The work by Amirm, Erin, and others on loudspeaker measurements and controlled listening deserves high recognition. What remains unclear to me is the reliance on music tracks for subjective listening: how can one ensure that a chosen piece reliably exposes a specific loudspeaker flaw? Meaningful listening of this kind cannot be done quickly and often accumulates into hundreds of hours. In my own experience, identifying one loudspeaker issue took weeks, and another response-related problem only became apparent after several months. After reading long discussions about the causes of unusual distortion behavior in an ELAC DF63 floor-standing speaker, I became convinced that simple, purpose-built test signals are needed to expose loudspeaker weaknesses before moving on to music.
Assuming the loudspeaker has already been measured and is of reasonable quality, listening should primarily answer whether small deviations seen in measurements are actually audible and objectionable under real listening conditions.
Creating test files is easy; generating sweeps or multitone signals is not the hard part. The real challenge is how to identify problems reliably without musical training or “golden ears.” For that reason, clear listening instructions are at least as important as the signals themselves. In practice, I find some form of measurement necessary; at a minimum, a smartphone app such as Spectroid is useful for identifying problematic frequencies and relative level changes.
Some examples of potentially useful listening signals:
The attached files (and translation) were generated with the help of ChatGPT. Please verify the files before use—you never know with AI.
My hope is that these files can be refined into more practical listening tools and expanded with additional signal types for evaluation prior to music listening. There are many variables, and even small changes (levels, markers, tone combinations) can significantly improve usefulness.
Assuming the loudspeaker has already been measured and is of reasonable quality, listening should primarily answer whether small deviations seen in measurements are actually audible and objectionable under real listening conditions.
Creating test files is easy; generating sweeps or multitone signals is not the hard part. The real challenge is how to identify problems reliably without musical training or “golden ears.” For that reason, clear listening instructions are at least as important as the signals themselves. In practice, I find some form of measurement necessary; at a minimum, a smartphone app such as Spectroid is useful for identifying problematic frequencies and relative level changes.
Some examples of potentially useful listening signals:
- Slow logarithmic sweep – Reveals whether low-frequency room or loudspeaker resonances produce a “multitone” bass character. It also helps assess the combined effects of placement, room modes, reflections, driver interference, and directivity at the listening position, assuming constant sweep level.
- Slow sweep with added 2nd harmonic (–40 dB) – A near-threshold harmonic makes the loudspeaker’s own harmonic distortion more audible. Frequencies where strong multitone artifacts appear should be noted.
- Slow sweep with added 3rd harmonic (–40 dB) – As above, but also reveals whether multiple harmonics coincide and reinforce the same frequency, which can be problematic with music.
- Bass timing using two-frequency transients – For example, 70 Hz combined with 2 kHz, with the high-frequency component delayed in steps (2–10 ms). The perceived point of time coincidence indicates whether bass is correctly aligned or delayed.
- Pink noise for level reference – While problems can be evaluated at any level, it is useful to know the approximate SPL at which an issue appears. This can be aproximated using band-limited pink noise (e.g., 500–2000 Hz) at 1 m and a smartphone SPL app, without changing volume between tests.
The attached files (and translation) were generated with the help of ChatGPT. Please verify the files before use—you never know with AI.
My hope is that these files can be refined into more practical listening tools and expanded with additional signal types for evaluation prior to music listening. There are many variables, and even small changes (levels, markers, tone combinations) can significantly improve usefulness.
