Ghosts n Goblins with 4 faults
Posted: October 15th, 2011, 10:43 am
I bought an untested Ghosts n Goblins board this week, knowing it would almost certainly be broken.
When it arrived, the game was running but the background graphics were a mess of repeated tiles and the sprites had horizontal jailbars. I took it upstairs to the test bench, left it running for a moment or two and then everything went black.
Checking the code ROMs there was no activity at all, in fact nothing on the board was pulsing. First thought was that this would probably be the crystal, so I replaced it with a known working one and the fault remained. The timing circuit on Ghosts n Goblins is on page A-2 5/8 of the schematics here on the left hand side:
http://arcarc.xmission.com/PDF_Arcade_M ... matics.pdf
As you can see it gets voltage from the chip in 3N which is a 74LS04 flip-flip. Replacing this chip fixed the clock signal. Important note here, this one actually kept me going for a while because initially the new chip didn't work. I tried a 7404 next and it was fine, with more testing it turns out that the circuit is quite fussy about which LS04 you have in there, changing resistor R44 for a slightly different value might sort that out. A brand new Fujitsu didn't work at all, but an old Motorola I pulled from a parts board worked just fine, as did a brand new Fujitsu 7404.
So I had the game running again, but the sprites had now vanished completely... I decided to sort out the backgrounds first, the second board in a GnG set is the video board, looking at it with the ribbon connectors at the top, the set of ROMs on the right are background tiles. Checking the address lines on these showed that address lines on pins 2-5 were stuck except for one with a signal so weak that the logic probe would just flicker for half a second and die. The ROMs are on B-2 9/9 on the schematics, showing that these address lines go back to a 74LS273 on page B-2 8/9, location 5A on the PCB. 273s are a larger flip-flop, and if you can see activity on the inputs and patchy or floating outputs, it's usually a safe bet that the chip has failed. Sure enough this looked dead, replacing it fixed the backgrounds.
Next I had to get the sprites back, so I started at the ROMs looking for activity, which are on page B-2 4/9 on the schematics, and bottom left of the board with ribbons at the top. The address lines were showing activity on them, but the CE (chip enable / select) lines which you would expect to see pulsing as various ROMs are accessed, were stuck high on two pairs and stuck low on the third pair. These are connected to a 74LS139 decoder near the ROM layout jumpers at the bottom of the board, but the inputs to that were also stalled. Going back to the schematic, the lines coming to the jumpers are AD8 (label missing) and AD9, so going to page B-2 3/9 we can see that these should come from a 74LS273 at position 8J. While the lines were stuck, they all looked solid. 273s take a clock signal, and checking pin 11 showed that the clock was dead, no signal at all. Following the schematics we can see that the clock line goes back to a 74LS139 at position 9K. This had healthy inputs, but one of the four outputs was dead - replaced and the sprites came back to life.
Finally there were just the jailbars to deal with. The jailbars were horizontal, and not changing position with the sprites which usually suggests a RAM issue rather than ROM access. The sprite RAM is at 12N and 7N on the PCB, pages B-2 5/9 and 6/9 on the schematics. While both showed activity on address and data lines, the data lines on 7N seemed a little quieter. I started checking the 74LS257 and 74LS273 chips attached to the data lines, and sure enough - the 273 was looking sick. Replaced and the game was finally fixed.
We can learn several things from this. One - the release of the schematics (quite recently) was an absolute god send. These boards have a mask that covers all the tracks, making diagnosis without them a very long and miserable process trying to work out which chips connect to which others. By now many with simple faults will have been binned because diagnosis was simply too time consuming. Secondly, always suspect the flip-flops. My first ever PCB fix was a broken flip-flop, and I've replaced many over the years. Multiplex/demultiplex seem to be the next most common TTL failure, maybe it's just luck.
When it arrived, the game was running but the background graphics were a mess of repeated tiles and the sprites had horizontal jailbars. I took it upstairs to the test bench, left it running for a moment or two and then everything went black.
Checking the code ROMs there was no activity at all, in fact nothing on the board was pulsing. First thought was that this would probably be the crystal, so I replaced it with a known working one and the fault remained. The timing circuit on Ghosts n Goblins is on page A-2 5/8 of the schematics here on the left hand side:
http://arcarc.xmission.com/PDF_Arcade_M ... matics.pdf
As you can see it gets voltage from the chip in 3N which is a 74LS04 flip-flip. Replacing this chip fixed the clock signal. Important note here, this one actually kept me going for a while because initially the new chip didn't work. I tried a 7404 next and it was fine, with more testing it turns out that the circuit is quite fussy about which LS04 you have in there, changing resistor R44 for a slightly different value might sort that out. A brand new Fujitsu didn't work at all, but an old Motorola I pulled from a parts board worked just fine, as did a brand new Fujitsu 7404.
So I had the game running again, but the sprites had now vanished completely... I decided to sort out the backgrounds first, the second board in a GnG set is the video board, looking at it with the ribbon connectors at the top, the set of ROMs on the right are background tiles. Checking the address lines on these showed that address lines on pins 2-5 were stuck except for one with a signal so weak that the logic probe would just flicker for half a second and die. The ROMs are on B-2 9/9 on the schematics, showing that these address lines go back to a 74LS273 on page B-2 8/9, location 5A on the PCB. 273s are a larger flip-flop, and if you can see activity on the inputs and patchy or floating outputs, it's usually a safe bet that the chip has failed. Sure enough this looked dead, replacing it fixed the backgrounds.
Next I had to get the sprites back, so I started at the ROMs looking for activity, which are on page B-2 4/9 on the schematics, and bottom left of the board with ribbons at the top. The address lines were showing activity on them, but the CE (chip enable / select) lines which you would expect to see pulsing as various ROMs are accessed, were stuck high on two pairs and stuck low on the third pair. These are connected to a 74LS139 decoder near the ROM layout jumpers at the bottom of the board, but the inputs to that were also stalled. Going back to the schematic, the lines coming to the jumpers are AD8 (label missing) and AD9, so going to page B-2 3/9 we can see that these should come from a 74LS273 at position 8J. While the lines were stuck, they all looked solid. 273s take a clock signal, and checking pin 11 showed that the clock was dead, no signal at all. Following the schematics we can see that the clock line goes back to a 74LS139 at position 9K. This had healthy inputs, but one of the four outputs was dead - replaced and the sprites came back to life.
Finally there were just the jailbars to deal with. The jailbars were horizontal, and not changing position with the sprites which usually suggests a RAM issue rather than ROM access. The sprite RAM is at 12N and 7N on the PCB, pages B-2 5/9 and 6/9 on the schematics. While both showed activity on address and data lines, the data lines on 7N seemed a little quieter. I started checking the 74LS257 and 74LS273 chips attached to the data lines, and sure enough - the 273 was looking sick. Replaced and the game was finally fixed.
We can learn several things from this. One - the release of the schematics (quite recently) was an absolute god send. These boards have a mask that covers all the tracks, making diagnosis without them a very long and miserable process trying to work out which chips connect to which others. By now many with simple faults will have been binned because diagnosis was simply too time consuming. Secondly, always suspect the flip-flops. My first ever PCB fix was a broken flip-flop, and I've replaced many over the years. Multiplex/demultiplex seem to be the next most common TTL failure, maybe it's just luck.