We’ll try to ease your audio production job by explaining what is dithering in audio processing.
It turns out that dithering in audio processing is not much of a concern if you apply the right settings. You should use dithering every time when reduction of a bit depth occurs, but keep in mind that your DAW might use 32-bit FP calculations. That means that even if you export a file with the same bit depth that your project had, you should still apply dither. Unless you are exporting a 32-bit FP file, this won’t make much sense unless further editing would be applied. Although noise-shaping is completely optional, you may as well still use it to reduce the overall noise level.
The chances are that if you stumbled upon this article, you have heard about dithering before and at least have a rough notion of what it does. The problem is that a proper explanation of dithering would involve a lot of math, and we have no intention of boring you. So in this particular article, we will try to give you the simplest explanation possible. So you could keep up your good work in these dire times without overloading yourself with mathematical equations.
Of course, if you are eager to gain as much knowledge as possible, we encourage you to investigate this subject further, and a few links will be left here and there to help you with that endeavor. But the best thing about our profession is that we can still achieve great results with very little background knowledge. So if you are a very practical person and value your time foremost, just read the next paragraph and leave it at that.
When do we dither?
You might be delighted to know that applying dithering is as simple as ticking a box on the menu. This process does not require any additional skills or manual labor. But the tricky part is to know when to tick the said box and when not to. You should definitely apply dither every time you are exporting your final mixdown unless 32-bit floating-point calculations were used, and this is also the bit depth of the mixdown that you are exporting. Only then you can leave dithering to a mastering engineer.
Even if your project’s bit depth is set to 24 bits and you export with the same bit depth, you should still apply dither. For internal processing, most modern DAWs use a bit depth that is higher than the one that you’ve set for the project. And dithering is used every time when bit depth is lowered to prevent noise from quantization error. So if you’ve mixed your project in 24 bits, as most audio engineers do, and intend to export it to 16 bits, you should apply dither with the addition of noise-shaping.
What is dithering?
To put it simply, dithering is a process of replacing noise from the quantization error with random noise. Usually, quantization noise is a very predictable form of distortion which could be removed with a simple algorithm. But just removing it would not fix the problem since human ears are very sensitive in terms of distinguishing different frequencies. So in place of quantization noise, the random noise is put, and due to its randomness, it’s almost undetectable by the human ear. Or at least, we do not pay much attention to it.
Essentially, dithering is a mathematical algorithm that is implemented within your DAW. Dither is applied every time when a signal is quantized, and most of the time, the process is fully automatic. The signal is quantized every time when the number of bits within it is lowered. If the reduction of bits is rather drastic, for example, lowering bit depth from 32 FP to 16, random noise from dithering could be noticeable. So to reduce this effect, a process that is called noise shaping is implemented.
Noise shaping is a filtering technique that is used to reduce the impact of the quantization noise or noise produced by dithering. It is established on the basic principles of human hearing. So essentially, noise shaping reduces the noise in frequency bands in which the person would notice it and increases the noise level in frequency bands where we would not perceive it. So, in other words, it is an algorithm that is used to make the noise less noticeable. And the question is, do you really have to use it?
Well, the simple answer would be – no, you do not. But, unfortunately, things aren’t that simple in the real world. Dither is used to convert the unpleasantly sounding distortion of quantization error to white noise, which could be perceived as a gentle hiss. Arguably, it is very similar to the hiss that is produced by reel-to-reel tape machines and other analog gear. It hardly could be used as a substitute for analog emulation VST plugins, but still, a little bit of noise could not hurt.
But the thing is, it is very hard to have pristine and clean recordings without any noise in the first place. So unless you are using very high-end virtual synthesizers exclusively, you still would end up with a number of different noises within your mix.
On top of that, a lot of different plugins with analog coloration could put you into a position where you can very much end up with a noisy mess of a mix. Even if it was your creative intention all along, generally, it’s not a very good idea to stack up various noises. So if you have a chance to reduce the perceived level of noise, you should definitely use it. This is where noise-shaping comes in handy.
Now when the things we actually should know are out of the way, we can dig a bit deeper into some theoretical knowledge about computing. Of course, bit depth is definitely a subject for a larger topic, so for now, we will try to be brief.
Bit depth displays how many bits are used to reconstruct a sample. When we are converting a continuous sound wave into a digital one, we have to limit the otherwise infinite number of points. So each time the sample is reconstructed, we deal with the number of points that are very approximated in relation to the original sound wave. This process of approximation is called quantization. When the point in the reconstructed sound wave does not match the point in the original one, we call this the quantization error.
Since bit depth is responsible for the resolution of the waveform and therefore its dynamic range, each time when the quantization error occurs, we will hear an unpleasant distortion. To get a better idea of how it sounds, just remember 8-bit electronic music that was widely used in 90’s video games. To avoid quantization errors or at least contain its numbers, we should use higher bit depths, but, although rather obsolete now, audio CD format required 16-bit audio. That means that in order to send your mixdown to be printed on a CD, you need to lower your bit depth significantly, which will result in a lot of quantization errors.
Freeze and Bounce
Although audio CD is not of concern anymore and streaming services and other music outlets are pretty happy with 24-bit audio, there are still some things that we need to keep in mind. So far, we’ve covered your general DAW settings and how you should export mixdowns. But there are very crucial internal processes. Although most plugins use the bit depth of your DAW, the actual processing is made with 32-bit floating-point calculations. This gives your audio processors an opportunity to be more precise and have headroom in terms of signal-to-noise ratio.
Actually, it is of no concern to us because after the processing is done, the signal is to be converted back to the bit depth set by the project. Essentially, you do not have to do anything about it since it is rather impossible to notice this process. But there’s a slight chance that you will get yourself in a bit of a pickle if you want to bounce or freeze individual tracks in your project. Usually, in such cases, dither is applied automatically. But generally, it’s a good idea to check your DAW settings and see if the dither is set to be applied when freezing tracks.