Note that this is an in-development extension, as noted by the EXTX moniker
instead of EXT. It's behavior is subject to change, and the extension string
will be removed (replaced with the official string once it's finalized).
Developers are discouraged from using this in production code, though feel
free to play around with it.
This method samples from the buffer so that it gets a time-correct 5khz stream,
which is subtracted from the original sample and has the high-frequency gain
applied, then added back.
A better method may be to average all the samples from the current one to the
one freq/5000 away, instead of bilinear filtering the two nearest freq/5000
apart. Processing cost will need to determine its viability
Here is a quick description of how the reverb effect works:
+--->---+*(4)
| V new sample
+-----+---+---+ |
|extra|ltr|ref| <- +*(1)
+-----+---+---+
(3,5)*| |*(2)
+-->|
V
out sample
1) Apply master reverb gain to incoming sample and place it at the head of the
buffer. The master reverb gainhf was already applied when the source was
initially mixed.
2) Copy the delayed reflection sample to an output sample and apply the
reflection gain.
3) Apply the late reverb gain to the late reverb sample
4) Copy the end of the buffer, applying a decay gain and the decay hf ratio,
and add to the late reverb.
5) Copy the late reverb sample, adding to the output sample.
Then the head and sampling points are shifted forward, and done again for each
new sample. The extra buffer length is determined by the Reverb Density
property. A value of 0 gives a length of 0.1 seconds (long, with fairly
distinct echos) , and 1 gives 0.075 seconds (short, indistinct echos).
The decay gain is calculated such that after a number of loops to satisfy the
Decay Time, a sample will be 1/32768th as powerful (virtually insignificant to
the resulting output, and only getting further reduced). It is calculated as:
DecayGain = pow(1.0f/32768.0f, 1.0/(DecayTime/ExtraLength));
Things to note: Reverb Diffusion is not currently handled, nor is Decay HF
Limit. Decay HF Ratios above 1 probably give incorrect results. Also, this
method likely sucks, but it's the best I can come up with before release. :)
