[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: (no subject) (3.3V & 5V environments)
> My PMC card is only 3.3V IO environment, and the PCI is 5V IO
> environment(PC), I want to debug my PMC card in PCI, so I have to
> design a
> adapter from PCI(5V environment) to PMC(3.3V environment).
> 1. Which chip can be used in this case? and have any successful
Intel makes PCI-PCI bridge chips, which are easily configured for
different signaling environments on both sides and make good PCI signal
level translators. Of course this forces your card to live behind a
bridge, which means you sacrifice latency, and I have heard of problems
booting certain devices behind a bridge on certain platforms.
If you get a PC with a 66 MHz slot, the slot SHOULD use the 3.3V
signaling environment (but be aware that there are motherboards that
blatantly violate this PCI spec). Any 66 MHz slot will run at 33 MHz if
a regular (non-66MHz) card is plugged into it.
> 2.I use TI SN74CBTD3861 as PCI buffer and voltage shifting in my
> but it doesn't work at all, why?
Don't know. In what way does it not work? Does the 5V pass through it?
Does nothing get through it? Or does your system simply not boot and
you don't know why? Please be more specific.
The FET switch approach (poor-man's level translator or limiter) looks
like it holds promise for this kind of usage, but it has some potential
disadvantages. It doesn't actually translate the voltages like a true
buffer does; it just becomes an open-circuit when the voltage exceeds
some point (rough vicinity of 3.3V), so 5V levels shouldn't go straight
through it. But the fact that it open-circuits the network is a little
odd, and MIGHT cause signal integrity problems in certain cases, so try
to have someone analyze it with SPICE or a similar program.
The PCI environment wasn't designed with these series FETs in mind. In
some cases the delays through the FET can be much greater than what they
quote on the data sheet. The series resistance does increase
dramatically as the signal voltage goes up. And the FET adds two more
IC pins of additional loading on the bus.
If these FETs were put on a PCI plug-in card, one could argue that they
cause a PCI Spec violation by having more than one device on each