N900 battery duration

Igor.Stoppa@nokia.com wrote:

> Sorry to again make these comments after you have run the measurement.
> But in your cases you are seeing an offset due to the speaker protection algorythms.
> Running the test with a headset plugged in would get rid of the offset.

Offset? Is that something that affects the MP3 in a different manner
than Ogg?

--
Tuomas

Tuomas Kulve wrote:

> Igor.Stoppa@nokia.com wrote:

>> Running the test with a headset plugged in would get rid of the offset.

> Offset? Is that something that affects the MP3 in a different manner
> than Ogg?

It should be the same, but since you are interested in the power
consumption of the decoding activity - if i understood correctly -
then you get an unknown term, which can also have 2nd order
effects due to the different timing and choices that both cpuidle
and cpufreq could make.

iow, if you have time, it might be interesting to run the test again
with an headset.

igpr

Igor.Stoppa@nokia.com wrote:
> It should be the same, but since you are interested in the power
> consumption of the decoding activity - if i understood correctly -
> then you get an unknown term, which can also have 2nd order
> effects due to the different timing and choices that both cpuidle
> and cpufreq could make.
>
> iow, if you have time, it might be interesting to run the test again
> with an headset.

I was more interested to know the difference between the mp3 and ogg.
But I'm also interested to know the difference between the headphones
connected and not connected.

--
Tuomas

Hello,

Igor.Stoppa@nokia.com a écrit :
> Tuomas Kulve wrote:
>
>> Igor.Stoppa@nokia.com wrote:
>
>>> Running the test with a headset plugged in would get rid of the offset.
>
>> Offset? Is that something that affects the MP3 in a different manner
>> than Ogg?
>
> It should be the same, but since you are interested in the power
> consumption of the decoding activity - if i understood correctly -
> then you get an unknown term, which can also have 2nd order
> effects due to the different timing and choices that both cpuidle
> and cpufreq could make.

But how the speaker protection algorythms can make a difference between
ffvorbis and libvorbis since the output should be exactly the same ?

Jean-Christian

Tuomas Kulve wrote:

> I was more interested to know the difference between the mp3 and ogg.

Then i think it's more appropriate to do it with headset.

Unless yourl typical usecase is really to listen from the speakers - in which
case i'd rather recommend to use the fm transmitter and larger speakers
from a radio.


igor

Jean-Christian de Rivaz wrote:

> But how the speaker protection algorythms can make a difference between
> ffvorbis and libvorbis since the output should be exactly the same ?

both cpuidle and cpufreq target the next state based on activity and
idleness thresholds.

if the algos add just enough to cross any limit that otherwise would not be passed, they can
have quite an influence.

especially if the performance of the 2 decoders is isgnificantly different.

Think about a thrshold-based tax system, if you want a (loose) analogy.

igor

Hi,

ext Tuomas Kulve wrote:
> Tuomas Kulve wrote:
>> Igor.Stoppa@nokia.com wrote:
>>> There is a bug in the cellmo SW related to absence of SIM: the current is higher when the SIM is not in.
>>> So in real life use case (i assume people who will buy it will actually use it as a phone) the use time will be longer.
>> I guess I need to do more testing then :)
>>
>> Although I'm not sure how to accomplish that as I need my SIM for the
>> daily use..
>
> The previous test was without a SIM card. The device was very idle and
> run for 93.75 hours (with a guessed few hour error marginal).
>
> After charging the battery I inserted the SIM card, rebooted and started
> a new test.

Why reboot?

I would expect most people just to charge without shutting down
the device.

(You may remember that e.g. N8x0 use less battery if it's put to
offline-mode instead of being shut down for the night.)


> I received two calls (hung up immediately), I received two
> SMS messages and sent one, and the notifier led was blinking for 45 mins
> at one point. Otherwise the device was idle and run for 93.62 hours. So
> pretty close to the test without a SIM card.
>
> I have a small shell script running there that writes a log every 60
> seconds. After writing the log it runs the "sync" command and starts the
> loop again. I'm assuming it doesn't affect the total run time much in an
> otherwise idle device.


- Eero

On Tue, 10 Nov 2009 15:56:13 +0200
Eero Tamminen <eero.tamminen@nokia.com> wrote:


> I would expect most people just to charge without shutting down
> the device.
>
> (You may remember that e.g. N8x0 use less battery if it's put to
> offline-mode instead of being shut down for the night.)

heh? why is that?

2009/11/10 Dieter Plaetinck <dieter@plaetinck.be>:
> On Tue, 10 Nov 2009 15:56:13 +0200
> Eero Tamminen <eero.tamminen@nokia.com> wrote:
>
>
>> I would expect most people just to charge without shutting down
>> the device.
>>
>> (You may remember that e.g. N8x0 use less battery if it's put to
>> offline-mode instead of being shut down for the night.)
>
> heh? why is that?
>

Well in the case of the N8x0 when booting up the screen brightness was
up to its maximum.
When booting up all the tablets chips and fishes uses maximum energy
until the PM kicks in and
takes control of everything.
This consumes A LOT of energy.
An N8x0 in offline mode (idle) consumes much less energy in a night
than its minutes-long start up.
--
anidel

Eero Tamminen wrote:
>> The previous test was without a SIM card. The device was very idle and
>> run for 93.75 hours (with a guessed few hour error marginal).
>>
>> After charging the battery I inserted the SIM card, rebooted and started
>> a new test.
>
> Why reboot?

So that I know the start scenario is the same. Not rebooting and doing
several tests in a row easily leads to increased memory consumption,
dying processes and weird behaviour.

> I would expect most people just to charge without shutting down
> the device.

Of course. I'm not booting the device in normal usage.

> (You may remember that e.g. N8x0 use less battery if it's put to
> offline-mode instead of being shut down for the night.)

How's that possible? Offline mode shuts down the wireless stuff but
still runs the cpu. And, afaik, shutting down the device shuts down the
cpu etc. as well (not sure what happens if you have alarms there). Do
you happen to have a pointer to some old discussions or wherever this
has been explained?

--
Tuomas