looks like the window size is defaulting to about 1/4
> the input timeframe.
> What is the rationale for that ?  1/4 seems a bit big
> to me.  Was that to get the desired frequency granularity ?
> Is there a standard window size for eeg coherence studies ?


To access low frequencies, you need large windows. For instance, with  
a 1 second window, using 3-cycle wavelets, the lowest frequency you  
can obtain is 3 Hz; and this is the type of frequency we might be  
interested in as theta (usually about 5 Hz) is just next door.
Typically we extract 3 or 4 second windows so for us this is not a  
problem; and it does not matter so much if our long data epochs overlap.


> When we cut our winsize in half the number of
> frequency bins was also cut in half, as per the
> code I saw.  Can we just compensate for this by
> doubling the padratio to get back the smaller sized
> frequency bins ?    Can we do this up until the point
> the bins would have the frequency width one would
> have gotten by just doing an unpadded fft on the
> whole frame (ie governed by the sampling rate of the
> input data) ?  Or, what are the limits of adjusting padratio ?


Sure. You will not be able to recover the lowest frequencies though by  
adjusting the padratio. There is no limitation to padding the data to  
my knowledge.

Hope this helps,

Arno


>
> thanks
> Kate
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu




More information about the eeglablist mailing list