aoTuV beta5.5 technical information The differences from the aoTuV beta 5... 1. The frequency domain width of M6 was revised. 2. For q-1/-2, a pre-echo reduction code was applied (M3). In addition, the M3 code was improved. 3. The floor setup parameters in the low bit-rate was changed. 4. The part including bug of noise normalization was rewritten. 5. libvorbis 1.2.0 were merged. Furthermore, a revision of Bug #300 and #1229 was applied. 6. The ATH curve of the high frequency area (more than 32kHz) was revised. ...and I tune up many parameters. 2008/03/30 Aoyumi ---------------------------------------------------------------------------- aoTuV beta5 technical information The differences from the aoTuV beta 4.51... 1. In order to work noise normalization correctly, the value of tone masking is compressed under certain conditions. A floor curve approaches the original audio spectrum line as a result. [New M4] 2. In order to cut down the bitrate, a part of M2 code of beta4 was deleted. @3. The ATH curve was changed slightly. 4. Pre-echo decreased slightly. This is based on change of M3 code. 5. In order to decrease artifact by collapse of audio energy, noise normalization processing of point stereo was extended. [32/44.1/48kHz only] 6. The threshold of channel coupling (lossless <=> point) is changed dynamically. This is a method based on a certain idea. [M6] ...and I tune up many parameters. 2006/10/24 Aoyumi ---------------------------------------------------------------------------- aoTuV beta4.51 technical information The differences from the aoTuV beta 4.5... 1. The changed part of "scale.h" of Xiph.Org SVN was reflected in aoTuV. This will solve the problem of the compile by gcc4. 2. A part of portion added by beta4.5 had caused memory access violation. Depending on environment, a problem arises in qualiy-1/-2. This problem was corrected. 2005/11/17 Aoyumi ---------------------------------------------------------------------------- aoTuV beta4.5 technical information The differences from the aoTuV beta 4... 1. M4 code of beta3 was deleted. M5 is added as what is replaced with it. The advantage of this method is that the result stabilized more is obtained. Instead, the change width of the bit rate becomes large rather than before. [32/44.1/48kHz only] 2. New M4 was added. There is this for the same purpose as M1. However, it is tuned up only in the low bit rate region according to work and individual parameters. [32/44.1/48kHz only] 3. M2 code was extended. This decreases a specific noise problem. [32/44.1/48kHz only] ...and I mainly tune up tone/noise masking and noise normalization parameters. The above change is applied below quality3. 2005/11/05 Aoyumi ---------------------------------------------------------------------------- aoTuV beta4 technical information The differences from the aoTuV beta 3... 1. Adoption of new hypot (point stereo). This works in general better than dipole/round hypot. And when residue is likely to decrease greatly at the time of a reverse phase, added code is made to increase the ratio of lossless. 2. Change of ATH curve. This is not a big change. 3. Residue used for threshold calculation of channel coupling is changed. This one seems to be good selection. 4. HF reduction code of beta3 was deleted. Now, this is not required. 5. The energy deficit of M4 code of beta3 -- correction of a bug Now. 6. The code for pulling downward floor of the frequency range which can be easy to be heard was added (New M2). This works by the low bit rate and has a comparatively good trade-off. 7. Change of threshold calculation of noise normalization. Moreover, it takes phase information into consideration. ...and I mainly tune up tone/noise masking and noise compander parameters. 2005/06/18 Aoyumi ---------------------------------------------------------------------------- aoTuV beta3 technical information The differences from the aoTuV beta 2 and Vorbis 1.1... 1. Dynamic impulse block noise control(q0-10). [32/44.1/48kHz only] Conspicuous pre-echo is reduced. 2. New ATH curve. 3. In order to cancel that a small sound becomes unstable, the noise level of low frequency is lowered under specific conditions. (Only below q3) There is a greatest effect on a piano etc. [32/44.1/48kHz only] 4. The rate of lossless stereo is increased in specific conditions. This improves many of conspicuous problems which a point stereo causes. (e.g drumus's hiss/voice, guitar and brass's distortion) [32/44.1/48kHz only] 5. The action of the noise normalization of point stereo is corrected. It was set to one of the causes of distortion in a long tone (e.g strings/oboe/high tone voice). 6. The HF reduction code was not working normally. This is corrected and it retuned up according to the present condition. [32/44.1/48kHz only] 7. The q-2 mode was added in all sampling frequencies. And nominal bitrate of q-1 are changed into a more nearly average value. Although nominal bitrate is more large at a low sampling frequency, average bitrate hardly differs from the former version. ...and I mainly tune up tone/noise masking parameters. 2004/11/21 Aoyumi