Hi everyone,

I was wondering about the energy consumption of my mobile phone while using its mp3 player function. At the same time I wanted to know whether the bitrate and the "vbr or cbr" fact were parameters that modify energy consumption. The first point may not interest anybody but I guess the second one yes, that's why I post this in mp3 and not in audio hardware

Therefore, I used a power meter to measure the current needed to play mp3's with my mp3 portable player (which is a mobile phone btw).

test conditions :
• mobile network conditions were the same during the test campaign (a "bad" network may increase energy consumption)
• the mobile phone is a Sagem myX-8.
• the same original sample (the 33rd first seconds of the song "master of puppets" by Metallica) has been used and converted to various mp3 files. 12 samples were tested :


• each time, one (and only one) file was added to the playlist and played in repeat mode during 3 minutes, while measuring the current. The first seconds of measure were not kept as we have to wait for the screen to be automaticaly turned off (a LCD On represents a huge consumption)
• for information, in idle mode the mobile phone consumpts 5.62mA with a standard deviation (s.d.) of 0,29x10^-3
• only one Lame version was used : 3.97a10.
• of course, the volume level remained constant.

I think that's all for the presentation, here are the results

Results :

____________________bitrate__________consumption (mA)

preset cbr 128__________128__________155.34
preset cbr 192__________192__________154.84
preset cbr 256__________256__________155.09
preset cbr 320__________320__________156.17

-V 6 vbr new__________134__________155.23
-V 2 vbr new__________214__________157.54

preset medium old vbr__________182__________156.23
preset standard old vbr__________220__________155.26
preset extreme old vbr__________263__________155.17

preset medium vbr new__________182__________154.21
preset standard vbr new__________217__________155.60
preset extreme vbr new__________250__________154.19

(note : all standard deviations are around 2*10^-3 / 2,5*10^-3)

mean and standard deviation
µ (consumption) = 155.4 mA
s.d. = 0.91

Conclusion :
Here I'm not able to see a correlation between bitrate and consumption, nor a difference between cbr and vbr. We can conclude the energy consumption is constant whatever the bitrate.
Since every softwares and hardwares are differents, I have no idea if this conclusion can be extended to any other portable mp3 players.


Other test that may interest you :

While respecting the same conditions given for the first tests, I thought it would be interesting to see whether there is a huge difference in energy consumption between the lowest and the highest volume level amplification (with headphones).
Of course, the same sample was used (preset cbr 128), and the test conditions were exactly the same, except for the sound volume.
(note : the software simply reads the mp3 file, there is no equalizer option or anything like this.)
Here are the results :
- lowest volume level, with headphones : 154.14 mA
- highest volume level, with headphones : 154.77 mA
=> relative difference : 0.41%.

Conclusion : there's no need to limit the sound volume for any other reason than ear protection, that will not affect (or so smally) energy consumption. Again, I have no idea if the other portable players behave the same way.


PS : the device can also play aac files, I plan to make some tests with this format, please tell me if anyone is interested in seeing the results
PS2 : I hope my english is understandable, please feel free to tell me the mistakes I've done.

Chapeau!

I dont think the test would hold true for a harddrive based player as much of the energy consumption comes from spinning the drive, which occurs more frequently w/ larger/higher bit rate files.

Great test.

It may be interesting to make an approximation of the measurement uncertainty you're dealing with, for instance by measuring the output for one specific condition a number of times (10 for example) and calculating the standard deviation.

Let's say the measurement uncertainty is 1 mA (seems reasonable to me), than all you're values are not statistically different.

You're right. Even in my case I don't even know which % of the energy is really used for decompression and not for anything else. Like if it wasn't enough complicated, my device has to deal with the mobile network and moreover I really don't know what the software is really doing while the mp3s are played.


• the uncertainty (i don't know it right know) from the measurement device is surely ridiculously low compared to what I'm measuring. It's a professional power meter which costs à lot (ok it's a bit childish to say that) !!
• The power meter measures the current every second, and my test lasts 3 min. I could prolonge the test, but that doesn't seem usefull when i see how stable is the consumption. So, standard deviations are quite low.

EDIT : english corrections

Ok, great.

Very interesting! I've always wondered what the impact of bitrate is on battery life.

By the way, I found this in the FAQ for flash players from the official iRiver website:



So, I guess this is probably not true, or at leat not significant.

edit: Off topic, but I find the "higher bitrate > higher quality" statement from iRiver also very interesting.

Mmmm I don't think we are able to say that. My poor knowledge in hardware (and software also) doesn't allow me to make such conclusion.

Which phone is it?
It is quite possible that your phone is using a decoder chip (as in NGage, as an example) and not doing the decoding in software. In this case, the power consumption should be quite constant, as the chip is optimized for this task.
However, for players using a software decoder ( many ARM based devides) consumption will change a lot more.

I have a small question that has bugged me for a few days. I always thought two parameters "-V 2 --vbr-new" and "-preset standard --vbr-new" are the same. But from the bitrate table in the first post, it does not look like so. It was also the case from my own tests a few days ago. Two parameters gave different files. So actually, what is -preset standard mapped to ? BTW I used 3.97 a11 in my tests.

As already said ( ) it's a Sagem myX8. Indeed it's possible there's a dedicated chip for this job, I can't tell. I guess I'll have to make tests with my old Creative MuVo 128Mb.