Rockbox.org home
release
dev builds
extras
themes manual
wiki
device status forums
mailing lists
IRC bugs
patches
dev guide



Rockbox mail archive

Subject: RE: Good news and bad news - Q about power saver
From: TP Diffenbach (rockbox_at_diffenbach.org)
Date: 2003-12-18


-----Original Message-----
From: owner-rockbox_at_cool.haxx.se [mailto:owner-rockbox_at_cool.haxx.se]On
Behalf Of c s
Sent: Wednesday, December 17, 2003 11:37 PM
To: rockbox_at_cool.haxx.se
Subject: Re: Good news and bad news - Q about power saver

--- TP Diffenbach <rockbox_at_diffenbach.org> wrote:
> it probably extends battery
> life by about 10%, perhaps as much as 15% if you
> just play long playlists
> without pressing any buttons on the unit. You can
> see some graphs here:
>
http://www.diffenbach.org/rockbox/ebl_test/results/ebl_results.html

> Is the graph an average of many tests?

No. Each graph is one run from full charge to auto-off because of out of
power. Each /point/ on the graph, however, is taken from six samples taken
10 seconds apart, with the largest and smallest valued discarded, and the
remaining four values averaged. This sample was taken once every five
minutes.

> How many different units tested?

Several, but because of an issue with Recorders, graphs were only processed
from the two tests shown. (The Recorder showed a bitrate ever when not
playing, making it harder to determine when the machines were playing and
when they were charging.)

> How many tests total?
again, I only compiled the data for these two. As for the testing
methodology, please see
http://www.diffenbach.org/rockbox/battery-patch-test-plan.html

> Did most or all units tested have individual results
> that resemble the graph and *always* (or at least very
> consistently) have similar percentage increases in
> total run time?

The extended battery build always ran longer than the baseline build.

> Also do you have a theory on why the recorder graph
> seems indicate that the batteries are drained faster
> during the first half of the run despite the fact that
> the power saving mode was always on? That doesn't seem
> right. Why would the batteries be depleted at a faster
> rate at any point on the curve for the "power saving"
> mode?

If you mean the "S" curve, then yes. The battery seems too fall off faster
the greater the charge. On the FMR, for instance, the battery falls from
4.25 to 4.2 volts about as quickly as it falls from 4.2 volts to 4.1 volts.
As the battery charge decreases, the rate of decrease also decreases, and it
plateaus. Then as the voltage level decreases still further, the speed of
decrease again speeds up.

However, when the disk spins up, there's an increased drain on the battery.
The reason I took six samples ten seconds apart, and threw out the highest
and lowest was, in part, to compensate for this temporary voltage drop
caused by disk spin up. At some point, there's not sufficient power for the
unit to remain on, and it auto-offs. Not surprisingly, this will usually be
when it is spinning up the disk. On the FMR the increased voltage drain due
to spin-up is approximately .6 volts. The FMR will auto-off at somewhere
between 2.15 and 2.3 volts, so anytime it's below 2.9 volts without spin-up,
it's vulnerable to auto-offing the next time the disk spins up. This
exaggerates the bottom half of the "S" curve.

> Sorry for all the questions, but that graph did more
> to raised questions than to provide answers for me.

It's worse than you think. As I mentioned, higher voltages fall off faster.
The obvious corollary to this is that, when charging, the more you charge,
the less marginal increase you get: as the charge increases, the rate of
charging decreases. For example, it takes a /long/ time to get the FMR up to
4.25 volts, and a reasonable amount of time just to get for 4.16 volts to
4.2 volts. As a result, "full" charges vary a bit, and this was apparent in
the raw data the graphs are based on. To compensate for this, I mark as the
start of each graph some time after the actual start of that run, matching
up the starting voltages. On The Recorder, graph, for example, the /real/
difference between baseline and test A is +7%, not +13%, but Test A started
with a little less voltage (.18 volts) than Baseline, even though the tester
used (as instructed) the /Archos/ firmware to charge.

The actual raw data for the start of the three Recorder runs was:
Baseline Test A Test B
Hours Mins Volts Hours Mins Volts Hours Mins Volts
0.01 0.60 5.63 0.01 0.52 5.45 0.01 0.60 5.50
0.09 5.27 5.57 0.09 5.27 5.44<-- 0.09 5.52 5.47
0.17 10.35 5.52 0.18 10.60 5.40 0.17 10.27 5.44<--
0.26 15.44 5.50 0.26 15.52 5.39 0.26 15.43 5.42
0.34 20.52 5.48 0.34 20.43 5.37 0.34 20.35 5.40
0.43 25.77 5.45 0.42 25.27 5.35 0.42 25.43 5.39
0.51 30.52 5.44<-- 0.51 30.43 5.34 0.51 30.52 5.36
0.59 35.44 5.41 0.60 35.77 5.32 0.59 35.35 5.35
0.67 40.27 5.39 0.67 40.27 5.31 0.68 40.77 5.34
0.76 45.60 5.38 0.75 45.18 5.29 0.76 45.52 5.33
0.84 50.36 5.36 0.84 50.68 5.28 0.84 50.27 5.31
0.93 55.52 5.35 0.93 55.68 5.27 0.92 55.10 5.30
1.00 60.27 5.33 1.01 60.69 5.25 1.01 60.77 5.29

To massage the data, I used the 30th minute of Baseline, the fifth minute of
Test A, and the tenth minute of Test B as my adjusted start times, because
each of those times the voltage was 5.44v, and was preceded five minutes
before by a voltage near 5.5v, and succeeded five minutes later by a voltage
near 5.41v.

The real times, without massage, were approximately 11:45, 12:20, and 11:55
minutes for Baseline, Test A, and Test B, respectively.

Also, note that you can see -- somewhat -- the decrease in the speed of
decrease in the raw data above. It's more apparent if you see the whole
series, but take a look at these values from hours 6 to 8; in two hours the
voltage drop is .11 volts, the same voltage drop we saw in the twenty
minutes between 15 and 40 minutes in the Baseline raw data, above (we will
see another .11 volt drop in the fifty minutes from 11:09 to 11:59) :

6.00 360.20 5.08
6.09 365.29 5.08
6.17 370.29 5.07
6.26 375.54 5.07
6.34 380.46 5.06
6.43 385.71 5.06
6.51 390.54 5.06
6.59 395.37 5.04
6.67 400.29 5.04
6.76 405.46 5.04
6.84 410.12 5.03
6.92 415.46 5.02
7.01 420.38 5.02
7.09 425.38 5.01
7.18 430.54 5.00
7.26 435.38 5.00
7.34 440.46 5.00
7.42 445.13 4.98
7.51 450.46 4.99
7.59 455.38 4.98
7.67 460.46 4.98
7.76 465.63 4.98
7.84 470.21 4.97
7.93 475.71 4.97
8.01 480.71 4.97

I am sure that by utilizing these graphs, someone will be able to make the
battery percentages displayed by Rockbox even more accurate. (Since my
Archos FMR seems irretrievably broken, it probably won't be me.)



Page was last modified "Jan 10 2012" The Rockbox Crew
aaa