Galaxian VRAM issues

DanP

Administrator
Staff member
vacBacker
Feedback
5 (100%)
Credits
2,232CR
Hi guys, after some pointers on where to start looking for VRAM addressing issues. Fluke test fails with R/W error with a random BTS. If you put it in loop the BTS changes faster than you can see it. Anyone got any hints or tips on where to start? I've changed the buffer at 5K, and the two 2114's themselves so far...

Dan
 

Macro

Active member
vacBacker
Feedback
4 (100%)
Credits
2,009CR
check out http://www.arcades.plus.com/galaxianfix.htm

if you test in the order given it goes via no buffers, then one, then both - helps narrow down what is involved when you get the fault.
 

DanP

Administrator
Staff member
vacBacker
Feedback
5 (100%)
Credits
2,232CR
Hi Mike, your pages are already well thumbed I can assure you.

"Video Ram

To access this ram, the CPU now has to go via an extra buffer chip (4J and 4K or 5J *) so if we do this last, we already know that the first set of buffers are OK. This occupies memory from 5000-53FF. A failure in this will usually be ram chips (3F and 3H) or the second buffer."

OK, I've replaced the buffer (actually changed it to a 245) and both ram chips. I'm certain I've got an addressing problem. Once you've verified these 3 where's best to start looking from there?

Cheers,

Dan
 

Macro

Active member
vacBacker
Feedback
4 (100%)
Credits
2,009CR
so, you have removed the chips at 4j & 4k, and fitted a 745 at 5J ?

did you put it in the right place, as that position has two sets of pins for different chips
 

guddler

Busting vectors like it's 1982!
vacBacker
Feedback
10 (100%)
Credits
4,058CR
I would normally take a looped value changing faster than, well, a fast thing to mean that more that one device is (incorrectly) enabled on the bus at once so I'd start by verifying that the address selection logic is all okay first.

Just in case you're not already aware, when probing out chip select lines, setting the Fluke probe to sync to address helps massively.
 

DanP

Administrator
Staff member
vacBacker
Feedback
5 (100%)
Credits
2,232CR
Macro, yeah I'm pretty sure I've got the 245's in the right place.

Gudd, any hints on how to do all the stuff you said? :)

Dan
 

guddler

Busting vectors like it's 1982!
vacBacker
Feedback
10 (100%)
Credits
4,058CR
It's pretty much covered in the basics of how to use a fluke and my paragraph above, but...

(Off the top of my head since I've not used my Fluke in a while).

  • Sync the probe to pod address access. How varies depending on whether it's a 9010 series or a 9100 series but it is only a case of pressing some buttons, nothing taxing. Check the manual.
  • Pick a RAM or ROM. Any will do and look up it's address range.
  • Read from an address in that range and stick it on a loop
  • While this is looping check the enable lines for all of the RAM / ROM. This is probably easiest done at the outputs of the select logic (which is probably 74LS138's or 139's). You should find that ONLY the chip that you are trying to read is enabled.
  • If any others are also enabled then suspect the 138 / 139
  • Rinse and repeat for other chips / buffers / etc.
  • I guess it could also be that a ROM or RAM had a stuck / duff enable pin which could also cause this. You're into a bit of remove / desolder or snip the pin and see what happens by the time you're at this point
You can do similar for writes in the case of RAM.

On boards like Missile Command, this same principle can apply to all the 74LS245's as well. There are quite a lot of them and you can get the situation where more than one is enabled at once when they shouldn't be. This isn't quite as simple to check for but the net result is that it's possible that (for instance) the board thinks it's reading from ROM / RAM when in fact it's reading from that and the trackball movement.

Otherwise known as 'The trackballs dumping all over the bus'
smiley1.gif
End result, game crashes or probably doesn't even run.

Not entirely sure the 245 buffers scenario applies to Galaxian (don't think it does), but the address select stuff most certainly does.
 
Top