Rockbox mail archiveSubject: Re: Testers wanted: up to 50% greater battery life
Re: Testers wanted: up to 50% greater battery life
From: Michael O'Quinn <michael_at_oquinn.info>
Date: Tue, 29 Jul 2003 09:49:27 -0700 (PDT)
On Tue, 29 Jul 2003, John Stevenson wrote:
> 4) Your testing would not be 'real life' based,
> people do not use their player in that way. Might be
> best to time over normal usage - more people could
> take part then. The normal stuff such as switching
> on, off, (which uses more battery power) creating
> queues and just playing with it, that way you will get
> a more realistic set of metrics.
That is not true. The purpose of his test is to compare battery usage
under a controlled set of conditions, and playing the way he described it
works. He will not be comparing this usage pattern against a more
"normal" usage pattern. He will be comparing it against other runs of the
same type, to determine which software has the longest run time compared
to the other versions USING THE EXACT SAME CONDITIONS FOR EACH TEST.
Doing it your way would be much more complicated to accomplish. THe
problem is that you would need to duplicate the exact same keypresses at
the exact same time to really get an accurate comparison.
Tom: You ARE planing to distribute all three versions to each tester,
right? I mean, given differences in battery life among seemingly
identical batteries (You use a 1 hour charger, I use an overnight charger,
so my batteries probably have a different run time from yours, that sort
of thing) you have to compare different runs on the exact same JB and
battery set to really get valid data.
To get decently accurate results, you should instruct the testers to run
one test first to fully discharge the batts, then fully charge them, THEN
run the first valid test to actually collect data.
I don't REALLY think this needs to be a blind test. It might not hurt to
have the user run the first test over when the other ones are done, just
to make sure the batteries are still behaving the same. This IS a
different than normal usage pattern, and the battery characteristics may
change over the life of the test.
That would be five runs minimum: One dummy to prep the batteries, three
actual tests, and one final test (re-run of the first actual test) to
confirm the batts haven't changed radically.
Could you make a counter that logs each time a song is played? This would
make collecting the data a lot easier, and should suffice for your
purposes. I don't look forward to sitting there with a stop watch.
If this really works, and doesn't break things, it'll be a very welcome
Received on 2003-07-29