home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: rec.audio
- Path: sparky!uunet!decwrl!world!DPierce
- From: DPierce@world.std.com (Richard d Pierce)
- Subject: Re: 20 or 24 bit digital recording?
- Message-ID: <BrsHFH.Ar7@world.std.com>
- Organization: The World Public Access UNIX, Brookline, MA
- References: <1992Jul20.155356.12452@phillip.edu.au> <1353@eagle.ukc.ac.uk>
- Date: Wed, 22 Jul 1992 12:02:04 GMT
- Lines: 89
-
- In article <1353@eagle.ukc.ac.uk> wgd@ukc.ac.uk (Warren Day) writes:
- >>Would increasing the resolution of CD increase its transparency?
- >
- >As it is unlikely that complete sonic transparency is just related to
- >resolution, I suspect the answer to your questio is no. Increasing
- >resolution will give more detail though.
-
- As it has never been demonstrated the detail is just related to resolution,
- I strongly suspect that this is not the case either.
-
- >Current vinyl is apparently about twice the resolution of CD (assuming
- >you've got a clean record, but we've talked about that). If you did
- >increase the CD's resolution you need to invent a new completely format
- >basically. Even though you are only changing the 16 bits to 24 bits you
- >need to increase the size of your CD or have only a 55 minute limit. Also
- >you need to completely change your error correction software. Increasing
- >your resolution just gives you more detail. However every so often people
- >come up with new plastics which can form more complex shapes and these can
- >be used to press records without changing your turntable.
-
- The discussion about "LP vs CD" resolution that has gone flying back and forth
- has to be one of the more amusing and ridiculous discussions I have recently
- been saddened to see on this newsgroup (I guess that's saying a lot!). I
- have no idea where the basis for your assertion about comparitive resolution
- comes from, but I would certainly be interested in seeing both some measurments
- and references to the relevent literature. But enough of that for the moment.
-
- >The 1-bit
- >players however use the 1 bit (delta sigma) system to reconstruct
- >the sound without having large errors in the resistors that multi-bits
- >can have.
- >
- >16384 ohms +/- 5% is about 800 ohms error. In a multi-bit player there
- >are also resistors of 1 ohm, 2 ohm, ... , 256 ohm , 512 ohm , 1024 ohm.
- >So the errors really corrupt the sound produced.
-
- Well, you've made an assertion here which is easily and unequivocably testable.
- That is that CD using parallel DACS ought to have whopping transfer errors.
- This should lead to easily measurable converter errors, even non-monotonic
- performance. The sorts of errors you describe here should be obvious and gross
- yet not a single CD, DAT or, for that matter DAC of any kind produced in the
- last 30 years exhibits the kind of errors you contend must exist.
-
- Most D/A converters do not use the binary weighted resistors you discuss
- here, they use what's refered to as a "R/2R" scheme. This prevents just
- exactly the sort of problem you describe here (some do use it, but they
- can be laser-trimmed to an accuracy far better than the 5% you assert).
- So instead of a ladder that looks like:
-
- +V ----+--32768--
- |
- +--16384--
- |
- +--8192---
- . . .
- +---4-----
- |
- +---2-----
- |
- +---1-----
-
- We end up with:
-
- +V----+--20K--
- 10K
- +--20K--
- 10K
- +--20K--
- . . .
- +--20K--
- 10K
- +--20K--
- 10K
- +--20K--
-
- (I may have the pair swapped, If so, fix it yourself, I ain't gonna do it
- in emacs at 2400 baud! :-))
-
- The type of gross non-linearity is a problem that was solved in DACS long
- before the CD player ever came along.
-
- There are other real advantages to single bit DACS (for example, no glitch
- at the zero crossing, etc.), but the issue you metioned is certainly not
- one of them.
- --
- | Dick Pierce |
- | Loudspeaker and Software Consulting |
- | 17 Sartelle Street Pepperell, MA 01463 |
- | (508) 433-9183 (Voice and FAX) |
-