The delay scale is roughly linear with respect to room size, however the
density is not linear with room size. The density is calculated by taking the
room size cubed, then normalized by some factor. Unnormalizing the density and
taking the cube root restores the original room size to use as a delay scale.
The patch also alters the delay and all-pass line lengths to be based on a 1
meter room size, so the the room size recovered from the density acts as a
direct multiple for the desired target length.
Note that the room scale range is unchanged (5m to 50m), so the minimum and
maximum delays are the same. It should also be noted that 50m may not be the
correct room size for a density value of 1. A density value of 1 corresponds to
an environment size of roughly 2.52m when converted from EAX (DENSITY_SCALE
should be 16 rather than 125000), but sizes that low result in undesirable
resonance in the feedback, indicating other changes are necessary for that to
work.
The latter is a bit more descriptive as f0 is often used to denote the
reference frequency of a filter, so f0norm indicates the normalized reference
frequency (ref_freq / sample_rate).
And fix the output filtering. The modulation code is still there since it's
(probably) technically correct, but the interaction with the feedback loop and
filtering on the output caused improper behavior which needs to be sorted out.
The core LateReverb_* functions are explicitly written out now, since the
tapping and blending done by the Faded version is a bit more complex and it's
not so easy to ensure proper optimizing on the Unfaded version.
Similar to the recent chorus and flanger changes, the modulation delay now
swings between -n to +n, where n is less than the delay length. This brings up
a slight issue with the linear interpolation, as modff doesn't produce the
correct fraction value for interpolation (it's inverted, with 0 being closer to
the next sample and 1 being closer to the base). So it's using nearest
interpolation for now.
Each 4 related all-passes now share a structure with one delay line, which uses
an interleaved sample history. Also fixes some potential rounding problems for
delay lines with interleaved samples.
This seems to be more in-line with the intended behavior, to allow build-up and
overlap within the reverb decay, rather than a pitch-shift on input.
Unfortunately there's no readily available implementation of this reverb model
that includes modulation to compare with, so a low depth coefficient is used to
keep it very subtle.
