Rapture/Dimension Voice Stealing Algorithm
OK, so the "Expression Engine" has a low limt in polyphony of 0 voices: one voice in monophonic legato mode. Polyphony of one sets it in monophonic, 16 voices is the default, and there's an infinite number of voices as an upper limit. These polyphonic limits are set per-Element, so dynamic allocation of voices in multitimbral mode isn't a concern. So, why even bring this up? I use polyphony limits as a performance device, and to impart some realism into emulating real-world instruments.
So, the voice-stealing algo is straightforward in the mono modes: new note takes priority over the last note played. At two voices, a low-note priority seems to take over. Beyond 3 or 4 voices, I have difficulty tracking the priority anymore. I guess that I could get all scientific here with some solid clips and listening tests (I'm winging this live), but it's be interesting to know the brief overview of the mathematics and decision-making behind the process.
Besides, I'm also wondering if I might be the lone soldier in this campaign. I know that this can be overridden (prioritized) in .sfz editing, so my real question would concern the default voice-stealing that occurs at polyphony limits (without any outside intervention).