The desktop teams would like to work with the speed program to develop some low noise power tests. The desktop teams have been working on creating minimized repros of problematic web pages: see the existing trivial_pages story set, and shrike@ is working on some more. We'd like the assistance of the speed program in making these tests have lower noise.

There are two mechanisms for power measurement: 
  1) battor - influenced by background tasks
  2) TBMv2 cpu_time. Includes power used to load the page. We only want power when the page is idle.

Could we get some assistance on improving (1) and/or (2)?