The same two lists again, with the durations taken down to the limit.
A segment of a hundred samples is a tone. A segment of ten is a shrill one. A segment
of one is no longer a shape at all: the list of amplitudes is being written straight
into the sample stream, and what was pitch becomes spectrum. This is the layer
yarrow is built on, and the reason it sits at the top of the piece.
The dial is scale. It multiplies every duration in
the list at once, so it sweeps the same material from a tone, through a rasp, to
broadband noise, without changing a single amplitude. Watch the segment length in the
readout as you move it: the sound changes completely between 40 samples and 1, and
nothing about the selection has changed.
Start with the gain low.
Reading the panels. The top panel is the
output at sample resolution, close enough in to count the steps while the segments are
long. The bottom panel is the same signal a dozen times further out, where the slower
shape is visible. Below scale 0.1 the top panel has nothing left to show,
because a segment is now shorter than a sample and the duration list has stopped
setting pitch at all.
The durations are drawn between a and b and
multiplied by scale, as in the material, where that multiplier is 0.1 and
the upper bound falls from 15 to 1 over the first forty-five seconds. sweep
does that fall by hand. Afterwards the stream is wrapped at a slowly moving threshold
and driven into a soft clip, which is what gives the layer its grain. The band stop
after it is the one the material sweeps across this layer; notchDepth
fades it in.