Detecting USB 2.0 port(s) on ASUS P4B266 motherboard

Henry Spencer henry-lqW1N6Cllo0sV2N9l4h3zg at public.gmane.org
Thu Mar 18 17:20:11 UTC 2004


On Wed, 17 Mar 2004, Taavi Burns wrote:
> ...If you feed a packet to the USB bus
> that is only ones, you will experience a 14% degredation in speed due
> to padding (a string of zeros does not suffer from this; I suppose the
> reasoning is that a string of zeros is "normal" and a string of ones
> is "less normal").

The most usual reason for that sort of thing is that a zero is coded as a
transition and a one is coded as no transition, and you need at least
occasional transitions to maintain clock synchronization.  A second common
reason is that a sequence of N consecutive ones is used as a marker for
things like packet framing, and so data has to have zeros injected to
guarantee that it never looks like a marker.  (Sometimes both apply.)

But yes, as I understand it, the reason for victimizing sequences of ones
rather than sequences of zeros is precisely that the former is thought to
be less common. 

> AFAIK, ethernet works the same way; 10bT runs at 10MHz, 100bT runs at
> 100MHz.  Hence the speed does not include sync, headers, or
> collisions.

Correct, although for large packets, those overheads are quite small.
(There is no sequence-of-ones issue for 10Mbit Ethernet -- it uses an
encoding that doesn't care -- and I don't *think* there's one for 100Mbit
either.)

                                                          Henry Spencer
                                                       henry-lqW1N6Cllo0sV2N9l4h3zg at public.gmane.org

--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml





More information about the Legacy mailing list