Typical tests should last 30-60s to encapsulate most low-level periodic wakeups[1]. All the tests are to be done on battery (obviously)


We need to profile the system when using the default applications shipping with Ubuntu. The easiest way is to start up the target application set and let the machine idle, then profile using powertop and latencytop.

Target application set

  • Openoffice (with a .doc and .ppt loaded)
  • Firefox (10 tabs: Gmail and some others)
  • Empathy (connected to gtalk, jabber)
  • Weather applet
  • Rhythmbox (with 100 .mp3 playlist) playing in the background
  • Evolution (IMAP with autofetch set to 10 minutes)

Idle tests

  • Idle after boot
    • Boot up default install, wait a minute or so to let all applets to settle down (e.g. WLAN) and then measure (full brightness and half brightness)
  • Idle after resume from suspend
    • Same as idle after boot
  • Idle after resume from hibernate
    • Same as idle after boot
  • Separate testing for server, desktop and UNR install
    • server is expected to consume the least and will supply the baseline for a specific HW platform

HW-specific testing

  • Radio tests
    • WLAN
      • Idle
        • Connect to AP, do network activity for a while and then idle
        • Does the WLAN device switch to low-power mode?
      • Disconnected
        • Disconnected from AP using applet and make sure we go back to base consumption (radio being turned off correctly)
        • Use rfkill switch and make sure radio is being turned off correctly
      • Active
        • Measure peak consumption and average consumption with activity over WLAN
    • 3G modem
      • Idle
      • Disconnected
      • Active
        • Similar to WLAN tests
    • Bluetooth
      • Idle
      • Disconnected
      • Active
        • Similar to WLAN tests
    • Additional INFO required: AP model
  • Display tests
    • Test idle with
      • full screen brightness
      • half brightness
  • Backlight measurements
    • Figure out usecase for backlight measurements (Useful for figuring out how agressive ALS should be)
  • Webcam
  • Ethernet

Usecase testing

  • Browsing over WLAN
    • Preconfigured with flash plugin at full brightness
    • 5-10 tabs loading a pre-cached set of webpages
      • Loading live sites changes the numbers (due to flash content and changing network conditions)
    • separate tests for browsing over 3G?
  • Music streaming over WLAN
    • Exercises the audio chip, important for ARM platforms that might use a separate DSP processor
    • Active
      • Peak and average consumption
    • Idle
      • Does the audio chip power off
  • Movie playback (longer test with DPMS disabled)
    • Run from DVD and from disk
    • Disk test should use some standard encoding formats (.mpg, .avi with H.264, ogg-theora, etc.)

[1] Longer tests can be considered if we have certain applications that have much longer cycles of periodic activity


AmitKucheria/UbuntuPowerManagementTesting (last edited 2010-09-24 09:41:47 by a91-154-124-12)