I am not interested in how bellows pressure is mapped to MIDI. I am interested in how it is mapped to sound, and I find the Roland sound synthesis to not respond to it in the manner I would find a satisfactory rendition of what I can achieve with the bellows on an acoustic accordion. As I said: for me, it ends up doing too little and too much.
Yeah, I think this is well put. I've an (inferior) FR-1xb, but I've also had opportunity to play the Bugari Evo (which is an 8x internally), and the way that expressiveness is mapped to sound, just seems so incredibly "fake" to me. There's such an abrupt drop in sound in response to minute drops in pressure, and if I adjust the "bellows curve" to remedy that (at least on my FR-1xb), then it immediately sacrifices expressiveness. (Perhaps there are better options for adjusting that in the superior 8xb or an Evo... but I doubt it, given what I've consistently heard from performers). Meanwhile, as
@dak says, the way it expresses "expressiveness" itself feels completely artificial: it sounds like you've turned down a volume knob (or pedal), as
@John M also said. There's no change in the sound's quality or timbre, which makes the expression seem much more hollow. It's long been my general impression of Roland synthesizers/keyboards as a whole, that they have a long history of being particularly bad at emulating acoustic instruments. It's genuinely weird to me that they went to the (admittedly, lesser) trouble of carefully emulating button/key squeak, and bass reed growl, while abandoning the (to me) much more critical need to accurately sample/model the sound of a reed across the varying levels of air pressure/flow that might be applied.
I've been hoping the upcoming Proxima Mia would be an improvement in the technology, but at least in this regard, the performance samples from
their marketing videos on YouTube don't exactly fill me with confidence.
I agree with
@Ventura that I don't see any reason why the MIDI side of things would be a problem, and this leaves open the possibility that someone could write a far-superior set of accordion sound models, driven by MIDI input from a V accordion (or other MIDI-capable accordion, but ideally with button sensitivity). But AFAICK no one
has, and there's little motivation in the industry to do so. And for independent software devs like me, the task of painstakingly sampling (or worse, modeling) individual reed response at numerous pressures and pressure-change velocities, is quite daunting.