Through the miracle of Google, I stumbled unto:
http://forums.afterdawn.com/thread_view.cfm/3724

The guy seems a bit overthetop, and I think somehow that it is in fact NOT transparent at 128kbps (technically impossible, imo)... but its an interesting one. I cant find anything anywhere to do with this.

Can anyone feed my curiosity? (and quite probably a dictionary)


EDIT: I found this http://www.lucent.com/press/0100/000124.cob.html
and this
http://www.sonicspot.com/audioveda/audioveda.html
... I'll have a play...

Ok, little bit of history..

Core of the PAC algorithm was developed basically by AT&T Bell Labs in early 90's (By JJ & friends).

PAC technology was derrived out of ASPEC algorithm, baseline for MP3 - but with pure 1024/128 MDCT (hybrid filterbank and crippled MS stereo were introduced to MP3 because of political reasons) and different scalefactor band mapping.

In 1994 MPEG requested proposals for new coding system (then called NBC, nowdays AAC) and most of the PAC design was used in MPEG-2 AAC codec (filterbank length, scalefactor band mapping, M/S stereo)

After 1994, PAC contined to be developed by Lucent, because that part of Bell Labs became Lucent. AAC also continued to be developed by MPEG (RM models) - with better quantization (from FhG) and coding (AT&T huffman coding proposal) and more improvements (TNS, pulse coding, etc..)

So, PAC and AAC were similar in original design but with different tools developed afterwards.

However, the best ePAC implementation (today's PAC) was worse than reference AAC implementation in JAES tests - Lucent did make a marketing trick with listening test performed with their ePAC and their AAC implementation, but their AAC implementation was inferior to the reference AAC implementation (done by AT&T, FhG and Dolby)

Latest PAC designs include more effective ultra low bit rate coding, but all of these technologies are inferior to the latest MPEG-4 designs (PNS, SBR and scaleable coding0

Thanks for the very quick and *very* comprehensive reply. I liked it.