auditus, -us (Latin): hearing, the act of listening.
Hear it, then see it: a hands-on course in sound for musicians, with its own analyser.
A hands-on course in acoustics and audio for musicians, built on a real home setup: a Steinberg UR22mkII interface, a condenser microphone, two desktop speakers and a guitar. Each lesson pairs a little theory with a listening test, using sounds generated by the computer or recorded.
First we learn to listen, then to record.
| # | Title | Status |
|---|---|---|
| 1 | Sound and hearing | done |
| 2 | Optimising the listening setup | to do |
| 3 | The numbers of sound: dB, dBFS, sample rate, bit depth | to do |
| 4 | Critical listening | to do |
A second part on recording will follow: microphones, miking an acoustic guitar, voice and guitar, basic editing, exporting.
Each lesson has a folder in lessons/ with a script that generates its test
sounds and a printable handout:
cd lessons/01_sound_and_hearing
python3 make_sounds.py
python3 handout.py
We are on Linux: files are played with aplay (package alsa-utils), from
the terminal, with no heavy player.
aplay sounds/01_a_110hz.wav # through PulseAudio: the normal way
aplay -l # lists the sound cards and their numbers
All the sounds of a lesson, one after another:
for f in sounds/*.wav; do echo "$f"; aplay -q "$f"; done
Without -D the sound goes through PulseAudio, which sends it to the default
output and mixes it with whatever is already playing (the browser, Spotify).
Ctrl+C stops playback.
aplay -D plughw:1,0 file.wav bypasses PulseAudio and talks to card 1
directly: it works only if nothing else is using it, otherwise it answers
Device or resource busy.
oscillum/ is the course's sound analyser: open a WAV, listen to it and
see it at the same time, in three synchronised views.
| View | What it shows |
|---|---|
| Waveform | the whole file, left and right channel, with the cursor |
| Oscilloscope | 5 to 100 ms around the cursor: the wave itself |
| Spectrum | FFT at the cursor on a logarithmic axis, in dBFS, with the main peak |
python3 oscillum/oscillum.py lessons/01_sound_and_hearing/sounds/03_a_harmonics.wav
python3 oscillum/oscillum.py --trace file.wav # and print what it does on the terminal
It speaks English and Italian: the system's language by default, or the one
asked for with --lang=en or --lang=it. The sentences live in
oscillum/i18n.py, one row per sentence and one column per language.
Space bar to play and stop, click on the waveform to move; when playback
ends the cursor goes back to where it started, like a tape recorder. "Fit"
zooms waveform and oscilloscope to the file's peak; the spectrum stays in
absolute dBFS. Only Tkinter and numpy; audio goes through aplay.
It is built the way the author's other Tkinter programs are
(Tkinterlite): an App holding a
Main frame, an Engine that owns its parts by composition (tools, wav,
player, log), one class per module, each chart a tk.Canvas subclass,
.format() strings and one exit per function.
cd oscillum
python3 -m unittest discover -s tests -v
The standard library's unittest, one file per module: translations, WAV
reading at every sample width, the log, the player's clock and the
arithmetic of the spectrum. The spectrum tests need a display and are
skipped without one.
In tools/, to analyse your own recordings:
| Script | What it does |
|---|---|
levels.py |
duration, peak and RMS in dBFS, clipped samples |
accents.py |
tempo, swing and accent per beat of a rhythm part |
chords.py |
chords per half bar, chosen among the given candidates |
midinfo.py |
tracks, instruments, tempo and metre of a MIDI file |
cd tools
python3 levels.py recording.wav
python3 accents.py verse.wav --channel 2
python3 chords.py song.wav --chords A,D7,E7,F#m,F
manual/ holds a short guide to the Steinberg UR22mkII on Linux (Debian,
PulseAudio, ALSA): panel, connections, gain, monitoring, everyday listening,
arecord/aplay commands, common problems, a safe order for +48V. The PDF is
regenerated with:
python3 manual/ur22mkii_manual.py
Python 3, numpy and reportlab (Debian packages python3-numpy and
python3-reportlab), Tkinter (python3-tk). For listening, aplay
(alsa-utils).
MIT, see LICENSE.