The PCBs arrived today, a little later than I hoped, but still with time to play with them.
For my first stab at a cartridge PCB, they're not bad. I went with the cheap non-solder-mask, non-silkscreen versions to test.
There were four problems with them.
The first problem was they were slightly too wide, it made for a very tight fit... I trimmed less than half a millimeter off the edge to solve that. I fixed this in the layout file.
The second was that the card edge was not beveled, but I knew this going in because ExpressPCB doesn't do beveling as far as I know. I used the dremel to bevel the edge as well. It's not perfect, but it works.
The third was that one of my traces didn't actually make it to the pad -- it was there, but was just barely not touching. Fixed that both on the board and in the layout file.
The fourth was an outright layout error - one of the EPROM control pins was left floating when it needed to be tied high. That needs a jumper wire, but I patched that up too.
The final problem I simply didn't foresee. I had intended to use ZIF sockets on the first board to more easily test the ROMs, and I hadn't realized that the ZIF sockets were MUCH larger than the raw ICs, not to mention they wouldn't fit in the Genesis. So, I extended the pins on one socket, and attached just it for testing.
The good news, is that Thunder Force 2 and the menu work perfectly on this, without the oddities I saw on the previous cartridge. (At least so far). I am going to independently test 3 and 4 with the single EPROM approach before I got for soldering one together permanently.
But, before I went that far, I did one more thing. I felt bad that Thunder Force IV was not region free while 2 and 3 were, so I undertook to fix that.
I found three parts that read the region register:
0482 move.b $a10001,d0 read version reg and.w #$80,d0 mask out region move.w d0,$f0b6.w Store REGION at $f0b6
0542 move.b $a10001,d0 read version reg and.w #$40,d0 mask out tv bne invalid if PAL, branch tst.w $f0b6.w check region bne $620 go ahead and run if not Japaninvalid: bsr $2444 failure?
third part (after Sega logo -- this is just deciding whether to say Technosoft or Tecnosoft):
c3a2 tst.b $a10001 test version register bmi $c3b4 branch if not Japan
So, presumably, all we really need to do is set the "bne invalid" in the second block to "bra $620". Then, of course, we update the three game checksums previously discussed.
They were pretty paranoid about region coding, eh?
Anyway, the branch is too far for a short jump, so we end up overwriting both the bne and part of the tst.w. But, this should be okay.
So, at $54C, replace "6608 4a78" with "6000 00D2"
At $15FF8, replace the checksum with "F9A3 23AF" - this should let you collect 1-ups
At $BFF8, replace the checksum with "BDE6 6EE0" - this should let you clear the second half stages
At $18E, replace the checksum with "0977" - this should let it boot.
I've re-attached the patch files here, this includes the new patch which is not 100% tested yet (but is tested against the known choke points.)