Can we ever send 25-100 Gb/sec signals over 24" line length of printed circuit board and still have mVolt signal at the receiver?

Abstract

We are going to show all the possibilities to achieve a 25-100 Gb/sec point to point signaling scheme for 24 inch channel length, which is made out of a printed circuit board (PCB), two packages, two sockets, and two back plane connectors. It was observed that a mVolt solution (given 1 Volt pulse signal at the transmitter and expected voltage at the receiver, after ISI noise subtracted off >1 mV) exists for 25 Gb/sec signaling scheme while using today's technology for 24 inch channel length. But for 50 and 100 Gb/sec signaling schemes, we do not have a mVolt solution with today's technology (24 inch). We have shown the required value of effective loss and the white noise required (10% of the signal amplitude) to obtain mVolt solution for 24"length. We have also identified a parameter impacting the I/O performance as we move toward a mVolt and /spl mu/Volt solution.