EEGLAB hardware and software recommendations
Open source spirit
EEGLAB is an open source software environment and is available free of charge to any user.
However, EEGLAB requires that you purchase and install the commercial Matlab software environment. Matlab version 7 is required; any version of Matlab 7 will run EEGLAB itself, though some new toolboxes require Version 7.6 or higher. Matlab and EEGLAB run under Linux/Unix, Mac OS X, or Windows. Purchasing Matlab can be expensive, though the Matlab student version is usually priced near $50 US.
We have attempted to remove all dependencies on add-on Matlab toolboxes. Thus, no additional toolboxes are necessary to run the EEGLAB (10.2 or higher) core distribution. However, some advanced plug-in toolboxes may use functions from a specific Matlab toolbox: see their documentation for details.
Note that all EEGLAB processing functions also can be run under the free Octave environment (v3.4) which may be useful when performing batch processing on a large number of datasets. However, Octave is about 50% slower than Matlab and its graphic functions are not yet fully Matlab code compatible (see here for producing figures under Octave).
Desirable hardware configurations for working with EEGLAB
Below we outline hardware and software requirements for EEGLAB and associated tools. Hardware requirements vary with the size of the datasets you want to process. We have outlined two levels of hardware needed for basic and advanced processing respectively. It is also possible to process data on (some) less powerful platforms but in trying to use them you may end up spending much of your time trying to avoid "Out of Memory" errors.
Basic processing requirements (and/or highly desirable features) for processing up to, say, 32 EEG channels per subject using core EEGLAB functions only:
- A 64-bit processor -- this greatly increases the amount of memory the system can address compared to 32-bit systems.
- Two or more processors -- this is common in new machines.
- 4 Gb (or more) of RAM (random access memory) -- the simplest solution for 'Out of Memory' problems when using a 64-bit processor.
- Windows 7, Linux, or Mac OS X. (Avoid Windows XP which does not work well with Matlab because of memory issues).
- At least 500 Mb of available hard drive space per subject for most simple psychophysiological experiment data.
- Matlab 7.x -- no additional Matlab toolboxes are required (but see below).
Advanced processing requirements (and/or highly desirable features) for processing 64 or more EEG channels and/or using the most recent EEGLAB toolboxes:
- 64-bit processors -- these greatly increase the amount of memory the system can address compared to 32-bit systems.
- Quad processors cores (4 or more ) are desirable - Quad-core machines are now standard in workstations class machines and laptops.
- 8 Gb or more of RAM (random access memory) -- the simplest solution for 'Out of Memory' problems when using a 64-bit processor.
- Windows 7 (or above), Linux, or Mac OS X. (Note: Avoid Windows XP which does not work well with Matlab because of memory issues).
- Note: Only Matlab versions 7.6 (or above) contain object-oriented programming tools used in the most recent toolboxes (SIFT, MoBILAB, BCILAB).
- At least 1 Gb of available hard drive space per subject for most psychophysics experiments
- The Matlab signal processing, statistics, and optmization toolboxes, whose functions are used in some advanced plug-in toolboxes.
- Although EEGLAB is free, the Matlab software environment that it runs on is a commercial product of The Mathworks.
- For Windows users: Avoid Windows XP which does not work well with Matlab because of memory issues.
- For advanced Mac OSX users: Matlab cannot access what OSX calls 'inactive memory'. To free it for use by Matlab, on a terminal window command line type
% du -su /
- For advanced Windows users: It is possible on some platforms to start up Matlab with an alternate memory manager using the "–memmgr" option. Look here for more information.
- For Ubuntu linux users: There are often graphics problems when Matlab uses OpenGL graphics under Ubuntu. To avoid them, at the beginning of each Matlab session enter
- For GPU users: EEGLAB functions do not use yet the capabilities of GPU (Graphic Processing Unit) boards, which currently have the highest performance/cost ratio. However, we have made first attempts at porting some functions to work with GPUs -- click here for more information.
Printing and editing EEGLAB figures
To edit pictures saved from Matlab figures so as to format them for publication, we advise the following procedure:
- Export the figure (myfigure) from Matlab as a postscript image (the .epsc format) by typing on the Matlab command line
>> print -depsc myfigure.eps >> print -dpdf -r300 myfigure.pdf
- Edit the resulting postscript or PDF image file using an editor such as Adobe Illustrator.
- If your vectorized image appears as an uneditable bitmap in the postscript editor, change the Matlab renderer to painter before saving the figure by typing on the Matlab commandline, before saving the figure,
>> set(gcf, 'renderer', 'painter')
- If you have images with transparent regions, export them using the Matlab SVG export tool (PDF above will not work).
- To ensure that the image you are seeing is the image that is being saved and printed, on the Matlab command line type
>> set(gcf, 'paperpositionmode', 'auto');
- To change the font of all elements in a figure, try
>> setfont(gcf, 'fontsize', 16);
Unfortunately in some recent versions of Matlab, saving vectorized version of figures has become difficult (artefacts in STUDY scalp topographies). Let us know if you find better solutions.