AAC Playback quality terrible Sansa Fuze V2

[rusty1404]

Hi

I have many Audiobooks which I have encoded from CD's. 
I used iTunes AAC encoder [64 (kbps), VBR, 22 050 kHz sampling rate, using voice filtering, optimised for MMX/SSE2] to encode my audiobooks.  I chose these settings because I wanted to match audible's enhanced format as closely as I could and I think the choice is the closest.  Perhaps it is not the closest?  The result is high-quality but small m4a (AAC) files.

The files produced play perfectly in iTunes, VLC, SUPER, KM Player, WMP etc.  They sound excellent, I can not discern any difference between the m4a files and the original source material.
The exact same files, when played on my Fuze V2 using RB sound horrible.  They sound like 16 (kbps) rips on RB.  The files sound exceptionally tinny, there is a crackling tinny echo which follows the audio like a bad smell.  It get's worse the higher the volume but is noticable from extremely low volumes.  Another m4a file encoded at 256 (kbps) (44 100 kHz sampling rate) sounds fantastic on my Fuze.

I am new to RB and just installed it on my Sansa Fuze V2.
The sansa fuze was running firmware fuze02.03.33.  I used this firmware to install RB.
RB Version: r27859-100822

MediaInfo says the following about a sample file which sounds horrible:

Audio
ID                               : 1
Format                           : AAC
Format/Info                      : Advanced Audio Codec
Format version                   : Version 4
Format profile                   : LC
Format settings, SBR             : Yes
Format settings, PS              : No
Codec ID                         : 40
Duration                         : 8mn 0s
Bit rate mode                    : Variable
Bit rate                         : 59.7 Kbps
Nominal bit rate                 : 64.0 Kbps
Maximum bit rate                 : 72.1 Kbps
Channel(s)                       : 2 channels
Channel positions                : Front: L R
Sampling rate                    : 44.1 KHz
Stream size                      : 3.42 MiB (99%)
Encoded date                     : UTC 2010-08-22 01:18:05
Tagged date                      : UTC 2010-08-22 21:09:42

Itunes says the sampling rate is 22 050  kHz, MediaInfo says 44 100 kHz.  I'm pretty sure iTunes is correct on this.

I just re-encoded 1 file as a sample using:
Test1.m4a - iTunes AAC encoder [64 (kbps), VBR, 41 000 kHz sampling rate, using voice filtering, optimised for MMX/SSE2]
Test2.m4a - iTunes AAC encoder [64 (kbps), CBR, 22 050 kHz sampling rate, using voice filtering, optimised for MMX/SSE2]

Test1.m4a sounds perfect
Test2.m4a has the same problem as the original files (which are VBR instead of CBR as in Test2.m4a)

So is the problem with the 22 050 kHz sampling rate?

The thought of re-encoding my collection and redoing the tags make me feel ill but the Sansa OF does not do AAC.  Help please.

[funman]

Try an encoding without SBR to see if it's the problem

EDIT: also you forgot to provide us a sample

[rusty1404]

Here is the sample file:
http://rapidshare.com/files/414520422/1-01_The_Kalahari_Typing_School_for.m4a

I just re-encoded 1 file as a sample using:
Test3.m4a - iTunes AAC encoder [64 (kbps), VBR, 22 050 kHz sampling rate, optimised for MMX/SSE2]

Test3.m4a has the same issue as every other file encoded with a 22 050 kHz sampling rate.

[Lear]

On a current simulator, there's periodic noise during playback (likely due to the SBR, as suggested). On a 3.6 simulator, it plays just fine, so I'd guess the fairly recent AAC optimizations are to be blamed.

As for bitrate, both applications are kind of correct. An AAC SBR files contains a standard AAC part, encoded in 22 050 Hz, and an SBR part, which is used to create the 44 100 Hz sound during playback.

[Buschel]

Ok, found the change which brought in the failure (r27225). Will now try to find a fix the bug itself.

Edit1: It is a change within specrec.c

Edit2: Baaad mistake... Fixed with r27868.

Thanks for reporting the failure!

[rusty1404]

Thanks for the response.  It's a pleasure to submit issues when the response is so positive and rapid.  I'm a developer myself, I appreciate your effort.

Just updated rockbox.  Checked version on my Sansa Fuze V2 - r27875 .  The issue is still there.  I cannot detect any difference between now and when I first reported the issue on r27859.

[Buschel]

Hmm, I have used the file you have linked (...Kalahari_Typing_School...) and decoded it on PCSim. There were heavy noise bursts (like Lear described) which do not happen anymore with r27868.
Can you provide a link to another problematic file? And can you test whether there is a difference in the current rockbox version versus rockbox 3.6?

[Lear]

I tested a current build on my Fuze V2. It still sounds bad, but it isn't the complete noise it was in the sim. There's still some "fuzzy" noise added to the sound, most of the time.

Haven't tested on the sim yet, since I pulled in the recent plugin changes, which causes link problems...

[Buschel]

Lear, does it sound fine on your Fuze V2 when using rockbox 3.6 or an older revision (e.g. r27224)?

[Lear]

Haven't tested an older version on target yet (and I can't do it right now), but I do hear the same noise in the simulator (I made it build; I'll commit the fix later today if no-one beats me to it).

[Buschel]

r27898 should finally fix this bug.