Digital Audio for DJs: Sample Rate, Bit Depth, Nyquist and Aliasing.

Digital Audio for DJs: Sample Rate, Bit Depth, Nyquist and Aliasing

The nature of digital audio: sampling and quantisation

An analogue audio signal is a continuous variation of air pressure or electrical voltage. A computer or DJ player cannot store infinitely many values over time, so it digitises the waveform through two operations:

1. Sampling in time: Measuring the waveform amplitude at regular time intervals.
2. Quantisation in amplitude: Assigning each measurement to one of a finite number of numerical values.

The accuracy of those two dimensions is described by sample rate and bit depth.

---

Sample rate and the Nyquist-Shannon theorem

Sample rate is measured in hertz or kilohertz and states how many samples are taken each second. A rate of 44,100 Hz — 44.1 kHz means 44,100 measurements per second.

Nyquist-Shannon sampling theorem

The mathematical basis of sampled audio states:

To reconstruct a band-limited analogue signal from its samples, the sampling rate must be greater than twice the highest frequency present in that signal.

If the upper limit of human hearing is taken as roughly 20 kHz, the minimum theoretical rate needed to represent that band is above 40 kHz.

Why 44.1 kHz instead of 40 kHz?

The 44.1 kHz Red Book format places the Nyquist frequency at 22.05 kHz. That leaves a transition band between the nominal audible limit and Nyquist so practical anti-alias filters can attenuate ultrasonic content without cutting abruptly into the audible band.

---

What aliasing is and why it degrades sound

If the input contains frequencies above the Nyquist frequency and they are not sufficiently filtered, the sampled system cannot represent them correctly.

Instead of simply disappearing, those components fold back into the representable spectrum as false, inharmonic frequencies. This is aliasing.

Aliasing in a DJ workflow

Aliasing can be introduced not only during the original A/D conversion but also during real-time processing:

• Pitch shifting and time-stretching: Large tempo or pitch changes can create new spectral components. Poor algorithms or insufficient oversampling may produce folded high-frequency artefacts.
• Saturation and digital effects: Distortion, clipping, aggressive EQ and nonlinear emulations create harmonics. Components above Nyquist can reflect into the audible range unless the processor oversamples and filters appropriately.

A higher project sample rate moves Nyquist upward, but it does not automatically make an algorithm good. Filter design and processing quality still matter.

---

Bit depth and dynamic range

While sample rate limits the highest representable frequency, bit depth describes the number of amplitude values available for each sample and therefore the theoretical quantisation-noise floor and dynamic range.

Dynamic range is the span between the largest representable signal and the noise or error introduced by quantisation.

The 6 dB per bit rule

In ideal linear PCM, every additional bit doubles the number of available values and adds approximately 6.02 dB of theoretical dynamic range:

$$\text{Dynamic range (dB)} \approx 6.02 \times N$$

• 16-bit: $2^{16} = 65,536$ possible values, giving about 96 dB.
• 24-bit: $2^{24} = 16,777,216$ possible values, giving about 144 dB.

These are theoretical figures. Analogue electronics, microphones, converters and the listening environment normally have a higher noise floor than ideal 24-bit quantisation.

---

32-bit float inside DJ software

Modern DAWs and DJ applications commonly process audio internally with 32-bit floating-point maths.

Unlike 16- or 24-bit fixed-point PCM, floating point uses a mantissa and exponent. This provides a very large numerical range for intermediate processing.

Practical implications

1. Intermediate headroom: A signal may temporarily exceed 0 dBFS inside a floating-point mix engine without being permanently clipped, provided it is reduced again before leaving that engine.
2. The final output can still clip: The physical audio interface and its fixed output range cannot reproduce unlimited values. If the signal reaches the final DAC above full scale, the output stage will clip or be limited.

Floating point protects calculations, not the loudspeakers or the final analogue path.

---

Myths and realities for working DJs

1. “96 kHz always sounds much better than 44.1 kHz in a club.” Not by itself. The master, PA response, room acoustics and level management are usually the dominant factors. High-rate files also use more storage and processing bandwidth.

2. “Digital audio is made of square steps.” Misleading. The stored samples are discrete, but a correctly band-limited signal is reconstructed by the output filter as a continuous analogue waveform.

3. “Saving an MP3 as 24-bit / 96 kHz restores the lost quality.” False. Up-converting a lossy file changes the container and file size but cannot recreate information removed during the original encoding.

---

Practical summary

• Distribution formats: WAV, FLAC or AIFF at 44.1 kHz and 16- or 24-bit are common choices, subject to player compatibility.
• Pitch and time-stretch: Starting from a clean, high-quality source reduces the chance that artefacts become obvious during large changes.
• Internal mixing: Floating-point processing offers large intermediate headroom, but the final hardware output must still remain below its clipping point.