Listen to Ed because he's speaking from true experience.
Keeping your diameter tighter seems to be better. Not sure if its because the weight is relatively lighter because of less material involved, or if its something related to leverage. But tighter, more compact diameters seems to be the better solution.
so unfortunately almost all of the academic simulations done concern magnets of a given volume, not a fixed block of a standard size.
I have been passively reading papers on this stuff for about 15 years now and i would strongly consider recommending that you build a 24 pole machine with a single layer of magnets 5mm thick, with the inside corners of the magnets almost touching each other.
the diameter will be 3 inches or 80mm for the inside radius of the magnets and 120mm diameter for the outside.
its the same repeating pattern of 4 magnets and 3 coils for a one layer machine.
if you build a mold to hold the wires while you press them in a press and glue the together then you can press the coils flat and get the air gap as small as practical.
since you're only looking at a coil about 5mm thick, 80mm id and 120mm od, its not a lot of copper to waste, if you decide that an even smaller, 12 pole machine with double stacked magnets is better.
in general, 6 to 10 poles of an inner diameter 1 third of the outer diameter gets you the best torque density when you define torque density by the total volume of the machine. but this required trapezoidal shaped magnets. your 1:2 ratio magnets prevent a T shape configuration from working well, and arranging them to be 20 by 30 mm is also not much better because this reduces the pole count to just 8 on each side. -this is why i'm suggesting 24 poles is worth trying.
but when the magnets are the constraining factor and copper is of no cost, the optimal dimensions of the machine approach almost laughable limits. think bicycle inner tube instead of crispy cream donut.
