| Author
|
## bit float...
|
orik
IsraTrance Junior Member
Started Topics :
45
Posts :
317
Posted : Dec 13, 2004 16:52
|
what exactly does the float stand for?
for example 24 bit float...sounds like it would have more valuse for some reason
but can anybody explain? does it have
anything to do with the ability of the point (10.00) to move and change the value? |
|
|
Hayez
Started Topics :
8
Posts :
393
Posted : Dec 14, 2004 00:10
|
I am not aware of 24 bit float type in use anywhere. 24 bit are usally refer to integer type.
here is a definition to float:
"Single-precision values with float type have 4 bytes, consisting of a sign bit, an 8-bit excess-127 binary exponent, and a 23-bit mantissa. The mantissa represents a number between 1.0 and 2.0. Since the high-order bit of the mantissa is always 1, it is not stored in the number. This representation gives a range of approximately 3.4E–38 to 3.4E+38 for type float. "
 "a new art came into my mind which only you can create, the Art of Noises, the logical consequence of your marvelous innovations." Russolo, 1913 |
|
|
|
mono mono
Onnomon
Started Topics :
5
Posts :
314
Posted : Dec 14, 2004 02:06
|
Quote:
|
On 2004-12-13 16:52, orik wrote:
what exactly does the float stand for?
for example 24 bit float...sounds like it would have more valuse for some reason
but can anybody explain? does it have
anything to do with the ability of the point (10.00) to move and change the value?
|
|
Yep, it's binary point that can move right or left, based on the exponent value. The mantissa, a 24-bit signed value (23 + 1 sign bit) contains the digits of the number. The exponent specifies where the point "." goes.
For digital audio, the practical upshot is that you don't really get any more dynamic range resolution in the end. However floating-point representation lets the DC level drift around and still gives you the 24-bit resolution, whereas a fixed 24-bit integer would start to clip once the DC value drifts too far negative or positive.
The useful part of float is in the intermediate calculations that the plug-ins, and virtual mixing does. Basically you can add or multiply lots of numbers together, and the exponent floats along as the number gets larger, generally retaining the resolution. Finally before the number goes out the door it's usually divided down (also called "scaling") which typically shifts the binary point back down to the integer value range. If you were to use integers the intermediate calculations might go past the largest allowable integer (a 24-bit signed integer can range from -8388608 to 8288607 in decimal). Although most integer processor arithmatic is 32-bit, these days the time it takes to do a floating-point operation is virtually the same as integer, so the software designers have decided to let the hardware handle the dirty work. Although some people might remember the infamous pentium floating-point bug years ago. I know Logic uses 32-bit floating-point, I'd imagine Cubase probably does too.
I realize this is complex and confusing stuff. It's a challenge to attempt to describe the floating-point concept.
-dean
|
|
|
|
|
