RE: [SI-LIST] : PCI topology

About this list Date view Thread view Subject view Author view

From: Ingraham, Andrew ([email protected])
Date: Wed Jan 31 2001 - 14:54:30 PST

Your topology descriptions (attachments?) did not make it through the mail

> Now, when I simulate with these trace lengths using idela 60ohm
> transmission
> lines the signal especially when either P1 or P2 is driving is
> non-monotonic
> at each of the receivers
That is not a problem. Most PCI signals (except clock and reset, that is)
are expected to be non-monotonic at most receivers, particularly those in
the middle of the bus or near the driver.

> and takes a long time for the signal to settle
> down.
That IS a problem, if the time exceeds about 10ns.

> 1) If in simulations the signals look monotonic, settle delay is good
> should
> I still worry about thing relevant to PCI because I know PCI is strict on
> the stub lengths and the total trace lenght etc?
That depends.

If you are designing something strictly for yourself, you can do anything
you want. You just have to make it work.

If you are designing PCI plug-in cards, which other people will use in
various different systems, then you do need to control things like the stub
lengths, so that your card will work in any system the customer puts it

If you are designing a motherboard that accepts PCI plug-in cards, the PCI
spec has no restrictions on total trace length. The responsibility is left
up to the motherboard designer to make his system work, with whatever trace
lengths he uses, with any number of PCI plug-in cards attached.

The reason for control over things like stub lengths, is that you have one
group of people designing motherboards, and another designing PCI plug-in
cards, and every plug-in card should work in every slot of every
motherboard, even though nobody has simulated or measured all those
combinations. By controlling the stub length, there are limits as to how
"bad" each plug-in card can be, which helps make all those combinations work
by default as long as everyone plays by the same rules.

Setting maximum stub lengths helps give everyone the flexibility to
interchange cards between systems. If you don't need that flexibility, if
you have a fixed design where nothing changes, then it doesn't really matter
whether it's "PCI" or not, as long as it works and you've analyzed it to
make sure that it works.

> 2) Also, I forgot to mention that I am using 10ohm resistor in series at
> each device and this does help me in achieving the extra trace length. Am
> I
> violating anything ?
If it's on the motherboard you have a lot of freedom to do what you want,
because that's your responsibility to make it work with any plug-in card.

If it's on a plug-in card, it might be a problem, because of the
interchangeability factor. Hard to say.

On the other hand, I believe CompactPCI uses 10 ohm series resistors on each
cPCI card, which helps it achieve improved signal integrity with more cards,
compared to conventional PCI.


**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected] In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at

About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue May 08 2001 - 14:30:47 PDT