Hmm, if that is so, maybe I can help just a bit. Here's Fade to Black Mettalica sample from Lame test sample page (http://lame.sourceforge.net/download/samples/ftb_samp.wav)
encoded with FhG AACDemo 2.2:
http://sivut.koti.soon.fi/julaak/ftb_samp-FGh_ref2.2.aac

Ok, here are more AACDemo 2.2 test samples:
Applaud
Blackbird
Castanets
Fatboy
Ftb_samp
hihat
mstest
http://sivut.koti.soon.fi/julaak/AACDemo2.2_test_samples.zip

Originals are at Lame test sample page:
http://lame.sourceforge.net/gpsycho/quality.html

*Cough* The samples were found on the net, from a reliable source..
(FhG doesn't allow posting of results of demo codec encodings.)
But these indeed are AACDemo 2.2 encoded.

Thanks, JohnV.



That's why I thought only people involved with AAC developing had had access to these samples.

Regards;

Roberto.

Encoder has a serious bug. The right channel contains from time to time strange low frequency
artefacts which makes it very easy to distinguish
between original and encoded file. For decoding
I used faad2 + AKG-401.

Can someone for example encode castagnets again
with switched channels (right <-> left)?

Unfortunately I don't think this is possible. I don't know any person who has access to AACDemo 2.2 at the moment.

Actually, AACDemo uses rather conservative cut-off filter that starts at 15 kHz and ends at 17 kHz.

Encoder is more or less state-of-the-art at 128 kbps. Fatboy is coded with small "bug" (this bug doesn't exist in Dolby AAC, but Dolby AAC encodes castanets with more distortion in low-freq domain)

It would be interesting to see how other AAC/MP3/... encoders encode same samples at 128 kbps CBR.

Ok, same set of samples encoded with PsyTEL AACEnc 2.15 - at 128 kbps.

I think it is a nice showcase showing how close the "independent" implementatioins have approached the state-of-the-art

http://www.psytel-research.co.yu/samples/psytel_test.zip

Note that those AACDemo samples were encoded in MAIN profile, and PsyTEL samples are encoded in LC (Low Complexity) profile!