RE: Approaches for PCI Overvoltage design using 3.3volt technology

[Ben Andresen <andresen@asic.sc.ti.com>]

I think 5V tolerant and 'universal' designs MUST include diode clamps to the 
Vi/o (PCI) bus.  As you point out this will necessitate a fair number of Vi/o 
pins to support the current load of the overshoot current.  There aren't any 
series resistors, though.  The resistance is in the source.  Your diodes just
have to be adequate to handle the current while maintaining a reasonable 
forward voltage drop.

When you build 5V or 3.3V/5V PCI interface in a 3.3V technology the actual
drivers work off the 3.3V rail which runs the core logic.  The diode clamps
are to the PCI Vi/o bus.  This meets the PCI spec just fine, but you now need
enough 3.3V core power pins to drive all the PCI outputs as well as 3.3V/5V or
5V PCI power pins to handle the switching transients.  This is a difficult
concept for some customers to accept, eg, one 5V power pin isn't enough.  The
drivers also must be 5V "tolerant".  This means the outputs do not have clamp
diodes to the core 3.3V supply; only to the PCI Vi/o supply.

Ben Andresen 


> From pci-sig-request@znyx.com Mon Feb 24 17:20 CST 1997
> Resent-From: pci-sig-request@znyx.com
> Resent-Date: Mon, 24 Feb 97 13:58:00 PST
> Cc: pci-sig <pci-sig@znyx.com>
> Subject: RE: Approaches for PCI Overvoltage design using 3.3volt technology
> Encoding: 37 TEXT
> X-Mailer: Microsoft Mail V3.0
> Resent-Message-Id: <"deNv01.0.yU2.7zW4p"@dart>
> X-Mailing-List: <pci-sig@znyx.com> archive/latest/4313
> X-Loop: pci-sig@znyx.com
> Resent-Sender: pci-sig-request@znyx.com
> To: Mailing List Recipients <pci-sig-request@znyx.com>
> Content-Type> : > text> 
> 
> 
> 
> From:  Andy Ingraham[SMTP:ingraham@wrksys.ENET.dec.com]
> 
>  --snip question on overvoltage in 5 volt signaling environment +   
> explanation about devices limiting voltage.
> 
> >> I was wondering how designers out there are handling this problem in   
> 3.3
> >> volt technology designs.  Are you designing only for a 3.3 volt   
> signaling
> >> environment?  Are you designing special I/O buffers which are actually   
> 11
> >> volt tolerant?  Are you ignoring this part of the spec altogether?
> 
>  --snip response about nobody designing only for 3.3 volts + explanation   
> of overvoltage waveform and clamp diodes for ESD
> 
> >Overshoot clamps, though not required in PCI's 5 V signaling
> >environment, are highly desirable.  The PCI Spec even points out that
> >they are probably necessary in practice when using 3.3 V components on
> >the 5 V bus.  You do need to be careful about how you implement them,
> >though; you can't just clamp to the 3.3 V rail.
> 
> >Regards,
> >Andy Ingraham
> 
> Are you saying that most designers use large series resistors (say on the   
> order of 1k) and protection diodes to the 5 volt rail to limit the   
> current and voltage respectively in their designs?  It would seem that   
> diodes alone would require a large number of 5 volt pins to sink the   
> current back into the board since the current would be on the order of   
> ~90mA. Per pin.  The addition of large series resistors would seem to add   
> problems elsewhere (undesirable propagation delays, current limiting on   
> the outputs etc.)
> 
> Alan
> 
> 
¶ÔÃ