Absolute Power


Anatomy of a Power Mystery

When we last left off, I was attempting to do a “product-level” power analysis of my iOS devices, with the goal of identifying the modules that consume the most energy during a viewing of the recent “Star Trek” movie.  To develop a minimum power baseline, I ran the devices (iPad, iPhone 4, iPhone 3GS, iPhone 3G) from full charge to shutdown, with brightness at minimum, sound off, and “airplane mode” on.  The results were impressive: the iPad’s Star Trek score was 7.75 (played the movie nearly eight times in a row), followed by iPhone 4 with 4.75, iPhone 3GS with 4.0, and iPhone 3G at 1.75.  Since the 3G was a couple of years old, I attributed the low Star Trek score to a drop in battery capacity over time, and removed it from further trials.

A few observations about the tests: the minimum brightness setting on the iPhones resulted in a very dim display, only reasonably viewable in a dark room.  The iPad minimum brightness setting was much better – all but the darkest scenes were viewable, but not spectacular.

As far as establishing baseline data, the outcome was very positive.  The iPad and iPhone 4 devices’ power indicators tracked remarkably linearly, just looking at battery power percentage indicator.  Except for some “stickiness” at 100%, probably either due to a slight over-voltage condition on a fully topped-off battery or a deliberate “reserve” behavior to make it look like the battery is “full” for a little longer, the Star Trek runs all came in within 1-2% of relative battery usage, all the way down through the 20% battery indicator warning.  The 3GS was a little less linear, so I opted to proceed with testing just on the two newest devices (with new batteries).  Based on this data, I short-circuited my data gathering methodology to a single run per configuration(repeated to verify) between 99% and 20% battery capacity on the iPad and iPhone 4 devices, rather than running tests from full charge to shut-down.  Thank goodness, because staying up all night watching Star Trek over and over gets old quickly…

Combining the Star Trek results with published battery capacity for each devices yields the “Star Trek Inefficiency” – that’s the amount of energy used to play the Star Trek movie on a device (higher is worse, like golf).  iPad: STI=3.2Wh, iPhone 4: STI=1.1Wh, iPhone 3GS: STI=1.0Wh.  Since the iPad and iPhone 4 have similar hardware configurations, the difference in energy is due mainly to the display – 2.1Wh is a pretty reasonable extra cost for viewing on the larger, brighter screen.  But note the relative cost of the display on the iPad – based on this data, at least 2/3 of the power used by the iPad while watching a movie is just for the display!

Next, the most obvious big-power item – juice up the display.  I ran tests in the same configuration, except with the display brightness at maximum.  The first unavoidable conclusion – the new displays are gorgeous!  The iPad had a big, bright display, compelling from many angles, making it a great platform for sharing movies and videos.  The iPhone 4’s new display was equally impressive – incredible resolution and clarity.  But here’s where size matters – no matter how you slice it, watching movies on the phone-sized form factor is a compromise, and a single-person activity at best.  The iPhone 3GS display was unimpressive – small, and lacking in contrast and clarity.  In one year’s time, this phone has gone from state-of-the-art to also-ran.  Things are tough on the bleeding edge.

Here are the results for the “high-brightness” test.  iPad: STI=6.2Wh, iPhone 4: STI=1.7Wh

Now the costs of the display are getting clearer (pun intended).  On an iPhone 4 with minimum brightness it takes 1.1Wh of energy to view Star Trek.  With similar hardware and software on an iPad with a large display at max brightness, 6.2 Wh of energy are required, suggesting that 85%-90% of the battery power is going directly to providing that big, bright, beautiful picture.

And what about all the other power drains on the device?  I played Star Trek on the iPad in “max movie” mode, with brightness AND sound at maximum.  No change in energy consumption.  Tried it again with my favorite headphones – a pair of Shure E5C’s.  Same result.  Sound is free?  Hmmm…more experiments necessary.  I watched Star Trek using a pair of Altec Lansing BackBeat 903 stereo bluetooth headphones.  Surely that would have some impact.  It did, but only a small impact.  1% more energy used on the iPad battery meter, bumping the STI up to 6.4, which is something like an additional 200 mWh over the whole movie, but the 1% is also at the accuracy of my measurements, so almost inconclusive.  Finally, my big shot – stream the movie rather than playing it locally!  I viewed Star Trek on the iPad streamed over wifi using the latest Netflix app.  Amazingly, the STI was still at 6.4!  What gives?  These scores are still pretty reasonable, suggesting better battery performance than what I generally see in real life (I guess I no longer regard watching Star Trek as real life!).  Where is all that power (energy) going?

Next time, some answers to the mystery, some suggestions on improving the battery performance on your iOS device, and some conclusions to our “product level” power analysis experiment.

  • del.icio.us
  • Digg
  • Facebook
  • Google Bookmarks
  • Print
  • Twitter
  • StumbleUpon