Hmm…
So, the noisy nature comes from a high feedback amount, try above 80 and higher if needed.
It sounds a bit squared, so I would go with a ratio of 2 on Operator A.
Operator A envelope is likely unchanged (e.g fully open), and the level is perhaps around 15-30 so that there is still tonal content in the timbre, but still noisy.
Operators B probably have at least one odd ratio - and before the fade we don’t hear any clearly layered sinewave (due to the equal mix of x&y) on top (e.g if B1 would be at 5, 9 …), so I’m assuming it’s at 2 or 1, and the second B is 9 or similar. Set the B envelope to a long attack and there you go.
I first thought that there was some detune applied due to the drifting in the noisy timbre, but I’m suspecting that it’s an LFO actually. Can’t hear much drifting in the B operators, so… Perhaps it’s not multi sampled? The drifting is affected by the pitch is what I’m getting at.
Just going off what I’m hearing, haven’t tried this on a unit, but it’ll probably get you close.