What Is Dither?

Dither is a deliberate application of low-level noise added to an audio signal prior to quantization in either analog-to-digital (ADC) or digital-to-analog (DAC) conversion. While adding noise to a signal might seem counterproductive, it actually improves perceived audio fidelity by randomizing the errors introduced by quantization. Without dither, the deterministic nature of quantization errors produces harmonic distortion and correlated artifacts that are far more audible than the benign noise floor added by dither.

The concept originated in the early days of digital audio and computer graphics, where engineers discovered that adding a small amount of random noise before rounding prevented the formation of structured patterns and tones. In audio, this technique transforms the otherwise harsh, signal-correlated quantization distortion into a constant, low-level hiss that the human ear can easily ignore. The key is that the noise level is carefully calibrated to be no louder than necessary to decorrelate the errors, typically around 0.5 to 1 least significant bit (LSB) of amplitude.

Why Dither Matters in ADC and DAC Conversion

Quantization is the process of mapping a continuous range of values to a finite set of discrete levels. Every time a signal is quantized—whether during an analog-to-digital conversion in a recording interface, or when reducing bit depth in a digital audio workstation (DAW)—an error is introduced equal to the difference between the true analog value and the nearest digital representation. In a undithered system, this error correlates strongly with the input signal, producing harmonic distortion and modulation noise that is especially noticeable in quiet passages or with low-level signals.

Dither ensures that the quantization error becomes independent of the input signal. By adding a known dither signal before quantization, the error is randomized, spreading its energy across the entire frequency spectrum as a uniform noise floor. This noise floor is far less obtrusive than the tonal artifacts that would otherwise arise. The practice is considered essential in any professional audio workflow, from recording and mixing to mastering and final delivery. Without proper dithering, even a seemingly pristine 24-bit recording can sound gritty or harsh after conversion to 16-bit CD quality.

Quantization Errors in Practice

Consider a quiet passage of music that drops to just a few digital steps above silence. In an undithered system, the signal may jump discontinuously between levels, producing a coarse, grainy distortion known as "granulation." Dither smooths this by ensuring that even signals below the LSB are represented as a random variation around the nearest level, preserving the subtle nuances of the original waveform. This is why dither is said to "linearize" the quantization process, effectively extending the dynamic range below the LSB.

Types of Dither

Not all dither is created equal. The probability distribution and spectral shape of the dither noise have a significant impact on its effectiveness and audibility. The three most common types are rectangular, triangular, and shaped dither. Each is suited to different scenarios and technical requirements.

Rectangular Dither (RPDF)

Rectangular dither uses a uniform probability distribution function (PDF) over a range of ±0.5 LSB. It is the simplest form to generate but also the least effective. Rectangular dither still leaves some correlation between the quantization error and the signal, meaning it does not fully eliminate distortion artifacts. It is rarely used in professional audio today except in very basic digital systems.

Triangular Dither (TPDF)

Triangular PDF dither is created by summing two independent rectangular noise sources, resulting in a triangular distribution over ±1 LSB. This type of dither completely decorrelates the quantization error from the signal, achieving true linearization of the quantization process. TPDF is the standard recommendation for most audio applications, especially when reducing bit depth from 24-bit to 16-bit. It adds only 1/3 more noise power than rectangular dither but offers significantly better performance. Most DAWs implement TPDF as their default dither algorithm.

Shaped Dither (Noise-Shaped Dither)

Noise-shaped dither goes a step further by manipulating the frequency content of the added noise to make it less audible to human hearing. Since our ears are most sensitive to frequencies in the 2–5 kHz region, shaped dither pushes the noise energy into the higher frequencies (above 15 kHz) or, in some cases, into very low frequencies where hearing is less acute. This technique can dramatically reduce the perceived noise floor, allowing for bit-depth reduction with minimal audible side effects. However, shaped dither is more computationally intensive and must be used carefully to avoid introducing pre-echo or other artifacts in the time domain. It is commonly employed in high-end mastering tools and dedicated dither plugins.

Flat vs. Shaped Dither

Flat dither (TPDF) maintains a constant noise power across all frequencies. Shaped dither uses a filter to modify the spectral distribution. The choice between them depends on the listening environment and playback system. For critical listening on full-range monitors, shaped dither can be advantageous, while flat dither might be safer for consumer devices that could play back through limited bandwidth speakers.

How to Apply Dither Correctly

Proper dither application requires an understanding of the signal chain and careful timing. Dither should only be applied once, at the final stage of the process, immediately before converting to a lower bit depth or output format. Applying dither multiple times will accumulate noise unnecessarily. Here are the recommended steps:

  1. Complete all processing first: All mixing, EQ, compression, and effects should be finished before dither is considered. Dither is added after any gain changes that might alter the bit depth.
  2. Set the output bit depth: Determine whether you are exporting to 16-bit (CD), 24-bit (standard for distribution), or any other fixed bit depth. For 24-bit output, dither is typically unnecessary because the noise floor is already extremely low.
  3. Choose the dither type: For most projects, TPDF (triangular) dither is the safe, proven choice. For mastering engineers working with 16-bit delivery, noise-shaped dither can provide a slight advantage, provided the shaping filter is well-designed.
  4. Apply at the last plugin or export stage: In many DAWs, dither is included as an option in the export dialog. If you use a dedicated dither plugin, place it last in the master channel. Ensure no other processing occurs after it.
  5. Verify the level: The dither noise should peak at around 1–2 LSB. Most modern DAWs and plugins automatically set the correct level. Over-dithering can raise the noise floor unnecessarily; under-dithering fails to eliminate distortion.
  6. Avoid double dithering: If your DAW offers dither at export and you also have a dither plugin active, disable one. Double dithering adds unnecessary noise and can degrade signal quality.

Dither in Recording vs. Mastering

During recording, ADCs typically apply dither internally, and engineers rarely need to think about it. However, when transferring analog tape to digital, or when using external converters, it can be beneficial to understand the dither algorithm your converter employs. In mastering, dither becomes a conscious decision: every bit-depth reduction (e.g., from 32-bit float to 24-bit, or 24-bit to 16-bit) requires dither. Mastering engineers often use specialized dither plugins with adjustable noise shaping to preserve as much dynamic range as possible while meeting delivery format constraints.

Common Myths and Misconceptions

Myth 1: Dither adds audible hiss. Properly applied dither adds noise at a level approximately 1–2 LSB, which for 16-bit audio corresponds to about -92 dBFS. This is far below the ambient noise in most listening environments and is completely masked by the program material in typical playback. In 24-bit systems, dither noise is effectively inaudible.

Myth 2: Dither is only for mastering. While mastering is the most common place where engineers manually apply dither, any digital audio conversion benefits from dither. ADCs use dither internally, and DAWs often apply it silently when rendering to a lower bit depth. Understanding dither helps in troubleshooting issues like low-level distortion or harshness in digital recordings.

Myth 3: You never need dither when staying at the same bit depth. This is generally true for integer formats (e.g., 24-bit to 24-bit) as long as no gain changes occur. But when processing in floating point (32-bit or 64-bit) and then converting to an integer format for output, dither should be applied because the conversion introduces quantization.

Myth 4: Noise-shaped dither always sounds better. Noise shaping can reduce perceived noise but may cause issues in certain playback systems, especially those with poor high-frequency response or digital amplifiers that produce intermodulation distortion from high-frequency content. Flat TPDF dither is more universally safe.

Practical Examples of Dither Usage

Example 1: 24-bit Recording to 16-bit CD

Most professional recordings are captured at 24-bit or 32-bit float. When producing a CD, the final master must be 16-bit, 44.1 kHz. The mastering engineer will apply dither (typically TPDF or noise-shaped) during the conversion. This ensures that the low-level details captured in the original high bit depth are preserved as a graceful noise floor rather than harsh quantization distortion. The result is a CD that sounds smooth and natural even in quiet passages.

Example 2: Mixing in 32-bit Float and Exporting to 24-bit

Many DAWs operate internally at 32-bit floating point, which has a huge dynamic range. When exporting a final mix to 24-bit for distribution, dither should be applied. Although 24-bit has a theoretical dynamic range of 144 dB, the act of truncating from float to integer introduces a new quantization step. Applying TPDF dither at this stage ensures the mix remains transparent.

Example 3: Sample Rate Conversion

When changing sample rates (e.g., 96 kHz to 44.1 kHz), dither is often applied as part of the SRC process to prevent artifacts. Some SRC algorithms include built-in dither; others require the user to add it after the conversion. Always check the documentation of your SRC tool.

Conclusion

Dither is a fundamental technique in digital audio that directly impacts the perceived quality of recorded and reproduced sound. By adding a carefully controlled amount of noise before quantization, dither eliminates the harmonic distortion and modulation noise that would otherwise corrupt low-level signals and quiet passages. Understanding the different types of dither—rectangular, triangular, and noise-shaped—and knowing when and how to apply them is essential for anyone working in audio production, from hobbyists to seasoned mastering engineers. Always apply dither as the final step in the signal chain, use TPDF for general purposes, and consider noise shaping only when the playback system is well-understood. For further reading, explore resources from Wikipedia’s comprehensive article on dither, Audio Science Review’s technical explanation, or the Sound On Sound article on demystifying dither. Proper dither usage ensures that your digital audio preserves the integrity of the original performance, delivering a listening experience free from unnecessary artifacts.