Hi,
I don't quite understand what is describe in ISO11172-3 in section D.2--Psychoacoustic Model 2.
It is written as below :-
I) Include absolute thresholds, yielding the final energy threshold of audibility, thrw
thrw = max(nbw, absthrw)
The dB values of absthr shown in Tables 3-D.4. "Absolute Threshold Tables" are relative to the level that a sine wave of ± ½ lsb has in the FFT used for threshold calculation. The dB values must be converted into the energy domain after considering the FFT normalization actually used.
Question:-
1) What does it meant by 1/2 lsb has in the FFT and FFT normalization?
2) The absolute threshold table also confused me. Can I know what is lower and higher index(line) ? How should I choose the absthrw for the above function?
Please give me some advice. Thanks!
Regards,
Pat
Someone once told me that the value of the lowest absthr ( around 4 Khz ) should be equal to the level of an input sine wave (4 Khz) of magnitude 0.5 (float) / 1/2 lsb..
I tried this out, and I think this is wrong.. It would only mean that the dynamic range at 4 Khz is 96dB.. Based on experiments, it should be much lower.. Maybe as low as 112 dB..
In Lame the ATH base is at -105dB. (without considering the dynamic adjustments)
Anyway you can not properly set the ATH level with only 1 pass.
-105 dB ? That would mean 18 bits audio ?
Our ATH formula is producing about -5dB at the lowest point, and it is scaled down by -100dB, so the based level is about -105dB.
But the vbr presets are adjusting the base level, and it is further adjusted during encoding by the ath level adjustment function (depending on the local loudness changes in the track)
-105 dB ? That would mean 18 bits audio ?
Considering the dithering effect, we cannot treat 1bit depth = 6dB dynamic range.