Re: pci compliant devices?

[tlyungha@collins.rockwell.com]

PCI compliant parts will drive all the way to the rails, even without a
reflected wave, it's just a question of how long.  When they get up to, or
down to, a particular voltage they are required to drive with a higher
impedance.  This slows the edge down, but then the reflection will come
along and push the waveform up to close to the rail.  The part would
eventually get there even without the reflection, but it would just take a
little longer.  When I look at the curves in the PCI specification for 5V
signaling, it appears that a PCI compliant device doesn't even have to ease
up on the output impedance until they are above (on the way up) and below
(on the way down) the TTL thresholds.  It seems like a PCI device could be
designed to be TTL compliant even without derating the timing for driving
TTL levels.

There are probably a lot of specs-man-ship games that can be played to
allow compliance with multiple technologies even without a configurable
output (which some devices support).

Tom Yunghans
Rockwell Collins
(909) 868-2694

.




Stuart Adams <sja@brightstareng.com> on 03/09/2000 01:09:26 PM

To:   pci-sig@znyx.com
cc:    (bcc: Thomas L Yunghans/Pomona/Collins/Rockwell)

Subject:  Re: pci compliant devices?





 What I don't understand is how FPGA/CPLD's claim
 both PCI compliance as well as TTL, LVTTL, LVCMOS
 compliance. PCI is reflected wave switched with the
 CLK->Q time (Tval) spec being the time from the
 clock edge to Vstep/Vtest.

 This in contrast to TTL, LVTTL, LVCMOS in which
 CLK->Q is the time from the clock edge to the output
 reaching Voh min or Vol max.

 For FPGA/CPLD's is seems that you just need to make
 sure you meet the PCI setup spec since the output
 driver is not relying on the reflected wave in order
 to reach Vih min.

-- Stuart