your trying to manipulate the direction of the flux so that it cuts through the copper at 90 degrees.
Actually the 90 degrees thing is not an issue. What matters is how much flux transfers from one side of the wire to the other, not what angle it is at with respect to the wire when it does so.
Efficient designs DO tend to have the angle near 90 degrees for most of the flux because that's a side-effect of packing things tightly so wire length is short and flux is dense.
====
The main problem with your genny is the thickness of the stator assembly. The stator is "air gap" and should be reasonably thin - both to increase flux density from the magnets and to keep the flux going through the whole assembly. The flux will want to fold over to the nearby opposite pole, rather than going through the coils, if the gap is wide compared to the pole spacing. The path through the coils to the magnets or pole-pieces on the return-path metal on the opposite side should be substantially shorter than the distance between the magnet centers.
With the illustration shown the flux from the magnets will mostly be folding back to the adjacent magnets. Only the top turns of your coils will be generating substantially and the rest will be mostly adding resistance. Shorten the core and wind the coils to fill the space. You'll both increase the field from the magnets and make more of it go through the coils.
For a given thickness of magnetic material you get less leakage flux if you have half the magnetic material on each side of the coils than if you have it all on one side. This is because, though the combination of the fields from the material on the two sides is linear (up to saturation), the material near the coils contributes more field than the material farther from them. If you take two thin magnets and put one on each side you have two magnets close to the coils. If you stack them both on one side you have one close and one farther away. The field is a dipole field so it falls off with the CUBE of the distance, so doubling the distance to one of the magnets divides the field from it by eight.
Air gaps in the magnetic path are like resistors in a current path: The more air gap, the less flux. Steel, or other (reasonably "soft") magnetic material, "conducts" magnetic flux by this mechanism: Unpaired electrons line up their spin so their magnetic field is aligned with the ambient field. The field from the electron gives as much boost to the field as it would weaken if the space weren
t occupied by the atom. If you have a piece of steel backing the magnets, the field goes from the back of one magnet to the back of the next within the steel, and if the steel is thick enough for there to be enough lined-up electrons in the cross-section to "conduct" the entire strength of the field, the whole field will stay in the steel and stay strong despite the longer path. If the steel is missing you can add the path of the field between the backs of the magnets (or through free space between the coils if there's no backing plate) to the total air gap in the path, and if the steel is too thin only part of the field gets the advantage and you have a middle ground. (You can check by seeing if paperclips will stick to the back of the plate. If they do, it's too thin and a lot of the field is taking a trip through the airl.)
Assuming the spacing between the magnet disks is enough smaller than the spacing between adjacent poles on the disk that the field mostly goes from disk to disk through the coils, the next questions are "How big should the air gap be?" and "How full of coils should it be?" to get the most out of the expensive magnets. (This isn't ALWAYS the most important design issue, but starting from optimum for that gives you room to trade things away selectively, later.)
We'll start with how full to fill the space with windings: With a given gap and field strength, and wire size, adding more turns raises both the voltage and resistance in proportion. With a load to produce a given current, both power and loss goes up in proportion: Efficiency stays constant and power goes up with the turns. Similarly if the wires are fattened, keeping the current density constant, or if additional windings are added with the same number of turns: Current goes up with added copper while voltage and loss percentage stays the same. So at a given gap, flux, RPM, and current density, the amount of power is proportional to the amount of copper in the gap. You want to come as close as possible to stuffing it solid with windings. (This assumes you don't stuff it SO full you compromise cooling and have to reduce current density to avoid burnout.)
So assuming the only air gap is between the rotors and you stuff it full of copper windings (and we'll assume the clearance gap between the moving parts is small and ignore it): How big should the gap be?
Starting from no gap (and thus no windings) and expanding the gap (and adding windings): At first the drop in field with added gap is trivial. The added power from added windings is almost directly proportional to the increase in the gap. As the gap gets wider the cancellation from nearby poles increases more with each unit of gap increase. At first the amount of power gained with each unit of increase is reduced, then levels off. When the gap is very large the field falls off so rapidly that the added resistance of more windings loses more power than the added windings generate. Somewhere in the middle is a gap, and an amount of copper, where the power is maximum. (Because the curve levelled off this "sweet spot" is wide. Being tens of percent off the optimum may make only a percent of difference in power.
It happens that, with the magnets evenly divided between the two rotors, the optimum is about where the gap-full-of-copper is about as thick as the thickness of the magnets on one of the two.
In addition to getting more out of your magnets by having half the material on each side of the coil assembly, you also have eddy-current losses if you drag the magnetic field through the return path behind the coils, but you don't have these if the backing plate turns with the field. So even if you put all your magnets on one side you want the plate on the other side to spin with them. If you're going to have a spinning plate, why not mount magnets on it?
So the optimum geometry to maximize power per unit of magnetic material is looks like a "short stack of pancakes", much shorter than it is broad, with everything co-rotating except the (coil) pancake at the center of the stack. The top and bottom pancakes are the moderately-thick support/return-path disks, the next pancakes inward are the magnets, and the center one, with the coils, is about as thick as each of the ones with the magnets.
(Edited: I had somewhere gotten the impression that the optimum gap/coil thickness was the total thickness of the magnets. While I was composing, Flux posted, and he's more expert on this issue than I am, so I edited my post to match.)
