[Gllug] re: SCSI vs Serial ATA

[Chris Bell]

On Mon 14 Jun, Andrew Halliwell wrote:
> 
> And verily, didst Jack Bertram announce to the hordes:
> > Indeed.  And at slow data rates, these complications aren't too
> > important (hence parallel printer cables being faster than RS232).  But
> > for any serious bandwidth, timing becomes so important that it's a
> > really big issue.
> 
> Yep... Just thought I'd also post a quick clarification for the
> non-electronics-aware...
> 
   When we talk about data rates we are referring to the rate at which real
data can be passed along a wire. This is not neccessarily as a nice square
wave, but it must be possible to retrieve the data from the signal.
   Calculations are often made based on a pure sine wave, the ideal shape
for elecrical AC mains. In fact it is only a true sine wave if it has always
existed unchanged and will continue to exist to infinity, and any change
results in a whole range of harmonics (multiples of the original frequency).
   This is also true for a wire that carries the signal, the terminations
try to emulate an infinite length of wire at each end of the real conductor,
anything else will result in either positive or negative reflections.
   A nice clean square pulse already consists of a whole range of harmonics,
each of which reacts differently, and any change in the sequence of pulses
(i.e. real data) causes even more, so the corners are rounded as the signal
travels. There can be fewer problems if the initial pulse is already rounded
by filtering to remove higher frequencies before sending any long distance.
   Anything that staggers out the other end then has to be separated from
the general background noise.

-- 
Chris Bell

-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug