Dither

F

fishkarma

New member
heres what i found on the net
but i dont really understand:

Dither—Dithering is a mathematical process where a random noise is added to the least significant bit of a digital word. With very low level signals, the quantization error becomes correlated to the signal level. This creates a measurable amount of distortion. By adding dither, the correlation between the signal level and the quantization error is cancelled, allowing the digital system to encode amplitudes smaller than the least significant bit. If you change the word size as a signal passes from one digital system to another, being able to add dither allows you to maintain a high quality signal.

does this mean that a random bit of noise is added to the least important part of a sample to stop distortion at the signal level? and thus allowing you to sample at differnt bit depths as you change system?
how does a quantization error create distortion? what is a quantization error?


please help i feel its important to understand everything thats thrown at me even if its not right away!
 
fishkarma said:
heres what i found on the net
but i dont really understand:

Dither—Dithering is a mathematical process where a random noise is added to the least significant bit of a digital word. With very low level signals, the quantization error becomes correlated to the signal level. This creates a measurable amount of distortion. By adding dither, the correlation between the signal level and the quantization error is cancelled, allowing the digital system to encode amplitudes smaller than the least significant bit. If you change the word size as a signal passes from one digital system to another, being able to add dither allows you to maintain a high quality signal.

does this mean that a random bit of noise is added to the least important part of a sample to stop distortion at the signal level? and thus allowing you to sample at differnt bit depths as you change system?
how does a quantization error create distortion? what is a quantization error?


please help i feel its important to understand everything thats thrown at me even if its not right away!
Click to expand...


I'm not going to attempt to go heavy into the science of it, as I don't know too much about it. It has to do with your bit depth. Say you record at 24 bit. When you go to a 16 bit cd, there will be some "distorting" of the audio because you're down converting it. IF you use dither, it'll make this sound much smoother. So, in all practicality, if you're down converting to a lower bit rate. Use it. You'll be happy.
 
Josh,

> Dither—Dithering is a mathematical process where a random noise is added <

Right, it's a tiny amount of noise. So tiny, in fact, that nobody can possibly hear the "benefit" unless you turn up the volume 40 dB beyond ear-splitting during a song fadeout or reverb tail or whatever.

Dither was invented by people who sell gear to distract you from what really matters - having accurate monitoring in a good sounding room.

--Ethan
 
Unless you plan on writing a dithering algorithm I wouldn't get too caught up in the details. Most DAWs come with decent dithering algorithms that you probably can't hear the difference between anyway.
 
fishkarma said:
what is a quantization error?
Click to expand...
I'll take a shot at explaining this.

An analog signal has an infinite number of possible values. A digital signal has a finite number of possible values. When mapping an analog signal to a digital representation (the process called A/D conversion), the signal level at each point in time must be mapped to the nearest possible sample value. The difference between the actual analog value and the resulting sample value is the quantization error.

Suppose you're sampling a 1 volt peak-to-peak sine wave signal using one bit of resolution. With one bit, each sample can have a value of zero or one. Let's say that the signal's voltage goes from -0.5V to +0.5V. Each time you sample the signal, you note the voltage, and record a zero if the voltage is less than 0V, or a one if the voltage is greater than 0V. If the voltage is 0V, just repeat the value of the previous sample.

Can you see that the result would look a lot like a square wave? In fact, you would not be able to distinguish our sine wave samples from a sampled square wave at the same frequency. A square wave has much richer harmonic content than a sine wave. We have created extra harmonic content through a deficient sampling process. This extra harmonic content is quantization distortion, and it's a result of our quantization error.

In reality, most of the quantization distortion is removed in the D/A conversion process, by the reconstruction filter. Dither allows you to preserve some of the information that is lost in the process of reducing bit depth. But, as Ethan pointed out, it can be pretty subtle. For the most part, you don't need to worry about it.

If your software offers the option to dither, use it. But only do it once, when you're reducing bit depth as the last step in producing a wave file for a CD.

If I've missed anything important in all this, I'm sure someone will be kind enough to point it out.

Don
 
Cheers DonF that actually makes sense now and is probably the best explaniation ive had to date!

edit
-i suppose then i should either dither with Dither.dpm or Maxim.dpm's dither then (not both or none)?

while im at it what damage can applying it twice do?
 
The thing to remember when reading up on dither is that when they refer to "noise", they mean it in an information sense, not necessarly in an audio sense.

It's not a question of adding low-level hiss which can barely be heard, its a question of randomizing the least signifigant bit information so that it masks otherwise more "colder" incorrect information caused by the simple chopping off of the extra bits.

G.
 
Ethan Winer said:
Dither was invented by people who sell gear to distract you from what really matters - having accurate monitoring in a good sounding room.

--Ethan
www.realtraps.com
The acoustic treatment experts
Click to expand...

maybe acoustical treatment was invented by people trying to distract us from buying the best dither algorithm available..... :p

{{kidding}}


+1 for Glen's point too.

fishkarma-
it sounds like you use Pro Tools (and if I remember right I've seen you discussing it around here before)...just create a master fader and add dither on the final insert before you bounce. Do this for a 16 or 24 bit session.
 
bennychico11 said:
maybe acoustical treatment was invented by people trying to distract us from buying the best dither algorithm available..... :p

{{kidding}}


+1 for Glen's point too.

fishkarma-
it sounds like you use Pro Tools (and if I remember right I've seen you discussing it around here before)...just create a master fader and add dither on the final insert before you bounce. Do this for a 16 or 24 bit session.
Click to expand...

ROFL hahaha
 
bennychico11 said:
fishkarma-
it sounds like you use Pro Tools (and if I remember right I've seen you discussing it around here before)...just create a master fader and add dither on the final insert before you bounce. Do this for a 16 or 24 bit session.
Click to expand...

yeah i got the 002 desk and a mac. lucky guess :p

i know you can put it on the master fader in fact the manual recommends it. i was wondering though if theres an advantage to using maxims dither instead would one be worse than the other?
as far as i know maxim only comes with the higher end digi stuff or you can buy it, where as the dither plug comes with pt. so what im asking is if the maxim algorithm is better than the regular dither?

cheers all who have helped on this one
 
Ethan Winer said:
Josh,

Dither was invented by people who sell gear to distract you from what really matters - having accurate monitoring in a good sounding room.

--Ethan
Click to expand...

Dither was invented for the exact reason the original poster quoted. Ethan if you don't understand these concepts you shouldn't respond with such a ridiculous answer. Perhaps you are trying to sell acoustical treatment?

Here's a rough analogy that helped me with the reasoning behind dither:

Spread your fingers out and put them in front of your computer monitor, notice how some of the screen is not visible (akin to quantization distortion). Now keeping your fingers spread move your hand left and right very quickly in front of the monitor. The areas that were not visible can now be seen (dither).

Dither is an essential part of quality A/D conversion and processing. Without it, digital sounds "colder" particularly at 16 bit word lengths. IMHO there is definitely an audible difference between the different type of dither. The choice is dependent on the sound that you want in the final product.

See the following link for examples of this:
http://www.24-96.net/dither/

Is dither (or the various types) a night and day difference? No. Is it important for a high quality product? yes.
 
Last edited:
bennychico11 said:
they're just different sounding. It's up to you which one sounds smoother

more interesting links:
http://www.earlevel.com/Digital Audio/Dither.html
http://www.24-96.net/dither/
Click to expand...

Sorry Benny,

I read your links after I posted, good stuff ...

Other good links:
http://www.cadenzarecording.com/Dither.html

http://www.national.com/an/AN/AN-804.pdf

Also remember that quantization error is cumulative with successive processing. While one process may not be extremely audible, the combination of volume changes, EQ, compression, etc. can add up making this type of distortion a definite issue if not dealt with.
 
Last edited:
You must log in or register to reply here.
Back
Top