I even made a numerical (though not scientific) test: I took PNG snapshots of the area you selected as photographed by X-T2 and by X-T3. Loaded them into Python and calculated the standard deviations of each channel (R,G,B,alpha):
import numpy as np
x_t2 = imageio.imread('x-t2.png')
x_t3 = imageio.imread('x-t3.png')
print('X-T2 noise: %s' % str(np.std(x_t2, axis=(0, 1))))
print('X-T3 noise: %s' % str(np.std(x_t3, axis=(0, 1))))
X-T2 noise: [ 8.65137779 8.13238693 10.14568645 0. ]
X-T3 noise: [7.92599595 7.53614226 8.89362844 0. ]
So OK, the noise seems to be a bit lower for X-T3.
(Please don't use my methodology for anything serious, it's crap.)
Last edited by katastrofa : Thursday 20th September 2018 at 20:09.