If I'm using pull-up resistors and current-limiting resistors on my board, and there's pull-up resistors inside the Amiga, there's going to be voltage divider effects. If I make the current-limiting resistors too high, the pull-down effect of the open-drain output won't pull the value at the Amiga side down enough to be read as a Low. Things were getting quite involved!

With a bit more information on the PIC24, I can produce a more concrete design for the device. The PIC24HJ12GP202 and PIC24FJ64GA004 have very similar pinouts. They only differ in the terminology for the DISV pin, which is a ground like on the dsPIC33, to which the VCAP pin must be connected through a 10uF tantalum capacitor. I'm not going to use the reduced-power ability of the PIC to run the core logic at a reduced-voltage supplied through VDDCORE. The upper left side has the ICSP PICkit and reset logic since that's where these pins are located. The 5V tolerant pins are towards the right side of the MCU, which hints that the right side of the board should be dedicated to the CD32 and the left side to the PS2, if I'm having one interface on each side. All the CD32 lines have current-limiting resistors and 10k pull-ups. The internal Amiga pull-ups from the schematic are 4.7k. In the lower left, I've added current-limiting resistors to the PS2 communication lines. Only the input lines from the PS2 pad will have 10k pull-ups to 3.3V, as required. The design is looking almost identical to the Ocelot, since that design was also driven by simplicity. The choice and placement of the ICSP and power regulation components is mostly fixed by the choice to use a PIC.