Spartan3 5V PCI

Bus switches may be necessary in a system that pushes the PCI spec to it's limits. What you can get away with on an eval kit is not what you can get away with if you are making something that is required to meet the PCI 2.2. Spec.

Note very carefully that you can get away with a LOT at 5V that simply won't work at 3.3V, and indeed vice versa. I had an 'improved' part some years ago from Intel where it quit working in an existing design at 3.3V. Apart from that, have you tried it with a unit that has 3.3V slots (to match your VIO)? That's a completely different test than you are doing.

You don't say what the board you are designing is: is it a drop-in PCI board, or a self contained unit that has a PCI interface between some units on it? If it's self contained, there are things you can get away with as you don't have to interface to the outside world (been there, done that).

Cheers

PeteS

It is a matter of luck. Most modern chipsets don't have 5V used in driving the PCI bus. Very often it is 3.3V so you won't see an issue until someone plugs in an old card, or uses an old motherboard, that maybe does drive something towards 5V.

Putting bus switches in the way technically fails the PCI spec but we have not had an issue reported with our FPGA development boards to do with having bus switches there. I don't think any other manufacturers have had either that I have heard of. If you do use this approach ensure the bus switch used does not have a diode to Vcc.

John Adair Enterpo> Hi,

PeteS, It is a PCI master/target card connected to an SBC(Single Board Computer) with 5V PCI. The SBC has only one edge (male edge) such that i will be designing the board to have an edge connector(female).

-yy