The same breakpoint waveform, with the tether removed. Each breakpoint walks wherever its own chaotic step takes it, and the only thing keeping the waveform inside a range is what happens when it reaches the edge. The amplitude bounds are not a limiter: they are part of the instrument, and each of the two can behave differently.
The dials are barrierLo and barrierHi.
0 reflects the walk back inside, 1 wraps it round to the other bound, 2 holds it
against the edge, and 3 throws it to a random place in the range. A wrapping bound
tears the waveform, a holding bound flattens it into a plateau, and a reflecting
bound keeps it continuous. Set the two differently and the waveform becomes
asymmetric, which is audible as a shift in weight rather than in pitch.
Start with the gain low.
Reading the panels. The top panel is one
cycle of the waveform between its two bounds, and the bounds are drawn where they
actually are, so a narrow range is visible as a narrow band. The middle panel is a
histogram of the steps the walk has actually been given, which is what
ampDist and ampDistP change. The bottom panel is the
audio.
ampDist 0–6 is linear, Cauchy, logistic, hyperbolic
cosine, arcsine, exponential and sinus, the distribution set of dynamic stochastic
synthesis. The steps are not drawn from noise but from the rotor's own angle, which
is why the default law already piles them at the extremes: a third of the steps land
in the outer eighth. Cauchy pulls them back toward zero, arcsine pushes them further
out, and sinus is not a distribution at all but one fixed step repeated. The duration
walk has its own law and its own bound, so the period drifts independently of the
shape.