Re: [iPAQ] H38xx power drain problems

Am Mi, den 21.01.2004 schrieb Phil Blundell um 14:30:
> On Tue, 2004-01-20 at 12:52, Nils Faerber wrote:
> > Strange.
> > Sure you have not had a user RTC alarm set?
> That was my first thought too, though I'm fairly sure I didn't.
> Certainly there wasn't any evidence of gpe-announce or anything like
> that being run.

Last night I had a drain and prepared suspend/resume scripts for logging
beforehand. Result is that a full resume that would have been logged has
not happened :(
So there has to be more to it.

> > I also think that the charging threshold, this 10mA thing, is utterly
> > wrong. First this does no good to LiIon batteries. They do not like it
> > very much to be fed up to the limit all the time. The last 5-10% (i.e.
> > 90-95%) are very critical. Most charging circuits reduce the charging
> > current to very low when entering the last 10% (to be charged). Likewise
> > it is also recommended that LiIon batteries must not be fully charger
> > again unless they have lost a certain amount of capacity; mostly the 10%
> > threshold is used again. This means that charging should only be done if
> > the capacity drops below 90%.
> The iPAQ charger certainly supplies only a very low current for the last
> phase of charging. This is where the 10mA trigger point for the LED
> comes from: if the battery is 95% charged by that time and you're in a
> hurry to use the ipaq, you don't necessarily want to wait around for an
> hour to get the final 5% charge.

Ah, I see.
Who controls the charging, i.e. is there an extra charger circuit? AFAIK
the Dallas chip is only a monitoring chip, right?

> > The effect you get when you always charge the battery is the "lazy
> > battery" effect, i.e. you get a reading != empty (even more than 50%)
> > but any higher than usual battery load, e.g. WLAN, would cause a battery
> > fault - this what happened to an older H36xx iPAQ now :(
> Ah. I wasn't aware that li-polymer batteries suffered from that
> effect. They certainly do wear out, and it doesn't surprise me too much
> to hear that the battery on your h3600 is shot, but I hadn't heard that
> top-up charging would accelerate the process.

At least that's my knowledge right now.
And telling from my oldest Mitac notebook which did exactly this and
which has only lost ~10% of its original capacity within five years I
would estimate that iPAQs could also gain from that technique ;)

> > Oh well... I see what I can do ;)
> Okay, cool. Let me know what you come up with.

Oh well...
last night I hacked on h3600_asic_battery heavily to avoid accidentially
enabling the charger. I moved the whole set_charger() calls to
h3600_xxx_battery_update() (where the 10mA thing is) and modified the
logic to *not* start charging if the battery is still more than 90%
full. I had the hope that the set_charger() calls scattered all around
the code could by accident under strange circumstances activate the
charger without AC - I was wrong :(
I had the drain again last night. But anyway I have this new charging
routine in place now. Let's see what happens after a full charge today.

Next I will try to see what happens at
        rmmod h3600_asic
and see if I can add parts of this to all the xxx_suspend() routines
because inside the asic code. I know that if I remove the asic module
the power drain (during normal operation) drops dramatically although
nothing important changes at removal (at least in my testcase I already
had frontlight off, LED off, etc...). Maybe this can also help.
If this does not help too I admit to have no further idea :(

> p.
CU
  nils faerber

-- 
kernel concepts          Tel: +49-271-771091-12
Dreisbachstr. 24         Fax: +49-271-771091-19
D-57250 Netphen          D1 : +49-170-2729106
--