Regarding battery usage on hosted platforms (like XDuoo X3ii)

[impig]

I thought I noticed a significant battery performance difference between sessions when I listened to my mp3 song collection vs times when I listened to longer radio shows (2-3h) encoded to AAC-HE. I searched rockbox site and found https://www.rockbox.org/wiki/CodecPerformanceComparison.html and thought I'd test how well would battery life improve if I try those same radio shows in ogg or mp3 instead of the mp4.

To my surprise there is basically no difference in battery life (cca 11h) when using:

MP4 AAC-HE Variable at overall 60 kbits/s
MP3 Variable at overall 65 kbits/s (vbr 6)
OGG Variable at overall  65 kbits/s (vbr q1)

Does codec performance not influence the battery usage?
Is it because the Rockbox on XDuoo X3ii is hosted and not native?
How likely are we getting a native port on XDuoo X3ii, is it possible?
Would the battery last longer on native port -  roughly to the ratio of the codec performance?

I appreciate the Rockbox firmware and your work very much I'm just trying to squeeze some more battery life out of the player and perhaps wonder if I should upgrade to a player that supports a native port like AIGO EROS Q / K.

Thank you!

PS: tested on version: f563fe54c2-250420

[speachy]

Does codec performance not influence the battery usage?
Is it because the Rockbox on XDuoo X3ii is hosted and not native?
How likely are we getting a native port on XDuoo X3ii, is it possible?
Would the battery last longer on native port -  roughly to the ratio of the codec performance?

I don't know if the X3ii's hosted platform implements dynamic CPU reclocking, but even if it does, it's not likely to make much measurable difference as that CPU is pretty modern and is quite power efficient with automatic internal clock-gating of unused components.  Consequently, the majority of the power used when playing back audio with the screen off is stuff that can't be powered down (eg DAC, power supplies, RAM, PLLs and other clocks, DMA engine, audio interface, etc.  That power usage is fixed no matter what playback audio codec is used, though it's _possible_ that lower playback frequencies will use marginally less power than higher frequencies (ie 44/48KHz vs 96/192KHz) because the audio path is being clocked more slowly.

As an extreme example of this phenomenon, on the original (native) X3 port, I implemented dynamic CPU reclocking to try and save power, and it made no appreciable difference in battery life (~6 hours on the old battery I had my device at the time) despite the clock speed being reduced to 1/3rd of what it was before (ie 540->180MHz).  Unfortunately on that device, switching clock speeds produced a faint but audible audio glitch, so in the end it was never enabled.

That said, despite the overall CPU utilization rate/clocking not making an appreciable difference, I'd expect a native port to last a little longer for one reason -- we'll have significantly (over 2x) more memory available for caches and buffers, which means we can leave the SD card powered down for longer stretches of time -- Those little buggers can be quite thirsty!  But I have no idea how much difference that will make in either absolute or proportional terms.

Meanwhile, as to the prospects of a native X3ii port -- While the underlying base port (ie Ingenic X1000) is quite solid, the specifics of how the various components are physically wired up are not known, and gaining that knowledge will have to come through potentially destructive inspection of the device's PCB.  Consequently, the main obstacle for a native X3ii port is that the X3ii only had a small production run, so there haven't been any available for purchase (even used) for several years now.  I think I'm the only [semi-]active developer who even has an X3ii, and I would very much like to keep it in proper working order.

(The ErosQ family is highly unusual in that years later it is _still_ being actively manufactured and even periodically revised)

EDIT:  Formatting and distinguishing between DAC and "playback codec"

[saratoga]

Even on CPUs made on ancient nodes that use way more power than modern stuff, decoding an AAC file takes 10 or 20 mA. From a 2000 mah battery, even an inefficient CPU could run 100 hours, perhaps much more.  You get a tenth that, so probably 90% of the power is going to other things.

Looking at the PCB, I see a relatively high power output amp that uses 8-10 mA just to be on with no output. Probably a lot more things like that with fixed power consumption. Presumably the design was unconcerned with power consumption and compensated with a huge battery. In that case the CPU power won't matter.

Of course it's always worth checking. Lots of high power components may have adjustable voltages or power gating for when not in use.

[impig]

Thank you, both, for your detailed answers! Much appreciated.