Recording & signals / A listening lesson

A digital waveform is a timed sequence of numbers

Separate taking samples in time from storing their amplitudes.

Choose the instants

Sampling records a signal’s value at selected times. In ordinary uniform digital audio, samples are evenly spaced. A sample rate of 48,000 samples per second means a spacing of about 20.833 microseconds. It does not mean the audio has 48,000 notes or that its highest representable sine frequency is 48,000 Hz. The timing grid and the signal’s oscillation rate are different quantities.

Evidence: 1

The numbers are not little sound blocks

A plotted staircase is a convenient way to show held sample values, but it is not a complete model of the reconstructed analogue waveform. Under band-limited conditions, a continuous signal can be reconstructed from sufficiently frequent ideal samples. Practical converters use filtering and finite precision, so their behaviour needs more than joining dots or drawing vertical steps.

Evidence: 2

Keep amplitude precision separate

Sample rate controls the temporal spacing. Bit depth describes how many amplitude codes are available in a specified storage representation. Increasing one does not automatically increase the other. A low-bit-depth signal can be sampled very often, and a high-bit-depth signal can be sampled too slowly for its bandwidth. These create different kinds of error.

Evidence: 3

Explore the ambiguity

In the aliasing lab, compare the input frequency with the selected teaching sample rate and inspect the plotted samples. Different continuous sinusoids can pass through the same sample values when the frequency range is not restricted. The browser still plays through its own actual audio sample rate; the lower teaching rate is a model used to explain the ambiguity.

Evidence: 4

Sample spacing

tₙ = n/fₛ and Δt = 1/fₛ. At fₛ = 48,000 samples/s, Δt ≈ 20.833 μs. An N-sample buffer lasts N/fₛ seconds under the usual playback convention.

Keep this in mind

Sampling chooses times; quantization chooses representable amplitudes. They solve different parts of digitization.

Make the connection

When samples tell another story

Hold the input frequency fixed and vary the teaching sample rate, watching the samples and alias readout.

Open the experiment ↗

Useful words

Sample rate

The number of sample frames per second. In stereo, each frame contains a value for both channels.

See in glossary ↗

Sources & further reading

These sources support the explanations and model assumptions. Recording credits are listed in the listening panels and the source register.

  1. Sampling and the frequency domain ↗Chris Donahue / Carnegie Mellon University · 2026
  2. The Nyquist-Shannon sampling theorem ↗Chris Donahue / Carnegie Mellon University · 2026
  3. Quantization and decibels ↗Chris Donahue / Carnegie Mellon University · 2026
  4. Aliasing ↗Chris Donahue / Carnegie Mellon University · 2026
Full source and rights register ↗