192khz .. playback fidelity lower than 44.1khz?

[Christian Borg]

Tony Hardie-Bick skrev 2012-03-06 23:56:
> On 06/03/12 21:47, Christian Borg wrote:
>> Is this a problem with 32bit float (24+8) only and not with fixed 24-bit?
>
> Fixed point is a completely different thing, it can be better, but it takes much more 
> coding effort. Using doubles is often a better solution.
>
>> I thought many of todays music mixing software used 32-bit float internally
>> because of the rounding errors being inaudiable?
>
> They do, and the errors are inaudible, mostly. The difference is indirectly noticeable 
> as lack of space in rich sounds (whether there is such a thing as a sound that's not 
> rich, is another discussion).
>
> This loss of information in the signal chain is cause for concern, because it cannot be 
> recovered.
>
> This is why most audio systems go beyond 24+8 when possible. Sharc DSPs, for example, 
> have an extended precision 40 bit FP format, which gives a great advantage, without 
> doubling the wordlength. In the DSP world, 24+8 is generally considered insufficient as 
> a processing standard (for storage and playback it's overkill, IMO), 40 bits is 
> workable, 64 bits is rarely gonna give you problems. You have to work hard to find 
> anything wrong there.
>

Thanks for the explanation Tony!
Much appreciated.

-- 
Christian B