I ran into a small problem last night when I went to build the actual cartridge. It turns out the EPROMs I have are 2MB, not 4MB like I was thinking and planning on. A quick search suggested that was in fact the norm across the board!

But I looked on and noticed that there were 3MB and 4MB cartridges - the only one of these I own is Lion King (3MB), so I popped it open to have a look. And sure enough, it has TWO ROMs, not one. A simply selection circuit made up of a 74LS00 Quad NAND gate was also present. I traced through this circuit, and it used !CE and A21 to activate the CE of the appropriate ROM, using the Quad NAND interestingly as two inverters and two NAND gates. Very clever!

So now I had to figure out what to do with my multicart. I didn't /really/ want the more complex circuit, and I wasn't sure I needed it. I also didn't want to sacrifice my Lion King cart.

But, there's only actually 2MB of game in my multicart - two 512k, and one 1MB, so if I could find enough space to squeeze in my menu, I could just use that. Suddenly my 43k menu felt really big, but first I needed to see HOW big it was by enumerating the blank spots.

After scanning a couple by hand, I wrote a little tool to look for blocks of 00 or FF bytes longer than 16 bytes long, report and total them. The results were not good. TF2 only has two blocks bigger than 1k, and a total of 9k free. TF3 has only 1.6k free. TF4 does "better", with a total of 19k free, but only four blocks over 1k. In other words, not a chance.

I could of course go for a dumb menu - plain text would squeeze into those spots, but no. Damnit, I want my fancy menu. So I'm back to square 1.

So this means two EPROMs, and I still haven't proven that my banking concept works. It would be nice to find a Genesis emulator that does NOT use Starscream for a change...

All that said, the banking circuit doesn't get much more complicated. I still tie in two lines to the address bus, and just use two flipflops. The difference now is that only one of the lines triggers bank select (selecting which 1MB half of a chip to use), while the other needs to trigger which EPROM to use.

The Sega cart tied this chip select in with the !CE line from the Genesis, but I don't really need to do that. These chips all also have a second select line called !G, which is normally tied to ground in a Genesis cart. By letting my circuit activate that line, and the Genesis trigger !CE, I should be able to get away with the bank selection without much more effort.

(Per the datasheet, !CE is the power control line, and !G is the output enable. So both chips will be powered up on Genesis select, but only one will output data. This should still be fine.)

For mounting the two EPROMs in a single ROM donor cart, well, I've done this before! It's possible to mount one of the ROMs /under/ the PCB, sharing the pins. I did this with my original Thunder Force 3 cart hack.

My only real concern is that this is a lot of soldering to do before I even know if it works. If it doesn't, I'm going to be grumpy and it will be hard to troubleshoot. Probably best to start with just the one chip containing the menu and TF2.