Digital Audio IV - Understanding Dither and Truncation

Digital Audio IV - Understanding Dither and Truncation

In previous posts, we followed the journey of sound from the analog to the digital world. We saw how ADC converters transform an electrical signal into a sequence of 1s and 0s, with a resolution defined by Bit Depth. But what happens when we need to change this digital resolution?

Imagine you mixed your song in 24-bit, but you need to deliver it in 16-bit for a CD or another format. This reduction process can create a subtle but destructive problem: truncation. And it's to solve this problem that one of the most specific (and important) tools in audio finishing comes into play: dither.

Today, we will demystify dither, understand what truncation is, and establish the golden rules of when and how to use this tool to ensure a professional finish for your tracks.

Truncation vs. Dither: The Problem and the Solution

First of all, let's understand the problem and its solution. They are opposite sides of the same coin.

Truncation (The Problem): This is the process of cutting (or "truncating") the bits of information from an audio file when reducing its Bit Depth. Think of a very high-resolution image that you need to save in a smaller version. You inevitably lose details. In audio, when going from 24-bit to 16-bit, the system simply discards the 8 "least important" bits of each sample. This generates quantization errors, which sound like a grainy and unpleasant digital distortion, especially noticeable in quiet parts of the music, like the tail of a reverb or fading piano notes.

Dither (The Solution): This is a very low-level noise, specifically designed to be added to an audio signal before truncation. It seems counterintuitive to add noise to improve the sound, right? But the magic lies in what this noise does: it masks the distortion artifacts of truncation. Instead of a digital distortion that is correlated with the audio, dither replaces this artifact with a smooth and constant noise floor, which is much more pleasant and natural to the human ear.

Where Dither Comes Into Play

Let's follow the process step by step to understand the practical application.

Scenario: You have finished your mix in 24-bit and need to export a final file in 16-bit.

Step 1: Truncation (The Digital "Decapitation") Without dither, the processor simply cuts off the extra 8 bits from each audio sample. The samples that previously had a precise location on the resolution grid are now forced to fit into the smaller 16-bit grid, causing rounding errors (quantization errors). The result is a loss of definition and the introduction of harsh distortion.

Step 2: Applying Dither (The Intelligent Noise) To avoid the problem above, we enable dither. It is the very last process in your mastering chain, applied exactly before the conversion to 16-bit. This added noise randomizes the quantization errors, transforming the digital distortion into a smooth, uniform hiss. Our brain perceives this hiss as something much more natural than the grainy distortion.

Types of Dither

Dither has evolved, and today we have different algorithms, many of which use a technique called "Noise Shaping." The goal of Noise Shaping is to "push" the energy of the dither noise into the frequency ranges where the human ear is least sensitive (usually above 15 kHz). This makes the noise even less perceptible.

The most common types you will find in plugins are:

• Type 1 (RPDF): This is a "flat" dither, without noise shaping. It is a safe and universal option, but the noise may be slightly more audible than other types.

• Type 2 (TPDF): Considered by many to be a slight improvement over Type 1, offering a slightly less perceptible noise.

• Noise Shaping (e.g., POW-r, UV22HR, etc.): These are more advanced and often proprietary algorithms (like those from Waves, iZotope, etc.). They move the noise energy to the high-frequency spectrum. They usually come in different intensities (e.g., Light, Medium, Aggressive), allowing you to choose how much "shaping" you want to apply, depending on the musical material.

When (and When NOT) to Use It

Here is the most important part. Using dither incorrectly can do more harm than good.

RULE 1: USE DITHER ONLY ONCE. Dither should be the last process applied to your audio and should only be done one time. Applying it multiple times simply accumulates unnecessary noise.

RULE 2: USE DITHER ONLY WHEN REDUCING THE BIT DEPTH. If you are working in a 24-bit session and are going to export a final file in 24-bit (to send for mastering, for example), DO NOT APPLY DITHER. The same goes for 32-bit float to 32-bit float. Dither is only necessary when "stepping down" the resolution (e.g., from 32 or 24-bit to 16-bit).

RECOMMENDATION: IN MOST CASES, LEAVE IT OFF. If you are a producer, musician, or mixing engineer, your work usually ends with exporting a high-resolution file (24-bit or 32-bit float) for the mastering engineer. In this case, you should never apply dither. Applying dither is the responsibility of the person creating the final master file for distribution. By delivering your mix without dither, you ensure that the mastering engineer receives the cleanest audio with the highest possible resolution. The same applies when exporting for the mix; it would just accumulate useless noise.

That's all for today!