Fluke testing Defender

guddler

Busting vectors like it's 1982!
vacBacker
Feedback
10 (100%)
Credits
4,058CR
I've got an early series Defender on the bench at the moment that crashes as soon as it comes out of rug test. I've read through the theory of operation and understand how the memory is working in relation to the screen etc., in terms of which 4bit values related to which positions on (and off) the screen etc.

I have put all new 4116 RAM in the board and still no dice. I've been over it and can't see anything glaringly obvious thats wrong so I'm trying to be a bit more clever about this and test it properly. I did spot a 74139 that I haven't checked yet so I need to do that, but can someone confirm if it is possible to test the RAM using the following, or if the address / bit mapping is more convoluted?

In particular, I'm not sure if using a Fluke is possible at all because of timing, or if I need to do something to setup the board for testing. Ie, do I HAVE to disable watchdog or can I just set the Fluke to ignore it (I've done both). Do I need to write somewhere to enable RAM access? (I don't think so, that's for the banked ROMs).

Bank 1 (Column 4)

0-3FFF, Mask 1 - 4L

0-3FFF, Mask 2 - 4M

0-3FFF, Mask 4 - 4N

0-3FFF, Mask 8 - 4O

0-3FFF, Mask 10 - 4P

0-3FFF, Mask 20 - 4Q

0-3FFF, Mask 40 - 4R

0-3FFF, Mask 80 - 4S

Bank 2 (Column 5)

Repeat for 4000-7FFF

Bank 3 (Column 6)

Repeat for 8000-BFFF

I'm aware that the sockets could be crap, but that's the point of wanting to do the RAM test so that I know which of the 24 chips I need to focus on if at all possible (if any, I really am NOT convinced that it's RAM at fault). At the moment, whatever RAM test I do, I get "Cannot Write" at the lowest and highest address, if I push through with 'continue' the RAM test will proceed after (I think) one more error.

I'm really not having much luck with repairs at the moment!
 

guddler

Busting vectors like it's 1982!
vacBacker
Feedback
10 (100%)
Credits
4,058CR
Oh, would also like to add that if I do a "FILL" (not the actual command) of 0-3FFF with a given 8 bit value then the expected area of the screen does indeed get filled with colour. Expected areas being left hand side, 4000-7FFF being middle, 8000-BFFF being right. Or thereabouts given 8K of the ram is work ram and therefore off the screen. So it's doing what my understanding of the hardware tells me that it should be doing.

That is, provided I've let the board run for a few moments first after boot so it's had the chance to stuff some colours in the palette.
 

philmurr

Active member
vacBacker
Feedback
48 (100%)
Credits
2,374CR
Are you intent on using the Fluke to troubleshoot - in which case I can't help
smiley5.gif
If not I may be able to help as I've repaired quite a few of these recently.

By "early series" I assume you mean the CPU board with only 1 decoder chip?

Have you been able to narrow down which board it is that's faulty (CPU or ROM) using a known worker??
 

guddler

Busting vectors like it's 1982!
vacBacker
Feedback
10 (100%)
Credits
4,058CR
Am I set on using the Fluke? Yes and No. I don't overly care, the initial goal is to get the boards fixed. But they are from the local arcade so they WILL come back again. Hence yes, I do want to get some proper understanding of how the hardware maps out to the software and write some Fluke scripts so that I stand a fighting chance next time.

By early board, yes, I mean the single decoder version. Sadly I don't have any other Defender boards. Well, I think I do in the loft but I know it isn't a worker. From what I remember it's deader than this one
smiley36.gif
 

philmurr

Active member
vacBacker
Feedback
48 (100%)
Credits
2,374CR
I can probably lend you a working ROM board after Christmas if that's of any use?

I know when troubleshooting it shouldn't make any difference which of the boards is faulty but while you're just setting up the scripts it's going to be a lot easier if at least one of them is a known worker...

The theory of operation is a great help, as is Sean Riddle's website
 

guddler

Busting vectors like it's 1982!
vacBacker
Feedback
10 (100%)
Credits
4,058CR
Well, just spent a couple of hours scripting CRC32's and gathering Fluke sigs and everything checks out OK in the ROM department. I think therefore that I can assume that the ROM board is fine, the connector and ribbon connector are good, the CPU data buffer is good and at least some of the address logic is good. In addition, at least some of the write logic is good and bank switching is good because in order to get all of those ROMs verified I had to write the desired bank to D000 and without at least some good write logic, none of that would work.

And yes, for anyone who likes to make notes of these things, those are the Fluke 9100 sigs for Defender Red ROM set. 9100 sigs are not the same as 9010 sigs.

defender_red.png
 

guddler

Busting vectors like it's 1982!
vacBacker
Feedback
10 (100%)
Credits
4,058CR
Well, my issues are now fixed. Afraid to say that I missed the obvious. The -5v fuse had popped in my test rig. Oops! So in all likelihood, the problem in this case is going to be with the cab, not with the PCBs, but I did manage to find out the following, and I intend to go a bit deeper into this as I still have a board of my own here as well.

In theory, you would think that you could have only one bank of RAM populated and test that individually of the rest. I don't know why, maybe someone that understands this stuff properly can explain. In theory when only the left most bank of RAM chips are in the board, you could do the RAM tests listed above (0-3FFF mask, 1 / 2 / 4 / 8 etc.). This does not seem to be the case. When I did that I would get failures at certain addresses.

So that makes you think that the RAM is not evenly spread out across the address range, and yet doing the RAM test appears to show that it is. Odd, but not entirely unexpected.

If you fully populate all the RAMs and then remove a chip then the chip you have removed will cause a failure of the bit mask you have removed. So the DATA values behave exactly as I have detailed above.

Bottom line then, there is more to the addressing distribution of the RAM than a straightforward 0-3FFF = Column 4, 4000-7FFF = Column 5, etc. That is what I need to work out in order to be able to write a proper 9100 program for testing the RAM.
 
Top