WE-DIG Meeting - Power Management

  • Article

Last night (or two days ago if you are reading this on Friday) the WE-DIG meeting topic was power management on mobile devices, not looking at operating system architecture, or driver development, but taking a slightly different approach, using operating system APIs or a dedicated hardware power monitor to provide details about a running device operating system, and review what happens to power consumption when using radios, running applications, back light etc... - There are two approaches to monitoring power in a system described in the session - the first is to use software APIs (GetSystemPowerStatusEx2), the second is to use a hardware monitor, like the one from Monsoon (~$800)

Using the Monsoon hardware monitor, Paul Yao used measured the power drain on the system by changing the back light settings on an AT&T Tilt

  • lowest level - 167mw [34 hours battery life]
  • highest level - 522mw [11 hours battery life]

Using the Software API (GetSystemPowerStatusEx2) , Paul got the following results.

  • lowest level - 188mw (delta of +21mw from the hardware results)
  • highest level - 522mw

Even though the session talked about power usage on mobile devices there was a discussion about monitoring the power usage on mains powered devices, which is interesting given the focus on power usage/costs within the home.

Intermec ran a power study on one of their rugged devices - They ran a device with the following settings/results (I didn't get the device/model number)

  1. Radios full on: runtime ~ 5 hours
  2. Radios on, 802.11 associated: runtime ~9 hours
  3. Radios off, runtime ~14.5 hours
  4. Display off, no radios, runtime ~20 hours (not very useful though!)

Its interesting to see the difference between WiFi enabled, not associated vs. WiFi enabled and associated. WiFi enabled and associated used considerably less power than WiFi just enabled.

Ron Van Der Putten talked about how badly written .NET Compact Framework applications could reduce battery life on a device - it turns out that it is quite hard to write a badly written .NET Compact Framework application (although I got the impression that Ron did manage to write some badly behaved code) - NETCF applications are event driven, which means you aren't typically sat polling (bad) in the application - Ron stated that writing a file to an SD card rather than the device file system draws 32 times the amount of power.

An interesting question from Ron was whether drawing black pixels to a screen would draw more power than white pixels - turns out that it doesn't make any difference (obvious, right?) - although, I wonder whether black pixels weigh more than white pixels?

Talking of Pixels - there's a video on YouTube from the Mythbusters crew that shows the difference between a CPU and GPU drawing to a surface - here's the link.

- Mike