[Eeglablist] [eeglablist] aliasing in pop_spectopo():

Anders Sand engiltigadress at hotmail.com
Thu Oct 6 02:14:23 PDT 2011


Hi,
I still can't tell what the source of my problem is. However, here is what I found from trying the three suggestions I received.
First, high pass filtering the data at 0.05 Hz seem to solve the problem. http://www.mypicx.com/10062011/hipass/ But I can't understand why.
Second, changing the zeropadding ('nfft') input in the spectopo function also seem to solve the problem - but only if I change the zero padding to 512 Hz which is my sampling rate. http://www.mypicx.com/10062011/zeropad/ But, Michael, you said that if the padding is the cause of the problem, then changing the amount to any non-integer multiple of the sampling frequency should solve the problem. This is not the case, for any amount except 512 the result looks as bad or worse as the default setting.
Thirdly, removing channel means by first recasting them as doubles did not work.
I appreciate any help in understanding the problem.Anders
CC: eeglablist at sccn.ucsd.edu
From: michael.ploechl at gmx.net
To: engiltigadress at hotmail.com
Subject: Re: [Eeglablist] [eeglablist] aliasing in pop_spectopo():
Date: Tue, 4 Oct 2011 19:27:43 +0200

hi anders,
let me guess: your data is recorded with a sampling frequency of 512 Hz and you are zero padding them to twice the sampling frequency, i.e. 1024 Hz. if this is the case there is nothing wrong with your data. the reason why the data look strange is illustrated in the attachment. the recorded samples are depicted by black squares and the underlying function that one would obtain by continuous fourier transformation is shown by the gray trace. when using discrete fourier transformation all samples except the one for the center frequency fall on the zero line and thus are not visible when plotting the non-zero-padded data. when you now zero pad the data to twice its length, the "artificially added" samples will fall exactly between the original samples and thus on the peaks of the side lobes. plotting this in matlab (i.e. directly connecting each sample with a line) will then look like a sawtooth pattern as it was the case in your data. to see if this really is the reason, just change the amount of zero padding to a non-integer multiple of the sampling frequency (e.g. from a default of 1024 to 1000) or disable the zero padding completely and check wether your data look better then.
i hope i could help,
michael
 
On Oct 3, 2011, at 3:21 PM, Anders Sand wrote:Hi all,
when I apply pop_spectopo(): (using the GUI) on any .bdf data I get what looks like aliasing (see http://www.mypicx.com/10032011/Aliasing/). I have tried importing the files via the biosignal toolbox and via converting (via polyrex) to .cnt. It looks the same for any .bdf file I open but looks normal for all example data attached with EEGLab. All frequency parameters are default.
Could somebody help me troubleshoot this problem?
ThanksAnders_______________________________________________
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
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20111006/b6880e32/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Picture 2.png
Type: image/png
Size: 25341 bytes
Desc: not available
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20111006/b6880e32/attachment.png>


More information about the eeglablist mailing list