Synth Arpeggio

This example demonstrates how to use BPFs to continuously change the pitch of different oscillators. It combines fixed pitch-class progressions with probabilistically shaped frequency and amplitude breakpoint functions, achieving evolving polyphonic structures with smooth filtering and stereo spatialization.

The script iterates through different pitch-class sets, applying continuous micro-retuning to arpeggio patterns. Each voice (defined by a waveform type) processes its own frequency and gain BPFs, which are subsequently passed through a lowpass biquad filter and spatialized across the stereo field.

synth_arpeggio.bell

setseed(96);
## Define four pitch-class sets to iterate through
$pcseq = (
    [0 4 7] [2 5 7 11] [0 5 7 9] [0 4 7 9] 
);
## Assign waveform types to separate voices
$waves = (saw cycle tri rect);
$numvoices = length($waves);
## Duration unit for each note event (in ms)
$unit = 150;
## Create a random shape for arpeggiation
$arpshape = dx2x(for $dx in 1...24 collect xrandom(1, 3) * xchoose(-1 1));
## Map arpeggiation shape to frequency range [4200, 9000] Hz
$arpeggio = scale(
    $arpshape, minimum($arpshape), maximum($arpshape), 4200, 9000
);
## Iterate through each waveform/voice
for $wave $voice in $waves do (
    ## Initialize time and data containers
    $x = 0;
    $freqbpf = null;
    $gainbpf = null;
    $pitchshift = 0;
    ## Assign unique panning to each voice
    $pan = ($voice - 1) / ($numvoices - 1);
    ## Iterate over repetitions and pitch-class sets
    for $n in 1...4 do (
        for $pcs in $pcseq with @unwrap 1 do (
            for $pitch in $arpeggio do (
                ## Retune pitch to nearest pitch class plus gradual shift
                $retuning = pitchdiff($pitch,  ($pcs + $pitchshift) % 12);
                $pitch += $retuning;
                ## Assign time-varying frequency and gain BPF entries
                $slope = xrand(0.8, 0.95);
                $freqbpf _= [$x mc2f($pitch) $slope];
                $gainbpf _= [$x xrand(0.125, 0.707) $slope];
                ## Advance event time and adjust shift
                $x += 1;
                $pitchshift += 0.01
            ) 
        ) 
    );
    ## Determine total duration based on length of BPF
    $dur = length($freqbpf) * $unit;
    ## Apply biquad filtering, then spatialize and transcribe the voice
    $wave(@frequency $freqbpf @duration $dur).process(
        biquad(
            ## random bpf for cut-off frequency
            @frequency for $x in 0...xrandom(1, 5) collect [$x xrand(2000, 6000) 0] 
            ## low-pass filter
            @type 0
            ## low resonance
            @q 1
        ) 
    ).transcribe(
        @gain $gainbpf @pan $pan
    );
    ## Rotate arpeggio to differentiate between voices
    $arpeggio = $arpeggio.rot(-6) 
);
## Apply global processing
render(
    @play 1 @process normalize(-3) freeverb(
        @roomsize 0.8
        @wet 0.1
        @damp 0.7
        @width 0.2
    ) 
)