Newcastle University

Any system that has to interact with other, independent, sources of data needs to know when incoming data has arrived in order to make sense of it. This may seem obvious, but it is not easy to do. It means that requests from an external time frame must be synchronized with the time frame of the receiving system, but unfortunately time is a continuous quantity, where the presence or absence of a request is digital. It’s either there, or it isn’t. This is a fundamental problem in which a continuous variable (time) has to be mapped to a discrete variable (request).

Synchronizers and arbiters are the circuits that attempt, imperfectly, to do the job, and the more systems there are that have to interact, the better we need to understand their imperfections and the resulting unreliability.

This seminar will deal with measuring the performance of real circuits, the trade-off between latency and reliability, and how decision making can be made more robust.

Refreshments will be available in E4 at 1.45pm.

Published: 25th September 2006