From: Magnus Holmgren <lear_at_algonet.se>
Date: Sat, 02 Jun 2007 12:24:38 +0200

Andrew Hart wrote:

> It sounds like the playing of aac/m4a in Rockbox isn't fully
> supported if only certain kinds of aac can be played or if they need
> to be optimised for streaming. Is this the fault of the aac codec
> Rockbox uses, a problem in certain aac encoders or is there something
> else going on I'm missing here? I don't use aac myself, but it's a
> popular format for MAC/Ipod/Itunes users. I'd be pleased if someone
> could educate me.

The support in Rockbox is a bit limited, yes (as is the case for Ogg
Vorbis, by the way). The codec as such isn't the problem (though we'd
like to have a smaller and faster one). It's a combination of how
MP4/AAC was designed and limitations in Rockbox.

The problem is that in order to play an MP4 file, certain information
(so called metadata) needs to be read first, before the actual decode
can start. In a file that isn't streamable, that data is located at the
end of the file. When playback of a track starts, that part of the file
might not have been buffered yet, so attempting to read that could force
the disk to spin up.

Once the so called "metadata on buffer" is implemented, it should be
possible to handle this in a good way. When reading other metadata
(such as track length, artist name, etc.), the metadata needed by the
codec could also be read, so that it is readily available when starting
the actual decode.

> I borrowed a couple of aac files from my brother last year when I
> first started using aac. At that time, the aac codec was in its
> infancy. It had no problem reading and playing the file, except that
> it would stop every 2-4 seconds for a moment. Since then, from what
> I've read, the Rockbox crew have optimised the AAC codec and so aac
> files are supposed to play without any stuttering nowadays.

Yes, AAC support has improved. Ordinary AAC is usable, but so called
HE-AAC (sometime called SBR) only plays without stops on Gigabeats.

> I've read material that seems to suggest higher bit rate aac is
> generally comparable to mp3 at a rate approx. 64kbit higher, e.g.,
> 192Kbit aac is roughly equivalent to 256 Kbit mp3 and 256 Kbit aac is
> equivalent to 320 Kbit mp3, etc.

When comparing two good encoders (like iTunes and LAME), the differences
are probably smaller than that, particularly at 128 kbps and above.

> However, if there aren't any decent encoders around (apart from
> Itunes), I won't waste my time looking into it again until a
> worthwhile command-line encoder is available. In the mean time, I'll
> stick with ogg, mp3 and flac.

Nero Digital Audio (the command line version) is considered to be pretty
good.

   Magnus
Received on 2007-06-02