| Author
|
If 24 bits do i still need to export the vsts to audio in a loud volume ( around 0 db) ?
|
Vermeee
IsraTrance Full Member
Started Topics :
108
Posts :
1069
Posted : Mar 20, 2010 03:50:45
|
If 24 bits do i still need to export the vsts to audio in a loud volume ( around 0 db) ?
or each track can stay in the original volume from the pre mix?
TBH in my experience i dont see why to raise volume to 0 db of each single track to export but im not a professional and im tryin to achieve a decent profissional mix....
 
http://soundcloud.com/bgos |
|
|
|
APriest
Started Topics :
2
Posts :
98
Posted : Mar 20, 2010 10:53
|
24 bit is integer, 16 and 32 bit are floating point, imho floating point is recommended if you do subsequent operations on the file.
If you export to 32 bit floating point the export volume isn't important because of the mantissa, 32 bit fp has same precision as 24 bit (24 bit for exponent and 8 for mantissa, same number of steps... not really) but fp gives you a wider dynamic range (volumes), 24 bit has about 150 dB of dynamic range, 32 bit has 1500/2000 dB of dynamic range.
Obviously you have 2000 dB of range but when you mix 2 sounds with 150 dB of difference the lowest sound is muted, too low in volume compared to the other volume.
32 bit is like "magical 24 bit", 64 bit daws have 53 bits plus 11 for the mantissa (wider dynamic range).
"Strange" behaviour to try:
Export a project without a limiter with volume over 0 dB (that clips) at 32 and 24 bit.
Now import to CoolEdit (or wavelab or other audio editor), if you lower the volume of the 32 bit file you will have an unclipped wave (no clip, like lowering the master fader in the daw), if you lower the volume of the 24 bit file you have a clipped file (not like lowering the master fader in the daw) with lower volume. |
|
|
|
PoM
IsraTrance Full Member
Started Topics :
162
Posts :
8087
Posted : Mar 20, 2010 18:18
|
Quote:
|
On 2010-03-20 10:53, APriest wrote:
"Strange" behaviour to try:
Export a project without a limiter with volume over 0 dB (that clips) at 32 and 24 bit.
Now import to CoolEdit (or wavelab or other audio editor), if you lower the volume of the 32 bit file you will have an unclipped wave (no clip, like lowering the master fader in the daw), if you lower the volume of the 24 bit file you have a clipped file (not like lowering the master fader in the daw) with lower volume.
|
|
do you mean than when exporting at 32 bit float the file don t clip?but at 24 bit it does ?
a good thing about cool edit/audition is that the waveform look like at the output of a DAC .
about the topic you don t need to raise the volume in fact it could be worst, if would advice you to export around -6 to -12 db peaks (or maybe even lower ,you don t need to boost the signal, just export it at the level it is mixed in the track)and if you can to never go louder than that in any chanels processing chain, won t make much difference but it s to avoid disortion that could hapen .i don t remmeber much about all that and i don t even aply that in my tracks but if i was trying to do things best i would stay lower than -10 db peak for every chanels if noisefloor is not a issue |
|
|
|
APriest
Started Topics :
2
Posts :
98
Posted : Mar 20, 2010 23:43
|
Quote:
|
do you mean than when exporting at 32 bit float the file don t clip?but at 24 bit it does ?
|
|
Yes, if you export a 32 bit (or 64 bit to exagerate  ) clipped waveform you can lower the volume in an audio editor and you don't loose anything.
Try to export a 32 bit file at +96 dB (96 dB of clip/distortion) or even more, you can come back to your unclipped/undistorced waveform by simply lowering the volume by 96 dB in an audio editor, exactly the same as lowering the volume fader on the master channel, with 32 bit you have nearly 2000 dB of dynamic range available, very different from analogue devices. |
|
|
|
Vermeee
IsraTrance Full Member
Started Topics :
108
Posts :
1069
Posted : Mar 21, 2010 01:02
|
my audio interface supports til 24 bits, but in cubase export options i have the options to export files to 32 bits float point.... will it work as 32 bits indeed? or my audio interface s the one that says the rules?
http://soundcloud.com/bgos |
|
|
|
Vermeee
IsraTrance Full Member
Started Topics :
108
Posts :
1069
Posted : Mar 21, 2010 01:09
|
[quote]
On 2010-03-20 18:18, PoM wrote:
Quote:
|
about the topic you don t need to raise the volume in fact it could be worst, if would advice you to export around -6 to -12 db peaks (or maybe even lower ,you don t need to boost the signal, just export it at the level it is mixed in the track)and if you can to never go louder than that in any chanels processing chain, won t make much difference but it s to avoid disortion that could hapen .i don t remmeber much about all that and i don t even aply that in my tracks but if i was trying to do things best i would stay lower than -10 db peak for every chanels if noisefloor is not a issue
|
|
well in fact my intentions was to export the files already in a pre mixed ( balanced volumes ) from the VSTS to go to the audio edition and post mix.... so i didnt want to raise the volume indeed. but i think i ve read in many places sayin to export the files to a very high volume so in case u need to raise the volume again u wont get the clips and anoyin noises from the bottom areas of the volume but i guess it s wrong then...
noisefloor are those noises that can show up in the bottom area from the volume of the file right ?
thanks for the support guys ......
http://soundcloud.com/bgos |
|
|
|
mquirk1
IsraTrance Full Member
Started Topics :
63
Posts :
384
Posted : Mar 21, 2010 07:00
|
|
the answer to your q is no. in 24bit you do not need to record as close to 0db as possible |
|
|
|
APriest
Started Topics :
2
Posts :
98
Posted : Mar 21, 2010 16:28
|
Sorry for my english.
In 16 bit 44100 (mono ) you have 44100 values in every second that represents amplitude, these values can range from 0 to 2^16 (power of 2) that is 65536 (65535 because we start from 0).
If you have a sound with a volume of 50000 and you sum with another sound with volume of 50000 the result will have a volume of 100000, but 100000 isn't representable by 16 bit integers (limited to 65535), so you have clip, 100000 is trimmed to 65535.
In Floating point you can sum what you want and the precision remains the same, the number of steps (the precision) is unchanged, what changes is the scale, a low volume has a small scale (like the height of steps, low volumes=low scale=low steps), an high volume has a big scale (high steps), i repeat, the number of steps is even the same.
The audiocard isn't related to bits of export.
It's normal that the audiocard has a fixed range of possible values (volumes), integers, like it's normal that CD are 44100 16 bit (integer) stereo.
When you output (audiocard, cd, dvd) data is integer. |
|
|
|
TimeTraveller
IsraTrance Full Member
Started Topics :
80
Posts :
3207
Posted : Mar 21, 2010 16:31
|
|
Colin OOOD
Moderator
Started Topics :
95
Posts :
5380
Posted : Mar 21, 2010 16:42
|
|
APriest
Started Topics :
2
Posts :
98
Posted : Mar 21, 2010 16:49
|
Another simple test to try:
Open audition, File->New and choose 16 bit mono. Generate -> Tones... -> OK (now you have a wave).
Effects->Amplify by 48 dB
Effects->Amplify by -48 dB
Repeat choosing 32 bit in place of 16 bit, really different results.
|
|
|
|
Shiranui
IsraTrance Full Member
Started Topics :
116
Posts :
1219
Posted : Mar 22, 2010 06:56
|
16 bit with a peak of 0 db has the same signal to noise ratio of 24 bit with a peak of 2^-(24-16) = 1/256 = -24 db (or maybe it's -48. I can never remember whether to multiply by 10 or 20)
so as long as your peak is above -24 db, you will have better signal to noise ratio than is possible with 16 bit. Of course the best SNR is achieved at peak of 0 db, but why risk clipping and losing dynamic range? -3 to -6 is a good range to aim for.
edit: apriest, you want to do the -48 db and then the 48 db, not the other way around. |
|
|
|
Shiranui
IsraTrance Full Member
Started Topics :
116
Posts :
1219
Posted : Mar 22, 2010 06:58
|
I tried to do that in audacity and it crashed  |
|
|
|
APriest
Started Topics :
2
Posts :
98
Posted : Mar 22, 2010 08:22
|
Quote:
|
On 2010-03-22 06:56, Shiranui wrote:
16 bit with a peak of 0 db has the same signal to noise ratio of 24 bit with a peak of 2^-(24-16) = 1/256 = -24 db (or maybe it's -48. I can never remember whether to multiply by 10 or 20)
|
|
Every bit is about 6 dB (6.02 perhaps, couldn't remenber), the difference between 24 bits and 16 bits is 8 bits so the difference in dynamics is about 6dBx8=48 dB.
Dynamic range of 16 bit is 16x6=96dB
24 bit= 24x6=144dB
32 bit having 24 bits of exponent is about the same as 24 bits (but "magic") ...perhaps the bits avilable are 25 because the precision of the mantissa is 1 bit larger than the bits it consists of.
64 bit float has 53 bits of exponent, 53x6=318dB ("magic" too).
Quote:
|
so as long as your peak is above -24 db, you will have better signal to noise ratio than is possible with 16 bit. Of course the best SNR is achieved at peak of 0 db, but why risk clipping and losing dynamic range? -3 to -6 is a good range to aim for.
|
|
That's true, i agree... above -48 dB.
Quote:
|
edit: apriest, you want to do the -48 db and then the 48 db, not the other way around.
|
|
I don't understand english very well (in italy teachers are communists).
Doing first the boost of 48dB you go beyond the representation possibilities of integers,
thats revealed when you cut.
Doing first the cut of 48 dB and then the boost the result seems the same but the precision is different, if you use integers 8 bits are lost, 8 bits because 48 dB are 8 bits .... if i'm correct , thats because floating point can change scale, integers have a fixed scale, a 24 bit integer waveform with a peak volume of -48 dB use a maximum of 24-8=16 bits in his peak. |
|
|
|
APriest
Started Topics :
2
Posts :
98
Posted : Mar 22, 2010 11:15
|
|
Mistake, where i've said "mantissa" i mean "exponent" and vice versa. |
|
|
|
