Improving Computer Scalability & Reliability to Enable Continued Technology Advancement

– Jerry Cox

The Clock-Domain-Crossing conundrum is this:  Where is the evidence of Clock-Domain-Crossing (CDC) failures that have occurred during a product’s service life?

We know there have been products that have had CDC failures during test and been re-spun, revised or abandoned before being put into service. Theory predicts there must also be products that have had CDC failures after test.

Knowledge of such failures is of utmost importance because we are becoming much more dependent on safety-critical electronics, the kind of electronics that will help you avoid colliding with that car in your blind spot, will deploy your airbag in an accident or will reduce the speed of the train on which you are riding as it negotiates a curve.

Number of CDCs per chip rapidly growing
Today’s System-on-Chip (SoC) products contain a large number of independent clock domains. This is because IP cores may run at a variety of frequencies and, in addition, it is impossible to satisfy the traditional synchronous design requirement of almost simultaneous clock-signal delivery to everywhere on a large chip.