Sun Mar 10 09:21:31 EDT 2013

Linear exponentials, conclusions up to now

- lin for radius: probably won't change much.
- exp(i lin) for phase: linear probably effects state too much

- straight exp(), or "lowpas pull" to exp(), or
- no correction?  solve it at upper level?

I think it is now understood well enough to start an implementation.
What is missing in the architecture is control rate operations.

Control rate ops act as state ops that are pushed out of the loop.
This needs some kind of syntax.  It could be done first for things
like computing 1/samplerate.