Super-thanks for the nice info, both!
So, in digital pot button mode if I support that, I'll have to count on Paula's internal pull-ups being up so the joy's button circuit then would only need to be button-to-ground? I saw a schematic which recommended a current limiting resistor to ground for the pot buttons. The multiway switch would need a third position for this but I think it's doable. I don't know what will happens if the joy is in the wrong mode (my analog mode) though.
If using a matrix, I'm assuming it'll also need a few diodes to prevent ghosting/snake-around like on keyboards. I wonder if driving the matrix would require less or more CPU than an analog decode (easy on an MCU but here it's complicated by the scan rate thing and calibration). I don't know what the frequency would be for scanning a matrix. Would also have to work with the MUX reads I guess.
Would this schematic make sense for a Matrix? Since buttons are normally high I'm assuming the in/outputs would sink, each in order?
The Pot trimpot works a bit like a resistor in a 555 circuit if I understand you correctly.
Edit: New schematic... everyone floats now. It seems to me that adding the digital pot button mode means that the XL&YR switches will have to be double contact or things get real complicated. Conductive rubber pads will kind of mush any traces together but I think a flex PCB might work as it allows for some trace isolation. A flex PCB can also be folded up 90deg for the LR buttons, unless those are cheap side mounted tact switches or micro switches. SNES-clone pads use a daughter PCB.
