Electrical Engineering
      and Computer Sciences

Electrical Engineering and Computer Sciences

COLLEGE OF ENGINEERING

UC Berkeley

   

Research Projects

Platform-based Mixed-Signal System Design

Xuening Sun, Alberto L. Sangiovanni-Vincentelli, Pierluigi Nuzzo and Alberto Alessandro Angelo Puggelli

Gigascale Systems Research Center

The increasing complexity of mixed-signal design due to extreme design constraints, increasing process unpredictability, and time-to-market demands require a new look into design methods. Traditional top-down design requires many design iterations and incremental changes to churn out suboptimal designs that may or may not meet design requirements. This task focuses on adapting the platform-based design paradigm to mixed-signal systems to facilitate IP reuse, high-level system design with guaranteed feasibility, and shortened time-to-market of innovative applications. Specifically, we focus on providing a set of tools and methods to support: 1. Efficient and accurate modeling of circuit (RF, analog, digital, and mixed-signal) performance space to create a comprehensive component library to serve as a platform for high-level system design. 2. Automatic physical architecture synthesis (component composition) of mixed-signal systems given application-specific signal processing algorithms and the component library to generate the set of all feasible architectures. 3. Global (architecture and circuit configuration) system optimization to determine optimal system design for the given application. The design driver for this task is a unidirectional UWB data communication system, where transmission occurs from multiple power-limited sensor nodes to a central coordinator. Since the transmitter is implemented in the simplest architecture to minimize power consumption, the main design challenge is recovering the weak signal on the receiver side with minimal power consumption.

Figure 1
Figure 1: Architecture synthesis flow.

Figure 2
Figure 2: Analog-PBD approach.