[Eeglablist] memory problems when reading in 128 channels Biosemi data

[Jim Kroger]

Hello, just to report our experiences, Matlab generally wants all data to 
be resident in RAM, and it uses 64 bit precision to represent the data, 
regardless that it is recorded in another precision such as 32. Then, 
EEGLAB will keep two copies of data, one working copy, and one for safe 
keeping, that does not include changes. Also, when loading a .bdf file, the 
process of translating from .bdf to .set file (EEGLAB format) can take a 
very large amount of memory. So, what may appear to be a small .bdf file 
can end up needing vast amounts of RAM. We struggled for months, not even 
being able to load a 1 gig data file.

In general, the correct solution is to use a computer with sufficient RAM. 
For example, our workstations have either 5.5 gigabytes or 8 gigabytes of 
RAM. You have to be running a 64-bit operating system to use more than 4, 
and Linux makes better use of RAM than Windows. This generally handles 
things. Second, the size of your data will have a big impact, especially 
when you are using 128 channels. We used to record for a couple hours, and 
sample at 1024/second. This created enormous data sets. We now do our ERP 
work at 256 by downsampling our raw data. For ERP approahces this is good 
enough. We keep our higher-sampled raw data around in case we want it for 
various time-frequency or synchrony analyses later. Second, you can split 
the data in several ways, then recombined the data after opening and saving 
it as a set file and perhaps doing preprocssing such as filtering and ICA, 
which can be impacted by large file sizes (we recently had an ICA on 32 
channel data take two weeks). When you epoch data, it makes (potentially) 
smaller files. You can only load in part of a .bdf file to work with and 
recombind later. You can just work with some of the channels, then work 
with others, then recomdine later. Feel free to email if you continue to 
have problems.

Jim


At 01:27 AM 5/22/2007, Miriam Gade wrote:
>Hi all,
>
>we recently bought a Biosemi EEG System and are now trying to analyse a 
>first oddball experiment  recorded with 128 channels. However, EEGLAB 
>continously reports memory failures when we try to read the data in, even 
>though the task manager under windows shows 500 MB of free memory. Has 
>anybody of you encountered the same problem? Do you know how to circumvent 
>it? thanks for any answers and greetings from italy miriam
>
>_______________________________________________
>eeglablist mailing list eeglablist at sccn.ucsd.edu
>Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu


--------------------------------------------
Jim Kroger
NMSU Psychology MSC 3452
220 Science Hall, Williams Street
Las Cruces,  NM 88003-8001
Tel:  (505) 646 2243
Fax: (505) 646 6212

Our website:
http://www.psych.nmsu.edu/~jkroger/lab/index.html
--------------------------------------------