Sujet Commentaires sur le test : Test du JD-08 de Roland

// Here is the description by Johannes Schultz:

// This file format is officially terrible.
// It contains two copies of most parameters - the original set of JD-800 parameters with slight modifications
// (such as using signed chars with -50...+50 range instead of 0...100 - fine by me), exclusively used for display in the UI
// - and ADDITIONALLY a set of precomputed parameters, which I assume are the ZenCore equivalents of the original parameters.
// You cannot just create a BIN file containing the display values that can be trivially converted from SysEx format.
// If you do that, all you hear is silence. No, you have to precompute every single ZenCore parameter, and many of them
// don't map linearly. Even better, for some envelopes, every time segment uses a different mapping!
// As a result of this design choice, you could actually create a file with a parameter display that is completely different
// from the sound that is actually generated until you start to wiggle some sliders, at which point the specific precomputed
// value for that slider will be updated.

// All bytes that actually differ between patches have been identified. The only remaining unknown variables are static,
// but not setting them to their expected values will break various features of a patch.
// Unknown variable names contain an offset relative to the first byte in PatchVST::CommonPrecomputed, and the expected byte value in hex.
// For structs that are repeated for all four tones, the offset is always for the first tone.