System-level Dynamic Power Management On Philips Nexperia Processor

Many multimedia signal processing embedded systems are intended to be hand-held and portable devices, such as cellular phones, portable consumer electronics and PDAs. Such devices work in power-constrained environments and are powered by rechargeable batteries. In order to extend the life of the battery, efficient power reduction techniques are required for the operation of these systems.Nowadays varieties of power reduction techniques have already been proposed and developed. From system-level view, there can mainly be classified into two categories: static and dynamic techniques. Dynamic or real-time techniques are the present research hot spots, which mainly consist of: Dynamic Power Management (DPM), Dynamic Voltage Scaling (DVS). In many cases, combinations of those various techniques are adopted in order to obtain the maximal power reduction for application systems.Philips Nexperia series media processors are dedicated to all kinds of multimedia applications. The advent of PNX1500 processor brings new features for saving power. To target the so-called "P4" mobile PNX1500 platform, the implementation of power reduction techniques on Nexperia media processor becomes critical. This master thesis investigates the feasibility of different power reduction techniques on Nexperia media processor. For implementation, I created a system profiler to statistically measure the application profiling information, including cycles, MIPS and workload for each running task. Based on that, a practical system-level power management program prototype was developed and implemented for a MPEG2 playback application on this media processor. I employed both DPM and DVS strategies to achieve maximal power savings. Afterwards, I evaluated the output results and estimating the ideal power consumption through a simplified power estimation simulation, as well as tracing the power reduction processes. Finally, I found that my program works well and does reduce power consumption for this practical application with tolerant AV quality. The physical measurement from Sunnyvale proved power reduction on PNX1500 processor applying such low power implementations.