ASL noise generator design

[trentgill]

at present the outputs are exclusively controlled by ASL, so it makes sense to implement noise as a sub-function of ASL.

ASL is centrally just about moving a voltage from {current location A} to point B over time C.

We can keep this notion of time C (ie the duration of the noise), but what is the meaning of the moving voltage?

noise by itself

in this case the voltage doesn't really matter, and a keyword / string 'noise' could be used instead of the voltage position

noise in sequence with voltages

here we can still use keyword 'noise' for a noisy segment, but there is no way of setting the amplitude of the noise in relation to the voltages. secondly, when the noise segment completes the voltage from which the next slope will run is a random value.

one alternative is to use a different to structure where 'noise' is just a 'shaper'. in this case noise would be scaled to the range of the current location A -> point B to 0. thus the voltage-gradient becomes a VCA curve over the noise amplitude.

while this is powerful & interesting, it's not necessarily what you want a 'noise' shaper to do. one could imagine wanting the 'noise' shaper to add noise to a path, rather than use the path as an amplitude sweep. ie. gradient+noise vs noise*gradient.

noise with varying amplitude / offset

the downside of the above 'shaper' idea is we can only scale the noise (with to destinations), but there is no way to offset the noise. notably it's not possible to generate bipolar noise (eg. +/-5V) which people may expect in the eurorack world.

we could introduce a totally new ASL keyword (like to), though it seems to be a very limited use-case to have to extend the DSL just to get it. of course this doesn't itself solve the question of "what is the voltage after the noise is done?" i wonder if there's another way?