Out of a lot of 12 boards we've had two now that have experienced input pad failures (shorts to VCCO). These are LVDCI_25 IBUFs with VCCO = 2.5V. They are being driven by 2.5V LVCMOS output buffers from a Marvell Phy (RGMII interface). The Phy and Virtex-4 I/O are powered by the same 2.5V supply. The traces are in the 1.5" to 2" range and are unterminated. The engineer responsible for the board suspects overshoot, and says he has observed overshoot of up to 500mV on these lines (which happens to be the absolute max for Virtex-4 2.5V inputs).
These traces and pins are not exposed so ESD is almost impossible as a cause of the failures. The Virtex-4 inputs are pretty beefy (well protected), and I personally have never seen an input buffer fail on any IC due to overshoot from a compatible driver (as opposed to being overvoltaged by an incompatible driver). So I have my doubts about overshoot causing the failures and want to make sure we understand this before building more boards. The current plan for the respin is to add series terminators to damp the signals. Even if the problem was not overshoot, the resistors are a good idea. I'm just looking for ideas and experiences on this.
Thanks, boys (and girls!).
Rob