I looked up some sources for GCR encoding, since that method is
one of the most efficient (short of DSP methods).

Apparently the "GCR" term (Group Code Recording) is not widely
used, since it got associated with tape drives.

"4b/5b" encoding is the most common form of GCR for network
applications.  FDDI (the fibre optic network protocol) uses it
with NRZI.  I think Fast Ethernet also uses it, but it combines
several other protocols together.

Some confusing terms:
source encoding - the way you translate your data before sending it.
                  This is "4b/5b" for FDDI.

channel encoding - the way the logic values are actually sent on the
                   wires.  This is NRZI for FDDI.

This web site has the table that you need for "4b/5b".  You just
break your data into nybbles, and lookup a 5 bit value to send in
place of each nybble.

http://www-dept.cs.ucl.ac.uk/staff/S.Bhatti/D51-notes/node51.html

Eric Engler