Trance Forum \| Stats \| Register \| Search \| Parties \| Advertise \| Login

There are 0 trance users currently browsing this page and 1 guest

Trance Forum Л Л Forum Production & Music Making - headroom Odb?

← Prev Page	1 2 3	Next Page →
First Page	1 2 3	Last Page

Author

headroom Odb?

Triplex
Triplex

Started Topics : 22
Posts : 134
Posted : Feb 27, 2005 20:04

jupiter : of course, but I said in 16bit audio. That is if you are editing in soundforge/wavelab in 16audio, or even inside cubase if you apply gain changes on a 16 bit audio and bounce.

http://triplex.trance.net

Spindrift
Spindrift

Started Topics : 33
Posts : 1560
Posted : Feb 27, 2005 20:23

Quote:

On 2005-02-27 19:50, Jupiter wrote:
Since all the host's like logic,nuendo,samplitude or whatever work internally with 32bit 'FLOAT'you don't loose anything,cause it floats and is not fixed.

So you telling me that a signal recorded with say 50% volume is containing the same information as a signal recorded at 100% volume.
And that the dynamic range could be the same even if you are only using half of it.

It's sure contradicting with everything I have learned, and I find it very hard to belive.
Care to elaborate or maybe have a link with more info?

(``З.И(``З.И(``З.ИИ.З`Д)И.З`Д)И.З`Д)
Ћ .....www.ResonantEarth.com..... Л
(И.З`Д(И.З`Д(И.З`Д``З.И)``З.И)``З.И)

http://www.myspace.com/spindriftsounds
http://www.myspace.com/resonantearth

Jupiter

Started Topics : 1
Posts : 50
Posted : Feb 27, 2005 20:50

Hi Spindrift,we are not talking about recording ,this thread is about mixing.Ofcourse if you record an external signal it's imortant to have a strong signal,since your converters are not 32bit float.
Internally if you bounce a file with a low level it has got exactely the same information as if you bounce it with 0db.

Nice grets
Fred

Jupiter

Started Topics : 1
Posts : 50
Posted : Feb 27, 2005 20:56

@triplex,even if you handle 16bit or 24 files in wavelab or cubase,the internal processing will allways be in 32bit float.
e.g. If you bounce a file with a very low level -30 or -40 and normalize it afterwords you have exactely the same result as if you bounce the file right away with 0db.
You can make the phase invert test to be sure.

Love and Light
Fred

Jupiter

Started Topics : 1
Posts : 50
Posted : Feb 27, 2005 21:08

A simple test for everybody.
1.Export a audiofile at 0db.
2.Export the same file at -30 db
3.Import those bounced files (x and y) in two different tracks .
4.Make sure the volumefader of this tracks is at 0db.
5.Make sure that the position of the files is exactely the same.(sample accurate).
6.Normalize both files
7.invert the phase on one of the files.
8.Playback both files at the same time.
9.Listen

hope i could help you
Fred

Triplex
Triplex

Started Topics : 22
Posts : 134
Posted : Feb 27, 2005 21:14

Quote:

On 2005-02-27 20:56, Jupiter wrote:
@triplex,even if you handle 16bit or 24 files in wavelab or cubase,the internal processing will allways be in 32bit float.
e.g. If you bounce a file with a very low level -30 or -40 and normalize it afterwords you have exactely the same result as if you bounce the file right away with 0db.

i am afraid that this is not true. maybe u tested at -30db and then normalized and probably you did not hear the difference, but do it at -80 to see what is happening. its the effect of bit-crushing.

sx handles routing and mixing in 32bit. this does not mean that it transforms all your audio files into 32bit when opening a project.

http://triplex.trance.net

Spindrift
Spindrift

Started Topics : 33
Posts : 1560
Posted : Feb 27, 2005 21:28

@jupiter
I still fail to see how a 32 bit floating stream at 50% volume can contain the same amount of information as a stream which is using the full scale level.

I do understand that you seem sure that you are right, but I cannot unfortunally take your word for it since you have not explained how it can be possible.
Where is the limit?
Does a sound bounced at at -100Db contain 32 bits of data still.
I'll be very curious to see if you have anything to back up your claims, because I sure haven't seen anywhere discussions about that being the case, but I have seen many claiming that you do loose bits when lowering the level.

What we are talking about here is the mixing process.
Included in that process is recording or bouncing down the mix as well.
If you are mixing with 50% of the volume and recording down that as a file you will have less information and dynamic range in that file.
Normalizing cannot bring back bits that have not been used.
A sound that is never recorded down can be theoretically how correct you like, but no one will hear it unless you save/record/bounce that file down.

About your test, I can say that it will not be a wole lot difference on a signal recorded at 24 bit at -30 Db, unless it has a very varied dynamic range, then you might be able to spot the difference in the very quiet parts.
If I did the test I would try it at -100 Db rather.

Colin OOOD
Moderator

Started Topics : 95
Posts : 5380
Posted : Feb 27, 2005 21:54

A single word of 32-bit float data has two parts - a 24-bit linear PCM value and an 8-bit mantissa that defines its relative scale. This means that a 'sound bounced at -100dB' still contains audio at 24 bit resolution, but scaled so that its output is at -100dB. The dynamic range of 32-bit float audio is, to all intents and purposes (ie. NOT literally, but as far as those intents and purposes relate to making music destined for a 16-bit linear medium) infinite.

Mastering - http://mastering.OOOD.net :: www.is.gd/mastering
OOOD 5th album 'You Think You Are' - www.is.gd/tobuyoood :: www.OOOD.net
www.facebook.com/OOOD.music :: www.soundcloud.com/oood
Contact for bookings/mastering - colin@oood.net

Spindrift
Spindrift

Started Topics : 33
Posts : 1560
Posted : Feb 27, 2005 22:06

Ahh yes, thanks for that Colin.
Nice and clear explanation.

But then still, at least with logic where you cannot record in 32 bit float, if you are mixing at say 50% volume, the resulting fixed 24 bit file will surely contain less information.
Or?

Colin OOOD
Moderator

Started Topics : 95
Posts : 5380
Posted : Feb 27, 2005 22:28

Yes, you're right. But then if you're mixing at 50% volume you obviously haven't learned YET how to do it properly.

Mind you, 50% volume is only -6dB, and some of my mixes are almost that quiet, so I have a lot of learning to do too. At least I use Cubase and bounce my tracks in 32-bit float...

Jupiter

Started Topics : 1
Posts : 50
Posted : Feb 27, 2005 22:38

sx handles routing and mixing in 32bit. this does not mean that it transforms all your audio files into 32bit when opening a project.

[/quote]
YOu are wright,maybe i shouldn't smoke so much all the time:-))))

Nice greetings
Fred

Colin OOOD
Moderator

Started Topics : 95
Posts : 5380
Posted : Feb 27, 2005 23:00

No-one said it does.

Spindrift
Spindrift

Started Topics : 33
Posts : 1560
Posted : Feb 28, 2005 00:25

Quote:

On 2005-02-27 22:28, Colin OOOD wrote:
Yes, you're right. But then if you're mixing at 50% volume you obviously haven't learned YET how to do it properly.

Mind you, 50% volume is only -6dB, and some of my mixes are almost that quiet, so I have a lot of learning to do too. At least I use Cubase and bounce my tracks in 32-bit float...

Ok, 50% could mean -10Db I guess depending if you think in logarithm or not

I meant using 50% of the bits, and that sure would be very slack mixing.

I must say that practially i find it quite irrelevant since it so extremly easy to make sure to use the highest possible volume without clipping in Logic.
Lika I mentioned before, using Ctrl-A you can adjust all the faders at the same time to get the master to peak at -0.1 Db without any effort at all.

I don't know if Cubase have the same function but for Logic users it sure makes life a lot easier.

Colin OOOD
Moderator

Started Topics : 95
Posts : 5380
Posted : Feb 28, 2005 00:28

You can link selected faders in SX and adjust them together in this way but DON'T DO IT because the faders won't get adjusted by the same amount. It's a well-known bug, and in this respect Logic makes things a little easier than Cubase.

thockin

Started Topics : 1
Posts : 114
Posted : Feb 28, 2005 00:53

Colin and everyone are right about float. You can massively reduce or increase teh volume, and you never lose precision (well, ok, it rounds very very little bits).

The instant you convert to int format you lose that flexibility.

As for clipping - floats don't clip. It's impossible by definition. Clipping is when you convert back to int, and the signal is over the max.

Look at it this way: Because float is essentially infinite, they had to draw the 0dB line somwhere. Usually that is 1.0. 1.0 is "full signal". When you convert a 1.0 float to a 16 bit integer, you get 65535 - full signal.

But float is infinite, so what happens if you convert 1.5 to a 16 bit int? You can't do that. You get clipping.

So it doesn't matter what levels you work with, as long as you raise it to a good level before you convert to int.

I tend to work with very low levels in a project, and make up for it at the end. Plenty of people use very hot levels and reduce at the end. As long as you stay float, it doesn't matter much.

Trance Forum Л Л Forum Production & Music Making - headroom Odb?

← Prev Page	1 2 3	Next Page →
First Page	1 2 3	Last Page