home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!pipex!unipalm!uknet!acorn!eoe!ahaley
- From: ahaley@eoe.co.uk (Andrew Haley)
- Newsgroups: rec.audio
- Subject: Re: Sony's SBM CD Mastering
- Message-ID: <1365@eouk9.eoe.co.uk>
- Date: 17 Aug 92 14:48:18 GMT
- References: <1992Aug14.224227.17426@fid.morgan.com>
- Organization: EO Europe Limited, Cambridge, UK
- Lines: 33
- X-Newsreader: Tin 1.1 PL3
-
- sethb@fid.morgan.com (Seth Breidbart) writes:
- : In article <1352@eouk9.eoe.co.uk> ahaley@eoe.co.uk (Andrew Haley) writes:
-
- [... deleted ...]
-
- : >A simple roundoff (with 1/2 bit dither to linearize the quantization)
- : >is the right thing to do in this case.
- :
- : Simple roundoff with dither is a good thing to do; I see no reason to
- : believe it's the best. Why do you think 1/2 bit (uniform?) dither is
- : best? I've heard that 1 bit (triangular) dither is better.
-
- Well, it's not so much 1/2 bit as 1/3 bit RMS (I can't remember what
- the frequency distribution should be). 1/3 bit RMS is near as dammit
- 1 bit pk-pk. I meant +/- 1/2 bit, you're quite right.
-
- : It seems obvious to me that intelligent processing must be better than
- : random dither: Of all the possible 16-bit conversions, pick the one
- : that is closest to the 20-bit (in some appropriate vector space).
- : This is computationally intractable, but there are obviously
- : heuristics that help.
-
- No, I don't think so. The dither is only intended to turn what would
- be distortion into noise. In the extreme case, if you've got a DC
- signal into the ADC (The 20-16 bit resampling is, in effect the same
- as an ADC) you will get the same DC voltage out of the DAC plus noise
- at the noise floor of your converter. You can't _possibly_ do any
- better than that. In theory, every ADC should add this dither before
- sampling, but in practice there's plenty of noise there anyway.
-
- : Seth sethb@fid.morgan.com
-
- Andrew.
-