Anelok: saving more power (2/2)

Werner Almesberger werner at almesberger.net
Thu Apr 30 12:23:24 UTC 2015


Felix wrote:
> Wow, 55uA is pretty impressive taking in account that you're using the RC
> oscilator :P

Especially since it has a "typical" current of 52 uA. I sometimes
worry that Anelok might accidently become a power source ;-)

Anyway, I already measured (a while ago) the system's minimum power
consumption to be 5.5 uA with rfkill and 5.8 uA without rfkill. So
that would be the real minimum.

We can't go quite so low in practice because we have to sample the
touch sensor every once in a while, though. That subsystem could
work autonomously without CPU intervention. So there may also be
some saving potential beyond what we have now.

> I was curious about total power usage, so with raw data and a bit of awk:
> 
>   cat session.gp | awk 'T=0; {SUM += $2*($1-T)} END {print SUM}'
>   > 8.72312e+10

Hmm, that would be the average current of that session multiplied
by session_duration / time_since_1970 ? 870 pA ?

For the average current (not very meaningful since the device would
normally spend much more time in standby), I would use something
like

bzcat session.gp.bz2 |
  awk '{if(!T) T = $1;SUM += $2*($1-T);T=$1;} END {print SUM}'

Please note that this includes date from a few aborted sessions
before the one shown, and some 40 seconds at the end after I
cut power - the plot I published only shows the "interesting"
part of the experiment.

So the 0.477 As from above would then become an average 1.9 mA
or 4.32 mW. Still not very meaningful, but more plausible :-)

- Werner



More information about the discussion mailing list


interactive