[FieldTrip] DPSS and montage

[Roemer van der Meij]

Hi Casper,

On the memory issues, using DPSS tapers on very large segments of data, 
while at the same time using cfg.foilim (instead of the more selective 
cfg.foi), results in a huge explosion of required memory. This is so 
because the amount of tapers for a specified smoothing increases with 
the length of the trials, and the same holds for the amount of estimable 
frequencies. In your case, it might be easiest to specify a cfg.foi, 
instead of cfg.foilim, e.g. cfg.foi = 20:1:200 should already greatly 
reduce the amount of memory needed. Increasing the amount of smoothing 
also greatly increases the amount of tapers needed. If you are looking 
for a higher smoothing, and you still don't have enough memory, you can 
try to cut your trials in smaller pieces, or use downsampling (the 
amount of tapers depends on the amount of samples used).

Best,
Roemer



On 12/10/2011 11:40, Casper van Heck wrote:
> Dear fieldtrippers,
>
> I've been using Fieldtrip as an Intern for two months now, and have 
> repeatedly run into a problem when using DPSS. When using the below 
> settings, the computer locks up within seconds (although it does seem 
> to do 'something'), stays that way for up to ten minutes (which is my 
> personal cutoff, where I force a reboot), and then reports an 'out of 
> memory'.
>
> cfg.method = 'mtmfft';
> cfg.output = 'pow';
> cfg.taper = 'dpss';
> cfg.foilim = [20 200];
> cfg.tapsmofrq = 1;
>
> The data fed to ft_freqanalysis is the result of roughly 30 seconds 
> continuous data of a single (originally bipolar) channel processed by 
> ft_preprocessing.
> When using 'hanning' it completes without a hitch, but DPSS doesn't. 
> I've tried the same thing using a different, more powerful computer, 
> which succesfully completed 30 seconds using the above settings, but 
> when setting the tapsmofrq to 3 the RAM was filled up in a period of a 
> few seconds, and the computer choked. The 'normal' computer has 2GB of 
> ram, and a dual-core intel running at 2GHz, and the more powerful 
> computer has 8GB ram with an quad-core AMD at 3.4GHz. What am I doing 
> wrong?
>
> Secondly, I was wondering about the 'montage' funcionality, since I'll 
> be making use of bipolar arrangements, but the parameters of the 
> montage-structure as defined on 
> http://fieldtrip.fcdonders.nl/reference/ft_apply_montage don't help me 
> much in making one of my own. Would it be possible to take a look at 
> an existing montage structure, such as the one referred to on 
> http://fieldtrip.fcdonders.nl/getting_started/neuralynx_fcdc?
>
> Sincerely,
>
> Casper van Heck
>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

-- 
Roemer van der Meij M.Sc.
PhD student
Donders Institute for Brain, Cognition and Behaviour
Centre for Cognition
P.O. Box 9104
6500 HE Nijmegen
The Netherlands
Tel: +31(0)24 3655932
E-mail: r.vandermeij at donders.ru.nl
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20111012/3d40da4e/attachment.html>