> 16-bit audio can perfectly reproduce audio as far as we can physically tell.
Imagine encoding a sort of real world dynamic range across 16-bits. This would go from 0db to 100db in volume. This would need more than 16-bits which yields an SNR of about 96db. The dB values are different and not comparable but you can see we don’t capture the full dynamic range of human hearing very well.
Humans don't have 100dB dynamic range across their full hearing spectrum. We're less sensitive to high frequencies, which means you can apply high-frequency dithering to improve the dynamic range without adding audible noise.
Imagine encoding a sort of real world dynamic range across 16-bits. This would go from 0db to 100db in volume. This would need more than 16-bits which yields an SNR of about 96db. The dB values are different and not comparable but you can see we don’t capture the full dynamic range of human hearing very well.