|
what is 32-bit float?? explain please
|
View this Thread in Original format
| Limit |
| I dont get it...why use it? |
|
|
| DigiNut |
"Float" means floating point, as opposed to fixed point. It's used on computers because there's almost no loss of precision, which in a fixed-point system would result in quantization noise.
Think of the hiss and overall graininess you get from tapes. That's partly precision loss because they're 8-bit fixed point. The 16-bit fixed-point you hear on a CD has almost no noise by itself, but if you process that several times in an editor you could start to get noise if it stayed in that 16-bit fixed form. Won't happen in 32-bit float.
Edit: I should point out that all sequencers use 32-bit floating point internally. If you're talking about bouncing a track to a 32-bit FP wave format, which some sequencers and editors allow you to do, then it's generally pointless and counterproductive, don't do it. |
|
|
| Diginerd |
Short version, 32 Bit float POTENTIALLY makes for better sound, but at the expense of CPU consumption.
32 Bit float is in the same ball park as 24 Bit fixed (As used in hardware based DSP systems), forgive the sweeping and innacurate generalization if you really understand what's going on..
24 Bit fixed & 32 Bit float are both VASTLY superior to 16 bit.
There are some fairly simple ways to prove this if anyone can be bothered to poke themselves. |
|
|
| camsr |
| quote: | Originally posted by dj_palm
in context of music i have no idea what it is but i know how to calculate from it to decimal or whatever. the 23 first bits are mantissas, the next eight is exponent, then the MSB is sign, then u can calculate the value using a formula who looks like this:
(1 + (mantissa / 2^23)) * 2^exponent and if sign is 1 the number is negativ. the number gets incredible exact but there are number that will never be exact, as 0.1.
no usefull info u can use on production tho. |
Unless you make a plugin called Float Crush :D :D :D |
|
|
| Limit |
Thanks...i just never saw the point to use it within cubase.. I always thought it was the same as 24 bit...I guess I was somewhat right.
Thanks again. |
|
|
| kitphillips |
| quote: | Originally posted by DigiNut
[COLOR=#99CCEE
Think of the hiss and overall graininess you get from tapes. That's partly precision loss because they're 8-bit fixed point. The 16-bit fixed-point you hear on a CD has almost no noise by itself, but if you process that several times in an editor you could start to get noise if it stayed in that 16-bit fixed form. [/COLOR] |
You mean that tapes (audio casettes) are digital?? I always thought they were analogue? Am I really misinformed or are we talking about different things (DAT)?
Oh yeah, and is 32 bit floating point not superior to 24 bit, and is 64 bit mixing (sonar) overkill. And is 24 bit floating point normally, if not why not... Think thats all.... sorry:p |
|
|
| DigiNut |
Yeah, I must have written that post about tapes after my nightly glue-sniffing. I don't know what the hell I was talking about. I think what I meant to say is that you get the same kind of artifacts from recording in 8 bits as you get from recording onto a cassette tape, which is not at all what I said, is it.
Try recording something in 8 bits and you'll hear what sounds like tape hiss. Obviously, tapes are analogue, so it's not due to the same mechanism!
Limit, you're always using 32-bit FP internally in Cubase. There's no way to turn it off. It's just the recording/exporting where you can control the bits. And incidentally, if you're going to bounce to 16-bit, as most of us do, you should always use a dithering plugin, because the truncation from 32 to 16 bits creates noise (and it actually is slightly audible on a clean track). |
|
|
| kitphillips |
| quote: | Originally posted by DigiNut
[COLOR=#99CCEE]Yeah, I must have written that post about tapes after my nightly glue-sniffing. I don't know what the hell I was talking about. I think what I meant to say is that you get the same kind of artifacts from recording in 8 bits as you get from recording onto a cassette tape, which is not at all what I said, is it.
[COLOR] |
:D Glad I wasn't going around saying that tapes are analogue all these years when they weren't! |
|
|
| Limit |
| quote: | Originally posted by DigiNut
Limit, you're always using 32-bit FP internally in Cubase. There's no way to turn it off. It's just the recording/exporting where you can control the bits. And incidentally, if you're going to bounce to 16-bit, as most of us do, you should always use a dithering plugin, because the truncation from 32 to 16 bits creates noise (and it actually is slightly audible on a clean track). |
Yeah I always dither when going from 24 to 16...Thanks Digi...you cleared it up some. |
|
|
| azndragon0613 |
| I figure if you have problems with your song clipping wise or you want to do some hard limiting with a boost, I'd export into 32 bit mode and do your compression in a external program like Adobe audition. |
|
|
|
|
