Recording & signals / A listening lesson

When different frequencies leave the same samples

Hear a high input fold into a lower apparent frequency in a declared teaching model.

Sampling loses the distinction

If a signal is not suitably band-limited before sampling, different continuous frequencies can yield the same discrete sequence. The reconstructed interpretation can then place energy at an unintended lower frequency. This is aliasing. It is not merely a blurry waveform: information about which original frequency produced the samples is missing.

Evidence: 1

Calculate one example

Choose a teaching sample rate of 4,000 samples per second. A 3,000 Hz sine lies above the 2,000 Hz Nyquist boundary and corresponds to a 1,000 Hz alias in the baseband representation, with the appropriate phase relation. Increase the input from 2,000 toward 4,000 Hz and the folded frequency moves downward toward zero. The direction reversal is a useful clue.

Evidence: 2

The display and sound have different roles

The lab calculates the samples and folded frequency at the selected hypothetical rate. Its audible output is synthesized at the browser’s actual playback rate to illustrate that computed result. It does not reconfigure your sound hardware to the teaching rate or claim to record a physical source. The table and plotted sample instants explain the calculation even if the example is silent.

Evidence: 3

Prevention must occur at the right stage

Before sampling an analogue signal, an anti-aliasing filter limits unwanted high-frequency content. During digital synthesis, methods such as restricting harmonic content or oversampling with filtering can avoid generating aliases in the final band. Filtering after frequencies have already folded together cannot generally recover which original components were present. That is why aliasing is a design issue, not simply a playback-volume issue.

Evidence: 4

Fold into the baseband

For f ≥ 0 and fₛ > 0, let r = f mod fₛ. The nonnegative alias frequency is min(r, fₛ−r), lying between 0 and fₛ/2. Phase must also be considered to reproduce the exact samples.

Keep this in mind

Aliasing is an ambiguity in sampled data; prevent it before unwanted components fold into the represented band.

Make the connection

When samples tell another story

Compare an input below Nyquist with one above it, following the folded frequency and sample table.

Open the experiment ↗

Useful words

Aliasing

The ambiguity in sampled data when different continuous frequencies produce the same samples. Band-limiting before sampling helps prevent it.

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. Aliasing ↗Chris Donahue / Carnegie Mellon University · 2026
  2. Sampling and the frequency domain ↗Chris Donahue / Carnegie Mellon University · 2026
  3. Sampling in practice ↗Chris Donahue / Carnegie Mellon University · 2026
  4. Filter types ↗Chris Donahue / Carnegie Mellon University · 2026
Full source and rights register ↗