Smartphone Power Consumption Characterization and Dynamic Optimization Techniques for OLED Display

摘要

Smartphones have emerged as the most popular and frequently used platform for the consumption of multimedia. Following the rapid growth of application number and the explosion of cellular network bandwidth, high power consumption, and limited battery capacity remain as the major challenges in smartphone designs. Therefore, lots of research is made to characterize and optimize the smartphone power performance.ududHowever, the existing research approaches on smartphone power characterization generally ignore the impact from the components' varying performance in different applications, as well as users' behavior during the practical usage. Hence, the power optimization techniques in the modern smartphone are inflexible to adapt to different application scenarios and user behaviors.ududIn this dissertation, I first proposed a new smartphone power consumption characterization and analysis approach -- ``SEER'', which was associated with both user ethological and smartphone evolutionary perspectives. The real-time power consumption is measured with a set of the most popular applications on different generations of Samsung Galaxy smartphones. And deep analysis is made to find how each smartphone component is utilized in different applications, and how the users' daily usage patterns impact on final energy consumption. The experiments show that some traditional power-hungry components, such as Wi-Fi and CPU, actually consume much less energy in practical daily usage. Meanwhile, OLED display panel is still the biggest power consumer in the whole smartphone system; even it's considered the most promising low power display technology.ududTo further optimize the display power consumption with OLED. I further proposed a set of dynamic power optimization techniques for OLED display, balancing the real-time power performance and the user visual perception experience. In this dissertation, the optimization is full-filled at three different levels: 1) Hardware based Optimization: Based on the traditional AMOLED display pixel driver, a novel DVS-friendly OLED driver design is proposed, which can minimize the display color distortion under aggressive supply voltage scaling. Correlated fine-grained DVS schemes (DiViCi) are also proposed to utilize the DVS-friendly driver into video streaming applications. 2) Software based Optimization: Despite the hardware modification, a dynamic OLED power model is built to evaluate the OLED panel power consumption and human visual perception quality assessment. A novel video category based dynamic tone mapping (DaTuM) technique is proposed for video streaming; 3) User Interaction based Optimization: The user interaction and visual perception during the display content capture phase are also taken into consideration, a novel OLED power friendly video recording application (MORPh) was also proposed.ududDedicated real-time management and reliability enhancement schemes are explored to promote the applicability of the proposed approaches . Experiments show that, with these power optimization techniques, the OLED display panel power performance on smartphone device is significantly improved with reasonable visual quality controllability.