karn@mouton.UUCP (10/18/84)
I have been following the discussion about increasing the CD's dynamic range with some form of "digital expander", and I have a few comments: 1. The dynamic range of the CD is approximately 96 db (16 bits @ 6db/bit). This is already better than many other elements of the audio chain, which now become limiting factors (e.g., microphone preamps, room noise). The noise floor on every one of my CDs seems to be determined by these factors and not by the quantization noise. For the same reason, it appears to be unnecessary to take specific steps to introduce dithering, since the recorded-in noise from the source performs this function. 2. Back when the CD was first introduced, I also wondered why they used a linear code instead of a logrithmically companded code which would give a quantizaton noise proportional to level (and thereby provide a relatively constant S/N ratio over a range of recording levels.) I guess Sony and Phillips opted for a linear code for these reasons: a) Good linear D/As are easier (i.e., cheaper) to build than log D/As b) Digital signal processing is easier, since no internal log-linear conversion would be needed c) 16 bits of linear are good enough anyway (see #1 above) Actually, I think it's about time we stopped worrying about improving the performance of the recording medium, because the introduction of the CD and related digital recording methods has suddenly taken the recording step from being one of the critical links in the chain to being one of the strongest. If you want a fertile area to work on, try the transducers (microphones and loudspeakers.) Phil