What is audio dithering?
At its core, dithering is just noise, and noise, by its very nature, is random. In the early days of digital audio, some clever engineers realized they could take advantage of random noise. By mixing it with the quantized signal, they could add enough variation to preserve the original signal.
Dithering must be applied each time the bit depth is reduced. If there is a reduction from 32 bits to 24 bits, the strength of the dithering is almost irrelevant. If you reduce the size to 16 bits (or less), a low to medium dither with some noise shaping is probably best.
The key point here is that the dither noise should not be completely related to the quantized signal, sometimes called “decorrelated”. When this condition is met and the interpolation noise level is correct, any given input sample can be rounded up or down depending on the value of the input signal. This not only helps preserve the signal but virtually eliminates distortion associated with its frequency content.
Dithering is a solution to one of the fundamental problems with digital audio, so if we want to understand what it does, we must first understand the problem. The problem is the amplitude resolution or the precision with which we can measure the level of the signal using ones and zeros.
Errors inevitably occur when we try to measure an infinitely variable analog source (our digital sound) using a finite number of digital values (those ones and zeros). Sometimes the analog level will be slightly higher than the closest digital value, and in other cases, it will be lower.
It’s like trying to measure someone’s height with a tape measure that only reaches to the legs. In digital audio, this rounding error is known as quantization distortion.
Using a 32-bit floating-point system, as almost all modern audio editors do, makes distortion so low that you really don’t need to worry about it (by the way, when ordering the mix and master in our studio, you will receive material in this lossless format quality).
However, as bit depth decreases, the level of this distortion increases. As you get closer to 16 beats, the sound can start to get quite noticeable and unpleasant in reverb tails, delays, and other quiet areas. This is because the number of bits determines how many discrete values it can store.
Sometimes you don’t need dithering if you’re using a particular VST plugin, because it (the noise) heals itself. This may be technically true, but only in some specific cases. Believe it or not, not all noises are created equal. Therefore, if the audio plugin you use does not have a specific dither setting, you should add it if you plan to reduce the bit depth.
Different digital audio workstations (DAW) work in different ways, but most offer some method of capturing a complex chain of sound effects to a file. If you haven’t explored the possibilities of doing this in your DAW software, it might be time to set it up. It’s best to use 32-bit or 64-bit floating-point numbers whenever possible, but if you’re forced to use 24-bit numbers, check to see if there’s an option to enable dithering.
We hope this helps you understand why dithering is so important to digital audio, how and why it works, and when it should be applied.