Mono101

The Mono101 is a one-voice bass and lead synth with one VCO, a sub-oscillator, a four-pole transistor ladder filter, one ADSR envelope, one free-running LFO and a constant-time portamento (polyphony.md).

import audioinstruments._rack.adapter as adapter

inst = adapter.create(48000, voice="mono101")
audio_out.play(inst.output)
inst.program_change(1)                           # Round Bass
inst.note_on(45, 110)

The controller section has portamento ("off", "on" or "auto" for legato-only glide), porta_time (5 ms to 3 s, exponential across the 0 to 10 slider), bend_range (semitones, 2 by default) and mod_vco / mod_vcf (how far the MIDI mod wheel sends the LFO to pitch and filter cutoff). The VCO has vco_range (16, 8, 4 or 2 feet, with 8’ at middle C on MIDI 60), vco_lfo (vibrato depth, squared across the slider up to 1.2 octaves so the lower third gives gentle vibrato) and pwm with pwm_src ("man", "lfo" or "env", sweeping pulse width from a 50% square down to 5% so the pulse never vanishes). The mixer blends saw, pulse, sub and noise on 0 to 10 sliders, with sub_mode choosing "square -1", "square -2" or "pulse -2" (a 25% pulse two octaves down, DC-centered and level-matched to the squares) locked to the VCO’s glided pitch CV.

The VCF is a 24 dB/octave transistor ladder (vcf_freq from 55 Hz to 16 kHz so 0 still passes bass fundamentals, vcf_res scaled to 0.98 at 10 for the edge of self-oscillation, positive vcf_env up to 7 octaves, vcf_lfo and vcf_kybd at 1 V/octave at 10). The VCA switches between "env" and "gate" (a 2 ms attack / 6 ms release gate contour) with a volume slider. One ENV (env_attack 1.5 ms to 3 s with the 0.51 peak correction, env_decay and env_release 1.5 ms to 12 s, env_sustain) runs with trig set to "gate+trig" (retriggers on every key), "gate" (legato notes do not retrigger) or "lfo" (gated by the LFO’s square wave). The LFO runs from 0.1 Hz to 30 Hz (lfo_rate) with lfo_wave set to "tri", "square", "random" (sample-and-hold stepped at each cycle) or "noise", crossfading across one block when switched so changing waveforms does not click. vel_vca and vel_vcf add velocity sensitivity (0 in a fresh patch). There is no shared effect after the voice, so inst.latency_samples is 0.

The twelve patches are Init Saw, Round Bass, Sub Bass, Squelch Bass, Vibrato Lead, Glide Lead, PWM Strings, Blip, Noise Snare, Pulsing Pluck, S&H Burble and Wind Sweep. The ten macros are vcf_freq, vcf_res, vcf_env, env_attack, env_decay, env_sustain, env_release, lfo_rate, pwm and porta_time. examples/mono101_demo.py plays all eleven musical patches in order and writes wav_exports/mono101_demo.wav; examples/poly_render.py mono101 checks the overlapping legato phrase for clicks. At the default 128-frame block with a note held, examples/poly_bench.py mono101 runs at about 14x real time on CPython (1.5 ms per 1024-frame buffer) and 10.6x on MicroPython (2.0 ms).

The Mono101 graph form

Wherever audiodsp has audiograph (CPython and the MicroPython and CircuitPython builds), create() gives you the Mono101 compiled to one C graph, as the Poly106 is (above); engine="rack" asks for the rack. Its three modules declare fragments: the voice, the LFO, and the MIDI module in the mono mode this instrument plays in. It plays the same panel, patches and macros. The rack and the graph agree to 64 counts at the worst and 6 rms over a note, and over a line played with keys over keys (legato, a retrigger, the return to a held key, a fresh note after a rest, a loud low note and a soft one), on Init Saw, Round Bass, Sub Bass, Squelch Bass, Glide Lead, PWM Strings and Blip (measured worst: 12 counts, 1.1 rms), and on Vibrato Lead to 19 counts on the line: tests/test_graph_mono101.py holds them there, with the rack given the board’s rounding, at block sizes of 64, 128 and 256. How the graph plays a mono line, makes the gap a retriggered envelope needs, and keeps the glide as exact as the rack’s is in architecture.md. Where they differ, and why:

  • The Noise Snare, the Wind Sweep and a noise added to a patch have the rack’s level and brightness, not its samples: the noise is the graph’s own generator. The S&H Burble’s random LFO is the graph’s too, so it draws its own steps of the same spread. A note of a second or two wanders by about a dB and a tenth of a brightness against the rack’s (the rack against itself with another seed does too); over thirteen seconds the Wind Sweep is 0.04 dB and the Burble 0.19 dB from the rack’s, and 4% in brightness at the worst.

  • The graph’s LFO runs a few parts in 100,000 of its rate away from the rack’s, which is about three samples of timing a second. A deep vibrato, or the Pulsing Pluck, whose envelope is gated by the LFO’s square, is 2 counts from the rack over a note of 64 ms, 160 over one of a third of a second, and the Vibrato Lead over the 4.4 s phrase tools/graph_phrase.py plays is 219.

  • The LFO crossfades across a block when you switch its wave, so the switch does not click; the graph’s LFO switches at once.

  • After a rest a voice wakes with its free-running oscillators where they were left. The voice mix parks a voice by the rack’s rule on the graph’s int16 wires, which can put a park a block from the rack’s when the tail is very quiet, and the oscillators then wake at another phase. The patches here don’t do it; a closed filter (a cutoff of 2 on the slider) can.

  • The voice takes the voice mix’s gain and the output’s scale into its VCA and rounds once, as the rack does (architecture.md): a quiet ring’s white floor is 0.33 counts rms against the rack’s 0.32 (0.93 when it rounded three times), and over the traits phrase Squelch Bass’s worst difference fell from 8771 counts to 82 on PWM Strings, a park that now falls in the rack’s block.