Fix: Reverb wrapping distortion caused by integer overflow in calculation about oxefmsynth

[djtuBIG-MaliceX]

I decided to have a look at this again, recompiled current sources and found the culprit. It's in the DC filter.

https://github.com/oxesoft/oxefmsynth/blob/master/src/synth/reverb.cpp#L118

The operation ou0 = b[i] - in1 + ((ou0*32674)/32768); is causing an integer overflow on calculation of ou0 * 32674. Casting ou0 with either float or double resolves the issue.

In addition, may want to consider using floating point buffers instead of integers, as DAWs can better handle 'extreme-value' loudness (eg: place a limiter in front of the signal chain), rather than have the synthesizer clip internally.

[djtuBIG-MaliceX]

If anyone wants to test, try these builds out.
OxeFMSynth-7jan2021.zip

[samboy]

Related: I think we can do better here than having a simple Schroeder reverb. As reverb guru Sean Costello has put it, the first thing reverb designers do is implement the Schroeder reverb. The second thing reverb designers do is realize the Schroeder reverb is no good and implement something else. A FDN is probably the easiest good sounding reverb to implement; a 4-line FDN can sound really good with careful tuning, and a 8-line or 12-line FDN can more easily sound good if we’re willing to take the CPU hit.