[Eeglablist] Biosemi rereferencing

[Jeff Johnson]

I've been trying out the new version of EEGLAB and was curious about this 
new message given when importing Biosemi BDFs:

"We highly recommend that you choose a reference channel IF these are 
Biosemi data
(e.g., a mastoid or other channel). Otherwise the data will lose 40 dB of SNR!"

Can anyone explain the nature of this message?  Is it just a general 
warning that BDFs should be rereferenced before data analysis (which to my 
knowledge is the typical practice with all EEG), or is it a warning that 
the BDFs should be rereferenced at the stage of importing into EEGLAB?  Is 
the loss in SNR due to the 24- vs. 32-bit difference between the BDFs and 
MATLAB?

In trying to get a grasp on this issue, I've compared these two procedures:
1)  Importing the data reference-free, then rereferencing to linked 
mastoids and exporting to ASCII.
2)  Importing the data referenced to linked mastoids, then exporting to ASCII.
When doing this, there were occassional differences in values on the order 
of .01 or .001.  Are these worth worrying about?  My exported values for 1 
channel after doing this were in the range of about -9900 to -8800, so I 
was wondering if these minute differences will eventually be transformed 
into larger differences in terms of microvolts.

Thanks,
Jeff