pisces7378
New member
What is dithering???
Track Rat
Just Your Average Sized Member
The short answer is, when recording at a longer word length (24 bit) and you need to get it down to 16 bit, you use dithering in the process. It's noise, sometimes shaped, added to the program material as it's being processed so information contained in the last 8 bits, such as decaying notes or reverb tails, isn't trucated, or cut off and lost. It basicly fools the 16 bit word length into recording a percieved higher resolution.
Slackmaster2K
Gone
It's all about resolution, or accuracy. Let's take a very simple example:
Take a simple decimal sample set that has an accuracy of 0.01 (1 100th):
5.55, 5.55, 5.55, 5.55, 5.55, 5.55
Now if we were to decrease the resolution of these samples to 0.1 (one tenth) by truncation (lopping off), we'd get:
5.5, 5.5, 5.5, 5.5, 5.5, 5.5
That's pretty close, but we're ignoring the fact that each sample was just as close to 5.6 as it was to 5.5!
Now if we were to use a "dither" technique, we might add a very small amount of noise to the original samples before truncating. Let's add noise that looks like this:
0.05, 0.00, 0.05, 0.00, 0.05, 0.00
If we add that to our original samples, we get:
5.60, 5.55, 5.60, 5.55, 5.60, 5.55
Now let's truncate these samples:
5.6, 5.5, 5.6, 5.5, 5.6, 5.5
There, that's a much more accurate representation of our original sample set. Now in the real life dither we're talking about, it's a thousand times more complex than this, but the principal is still essentially the same.
Another way I've heard dither described that's pretty cool is this:
Hold your hand up in front of your monitor with your fingers spread. You'll obviously notice that much of your monitor is blocked and you no longer have a good view of the screen. Now move your hand back and forth very rapidly. All the sudden you've got a much better view of what's on the screen, even though your hand is still blocking it!
Slackmaster 2000
Take a simple decimal sample set that has an accuracy of 0.01 (1 100th):
5.55, 5.55, 5.55, 5.55, 5.55, 5.55
Now if we were to decrease the resolution of these samples to 0.1 (one tenth) by truncation (lopping off), we'd get:
5.5, 5.5, 5.5, 5.5, 5.5, 5.5
That's pretty close, but we're ignoring the fact that each sample was just as close to 5.6 as it was to 5.5!
Now if we were to use a "dither" technique, we might add a very small amount of noise to the original samples before truncating. Let's add noise that looks like this:
0.05, 0.00, 0.05, 0.00, 0.05, 0.00
If we add that to our original samples, we get:
5.60, 5.55, 5.60, 5.55, 5.60, 5.55
Now let's truncate these samples:
5.6, 5.5, 5.6, 5.5, 5.6, 5.5
There, that's a much more accurate representation of our original sample set. Now in the real life dither we're talking about, it's a thousand times more complex than this, but the principal is still essentially the same.
Another way I've heard dither described that's pretty cool is this:
Hold your hand up in front of your monitor with your fingers spread. You'll obviously notice that much of your monitor is blocked and you no longer have a good view of the screen. Now move your hand back and forth very rapidly. All the sudden you've got a much better view of what's on the screen, even though your hand is still blocking it!
Slackmaster 2000
cordura21
New member
great explanation!