Rockbox mail archiveSubject: Re: alarm
From: Jens Arnold <arnold-j_at_t-online.de>
Date: Sat, 23 Sep 2006 10:03:37 +0200
On 17.09.2006, Paul Louden wrote:
> I think it might be a good idea to abstract the alarm concept
> just a little bit more, so that on non-RTC targets you could
> say "X hours and X minutes from now" so that even without an
> actual clock you could still set a moderately reliable alarm.
> As long as it hits within 10 minutes of the mark, you'd be
> fine (just always set it 10 minutes earlier than necessary,
> assuming that horrible a degree of accuracy, you can live
> without that last 20 minutes of sleep if it happens to slide
> in the wrong direction, right?)
No I can't ;)
In fact I completely agree with you on that generalised alarm
RTC-equipped targets should get the true RTC alarm feature
ported (see my other mail) in order to not waste battery.
Non-RTC targets could get a count-down alarm. The timer tick
is precise enough for this, even on coldfire with constant
boosting/ unboosting the error is just a few seconds per day.
In order to save battery here, I wanted to do some experiments
with reducing power consumption as much as possible (low-power
operation): *switch off* LCD, SDRAM (!) (the alarm code would
run in IRAM only), and clock the coldfire down to either 11MHz
(PLL disabled) or the lowest possible clock with PLL (iirc
6.25MHz), whichever saves more power.
Received on 2006-09-23