I did a lot of research into ways of somehow producing direct MIDI output from the existing LSDJ work, but in the end just tracked two channels' worth of the song from scratch as MIDI sequences. My programming style was pretty conservative, since I was operating in the dark about the nuances of the "instrument" — favoring longer sustained notes over short ones or rapid arpeggios, and avoiding pitch-wheel commands because, as I understand it, there's not a uniform pitch-bend range implemented across MIDI instruments (standard seems to be a lame 2 half-steps from what I can tell) and I had no idea whether the bend range could be user-defined through CC commands or what (and even if they could, whether a non-standard instrument like this would honor that sort of command). Anyway I tracked the song out, doing my best to approximate the original song while also adhering to the Tesla coils' limited pitch range. It was definitely a shot in the dark. I noticed that some of the notes I programmed rarely seemed to fire; these notes tended to be A: at the lower end of the stated acceptable pitch range, B: fairly short (eighth notes maybe), and C: routed to the red Tesla coil. Don't know if it was any one of those factors in particular, or some combination of the three, that made them seem prone to getting missed, but I definitely noticed their omission. Thankfully they seemed relatively nonessential and the track seemed to come off OK in the end. Definitely curious to know more about the responsiveness of the coils.
Would love to hear the sampled Tesla arc demo of your track!