Accelerometer question

Andy Green andy at
Fri Oct 10 13:03:33 CEST 2008

Hash: SHA1

Rui Miguel Silva Seabra wrote:
> On Fri, Oct 10, 2008 at 08:33:19AM +0100, Andy Green wrote:
>> Rui Miguel Silva Seabra wrote:
>>> 	The rate at which they feed me data only starts up when the file
>>> 	handle is opened, and stops when it's closed?
>> This is the case, the interrupts are quenched when no file handles are
>> open on them.
> The interrupts are quenched, but is it still draining battery life?

Not in any significant way.  What does that damage is CPU reacting to
200 interrupts and input subsystem events a second, when there are no
handles open on the accel, it gets put to a static low power mode which
draws almost nothing.  In suspend, we even pull its power rail.

>>> I'm just wondering, because my 1 year old Nokia 2670 just lasted a whole
>>> week in idle on an 700mAh battery, and OpenMoko lasts only about 8/9
>>> hours in idle with a 1200mAh battery. It is nuts, we all know that, but
>>> I'm more interested right now into extending battery life, as it is my
>>> single greatest problem with OpenMoko right now.
>> The issue is "what does idle mean".  Suspend for Freerunner should last
>> a week.  Sitting there with CPU up even with backlight down is 60-90mA
>> on battery, suspend is 5 - 10mA and that's why you get battery life
>> reduced accordingly if you stay out of suspend.  That's a kernelspace
>> issue on the one hand (suspend is not entirely reliable) but also a
>> userspace issue (there have been bugs negotiating entry to suspend
>> instead of sitting there with CPU up at 400MHz and backlight down).
> AFAIK (which isn't that much really) this old Nokia phone does not
> actually suspend (or if it does it's mindbloggingly fast on recovery).
> 5-10 mA would mean at worse case about 5 days. That would be very much
> acceptable in terms of usage, but sill very low... it would last that
> long in suspend because it's a 1200mAh battery! :)

Most of that is going on the GSM side, we eat ~1mA for the board itself.
 That's not great, but to put it in context the DRAMs eat >500uA of it.
 Literally what the calypso is doing with the rest of it is out of our

> Anyway suspend is definitly not reliable :) which I suspect in great
> part has to do with the blasted NDA over the graphics card documentation.
> Remember the White Screen of Death?

Yes, I vaugely recall something of the sort :-O

> Maybe OM inc guys need help with the X driver, but since it's not on
> their priority, suspend will hardly work reliably anytime soon... I fear
> :)

Wah it's actually my priority.  Despite what it seems like there is
ongoing progress with it up on 2.6.27 branches.

- -Andy
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora -


More information about the community mailing list