Which audio format do you use with Rockbox?

[prankstare]

I own a Sandisk Sansa Clip Zip and, since Rockbox supports a wide range of audio formats, I decided to tackle onto Musepack again after a few tests among some lossy audio formats. So I am currently transcoding all my .flac collection to .mpc -q10 (~ 350kbps) and sounds great. I know some of you might say this is crazy and there's not much of an audible (perceptive) difference between MP3 -V 2 and whatever other lossy format at such "transparent" specification but, if this is the case, then I'd rather ask ourselves why we prefer original CD's or any other lossless format rather than CBR320kbps or -V 2 LAME MP3 for instance. Well, personally, Musepack DOES have a different feeling to the sound, especially at very high bitrates (maybe it's just quality headroom, or maybe it has something to do with decoding, I have no idea really).

How about you? Which audio format are you using with your portable player?

[saratoga]

Well, personally, Musepack DOES have a different feeling to the sound, especially at very high bitrates (maybe it's just quality headroom, or maybe it has something to do with decoding, I have no idea really).

If a format sounds different at high bitrates then something has gone quite wrong.  The idea is that at high bitrates they should sound identical to the source.  In the case of MPC, I would expect such high bitrates to sound identical to FLAC for virtually all files.

[prankstare]

If a format sounds different at high bitrates then something has gone quite wrong.  The idea is that at high bitrates they should sound identical to the source.  In the case of MPC, I would expect such high bitrates to sound identical to FLAC for virtually all files.

I said it gives a better 'feeling', which would have to imply that the sound itself has to be different somehow, yes and no. We're talking very subjective things here, but I still believe there must be subtle changes in sound (could be the decoder or the EQ or something, I don't know since we have to take in consideration many variables here - in this case in special because I do have to use EQ and other effects on). With EQ and some other tweaks on, I could blind-test MPC (q10) Vs. MP3 (-V 0) with some difficulty yes, but could do it. Yet, I am more than aware that my methods for blind-testing isn't really ideal, but it's just how I listen to music with my Clip Zip everyday so I am happy with it.


PS: If we let us take this matter onto a more philosophical perspective, I'd take the liberty to ask myself what is reality anyways... is it only the things we can see, touch or hear through our senses? There's a disparity on this cheesy conclusion IMHO. Have you guys heard about binaural sounds? Given a difference in pitch for left and right channels separately, it's something our brain can't quite decode and turns out that it creates its own beating sound inside our heads... it's just one of those funny things you know... hehe...

[saratoga]

I said it gives a better 'feeling', which would have to imply that the sound itself has to be different somehow, yes and no. We're talking very subjective things here, but I still believe there must be subtle changes in sound

Well theres two possibilities:  something wrong with MPC, or that you're wrong.  I suggest using an ABX test to determine which:

http://wiki.hydrogenaudio.org/index.php?title=ABX

If you can pass the test, then you know its a real problem.  If you can't, its probably imaginary.

Have you guys heard about binaural sounds? Given a difference in pitch for left and right channels separately, it's something our brain can't quite decode and turns out that it creates its own beating sound inside our heads... it's just one of those funny things you know... hehe...

I think you misunderstand what a binaural beat is.  Your brain can certain decode the difference signal between the two frequencies.  Check out the wikipedia article:

http://en.wikipedia.org/wiki/Binaural_beats#Acoustical_background

[wilsonsamm]

What I do is I copy my files to my rockbox player, and then run a bash script that transcodes all the files it finds. Flac and WAV files become Ogg Vorbis files, but MP3 files are just downsampled using LAME, because I have read advice against transcoding from one lossy format to another.

This transcoding makes a big difference to the filesize. Even with reasonable quality on the resulting files, after running this program I often end up going from, say, 1.5 gigabytes in use to less than 700 megabytes. That's important when your player has only a very small capacity of 1.9 gigabytes.

If you want to see my script, just let me know and I'll post it here.

[gevaerts]

MP3 files are just downsampled using LAME, because I have read advice against transcoding from one lossy format to another.

What's a bad idea is transcoding from lossy to lossy. It's not better if both ends use the same codec.

[wilsonsamm]

MP3 files are just downsampled using LAME, because I have read advice against transcoding from one lossy format to another.

What's a bad idea is transcoding from lossy to lossy. It's not better if both ends use the same codec.

That advice I read (I can't find it again now), but it said that transcoding between lossy formats is worse because (for example) Ogg and MP3 will each be leaving out different data. In that way, an MP3 made from an Ogg file will have more loss than an MP3 that's made from an MP3. Is this wrong?

[saratoga]

That advice I read (I can't find it again now), but it said that transcoding between lossy formats is worse because (for example) Ogg and MP3 will each be leaving out different data. In that way, an MP3 made from an Ogg file will have more loss than an MP3 that's made from an MP3. Is this wrong?

Yes that is wrong.  Its equally bad to do MP3 > Ogg as MP3 > MP3. 

FWIW I used to transcode high bitrate lossy to low bitrate for portable use to fit more files, but now theres no point given how cheap flash memory is.

[prankstare]

I think you misunderstand what a binaural beat is.  Your brain can certain decode the difference signal between the two frequencies.  Check out the wikipedia article:

http://en.wikipedia.org/wiki/Binaural_beats#Acoustical_background

http://www.youtube.com/watch?v=IrdVEBbskZQ

I am no audio engineer but if you take a little bit of your time to watch the video (jump to @3:45 to cut the blablabla thing), it shows exactly what I mean. I may not fully understand, theoretically what binaural beats are, but we can have a glimpse through actual experience. Just put your headphones, go to 3:45 in that video and listen carefully. You'll notice that when he feeds some Hz difference in each channel, the resulting sound is different, but only inside our minds (if you remove one side, you'll notice it's still pure signal sound and no beats). I know this isn't the best place to discuss such a subjective/philosophical thing but, all I am here to say is that we can look at graphs, numbers, sound waves or even listen to actual audio, but what reality is we can never fully grasp or understand.

FWIW I used to transcode high bitrate lossy to low bitrate for portable use to fit more files, but now theres no point given how cheap flash memory is.

Agreed. I just don't use lossless on my Clip Zip cause for a 32GB flash memory card it's still a bit spacey, but lossy Musepack @350kbps gives enough 'quality headroom' to make it up (on an sub-conscious level, perhaps). Haha!

[Nick H.]

FLAC.

[Mach-X]

Vorbis q5. Maintained by the same folks as flac. And smaller filesize at recommended transparent settings than lame ie v2.

[Julian67]

I use Ogg Vorbis for my music collection, both on PC and on Rockboxed players.  If I download podcasts or similar and they are mp3 or m4a or any lossy format Rockbox can play I don't transcode as it seems (at best) pointless; anything that plays is OK with me.  I notice that in recent Rockbox versions decoding the popular formats is so efficient that there's little or no difference in battery life between several different codecs.

Using Ogg Vorbis -q 7 I've not, in several years, encountered an artefact I can identify even with notorious killer samples such as trumpet or eig.  I've not tried musepack with killer samples but I have compared LAME, Vorbis, Nero AAC and WavPack Hybrid and remain content that Ogg Vorbis is as good as or better than other lossy codecs in terms of audio quality (i.e is, as far as I can tell, transparent at same or lower bitrate) and is also for me the most convenient.

Good links for killer samples: http://www.hydrogenaudio.org/forums/index.php?showtopic=95552&mode=linear

btw killer samples are sections of audio that present real difficulty for lossy encoders, resulting in obvious artefacts, easily apparent even to the untrained ear, sometimes even at very high bitrates.  If you want to know how awful a lossy encoder can be then do a series of tests with trumpet and eig....it's horrible enough that you might join the Neil Young fan club.  On the other hand you'll also find that even the most challenging samples can be encoded to sound identical to the originals but it may require higher bitrates than commonly used.

mpc -q 10 (~ 350kbps) looks like overkill to me.  I think if you do some real abx tests you'll find Ogg Vorbis or LAME are transparent at substantially lower bitrates than 350, even with the most difficult samples.  Once you are aware of a variety of lossy codecs which meet your audio needs then you can make an informed choice based on all the other stuff that matters: gapless support, metadata handling, device/software compatibility, encode/decode speed, CPU useage etc.  Experience suggests that any modern lossy codec which uses a psychoacoustic model is going to achieve transparency well below 350 kbps.  This excludes WavPack Hybrid because it doesn't use a psychoacoustic model and can have audible artefacts at high bitrates, and it excludes faac because it's a long time since it was modern....LAME and Ogg Vorbis really make a lot of sense and there are good reasons why they well liked by interested people as well as by casual listeners/consumers.

then I'd rather ask ourselves why we prefer original CD's or any other lossless format rather than CBR320kbps or -V 2 LAME MP3 for instance.

Your "why we prefer.." is not the universal fact that it purports to be: it's an unfounded assumption, a personal preference, a mere assertion.  As a source for transcoding then I prefer losslesss and don't care if it's on CD or flac or some other lossless file.  As far as listening goes then if I can't distinguish one source from another then, in terms of sound quality, the format or medium isn't important.