hi john,
regarding the recent NRAO memo on word boundry detection (in your link below): i think the early ATA engineers tested prototype hardware that didn't use 8/10 encoding in high speed serial links. (since the data is noise like, one shouldn't need two extra bits to insure sufficient transitions and detect word boundries), but they encountered problems, and abandoned them in favor of 8/10 encoding. philosophically, i prefer industry standards, like 8/10 encoding. one can transmit test patterns or any data one wants; it's more flexible and upgradable, and doesn't cost much more - it's sometimes cheaper to stick with industry standards, especially if one includes NRE costs. and SERDES's have tricky stuff inside - many of them don't gaurantee specs if you don't 8/10 encode. best wishes, dan John Ford wrote:
Thought this might be of interest: New Electronics Division Technical Note EDTN No. 213 Title: Word-Boundary Detection in a Serialized, Gaussian-Distributed, White-Noise Data Stream Authors: Matt Morgan, Rick Fisher Date: October 13, 2009 http://www.gb.nrao.edu/electronics/edtn/index.html