Re: OT question about sound cards/chip-sets and high-end music systems
On Mon, 05 Oct 2009, Boyd Stephen Smith Jr. wrote:
> On Monday 05 October 2009 13:20:14 Johannes Wiedersich wrote:
> > Boyd Stephen Smith Jr. wrote:
> > > It is purely digital. 16-bit (not sure if this is floating- or
> > > fixed-point), stereo, 44.1 kHz samples, IIRC.
> > What's the difference between 16-bit floating-point and 16-bit
> > fixed-point? I always thought those are just integers.
On floating-point, you can encode a higher dynamic range, but the precison
varies. Fixed-point has the same precision on the entire range.
But you could have a function to map the integer range to a larger dynamic
range too (this is called quantization). This sort of encoding should not
be confused with regular floating-point.
CDs use PCM, which _does_ have a quantization function.
> Some audio codecs use floating-point, which is like a "float" or "double" or
> "long double" in the C programing language. Some bits are the exponent
> (usually with a bias) and some bits are the mantissa. Other audio codecs use
> fixed-point, where the 16-bits simply a signed integer (or possibly an
> unsigned integer modified by a bias).
No codec worth its salt will use 16 bits, be it fixed point or floating
point. Usually floating-point codecs will use standard precision or long
precision floating point (which is >80 bits, I think). And most integer
codecs will use 32-bit or 64-bit fixed-point arithmetric. They use that to
decode the PCM stream (which is 16-bit).
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot