Each band can have an associated number of tones. Each tone has one envelope and a collection of sine waves.

The envelope is mainly a rectangular window that gets ramped in an out over four samples (start and end position of the ramp is specified in 4-sample-units). The ramp is sin^2-like over four (essentially three, as the first ramp point is zero) samples according to ramp[x] = sin^2(x*pi/8). The zero of the ramp is at the start position of a start ramp or directly before the stop position of the stop ramp.

Each of the sine waves has two amplitude values (one linear, optional, the other one logarithmic (+4 means double amplitude)), the period specified in pi/1024 per sample and a phase, specified in a resolution of pi/16).

Tones are specified for one block reaching possibly into the next. If two overlapping tones are specified (they overlap only if the first one is not ramped out before the second one is ramped in), they are blended using sin^2 weighting over the whole 128 samples.