Fine-grained Energy Profiling in Mobile Devices 

Research output: ThesisLicenciate's thesisTheses


Mobile phones have several use cases such as making a phone call, sending an SMS, browsing the Internet, or playing a game. Mobile phones are also equipped with a wide variety of hardware allowing these activities. The overall energy consumption is easily measurable but it does not explain how much power is dissipated on particular activities and which devices are mostly responsible for the consumption. Thus, this thesis defines the concept of fine-grained energy profiling where the total energy consumed is broken down into subsystems. This thesis examines mobile phone power dissipation and energy consumption analysis. It describes how to measure overall total energy consumed by the device and develops methods that allow breaking down the energy consumption to components or subsystems. Specifically, the tools developed in this thesis allow studying the energy consumed by the CPU, GPU, display, WiFi, Cellular 3G and SSD disk. The device studied here is the Nokia N900 Maemo/Linux phone. This thesis describes a full-system power measurement setup including a fake battery and DAQ acting as an electric power meter. The thesis also develops logging tools that monitor the subsystem load and provide a method for formulating a linear regression model between these two through a set of microbenchmarks. This model allows estimating the subsystem energy consumption and total energy consumption based on observed loads of the subsystems. The model achieved 80% accuracy when compared with measured total energy consumption to the total power dissipation predicted by model.
Original languageEnglish
QualificationLicentiate's degree
Awarding Institution
  • Aalto University
  • Soisalon-Soininen, Eljas, Supervising Professor
  • Hirvensalo, Vesa , Thesis Advisor
Publication statusPublished - 2016
MoE publication typeG3 Licentiate thesis


  • Computer science energy
  • regression
  • Phone


Dive into the research topics of 'Fine-grained Energy Profiling in Mobile Devices '. Together they form a unique fingerprint.

Cite this