Raising the level of abstraction: A signal processing system design course

Signal processing systems are inherently heterogeneous. They often contain a mixture of ltering, communication, and control algorithms implemented by a variety of technologies such as digital hardware, software, and analog circuits. Modern system-level design manages heterogeneity by rst decoupling speci cation from implementation and then decomposing the speci cation into a hierarchy of simpler block diagrams. Each block diagram can be associated with a di erent set of formal rules governing its behavior, so the hierarchical composition becomes heterogeneous. The same system speci cation can be executed for simulation or synthesized into a variety of implementations. In this paper, we describe a graduate-level signal processing system design course that presents this modern approach. The course covers block diagram modeling, algorithm speci cation, system simulation, and system synthesis. Students gain hands-on experience by modifying university and commercial systemlevel CAD tools. This paper proposes steps to transition this course into the undergraduate curriculum.