> On Tuesday 04 April 2006 09:36, Erik Hofman wrote:
>> Timothy Miller wrote:
>> > On 4/3/06, Justin Findlay <[EMAIL PROTECTED]> wrote:
>> >> I'd say more than 24, perhaps 32 bits per channel.
>> >
>> > Ok.  I thought about saying 32, but I didn't want to look absurd.  :)
>>
>> Internally you would have to use 32-bit floating point values,
>> everything else will be obsolete rather soon. 24-bit D/A conversion
>> would be nice and  32-bit D/A converters would be great.
>> Or forget about an internal D/A converter all together and use digital
>> output only.
> There are problems associated with 32bit floating point signal processing
> in
> audio especially if you are trying to do DSP on the card. The easiest
> example
> of this is low freqency/high Q filtering which needs simultaneous use of
> large and small values, something floating point is not good at. A double
> precision 24bit fixed point DSP would be much better. You get a dynamic
> range
> of 288dB plus more predictable results from your algorithm.

What is "double précision 24 bits DSP" ?

32 bits flotting point unit use 24 bits mantissa. The precision is the
same as a 24 bits fixed integer unit. So i don't understand you're point.

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)

Reply via email to