| Author
|
Lets get into the details of wordlengh..
|
Kitnam
Mantik
Started Topics :
110
Posts :
1151
Posted : Nov 23, 2004 15:53
|
Hi all you masters of audioengeneering..
I am realy confused.
Usually I burn my Tracks on CD with a wordlengh of 16 bit (which is the maximum for burning on CD). A .wav-file with a 16bit format contains more quality then a wavefile with 20/24 or 32bit (...says the masteringprofi)
Seems logical to me: 32bit is 2 byte more than a 16 bit word.
1001 1110 1110 1110 (16 bit)
1101 1000 0010 1100 1001 1011 0000 1100 (32 bit)
so everytime you edit-audioinformation use the highest bitrate you can handle and reduce it to 16 bit format at last step of mastering! (and use dither)
but what the hell is a dither...
i dont realy know yet...
as far is i understood is a dither something like a low-db-noise-generator which makes the deep of a track having more audience and reduces the lack from analog-wave to digital-(stairlike)-wave. this has something to do with a tricky algorithm.
you find dithering in plugins from steinberg or wavelab (L1, L2 ultramax.) dithering should be used only once on a track.
i only new a few parts of this knowlegde..
so if you have something to correct here or to add please post so we can build up the house and reduce the myths in our head. |
|
|
FluoSamsara (Oxygen)
IsraTrance Full Member
Started Topics :
84
Posts :
1164
Posted : Nov 23, 2004 17:30
|
Quote:
|
On 2004-11-23 15:53, Kitnam wrote:
A .wav-file with a 16bit format contains more quality then a wavefile with 20/24 or 32bit (...says the masteringprofi)
|
|
What? That makes no sense to me....16bit better quality than 24bit??Maybe I'm getting something wrong...is that what u're saying? |
|
|
|
orik
IsraTrance Junior Member
Started Topics :
45
Posts :
317
Posted : Nov 23, 2004 17:42
|
no thats not right!
24 bit has more resolution then 16 bit,
think about it- a 24 bit word has more info than a 16 bit word right? therfore it contains more info (or detail if u want)
....dither means adding noise - or white noise when you sample an analog signal to
digital form,
wich would put each sample you sampled within the sound wave over more acurate valuse therefore making it more realistic
or more similar to the analog source...
|
|
|
|
UnderTow
Started Topics :
9
Posts :
1448
Posted : Nov 23, 2004 22:32
|
I think he just misswrote that sentence. 
To answer your question,
How dither works:
=================
The easiest explanation I have read is the one in Bob Katz' mastering book. Here is a paraphrase:
Imagine a simple A/D D/A converter in a box that is scaled in such a way that 1 Volt = 1 bit. If you feed a constant signal of 0.3 Volts to the A/D converter, the first bit will never be triggerd so the output of your A/D converter would be a constant row of 0s and the output of your box (after the D/A stage) would be 0 Volts.
If you feed the A/D converter with 0.8 Volts, the output would be a constant row of 1s and the output of your box would be a constant 1 Volt.
Now imagine that you feed random noise to the converter so that half the time the voltage is below 0.5 Volts and half the time the signal is above. (To be exact, half the time the voltage is between -0.5V and 0.5V and half the time the voltage is greater than this).
Now you get a stream of 1s and 0s. On average you have as many 0s as 1s.
In other words, you have noise on the output of the box but the average signal is still 0 volts.
Now mix the 0.3 Volts signal and the noise and feed it to the converter. You still have noise on the output converter but the AVERAGE signal is 0.3 Volts! And presto! You can now sample a signal that is smaller than the 0.5 volts resolution of your converter!
Noise Shaping:
==============
To make things more pleasant on the ears, smart people came up with the idea of shaping (filtering) the noise so that there is less of it in the frequencies that we are sensitive to (mid-range) and more where we are not so sensitive (very high and very low). As long as the average amount of noise is just enough to trigger the first bit of your converter (or the first bit of the bit depth you are changing to), everything is fine.
About only adding dither once:
==============================
One should really dither each time one lowers the bit depth. Alot of plugins do this automaticly. (You can't really hear this though. If you are going from a 64 bit (float) plugin down to 32 bit (float) any dither added is at arround -385 dbFPS. Inaudible. )
I know that Sonar automaticly adds dither when it goes from 64 bit (all the busses) down to whatever your soundcard supports. You can turn this off in the options if you want but it is best left on. I don't know how Cubase and Logic handle this ...
Having said all that, you can't actually add dither yourself anywhere but at the last stage when you go from 32/24 bits down to 16 bits. And indeed, you should only add it once.
I hope this demystifies dither a bit.
UnderTow |
|
|
|
UnderTow
Started Topics :
9
Posts :
1448
Posted : Nov 23, 2004 22:33
|
EDIT: Double post deleted. |
|
|
|
orik
IsraTrance Junior Member
Started Topics :
45
Posts :
317
Posted : Nov 24, 2004 13:43
|
[quote]
I know that Sonar automaticly adds dither when it goes from 64 bit (all the busses) down to whatever your soundcard supports. You can turn this off in the options if you want but it is best left on. I don't know how Cubase and Logic handle this ...
alastair!
what does that mean then all the busses?
the sequencer encreases the resolution acourding to the number off busses you add
in your sequencer (from the sound card busses)? |
|
|
|
H2O
IsraTrance Junior Member
Started Topics :
16
Posts :
352
Posted : Nov 24, 2004 14:29
|
|
UnderTow , how 32bit registers can handle 64bit data ? Are you all talking about word processing or resolution of sound ? Please explain . |
|
|
|
ZilDoggo
Started Topics :
4
Posts :
663
Posted : Nov 24, 2004 15:09
|
"how 32bit registers can handle 64bit data"
modern cpu's have SIMD ops.
these alu's will handle 8, 16, 32, 64 and 128 bits data for arithmetic like adding and multiplying,. i think even 256 bits these days.,
Undertow,
"(You can't really hear this though. If you are going from a 64 bit (float) plugin down to 32 bit (float) any dither added is at arround -385 dbFPS. Inaudible. ) "
shouldnt that be -192dbFPS ?
i mean, the dither will be added to the 32 bit signal, not the 64 bits signal.,
and 32 bits would have -192dbfs bandwidth..
but you would not hear it anyway
greets.,
aka., |
|
|
|
ZilDoggo
Started Topics :
4
Posts :
663
Posted : Nov 24, 2004 15:16
|
kitnam,
"A .wav-file with a 16bit format contains more quality then a wavefile with 20/24 or 32bit (...says the masteringprofi) "
this is ONLY true if your sequencer renders to 16 bits,.
(so then you would upsample it to 24 bits and with that possibly reduce the quality)
if your sequencer works internaly in 32 bits floating point and you can render the track to 32 or 24 bits this will have the best quality for mastering.,
16 bits is still ok .,. but never make a 16 bits pre-master into a 24 bits one before you send it to mastering.,,
so this 'profi' advice is a little bit old
greets.,
aka., |
|
|
|
H2O
IsraTrance Junior Member
Started Topics :
16
Posts :
352
Posted : Nov 24, 2004 17:19
|
|
Should the dither from 16bit to 24bit be on mixing stage (pre FX insert form) ? also what would you advice as best dithering software ? |
|
|
|
UnderTow
Started Topics :
9
Posts :
1448
Posted : Nov 24, 2004 18:18
|
Quote:
|
On 2004-11-24 14:29, H2O wrote:
UnderTow , how 32bit registers can handle 64bit data ? Are you all talking about word processing or resolution of sound ? Please explain .
|
|
I am assuming that they use double precision float. Which means I forgot the mantissa when calculating the noise level. But I am not entirely sure how the 64 bits is implemented.
Zildoggo, actually you are right. Although the dither noise is added to the 64 bit signal before it is converted (otherwise it is too late and pointless), it has to trigger the 1st bit of the lower bit depth. So indeed , the noise is at the lowest bit of the new format.
UnderTow
UnderTow |
|
|
|
UnderTow
Started Topics :
9
Posts :
1448
Posted : Nov 24, 2004 18:30
|
[quote]
On 2004-11-24 13:43, orik wrote:
Quote:
|
I know that Sonar automaticly adds dither when it goes from 64 bit (all the busses) down to whatever your soundcard supports. You can turn this off in the options if you want but it is best left on. I don't know how Cubase and Logic handle this ...
alastair!
what does that mean then all the busses?
the sequencer encreases the resolution acourding to the number off busses you add
in your sequencer (from the sound card busses)?
|
|
No these are the internal buses in Sonar. Not your sound card buses. You can use them as mix buses or send (aux) buses. All the tracks work at 32 bit and the buses work at 64 bit.
The idea is that you get near infinite headroom when mixing down tracks through the buses. As long as you pull the fader down so that the meter doesn't go over 0, it doesn't clip.
But I am a bit old school and I mix in such a way that it doesn't go over 0 anyway. I don't think it can hurt to go over 0 in the buses as long as the final output to the sound card is below 0 but I know it certainly doesn't hurt to be carefull.
What I say above is valid for some other programs but certainly not all! It is safe to stick to the traditional method of keeping everything below 0 dBFS. Also this means your techniques will translate well to analogue mixers and fixed point maths systems like ProTools and Mackie D8B.
UnderTow
|
|
|
|
ZilDoggo
Started Topics :
4
Posts :
663
Posted : Nov 24, 2004 19:11
|
undertow,
interesting topic.,
i always was convinced that dithering was aplied afterwards and only to >mask< the lsb error (so not providing any more information, actually reducing the dynamic range by a bit but sounding more natural cause brain doenst hear errors only natural noise).,
but i do remember that there are some forms of dithering (uv22 ?) that claim to include information 'beyond' the final bit depth., so you can 'hear through' the bits ., (this is THEIR description  )
i never understood this, but as you explain, when it is applied beforehand it can offset the dithering pattern of the final result and thus provide some information from stuff removed..
cool
greets.,
aka., |
|
|
|
UnderTow
Started Topics :
9
Posts :
1448
Posted : Nov 24, 2004 23:17
|
Quote:
|
On 2004-11-24 19:11, ZilDoggo wrote:
undertow,
interesting topic.,
i always was convinced that dithering was aplied afterwards and only to >mask< the lsb error (so not providing any more information, actually reducing the dynamic range by a bit but sounding more natural cause brain doenst hear errors only natural noise).,
|
|
No no no! See my explanation above.
Quote:
|
but i do remember that there are some forms of dithering (uv22 ?) that claim to include information 'beyond' the final bit depth., so you can 'hear through' the bits ., (this is THEIR description )
|
|
This is true and is valid for all dithering. It DOES increase the resolution beyond what would otherwise be possible with any particular bit depth. Of course some dithering is better than others and some works better on different materials.
But with psytrance being at such high levels, any good dithering will work as you won't hear the noise anyway.
Quote:
|
i never understood this, but as you explain, when it is applied beforehand it can offset the dithering pattern of the final result and thus provide some information from stuff removed..
cool
|
|
Very cool indeed. Very smart people have come up with this stuff.
UnderTow |
|
|
|
Kitnam
Mantik
Started Topics :
110
Posts :
1151
Posted : Nov 24, 2004 23:49
|
thx dudes...
@zil
i just swapped 16bit with 32, for sure 32 bit is more... as far as i know mixing down audio from cubase in 32 floating makes not so much sence because a lot of plugins doesnt generate high-resolution formats....
myth? |
|
|
|
