Re: [ReZound-users] suggestion and question
Status: Beta
Brought to you by:
ddurham
From: Marc R.J. B. <mr...@dn...> - 2003-07-15 11:23:31
|
On Mon, 14 Jul 2003, Gian Paolo Mureddu wrote: > Davy Durham wrote: [32 bit floating point vs. 16 bit integer] > >After LADSPA I'm thinking this is the next thing to address. > Hey, why not having the option to make a recording in either one, 32-bit > float or 16-bit integer? That might take longer to compile or even a > larger program, but I thought that could be great! (I am not a coder, so > I wouldn't know of any problems at code level) Once code works with float32 internally, there is not much of a point in keeping integer code. I understand that currently the main issue is worries about performance, mainly due to increased disk I/O. Why not optionally do the on-disk processing at float24 rather than float32? This would give only half the disk IO performance loss compared to float32, while already giving much better fidelity than the 16 bit stored in file (e.g. is virtually lossless). If a file was originally stored in 24 bit or better, it is always possible to use 32 bit work files for those. Of course all math can still be carried out in 32bit float even when using only a 24bit float work file. I'm not familiar yet with the native file format of rezound, but an audio program with its own native format could do completely lossless editing by simply storing the original files plus the edit operations they were subjected to. (a bit similar to the format photoshop uses for saving images). When saving to wav, mp3 or any other format, this data could be processed to output the correct sample values. grtz MRJB |