Sound dithering

Music Manic · Post by **Music Manic** » Tue Jul 23, 2002 2:31 pm

Was just wondering how Creamware audio engine works compared to progs like Cubase,samplitude etc.
Seeing that a project can load ASIO,wav drivers plus route them through the mixers,I would like to know how the final stereo output compensates all this mixture.
Does the mixer "alter" the sound as with the drivers;regarding Realtime signals or Rcorded wav files that is.
Heard avout dither bugs in some of plugins,some of which are fixed.
Would like more insight if there's any please.
Thx!
P.S. I ask this because Creamware gives overall nice sound,but would like to know where faults lay!

Eurocide · Post by **Eurocide** » Wed Jul 24, 2002 12:22 am

I am also interested in this theme.
Especially how the different ASIO devices work.

E.g.: where is the significant difference between the ASIO 32bit and ASIO 32Float device.
Cubase can record *TrueTape* with both (even with the 16 and 24 ASIO devices) and for my ears the sound isn't influenced.

DJATWORK · Post by **DJATWORK** » Sun Jul 28, 2002 7:09 pm

I see a lot of confusion here...

Firt at all, the ASIO, WAVE, GSIF, EASI, etc, are the drivers used to comunicate the Software (Cubase, Logic SoundForge, etc) with the Hardware (Pulsar, SoundBlaster, MOTU, etc).
Its like a way to transport the digital signal from the card to the software, like a bridge or something like that. That has nothing to do with the sound.
The information that travel on those drivers, has nothing to do with the way that drivers "transports" it.

Thas the reason because there is not any "compensation".

Is similar to say ADAT Lightpipe and TDif, or AES/EBU and SPDif. Those are diferent protocols to transfer the digital audio signal. You don´t use any compensation between those protocols, because there is no sound modifying.

OTher diferent thing that has nothing to do with that protocols is the bit depht.
The bith depht can be 16 Bits, 24 Bits 32 Bits and so. And theree are other different that are the "Floating poit". IN simple words, and for example a 32 Bits Floating Point is like a 24 Bit (real) Bith depht PLUS 8 bith encoded (not real, but Floating Point). Adding 24 Bits plus 8 "encoded" bits, you have those 32 Bits FP, what in facts is "better" than only 24 Bits.

Whats the difference between 32 Bits and 32 Bits Floating Point?
The conversion process, quality and weight of the file.

Dithering...
In a A/D conversion (and D/A too) there are several "digital errors" introduced in that process by the converters, appart of the converter quality.

One of those digital errors is a kind of distortion ocurred when a low level signal is recorded in the lower bits of the digital signal. So if you add another signal to that low level signal recorded (a kind of noise), you guarantee that any signal is never gonna be lower than the noise.
If the noise is in the limit of the distortion, you can be sure that no audio signal is going to be recorded under the distortion level. (too much complicated?)
The signal introduced is called DITHER, and there are different theories and different algorithms for that dither. (Apogee UV22 for Example).

When you convert a 24 Bits signal to 16 Bits signal there is no A/D or D/A conversion, but because of the bit depht substraction, many audio that in the 24 Bit signal where "not low signal", in a 16 bit audio signal can be under that distortion level. Thats why DITHER is added too in the 24 to 16 bits conversion.

I hope it had help and clear more than complicated...

any doubt, ask and if I can explain you more, I´ll be gratefull to do that!

DJATWORK

_________________
Luis Maria Gonzalez Lentijo
DjatWork! Optimizaciones
Buenos Aires
Argentina

<font size=-1>[ This Message was edited by: DJATWORK on 2002-07-28 20:17 ]</font>

Post by **hubird** » Sun Jul 28, 2002 8:53 pm

good post Djatwork! very short but also very to the point.

One more thing: one should avoid using dithering two times on the same file.

This keeping in mind you could change your save stratagies with mix down and mastering files.

Nowadays I keep my 'definitive' mixes undithered as long as possible (24bit).

It allows me the possibility to edit that later not so definitive file, f.i. with new audio, than finish again with limiting plus dithering.
(In Waveshell's +L1 Ultmax you have all the dithering options).

Question: to burn that file to cd you have to change to 16 bit.
As you say there is dithering used too, and far as I know this is part of the process, you can't change this.
At the end you allways use dithering twice.
Isn't this a problem?

Let There Be Music!

at0m · Post by **at0m** » Sun Jul 28, 2002 10:04 pm

Very nice post DJATWORK, do you want to repost it in Study forum? Would be cool

On 2002-07-28 21:53, hubird wrote:
Question: to burn that file to cd you have to change to 16 bit.
As you say there is dithering used too, and far as I know this is part of the process, you can't change this.
At the end you allways use dithering twice.
Isn't this a problem?

How comes you'd dither twice? After mastering/mixdown, you can still export the audio from for example Cubase and dither only once during final mix 16bit wav writing?

As I understood from DJATWORK, 16bit Wav Dests don't incorporate dithering. This would be nice tho, if it were done automatically. Cos it's quite obvious that one wants to record 16bit. Or we'd have selected 24bit wav drivers. But then still, all dsp works is 32bit.
On the other hand I've almost never used dithering and my 16/24bit recordings only crackled a couple of times. How comes?

I don't see when or where to actually put a dithering device in my setup. I know on mixdown to inside the multi-tracker, I'd apply dithering before writing it to cd.

Music Manic · Post by **Music Manic** » Sun Jul 28, 2002 11:42 pm

So the drivers carry the signal to the software,and the software deals with the signal,by using its audio engine.This is what creates a difference in sound quality,correct.

[quote]
On 2002-07-28 20:09, DJATWORK wrote:
I see a lot of confusion here...

Its like a way to transport the digital signal from the card to the software, like a bridge or something like that. That has nothing to do with the sound.
The information that travel on those drivers, has nothing to do with the way that drivers "transports" it.

at0m · Post by **at0m** » Sun Jul 28, 2002 11:53 pm

Then why do we have 16/24/32/32flt versions of drivers?

DJATWORK · Post by **DJATWORK** » Mon Jul 29, 2002 2:23 am

Thanks for your thanks!
I´m new here, and it so cool to know that I can "learn" a lot and "give" to other my knowledge!

There are a lot of things to answer...

This Forum is great... about the STUDY forum, where is it?

Thanks!!

DJATWORK

DJATWORK · Post by **DJATWORK** » Mon Jul 29, 2002 2:49 am

Question: to burn that file to cd you have to change to 16 bit.
As you say there is dithering used too, and far as I know this is part of the process, you can't change this.
At the end you allways use dithering twice.
Isn't this a problem?

Let There Be Music!

First at all, Dithering ocurr ALWAYS when changing the bit depth from bigger to smaller quantity.

You don´t want to change that because if that no happens, you can hear that distortion during low signal, for example in Classic music, or in a fade out tale.

I think that you are seeing Dithering like something bad, but is the oposite.

A thing that wasn´t say here until now, is that there are lots of Dithering ALgorithms.
The manufacturers don´t give to match information about each one and how exactly they act introducing frequencies, but all of them make the good as they can to make that noise signal "invisible", the technical therm is "masked".

For example the well heard algorithm "Apogee UV22" is based in a noise signal of 22 Khz (the human years limit is on 20 Khz) so theorically that noise is unsignificant for any of us. THere are another signals too (lower than the principal in 22 Khz) and there appear subharmonycs too, but anyway, in any case, the dithering is dificulty heard.

Think this: if in a recording, in a low level moment you "hear" the Dither, if the dither where not there, you where hearing a distortion in your instrument, or final mixing.

Note that sometimes you can hear any "noise" but could be produced by the cables, amplifier, mixing console, etc etc, not necesarily is Dither.

Well...

DJATWORK

DJATWORK · Post by **DJATWORK** » Mon Jul 29, 2002 3:05 am

Atomic:
[/quote]

How comes you'd dither twice? After mastering/mixdown, you can still export the audio from for example Cubase and dither only once during final mix 16bit wav writing?

As I understood from DJATWORK, 16bit Wav Dests don't incorporate dithering. This would be nice tho, if it were done automatically. Cos it's quite obvious that one wants to record 16bit. Or we'd have selected 24bit wav drivers. But then still, all dsp works is 32bit.
On the other hand I've almost never used dithering and my 16/24bit recordings only crackled a couple of times. How comes?

I don't see when or where to actually put a dithering device in my setup. I know on mixdown to inside the multi-tracker, I'd apply dithering before writing it to cd.

[/quote]

Again, Dither is a process that take place EVERY time the Bit Depth of a signal is reduced. Is important to understand that, but it does not necesarily afect in a audible way the sound.

I assume that the Pulsar MIxer (all of them) works completely in 32 Bits. (since the inputs, faders, efects, Eq, etc). So when you enter the signal nothing happens, BUT when you get it out of the mixer, to a smaller Bith Depth destination, Dither take place AUTOMATICALLY.

Thas why you "had never used dithering", its simply part of a process that you doesn´t necesarily know that its happening.

If you MIX signals (for example Export a proyect to a MIxdown file) and the final mix file has the same bith depth than the original traks of the proyect, there is not need of any conversion, and then there is no need to ad dither.

There is a confusion thing, now is like a "fashion" that several PlugIns include Dithering Options. I really don´t understand that... its like redundant, but I think that is because the PlugIn make the processes in 24 or 32 bit, and offers that option to convert the file in the outputagain to the original format.

I really think that is a marketing thing, because the 99% of the people that use that plugins don´t really understand exactly what is that for (i´m on that 99%)...

Again, don´t worry, the dither take place "automatically" if it´s needed, and we´ll never know it was added.

DJATWORK

DJATWORK · Post by **DJATWORK** » Mon Jul 29, 2002 3:16 am

On 2002-07-29 00:42, Music Manic wrote:
So the drivers carry the signal to the software,and the software deals with the signal,by using its audio engine.This is what creates a difference in sound quality,correct.

NO. There is NOT sound quality difference between an ASIO and a WAVE driver. The diference is the channel quantity, SPEED, etc... and now for example there is a "new" driver from Microsoft for Windows, the "WDM", Windows Driver Model. This lets the software (SONAR in Example) transfer audio data in les than 4 miliseconds.

If you REALLY found a quality difference between one and the other, you should e-mail CW and ask for explainations.

It´s like buying a new Digital Mixer and found that the inputs 1 to 8 are lower in quality than 9 to 16.

That CAN´T happen, and if you really found that, is a big trouble for CreamWare and the Driver developer (ASIO= Steinberg - EASI= Emagic - GSIF= NEmesys, now Tascam, etc).

DJATWORK

DJATWORK · Post by **DJATWORK** » Mon Jul 29, 2002 3:23 am

On 2002-07-29 00:53, at0mic wrote:
Then why do we have 16/24/32/32flt versions of drivers?

That´s to allow the aplication (Cubase, Logic, Sound Forge) to transfer signals with that different Bit Depth.

The most convenient is to let a "headroom" for make the process. Pulsar make the audio process at 32 Bits, so you should work in 24 Bits as longer. Anyway, it´s your choice.

There are another cards that make the internal process in 48 Bits and Y had heard from some that works in 64 Bits.
It´s too match for me. sincerelly it has no sense.

The biggest joke around all of this, is that all this "new" bith depths are made watching the future of DVD producing...
And DVD encodes the music in a very very similar way than MP3 does, so... we are recodring, processing and mixing in a super spectacular quality, and then the Final Mix is compressed 10 to 14 times to fit into an MP3 file... I finally prefer 16 Bits or 24 as longer and then CD Mastering...

DJATWORK

Eurocide · Post by **Eurocide** » Mon Jul 29, 2002 5:55 am

Thank you Djatwork for solving the confusion.

But I've got another questions regarding bit depths and ASIO drivers.

Cubase SX 1.02 has a display problem with our Creamware 32Float device: the Masterlevels are always maxed out.
When using "only" 32bit ASIO this display error is gone.
I always record in Cubase with its 32bit TrueTape function.

Question 1: is there a quality difference when using 32bit ASIO (not 32 Float) when recording with TrueTape?

Question 2: When I analyze the recorded .wav-file Cubase says in its "statistics" that the estimated bit depth is 14 oder 15, sometimes 25 bits, although there are wide dynamic changes and I set the recording levels correctly.
Confusing for me.

More confusing: When zipping such a 32bit-file (which doesn't contain silent parts!) with ZIP the file size is reduced to about 50%. For me this would mean, that a LOT of audio data contains ZERO.
So i suppose, that 16bit are not used at all during the recording. Right or not?

My A/D converter is a A16 (1st version) with internal 20bit resolution.

Thanks in advance,
Eurocide-HQ.

at0m · Post by **at0m** » Mon Jul 29, 2002 6:27 am

(remark) Zipping .wav files reduces their audio quality, look for a small program 'wavzip' or something that takes care of correct conversion to .zip file. The zipfile made w wavzip can be unzipped w any zip program.

A little like usinf .sfark for compression of soundfonts.

Music Manic · Post by **Music Manic** » Mon Jul 29, 2002 9:50 am

Sorry you missed my point.
What do the audio engines of programs like Cubase/Logic etc do with the audio drivers when they receive signal!
Thx for info BTW

DJATWORK · Post by **DJATWORK** » Tue Jul 30, 2002 3:42 am

On 2002-07-29 06:55, Eurocide wrote:

Question 1: is there a quality difference when using 32bit ASIO (not 32 Float) when recording with TrueTape?

DJW: There is always a techincal diference when recording with 32 Bits or 32 Bits Float. I actually use Nuendo, and in Nuendo the TrueTape does nothing to do with the Bit Depth or the sound quality, it´s just the way that Nuendo manage the new files when recording, and stuff like that... Cubase SX should be the same, or not??? (pressing F1 you got an ascellent HELP, and you can search TrueTape 32, and see what it says. THen Tell us!!)

Question 2: When I analyze the recorded .wav-file Cubase says in its "statistics" that the estimated bit depth is 14 oder 15, sometimes 25 bits.

DJW: That should be an "average" value, of the Bit using compared with db. Every single Bit gives you 6 db of dynamic range, so 16 Bits: 96db, 24 Bits 144db, etc...

That maxium value can correspond to the converter that you are using. Make this experiment... Make a SQUARE waveform in Sound Forge, route it out of your PC by any output, make it IN from A16 Converter only with 1 or 2 db to clip, then connect the AdAT input DIRECT to an ASIO 24 Bits, and record that mono signal in Cubase. THen see what happens with the Bith Depth average value... Then with that data we´ll think how it works..