PCB Repair: Narc
Problem
ROMs U26-U40 reported bad.

Diagnosis
The ROM board tested OK with a known-good CPU board, so the problem was somewhere on the CPU board.
The bad ROMs are Image Memory ROMs, which are accessible by the DMA chip on the CPU board. The image memory data bus is 32-bits wide, with ROMs U25-U40 occupying bits 24-31:


The image memory data bus travels from buffers on the ROM board straight into the DMA chip on the CPU board:

At this point we might assume that one or more bits of the image data bus, specifically 24-31, are bad. A faulty DMA chip was ruled out by swapping in a known-good chip, producing the same result. All bits of the image data bus tested for continuity and none appeared to be shorted to other signals. The DMA chip PGA socket was cleaned for good measure, to no effect. Perhaps I was looking in the wrong place?
To test the image memory ROMs, the game uses the DMA chip to copy the ROM data into video RAM, 32-bits at a time. The data is then read from video RAM and checksummed. All video RAM was reported as OK, so that wasn’t the problem. It dawned on me that perhaps there was a problem with the bus between the DMA chip and the video memory. I needed to check the two video RAMs connected to bits 24-31 of the Video Data Bus:

Sure enough, VD29 had become disconnected from the DMA chip. I couldn’t see any visible damage to the PCB so I’m not sure where the break occurred. As the pins of the DMA chip aren’t shown in the schematic (PGA numbering is generally non-standard), I used my working CPU board to determine which pin of the socket corresponded to VD29:
I ran a wire from that pin to the closest VD29 signal I could find: pin 16 of U15. To great relief, this fixed the problem:
Fix
Fix trace leading from DMA chip Video Data Bus bit 29 to U15.



Nice find Phil! -John