71 dB
Headphoneus Supremus
Dynamically over-compressed brick-walled music uses effectively all the bit depth and that's why lossless endoders struggle to find opportunities of getting rid of redundancy. Sacrificing bit depth can make lossless files smaller as I have demonstrated earlier:With how brick walled, much modern music is when encoded in Flac the average bitrate is 1000 ~ 1415kbps. With TAK codec some albums even reached 1950kbps despite being 16bit at 44.1 kHz!
This is how to do it correctly:
1 - Generate normal 16 bit dither of your choice for the duration of your original track to be (re-)encoded losslessly.
2 - Multiply this dither by 2^n, where n is the amount of bit depth to be sacrificed, for example n = 3 => 2^n = 8.
3 - Add the dither to your track in floating point mode so that signal clipping doesn't happen.
4 - Divide your track+dither by 2^n.
5 - You are good to go. Export as a lossless file.
These steps do not reduce sound quality at all apart from raising the noise floor by 20*log10 (2^n) dB. There is no increased distortion. Only increased noise level, but with "brick-wallet" music that is super loud all the time you don't need much dynamic range, do you? Since the most significant bits aren't used at all, the files are quieter and volume needs to be raised a bit, but it helps they sound typically super-loud to begin with (the motivation for brick-walling).