In Reply to: RE: Not Really........... posted by J. Phelan on July 14, 2009 at 17:47:09:
I didn't know anything about this until your original post, but let me see if I understand what the 20-bit thing from the 1990s was all about.
If you do a sequence of arithmetic calculations and round to the nearest, say, 0.001 after each calculation, you'll eventually end up with accumulated error of 0.003, 0.004, etc. If you chain enough of these rounded calculations, you'll lose an arbitrary amount of precision in the result. But if you round to the nearest 0.000001 after every step and save the rounding to 0.001 for the final result, you'll almost certainly get the one and only right answer. This is why calculators have internal precisions that are several digits longer than the number of digits they can display. In your words, it's like "headroom".
I'm guessing this is what 20-bit processing is -- adding 4 extra bits for the number-crunching parts so that you can chain a bunch of processing steps, and when you round to 16 bits at the end of the day, you get the right answer.
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- 20-bit processing - ephemere 07/14/0919:30:52 07/14/09 (48)
- digital photo analogy - mls-stl 20:40:49 07/14/09 (11)
- RE: digital photo analogy - J. Phelan 20:56:55 07/14/09 (10)
- recording fads and fashions - mls-stl 06:43:37 07/15/09 (9)
- RE: recording fads and fashions - J. Phelan 10:09:50 07/15/09 (8)
- I'm open to all kinds of music - mls-stl 12:48:55 07/15/09 (7)
- RE: I'm open to all kinds of music - J. Phelan 17:43:16 07/15/09 (6)
- an example - mls-stl 20:43:41 07/15/09 (5)
- RE: an example - J. Phelan 21:31:35 07/15/09 (4)
- Unfortunately... - mls-stl 05:48:00 07/16/09 (3)
- RE: Unfortunately... - J. Phelan 08:22:44 07/16/09 (2)
- RE: Unfortunately... - mls-stl 20:23:39 07/16/09 (1)
- RE: Unfortunately... - J. Phelan 21:02:55 07/16/09 (0)
- RE: 20-bit processing - Todd Krieger 19:47:20 07/14/09 (34)
- RE: 20-bit processing - ephemere 21:25:22 07/14/09 (33)
- RE: 20-bit processing - Todd Krieger 23:06:49 07/14/09 (32)
- RE: 20-bit processing - J. Phelan 01:03:22 07/15/09 (31)
- RE: 20-bit processing - Todd Krieger 01:14:16 07/15/09 (30)
- RE: 20-bit processing - J. Phelan 01:37:18 07/15/09 (29)
- RE: 20-bit processing - Todd Krieger 11:38:27 07/15/09 (28)
- RE: 20-bit processing - J. Phelan 17:45:07 07/15/09 (27)
- RE: 20-bit processing - Todd Krieger 23:56:02 07/15/09 (26)
- RE: 20-bit processing - J. Phelan 00:56:26 07/16/09 (25)
- RE: 20-bit processing - Todd Krieger 18:39:20 07/16/09 (0)
- RE: 20-bit processing - Tony Lauck 10:10:05 07/16/09 (23)
- RE: 20-bit processing - J. Phelan 12:39:52 07/16/09 (22)
- RE: 20-bit processing - Tony Lauck 13:51:03 07/16/09 (21)
- RE: 20-bit processing - J. Phelan 15:31:50 07/16/09 (20)
- RE: 20-bit processing - Todd Krieger 18:55:11 07/16/09 (1)
- RE: 20-bit processing - J. Phelan 19:09:31 07/16/09 (0)
- RE: 20-bit processing - Tony Lauck 17:48:57 07/16/09 (17)
- RE: 20-bit processing - Werner 01:21:17 07/17/09 (1)
- Why 32 bit converters - Tony Lauck 08:31:03 07/17/09 (0)
- RE: 20-bit processing - J. Phelan 18:02:57 07/16/09 (14)
- RE: 20-bit processing - Todd Krieger 19:18:11 07/16/09 (0)
- RE: 20-bit processing - DaveK 18:27:01 07/16/09 (12)
- RE: 20-bit processing - J. Phelan 19:05:48 07/16/09 (6)
- RE: 20-bit processing - DaveK 20:39:33 07/16/09 (5)
- RE: 20-bit processing - J. Phelan 21:11:17 07/16/09 (4)
- RE: 20-bit processing - DaveK 21:39:57 07/16/09 (3)
- RE: 20-bit processing - J. Phelan 22:20:39 07/16/09 (2)
- RE: 20-bit processing - Tony Lauck 19:55:26 07/17/09 (1)
- RE: 20-bit processing - J. Phelan 20:17:18 07/17/09 (0)
- RE: 20-bit processing - Tony Lauck 18:47:12 07/16/09 (4)
- RE: 20-bit processing - J. Phelan 19:12:41 07/16/09 (3)
- Many others ? - Metralla 04:26:22 07/17/09 (2)
- RE: Many others ? - J. Phelan 05:45:14 07/17/09 (1)
- RE: Many others ? - Tony Lauck 20:49:38 07/17/09 (0)
- RE: 20-bit processing - J. Phelan 19:34:08 07/14/09 (0)