Re: Flower Cultivation

[ross@teraflop.com (Ross Harvey)]

  > 
  > Do people really bypass +3V on a +5V-only board as the
  > spec (2.1) recommends, or does that fall in the general
  > category of "lilly gilding?"
  > 
  > Been-there-done-that, srd
  >


This was discussed at incredible length, but it was quite some time ago.

In a way, this is a really good question. It brings up how you really
have to follow the entire spec, even the parts you don't understand.
You can't just implement those parts of PCI you see the reason for, and
blow off the rest. (You don't have to do EVERYTHING, because lots of
stuff is optional, but this isn't one of those.) The PCI spec is
remarkably well-written, with a simple core, layered features, and
extensive attention to signal integrity.

In this case, that's why: it's a signal integrity issue. Those pins are
NOT just to supply DC power, they are also an AC signal return. The
pinout cleverly interleaves power and signal pins to control the Z (and
hence the t) of the data and control lines. You need those caps to
control the Z of the adjacent signal lines and to prevent crosstalk
EVEN IF YOU NEVER USE THE DC VOLTAGE THEY ARE SUPPLYING. And, you have
to locate them as specified.
dhW