From: Walter Stafford <STAFFORD@bbri.org>
  To  : rasmb@bbri.org
  Date: Fri, 30 Apr 1999 6:57:50 -0400 (EDT)

RE: Molecular weight (Mw)

To:	SMTP%"hyarabe@unix1.sncc.lsu.edu"
CC:	STAFFORD
Subj:	RE: Molecular weight (Mw)

Dear Yarabe,

	The diffusion coefficient can be easily calculated from g(s*) curves 
by fitting to a single gaussian. The diffusion coefficient is related to the 
standard deviation of the g(s*) curve as follows:

		D = {[Rm x stdev x w^2t]^2 }/{2 x t}

Where Rm, w^2t and t are gotten from the top few lines of the g(s*) output 
file.

If you put this value of D into the Svedberg equation to calculate M you
should get a value of M that agrees with the value you got from sed eq, IF the
sample is homogeneous and the time interval wasn't too big (i.e. you used a
small number of traces). These are about the only caveats. There is a PDF file
at the RASMB site which contains a rule of thumb explaining how to choose the
maximum number of traces to use without introducing too much artificial
broadening. Remember that stdev will have units of Svedbergs and must be
multiplied by 1 x 10^-13 before inserting it into the above relation. 

ftp://rasmb.bbri.org/rasmb/spin/ms_dos/dcdt-stafford/rule_of_thumb.pdf

Let me know if that helps.

Walter