MP4 AAC Encoding at 128kb/s

 

I tested both the PsyTEL and MBSoft AAC encoders as there is no clear cut winner in the field yet. Both codecs produce very similar results. I used the MBSoft plugin for WinAmp to do the decoding in both cases. The Homeboy plugin gives essentially the same results.

MBSoft128kAAC.gif (4377 bytes)
MBSoft 128kb/s AAC Frequency Response

MBSoft128kS.gif (39674 bytes)
MBSoft 128kb/s AAC Spectral View

PsyTEL128kAAC.gif (4359 bytes)
PsyTEL 128kb/s AAC Frequency Response

PsyTEL128kS.gif (38779 bytes)
PsyTEL 128kb/s AAC Spectral View

Both encoders produced virtually identical graphs for the 128kb/s data rate. The frequency response is identical to the source up to 10kHz but above this is suppressed when compared to the source and rolls of gently to about 19kHz.

The spectral view shows the presence of considerable amounts of high intensity material right up 19kHz but heavy pruning of lower intensity high frequency components is obvious right down to 10kHz. This gives the spectral view a chunky appearance as compared to MP3 encoded material.

In listening tests I felt that the MBSoft encoder produced a warmer sound than the PsyTEL encoder but the fine detail was slightly fuzzy. Overall I think the PsyTEL encoder produces a slightly better result. I did have a problem with the PsyTEL encoder in that it would not function correctly under NT4 and had to run it under Win95 to encode the file.

The 128kb/s AAC file definitely sounds better than a 128kb/s MP3 file. There were no obvious artifacts and the sound has good presence. On the whole though I feel that 160kb/s MP3 files sound better. As there is still a lot of refinement to be done on the AAC codecs I may have to change this opinion in future.

The shocker comes when you see the resources needed to play back the file. On my machine it consumed a whopping 55% of CPU time to decode the output on playback. This reminds me of the early days of MP3 players and we will most probably see a big drop in CPU utilization when the codecs are refined.

At this stage I do not feel that AAC encoding is ready to replace MP3 for general purpose use. The much longer encoding times and higher CPU utilization on playback overshadow the benefit of better sound quality than MP3 at 128kb/s.

An alternative encoding scheme to AAC in the MP4 stable is NTT and Yamaha's TwinVQ encoding technique. We examine this in the section below.

[Back] [Next]

1999 Peter Miller for MP3'Tech - www.mp3tech.org