This is dithering – and so you use it

By | April 6, 2021


In the beginning…
When we record digitally, we have the choice to choose which quality of sound we want to use. The quality is determined partly by bit depth, usually 16 or 24 bits and partly by the sampling frequency, usually 44.1 kilohertz. All in all, this is called resolution in the world of recording – the higher the resolution (just like in the world of image processing) the better, at least on paper.

Exactly which resolution to use is, however, more unclear and has given rise to countless discussions in recent decades. On the one hand, there are those who advocate as high a resolution as possible, 24 bits / 96 kilohertz or more, while others believe that 16 or 24 bits / 44.1 kilohertz is more than enough.

However, it is not obvious that higher sampling rates than 44.1 kilohertz actually sound better. The reason for this is the so-called Nyquist-Shannon sampling theorem, which can be explained in a somewhat simplified way by the fact that humans do not perceive frequencies higher than 20 kilohertz and then 44.1 kilohertz should be more than sufficient. Despite this, many claim that higher sampling frequencies still provide better sound quality and point to various more or less scientific theories, as well as their own golden ears.

The pieces fall into place
Regarding bit depth, pretty much everyone agrees that 16 bits is good, 24 bits are better and that everything above this is superfluous. It can be said that each “bit” corresponds to 6 decibels of dynamic range. Thus, a 16-bit signal has a range of 96 dB (8×16 = 96) and 24-bit signal 144 decibels. With the above in mind, I therefore always recommend the resolution 24 bits / 44.1 kilohertz when recording and mixing. The exception is if you work with film music, where the tradition is 48 kilohertz.

Dithering – CD and streaming
However, the problem starts when you need to convert your pre-mixed music from 24-bit to 16-bit, for printing CDs (which only allow 16-bit at 44.1 kilohertz) or various streaming services such as Spotify and iTunes, depending on the upload service you use. Unfortunately, it is not as simple as mixing in 24-bits and then exporting in 16-bits, because the reduction of eight bits (24-8 = 16) means that the dynamic range is reduced, but in a way that is not transparent, that is, inaudible. A kind of (quantization) distortion of the sound occurs.

Due to this, dither is one of the most misunderstood areas in the digital audio chain. Many people simply do not know what it is, when to use it and above all not how. What dithering does is completely add a barely audible noise to the process to mask the distortion (distortion) that occurs in the 24- to 16-bit conversion itself.

But is it really heard if you ignore dithering? Maybe not for an untrained ear. But just like the more or less subtle differences between a wav file and an mp3 file, it is guaranteed to be there, even at normal listening volume. This so-called quantization distortion can then be perceived as thin, grainy and with a worse stereo image. In comparison, a properly performed dithering sounds fuller, warmer, smoother and more analog.

Dithering must be applied absolutely last in the signal chain before final export and is built into most limiters and music programs. Often it is as simple as clicking in 16-bit and settling for it. However, my personal dithering plug is independent and is called Kazrog master Dither. As easy to use as it is to start your music program, well almost. And so it is with that thing.

Feel free to ask questions or comment on the article in the comment field below, and we will spin the topic together. Or if you prefer to discuss mixing in the “Mixing and mastering” section of the Studios forum here!

Fredagstipset is a recurring series where Studio writer Jon Rinneby shares tips every Friday in, among other things, recording and mixing. Here you will find all Friday tips.