Re: [SI-LIST] : Signal Integrity of 500 MHz on PCB

D. C. Sessions ([email protected])
Thu, 06 Nov 1997 14:39:59 -0700

Mike Jenkins wrote:
> On the question of propagating a 500 MHz clock across an FR4 board,
> Fibre Channel and Gigabit Ethernet designs ship sub-nanosecond
> pulses across boards routinely. The fastest continuous frequency
> in such a data stream is, in fact, slightly over 500 MHz. What
> makes this work is 1) differential signals, 2) point-to-point
> signal paths only, and 3) doubly terminated nets. If your
> application can afford these constraints, I believe propagating
> a 500 MHz clock is feasible.

The original question was wrt discretes -- more of an RF question
from the way it sounded. Otherwise, there's no special reason why
frequencies much higher than 500MHz shouldn't go over FR4, since
the main problems are in other areas. We already ship stuff around
at better than 700 MBaud routinely, and Rambus hits near a GBaud.
Package parasitics and device speeds are more of a problem than
the copper is, so long as you can avoid microstrips and similarly
suboptimal tactics.

Now shipping a GHz CLOCK is ugly, but that's mostly because the
delay is hard to contol.

D. C. Sessions
[email protected]