But is it missing more than that...
?
Maybe you indeed have, but the vast, VAST majority of people who claim modeling is deficient in some appreciable way are again, naïve Braydens comparing apples to oranges, not über-engineer-golden-ear Stans who understand multiple levels of context and have the knowledge and experience to properly compare apples to apples.
"Oh, I just played it and something's missing" is not a rigorous (or even helpful) exercise, especially given that 95% of the time, the critic is using a different playback system at a different volume at a different position in the room, which is more than
50% of one's tone and experience.
Of course something's missing—at least half of it's missing!—but almost all of the time, it's one of those things (or user error, or an impedance mismatch, or gain staging, or a dozen other things), not some overarching, pervasive deficiency in modeling technology. Don't get me wrong—there
are remaining deficiencies in modeling technology if the goal is a perfect facsimile of the sound and feel of the source amp's closed systems, but for almost every user, the deficiencies of everything outside those closed systems (that is, things the modeler can't be held responsible for) are
deafening in comparison.
I mean, you sit in YGG Studios and watch one of LA's top electric guitar recording engineers play, trying to figure out when the computer might be switching between the real Soldano preamp and the modeled Soldano preamp. They turn to you and ask "When are you going to switch in the model?" and you say "The model switched in ten times." Now if that engineer or someone else in the room were to say something's missing, yeah, we take that very seriously. And they sometimes do, and we tweak before releasing it to the public.
Thought about this a bit more, and it's a really good question. Not sure, to be honest. There are additional DSP blocks that can be added to Helix amp model tools that describe more esoteric or hard-to-nail behavior (thanks, Helix Core!), and the models would indeed get bigger. At a certain point, however, the architecture might not be ideal for taking it beyond those additional DSP blocks, so the next step would be to scrap the whole thing and come up with a different modeling methodology altogether, one that allows for much easier scalability. I suspect it'd start out with a much bigger DSP footprint, but I couldn't begin to guess how much bigger.
If you pointed a gun at my head, I'd guess that... 250-300%?... more DSP per amp model might hit a point where we're at truly diminishing returns to appease 99.5 % of hardcore Stans. Anything beyond that and we're waxing poetic about crystal lattices and such. But also keep in mind that plenty of Stans are perfectly happy with where we are now. Some amps' behavior and quirks are harder to nail than others, so one might appease a hardcore Stan at 80% of where we're at and another might require 300%.