[SI-LIST] Re: PCI Bus Routing
- From: James_R_Jones@xxxxxxxx
- To: ira@xxxxxxxxxxxx, si-list@xxxxxxxxxxxxx
- Date: Thu, 29 May 2003 13:27:38 -0500
I believe that the best way to determine the answer to this problem is to
brute force simulate the problem and guarantee the PCI timings (setup and
hold) at each device.
That being said, there is a trend that I personally have noticed regarding
PCI topologies that I will share. I may say that the issue that you are
really dealing with is that of settle delay. So the total etch length is
the key, and guaranteeing setup time becomes an issue.
Imagine a daisy chain configuration. The total length of the trace will
dictate flight time and settle delay when PCI traffic goes from end device
to end device. Also total length dictates settle delay from end device to
middle device, because the waveform at the middle device will not settle
until the reflection has propagated to the end device and back. So the
total etch length will dictate whether the bus will meet timing.
Now imagine a tee topology with three total devices. Now you have a stub
hanging off to each device. When PCI traffic goes to one device, the
waveform will not settle until the reflection comes back from the other.
So, once again total etch length is dictating whether the bus will meet
timing.
Both reflective cases must take into account the length of the stubs and two
times the stub propagation latency. All devices must be able to talk to
each other according to PCI spec. So now I think you would have to look at
the placement of your comoponents and theoretically determine each
combination of driver/receiver and then figure when the worst case
reflection will settle. This will give you a theoretical "better" between
the two topology options.
With the complexity of your topology, there really is no substitute for
simulation, however, because the number of stubs on your tee will cause an
almost untrackable number of reflections. Whatever you decide, you would be
well advised to check in simulation. Also simulation will give you an
indication of how xtalk will affect the topology, which I don't know of any
theoretical way to determine these effects on your complex topology.
James R. Jones
Dell
-----Original Message-----
From: Ira Kravitz [mailto:ira@xxxxxxxxxxxx]
Sent: Thursday, May 29, 2003 6:48 AM
To: si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] PCI Bus Routing
Hi,
Should a PCI Bus (32 /33M) going to 9 devices be routed in a daisy chain
fashion (will be about 16 inches long) or as a bus with stubs hanging off
it.
Thanks
Ira Kravitz
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
