RE:Wanted...PCI bus 3V driver/receiver models

From: Ben Andresen <andresen@asic.sc.ti.com>
Date: Wed Feb 28 1996 - 15:32:33 PST

Jeff, I agree that 5V chips powered from 5V and 3V chips powered from 3V can
both talk on a PCI bus, but it isn't as simple as you show it to be. The
motherboard defines the signaling environment, EITHER 3V or 5V. All cards
must meet the specs for that signaling environment, and the characteristics
are different. The 3V signaling environment input thresholds are much tighter
than for 5V so an input buffer conforming only to 5V specs will not be 3V PCI
compliant. The output buffer V-I characteristics change between 3V and 5V
operation. Also, there MUST be a Vcc clamp diode for 3V PCI (PCI Local Bus
Spec Rev 2.1, section 4.2.2.2) which will cause outputs driving 5V rail to
rail severe problems in a 3V signaling environment.

Bottom line is you must use I/Os specifically designed for the signaling
environment defined by the motherboard to be PCI compliant. You CAN'T
arbitrarily mix 3V and 5V signaling components and expect it to work. This
doesn't say anything about the supply voltages and technologies of the
various components on the bus which can be almost anything. I'm talking only
about the PCI signaling environment as defined by the motherboard.

If the final signaling environment is undefined the solution is to use
"universal" PCI I/O buffers. These sense the PCI signaling environment being
used by measuring the PCI I/O designated power pin voltage, and adapt their
characteristics to fit the appropriate specs. These CAN be used as without
reservation as described in your msg below.

Ben Andresen

> From owner-ibis@vhdl.vhdl.org Wed Feb 28 15:23:37 1996
> To: gdpeter@sandia.gov
> Cc: chu@rock.enet.dec.com, ibis@vhdl.org
> Subject: RE:Wanted...PCI bus 3V driver/receiver models
>
> Gary: I believe your situation as pictured below
>
> 5V 3V
> | |
> | |
> ------ PCI bus ------
> | 5V | --------------- | 3V |
> | chip| | chip|
> ------ ------
>
> is perfectly valid under the rules of the PCI protocol provided no
> additional power is drawn from the 5V when it is driving the bus.
>
> FYI: Digital and others have been doing this successfully for years.
> I'll be glad to provide you with any infomation you may need for the
> success of your project.
>
> Jeff Chu
> Digital Equipment Corp.
> ======================================================
> ======================================================
> Date: 26-FEB-1996 18:55:05.73
> From: CRL::"gdpeter@sandia.gov" "Gary Peterson"
> Subj: Wanted...PCI bus 3V driver/receiver models
> To: si-list@silab.Eng.Sun.COM, ibis@vhdl.org
> CC:
> I am trying to get a feel for how one might mix 3-volt level signaling
> and
> 5-volt level signaling PCI bus driver/receivers on a common PCI bus. I
> have
> IBIS models for 5-volt level signaling driver/receivers from the IBIS
> library courtesy of Intel. I would like to get models that are 3-volt
> level
> signaling and "5-volt compliant" (I think this just means no VCC
> clamping
> diode so the parts can be used with 5-volt supply) and models that are
> 3-volt level signaling ONLY (cannot be used with 5-volt supply because
> they
> have proction diodes?).
>
> I am modeling the behavior of a specific PCI bus design using Quad XTK.
> We
> will run at 25MHz instead of 33MHz so have more than 10ns for data to
> settle.
> IBIS models are preferred as they require less trouble on my part to
> translate
> into useable Quad models (I'm one of those u-s-e-r-s who wants models
> that
> plug-and-play without any effort on my part...ha). I can use Hspice
> models,
> if that's all that is available, and will establish non-disclosure
> agreements
> as necessary to get them. Thanks to any and all who are willing to
> help.
>
> P.S. I'm placing an order with ZEELAN ASAP in order to get as many
> plug-and-play models as I can afford. I've finally bit the bullet
> and decided I'll have to pay time and $ to get the models I need.
>
> Gary P.
> ---
> Gary D. Peterson
> _/_/_/ _/ _/ _/ SANDIA NATIONAL LABORATORIES _/_/_/
> _/ _/_/ _/ _/ P.O. Box 5800, M/S 0503 _/_/
> _/_/_/ _/ _/ _/ _/ Albuquerque, NM 87185-0503
> _/_/_/_/_/_/
> _/ _/ _/_/ _/ Phone: (505)844-6980 _/ _/_/
> _/
> _/_/_/ _/ _/ _/_/_/_/ FAX: (505)844-2925 _/ _/_/ _/
> E-Mail: gdpeter@sandia.gov _/_/_/
>
>
>
> % ====== Internet headers and postmarks (see DECWRL::GATEWAY.DOC) ======
> % Received: by easynet.crl.dec.com; id AA29668; Mon, 26 Feb 96 18:52:15
> -0500
> % Received: by crl.dec.com; id AA06904; Mon, 26 Feb 96 18:45:00 -0500
> % Received: from sass165.sandia.gov (sass165.sandia.gov [132.175.109.1])
> by vhdl.vhdl.org (8.7.3/8.7.3) with SMTP id PAA15733 for
> <ibis@vhdl.org>; Mon, 26 Feb 1996 15:29:39 -0800 (PST)
> % Received: from sass191.ess.sandia.gov (sass191.csu891.sandia.gov
> [134.253.21.54]) by sass165.sandia.gov (8.6.12/8.6.12) with SMTP id
> QAA27862; Mon, 26 Feb 1996 16:21:50 -0700
> % Received: from sass1365.sandia.gov by sass191.ess.sandia.gov
> (4.1/SMI-4.1)id AA04905; Mon, 26 Feb 96 16:23:13 MST
> % Date: Mon, 26 Feb 96 16:23:13 MST
> % From: gdpeter@sandia.gov (Gary Peterson)
> % Message-Id: <9602262323.AA04905@sass191.ess.sandia.gov>
> % To: si-list@silab.Eng.Sun.COM, ibis@vhdl.org
> % Subject: Wanted...PCI bus 3V driver/receiver models
>
>
>
>
Received on Wed Feb 28 15:47:36 1996

This archive was generated by hypermail 2.1.8 : Fri Jun 03 2011 - 09:52:29 PDT