RE: Right angle adapter stub length ?

["Ingraham, Andrew" <Andrew.Ingraham@compaq.com>]

> We bought an off the shelf PCI right angle adapter board that is 1.75"
> high. Will using this adapter create a stub that is longer than the PCI
> spec or cause the PCI bus to work unreliably ?
> 
Any passive PCI extender or right angle adapter (that plugs into an ordinary
PCI slot) technically violates the PCI spec.

The total length of the CLK trace, from the motherboard's PCI connector to
the PCI component, must be 2.5 inches +/- 0.1 inch.  You should already have
that much length on the PCI board you plug into the extender, so any length
on the right angle adapter is too much.

If the adapter provides more than one PCI connector, you would also have
multiple loads on that CLK which would further skew it and could make it
unreliable (glitches and false clocking).

Non-clock trace lengths should be no more than 1.5 inches total from the
motherboard's PCI connector (2.0 inches for the upper half of a 64-bit bus).
An extender that is 1.75" high adds at least that much length to the stubs
and would certainly violate all the stub length rules too.  (*)

But will it work?  Maybe.  Reliably?  No.  33MHz PCI is robust enough that
it might work, despite the excessive clock skew and other spec violations.
But there are no guarantees and in many cases it won't.  If this is just for
your in-house debugging and you don't care if it crashes, go for it, as long
as you understand the possible consequences.

To not violate PCI specs, it must be an active adapter with a PCI-PCI
bridge.

Regards,
Andy


*  A case might be made that it wouldn't add stub length if plugged into the
"last" PCI connector on the motherboard's PCI bus, if in fact one of the
connectors was on the physical and electrical end of the bus.  But since the
motherboard wasn't designed for that configuration, all bets are off.