How much noise and distortion can we hear in complex signals?

Listening tests with a multitone and a two-tone signal (100 Hz + 1 kHz).

Jan. 23, 2026

Distortion spectrum of slightly hard clipped two-tones and multitones is similar to that of white noise. This means that masking of distortion components can be approximated by using white noise for the masked signal, and both SINAD and SNR thresholds can be evaluated in the same test. (Note: noise approximation doesn't apply to pure tones and closely spaced two-tones. It also doesn't apply to two-tones with large separation when distortion is greater than 1%.)

These tests also demonstrate how masking, and as a result of masking the audibility of distortion and noise changes with volume. Distortion and noise is less audible at very high volume (above ~80 dBSPL).

Test method: the signal was mixed with white noise of varying noise levels.

Notes & tips:

To avoid immediate hearing loss... Start the test at a low volume and adjust to the desired level (preferably with a global volume control).
Of course, the test is harmless, the audio samples are normalized to a level slightly higher than the comfortable.
Headphones are recommended to keep other living creatures out of the test, especially at midnight.
The test takes about 5 minutes (but the experience is priceless).
Don't expect a musical experience... (Maybe next time I add some vibrato to the signals...)

Audibility of noise with octave harmonics

Signal (masker) frequencies: 100 Hz, 200 Hz, 400 Hz, 800 Hz, 1.6 kHz, 3.2 kHz.

This is a "partial" multitone signal with octave intervals and an organ-like sound. The harmonics at 6.4 kHz and 12.8 kHz are missing, as they would give the signal a very unpleasant sound. The threshold for the "full" multitone is slightly different and independent of level.

Expected threshold (SNR): 35 - 45 dB. The noise with an SNR of 30 dB should be audible.

Reference (signal only)

SNR = 30 dB

SNR = 35 dB

SNR = 40 dB

SNR = 45 dB

Expected results (just a guide):

SNR [dB]	"Medium level"	"High level"
30	audible	audible
35	audible	barely audible
40	barely audible	not audible
45	not audible	not audible

Audibility of noise with 100 Hz and 1 kHz two-tone

Signal (masker) frequencies: 100 Hz, 1 kHz.

Reference (signal only)

SNR = 50 dB

SNR = 60 dB

SNR = 60 dB (noise only)

Expected results:

SNR [dB]	"Medium level"	"High level"
50	audible	not audible
60	not audible	not audible

In these files SNR is not A-weighted (result is in dB, not dBA), though with white noise the difference between A-weighted and unweighted measurements is negligible (~2.5 dB).

Masking threshold graphs

(2026.02.13.)

Approximate masking threshold for octave spaced tones (red: masking threshold, green: absolute threshold of hearing):

Approximate masking threshold for octave spaced tones with missing harmonics above 2 kHz:

Csaba Horváth