Re: [SI-LIST] : PCI BUS: UNEQUAL LOADING

About this list Date view Thread view Subject view Author view

From: Chandan (chandan_career@yahoo.com)
Date: Tue May 22 2001 - 02:06:45 PDT


Hi Steve,

Lets consider a case in which the PCI bus length is
long enough to create such effects. Let us
also leave out the devices at the end of the PCI bus.
Then I would expect the following effects:
1. The drivers would then be over-loaded because the
loaded characteristic impedance could be
lower than the impedance that it was designed to
drive.
2. There will be a mismatch between the driver's
output impedance & the transmission line
impedance, thus creating other transmission line
problems.
I assume that you had these two points in mind when
you suggested using polite drivers & terminations.

I have some other questions
1. By damping resistors, I assume that you mean series
resistors. Do you suggest usage of series
resistors (or AC terminations) for all devices on the
PCI bus? Have you tried this on a design before?
PCI is a bus that makes use of reflected wave
switching, and hence does not use any form of
termination. I have not seen any design where
terminations (series, parallel,or AC) have been
used.
2. PLease explain "unequal loading is frequency
dependent" (in more detail).

Thanks,
Chandan

--- "S. Weir" <weirsp@atdial.net> wrote:
> Chandan,
>
> Unfortunately the statement ""see" a line of
> characteristic impedance that
> is half..." is not really accurate. This is true
> for spectra that is
> significant compared to the physical path on either
> side. This means that
> the unequal loading is also frequency dependent, (
> joy ).
>
> So, what happens is the devices that are near one
> end of the bus or the
> other are somewhat impacted, while the device(s) at
> the center of the bus,
> and assuming the bus is long, are the most impacted.
> The original idea of
> the PCI bus was to limit its length so as to
> minimize the effects we are
> talking about to an "acceptable level".
>
> The typical resolution begins with using drivers
> that are "polite". If
> this is not adequate, then the next typical step is
> damping resistors. In
> the extreme case AC terms are placed at both ends of
> the bus.
>
> Don't add stubs to try to balance the bus. Stubs
> never improve signal
> performance.
>
> Regards,

> Steve.
> At 09:23 PM 5/21/01 -0700, you wrote:
> >Hello everybody,
> >
> >Assuming that devices on a PCI bus are connected in
> a
> >daisy chain, all the drivers - except those at the
> >ends of the PCI bus - "see" a line of
> characteristic
> >impedance that is half the value of the loaded
> >characteristic impedance.
> >
> >1. What are the adverse effects of this unequal
> >loading of the PCI bus drivers?
> >2. Is there any other way to avoid this situation?
> >
> >In order to ensure that the PCI devices at the end
> of
> >the bus also "see" the same value of characteristic
> >impedance, I would like to extend the PCI bus on
> both
> >ends (i.e., I would like to add an extra inch prior
> to
> >the first device and another inch of trace after
> the
> >last device).
> >
> >Kindly comment. Is it worth the trouble?
> >
> >Thanks,
> >Chandan

__________________________________________________
Do You Yahoo!?
Yahoo! Auctions - buy the things you want at great prices
http://auctions.yahoo.com/

**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. 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 http://www.qsl.net/wb6tpu
****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Thu Jun 21 2001 - 10:12:04 PDT