Quantizer levels need not be powers of two when entropy coding is used
Evidence
Although quantizers very often have a number of levels, which are powers of 2, such as 16, 256 or 65,536, because this makes best use of a binary word, there is no reason that other numbers of levels cannot be used, especially if entropy coding is going to be employed. — David M. Howard, Jamie Angus, Acoustics and Psychoacoustics, 2656
Categorization
- QUESTION: Worth chasing how non-power-of-two quantizer levels interact with entropy coding in perceptual codecs like MP3 and AAC.
- corroborated: 2 independent passages support this claim — David M. Howard, Jamie Angus, Acoustics and Psychoacoustics, 2656
- XREF: connects to: Entropy coding trade-offs limit how close code rates approach information-theoretic limits; Unequal symbol probabilities always reduce a source’s entropy below its bit capacity; Arithmetic coding approaches source entropy better than table-based Huffman methods — David M. Howard, Jamie Angus, Acoustics and Psychoacoustics, 2644 (+1 more)
Broader Connections
- Quantization always injects noise that lowers audio signal-to-noise ratio
#qualifies(0.72) — David Howard, Jamie Angus, coustics and Psychoacoustics, 2651