The paper propose to use 48 bits world with 24 bits DSP. They use 2 registers for an audio sample. They basically said that manipulating 24 bits audio sample with 24 bits arithmetic raised the noise too high... What a surprise :)
They propose to use 32 bits audio sample, plus 8 bits in the low end. So it's like using 40 bits. Then you add 8 bits at the top to manage overflow (which have the same roll as exponent). If you want to use flotting point arithmetic, you need 40 bits mantissa. Maybe it's easier to have 48 bits integer (or more) rather than having 48 bits flotting point unit. > As I understand it you basically pair 2 x 24bit DSPs to give yourself a > 48bit > fixed point DSP. > > The benefits are explained here. > http://www.jamminpower.com/PDF/48-bit%20Audio.htm > > I know it would be more expensive but considering that the ear is very > sensitive to "pumping" of audio I think it would be worth it if you were > going to be doing filter on the card and selling it as a pro-audio card. > > Regards, > > Paul Mullen. > >> Nicolas Boulay _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
