Hi,
I have included a few samples encoded using the MPEG 4 AAC Gain Control tools.
However, I disabled block switching and also extended gain-control compensation to the lower subband as well to meet ATRAC3 specs. As a result it is not MPEG4 compliant. However, this is just a demonstration of the capability of the Gain Control tool.
The audio quality may face some degradation due to the not well tuned psychoacoustic model. Also, as has been pointed out, there seemed to be some problem with low frequency masking threshold for certain clips such as fatboy and castanet as well which I believed is not due to temporal processing problem..
http://www.kturn.com/store/dl/rushing original.pcm
http://www.kturn.com/store/dl/rushing_l.pcm
http://www.kturn.com/store/dl/hihat original.pcm
http://www.kturn.com/store/dl/hihat_l.pcm
http://www.kturn.com/store/dl/sidestick original.pcm
http://www.kturn.com/store/dl/sidestick_l.pcm
Please note that, they are all encoded in long_block mode only, TNS disabled and the pre-echo spread is still reduced to less that 1 msec.
However, I am not able to compensate clips where the attacks are closely spaced..
eg, one attack each on each half of the mdct analysis window. The Gain Control specs allows up to 7 Gain change points, which I use only 1 gain point.. How to make use of the rest of the gain change points remains a very tricky enginnering tasks. Also, there is suspicion on the accuracy of the mapping of the psychoacoustic masks to the gain-changed mdct level.
Please feel free to try out and comment on the results.
kww