What is the purpose of dithering?
Dither is low level noise which is intentionally added to digital audio when ever it is converted to a lower bit depth. Dither noise is introduced to achieve two things:
- It preserves the information that would otherwise be lost when you convert to a lower bit depth.
- It eliminates the distortion that occurs as a result of lowering the bit depth.
Eliminating quantization distortion
Bit depth defines the number of values available to represent the amplitude of the samples which make up a digital audio file. When you convert an audio file to a lower bit depth, you reduce the number of values that are available to represent the amplitude. As such, the bit value of some samples will be ‘quantized’. This means that they are rounded to the nearest available value. These rounding errors cause something called quantization distortion. Quantization distortion can introduce harmonics and other distortion types which are correlated to the original signal.
Dither noise is generated by switching the last bit in the data on and off at random. This creates a smooth, low level white noise/hiss. Switching this bit on and off at random randomizes the quantization error. As such, it decorrelates the errors from the original signal, in turn eliminating quantization distortion.
At the kind of bit depths that we are generally working with, i.e. 16 and 24 bit, this distortion is extremely quiet. It becomes much more apparent at lower bit depths. But even at 16 bit resolution, it is possible to hear quantization distortion. This usually happens at times when the music that would otherwise mask it is very quiet. For example, You might hear quantization distortion during a fade out or at the end of a reverb tail. In fact, even if you don’t hear the distortion itself, its presence can make your audio sound harsh. As such, it’s important to use dithering to eliminate quantization distortion.
Retaining low level detail
In addition to using dither to prevent quantization distortion, it is also used to preserve low level details which would otherwise be lost when you convert to a lower bit depth. When you convert a digital file to a lower bit depth, you ‘truncate’ (i.e. shorten), the number of bits being used to represent the digital audio.
Say for example you recorded your song at 24 bit resolution, but you need to bounce your song at 16 bit resolution so that it can be burned to an audio CD. If dithering was not applied before you converted to the lower bit depth, then the information in bits 17 to 24 would be lost. This would result in the loss of very low level details in your audio. However, by applying dither, the low level details from the truncated bits are effectively added to the dither noise which is then encoded into the 16 bit version. As such, the low level details are retained.
When should I use dithering?
When to apply dither is a hotly debated topic amongst producers. Different published music production books and resources give different answers…
The approach which makes the most sense to me, and the one which I follow, is one which I learned in this video by mastering engineer Ian Shepherd. In his video, Ian states that you should apply dither when ever you bounce to a 16 bit or a 24 bit file.
The reason for this is as follows… Quantization distortion and the loss of low level detail as a result of truncation both occur when ever you convert audio to a lower bit depth. All modern DAW’s process your audio internally at 32 bit floating point resolution. So, whenever you bounce to a 16 bit or 24 bit file, you are lowering the bit depth. As such, dither is required. That’s the case even if it means applying dither to a song more than once, i.e. once when you bounce the mix for mastering and then again when you bounce the final mastered track for release. The only time that you don’t need to apply dither is if you’re bouncing a 32 bit floating point file.
How do I add dither?
There are a few different ways to apply dither. Most mastering plugins give you the ability to apply dither. The bounce dialogue in some DAW’s also gives you the option to add dither. Some limiter plugins have a dither option. Or alternatively, you can use a dedicated plugin which exists solely for dithering.
There are a couple of rules to follow when you apply dither to your session. Firstly, if you’re adding dither via a plugin, be sure to add it as the last plugin on your master fader. There should be no other processing after it. Also, you only need to apply dither once to each session. So, add dither as the last plugin, only once per session, when ever you bounce to a fixed point bit depth.
Many of the options that allow you to apply dither will give you the option to use ‘noise shaping’. Noise shaping focuses the frequency content of the dither into areas of the frequency spectrum which are harder for us to hear. You can use this as a means of making dither noise more transparent.
At first, dither can seem pretty confusing. Not least of all because it involves intentionally adding noise to our audio, which is usually something that we would look to avoid. Hopefully, this lesson has given you an understanding as to why adding dither noise is actually very beneficial. When you lower the bit depth of an audio file, you introduce quantization distortion. You also lose low level details. Through the use of dither, you can not only eliminate the distortion, but also preserve the low level details too. As such, an audio file with dither applied should sound the same at a lower bit depth as it did at a higher bit depth, but with the addition of very low level dither noise.
Are you sure to always apply dither when you reduce the bit depth of your audio files? At what points do you typically need to reduce the bit depth of your audio files when you produce music? Leave your thoughts in the comment box below.
Get the best results from EQ, compression and vocals with the FREE 'Home Studio Bundle'