Hi, I'm editing the 64b/66b page and noticed that a reference to it was deleted from 8b/10b. I disagree with the reasons for deletion.
64b/66b was specifically designed to put statistical limits on DC balance and run-length. It performs in both these regards better than
SONET which was designed to drive AC-coupled laser links. I've modified the 64b/66b page with references to the draft proposal that
describes the performance of the code in these regards. It is true that 8b/10b puts hard, short limits on DC balance. 64b/66b instead
says that with a .001uf coupling capacitor, you'll only exceed a 2.5% DC walkoff in 10^22 bits (that's 31,000 years). The main observation
with 64b/66b is that the hard limits of 8b/10b weren't worth at 25% code overhead, and that solid, robust systems could be designed using
reasonable statistical methods. At some point, the real risk of a nuclear war, meteoric impact, or a power supply failure dominates both 8b/10b and 64/66b reliabilities.
kind regards, -- Rick Walker Richardcwalker (talk) 02:30, 27 April 2012 (UTC)
Please forgive me if I'm not using wikipedia properly. This is my first attempt at joining the party.
- Hi Rick,
- I was mostly referring to DC balance. 64b/66b encoding really doesn't do anything in that regard, if your data is biased in some way, your link will drift out of balance. 64b/67b was added to address this and is used for e.g. Interlaken (networking).
- Now, run-length is indeed limited to 65 bits, but - you say it yourself, 8b/10b encoding places hard limits. I'd simply say that their goals are remotely the same, but not similar, how about that?
- (also, the section heading "Exceptions" is totally misleading - there's no rule and no exception...)
- --Eqvinox (talk) 08:52, 13 July 2012 (UTC)