Atari Mathbox Troubleshooting

[guddler]

Can anyone share any nuggets of wisdom on this please? Right now, specifically Tempest but BZ is very similar and I have still not got my BZ up and running that I got in Feb 2012

OK, Tempest Aux board for the Computer Museum in Cambridge (which I am not making any money from repairing if that influences if anyone will help).

Test mode comes up fine, with a Math error. Game will not run. Ordinarily I would use SA to troubleshoot this but the problem is that I cannot get a stable gate out of test 1 in the troubleshooting guide. I've been all over the board and because it is essentially (from what I can see) a big loop where the outputs eventually feed back as inputs I'm finding it really hard to work out why I'm not getting a good clock to test my sigs for test 1.

Although it's been a month or so now, I thought I did get enough out of the other tests to come to the conclusion that the proms and sockets were OK.

 

[DanP]

Hi Gudd,

Are you following the instructions for the Cat Box method of SA with Tempest/BZ mathbox? Did you make the harness that it needs to work with the Cat? Not sure how you'd apply it without a Cat box but presume you need the harness to get stable sigs.

Second thing, have you verified that the 4 ALU's are all OK? I always eliminate those first cos without them nothing is ever going to work.

Cheers,

Dan

 

[guddler]

Hey Dan, cheers for the reply.

The sa guide is for either cat box or HP Signature Analyser. It has sections for both so of course I'm using the HP. I've made up the cable. And that's what I mean by the test numbers. There are tests you do with each of the three plugs attached. When plug #1 is attached it's meant to force the board to provide its own clock driver. But I get no generated clock.

When you say ALU do you mean the bit slice processors? If so, yeah, they should be good as either plug 2 or 3 tests check them. But from memory I don't remember which. I'm just stuck because with plug 1 connected I get no gate on my analyser so the clock is locking up.

 

[guddler]

PS: I do need to double check what I've posted when I'm not on my phone and next up the shed which may not be tonight...

 

[Mitchell Gant]

I've got a HP 5004 too and I also could not get test 1 to work properly either. I assumed I was doing something wrong, as it was all rigged with croc clips rather than building a proper test harness.

I'd like to know what the problem is now!

Test 3 did work though and found a problem on one BZ board.

I've fixed a few BZ Aux boards now, all had dirty PROM legs that caused extra trouble, one had a bad 2901 bit slice and POKEY, another had several TTL chip faults. But the worst one to fix for me was showing a bad level on a data line. The signature analyser was no help at all in this case. Turned out to be a short between traces underneath one of the 2901's thsat looked to me like bad tin plating, as if it had been there from manufacture! Maybe it was as I bought the board 'untested'.

The 2901's don't seem to read voltage levels like TTL on my scope either, so a lot of signals that looked suspicious to me were actually OK when compared to a working board. I seem to remember there's a 74S03 open collector output gate on that board too, that doesn't read like a nice clean signal either.

In the end shotgunning most of the TTL logic, swapping the 2901's around and copious PROM leg cleaning were what fixed them rather than the signature analyser!

I've got a BZ on my bench right now too, so I'd like to help verify any tests or suggestions about getting test 1 to work properly on the HP.

 

[guddler]

That's interesting! I could have sworn that I've had test 1 working before. Can't remember if that was BZ or Tempest mind. I've got so many non working BZs I don't know where to start!!!!

Thinking about it, I still need to test the Tron power too

 

[ColinD]

You put me off troubleshooting the BZ Mathbox years ago when you said the sigs were wrong, but apparently on later drawings this has been fixed. James at RGP has managed to get the sigs running afaik.

 

[guddler]

OK, I have a spare working Tempest set (not for sale) so my interest is now piqued. When I'm next up there I will see if signature analysis works on that one.

 

[Mitchell Gant]

I briefly tried the Test 1 on a BZ Aux this afternoon. This time I spotted that I need to add ground links to the test header pins 3, 4 and 5, shown on the test harness #1 drawing.

I see now that pin 3 is input 'M START' and pin 4 is input 'STOP'. Pin 5 is shown on the schematic (sheet 3 side B) as input 'J' but on the drawing near the harness (sheet 1 side B) J is shown as pin 6, with pin 5 unlabelled. Must be a misprint on the drawing. I'm using a PDF of DP-156 2nd printing.

This gave me a correct signature CC34 when probing +5V now, unlike last time I tried it.

I probed PROM M1 and got matching signatures. But when I moved to chips D5 and E5 I read unstable signatures. D4 and E4 were OK, but that's all I had time to try out.

So I've confirmed that, when correctly wired to a HP5004 you can indeed get some correct signatures, but need to know why the D5 and E5 chips were not matching.

It could be simply that the Aux board is faulty. The game plays OK but during gameplay and at specific repeatable points during attract mode the vectors on screen do jump about and cause drawing offsets to certain objects.For example the moon image will jump lower on the screen for a few frames. So maybe that's the cause of the incorrect signatures, though I thought I remember swapping the Aux board and the same problem persisted, meaning the fault is on the main PCB.

Next time I get chance I'll try a know working Aux board and see if I can verify further, but I've made some progress at least.

 

[guddler]

Really need to take a look at this. Will try to get up there over the next few days.

 

[RGP]

Hey, anyone solve this? I've just done about four BZ sets and had the SA working just fine but on the last one i keep getting random values constanly on the screen of the SA.

I was reading the manual for the 5004 tonight and have come to the conclusion that either the rev of the aux board is different and has a pinout change or that it is not generating the right clock to the SA to latch to in the first place.

I can confirm that 2nd print DP sigs work apart from one pin on one chip where you should see CC34 but instead get 0000 (E5 I think).

I'll know tomorrow when I go back into the unit. My SA passes its own self tests.

I also find that depending on ROM version on the main board you sometimes only need to ground R129 3 times instead of 5.

I have also never on any set needed to ground B6-1 to constant reset the CPU as per the manual - if I do then it definitely doesn't work.

Will report back after a bit more probing tomorrow.

 

[RGP]

As promised i've looked into this and found a solution that seems to work for me.

As the SA is clock driven and the board is supposed to generate the SACLK signal which the SA is hooked to I figured that the problem might be in that circuit.

I followed the SACLK signal back to the LS161 at D1 on the aux board and found that one of the pins was not putting out the right frequency.

On the scope you can clearly see square waves for signals coming out of QA, QB, QC and QD but unless you engage frequency mode you won't see the issue - a casual glance would just say yes this is working.

I had 23.8khz coming out of one pin (sorry I don't remember which pin) which should be 11.8khz.

Also you need a 3.01mhz signal on SAEN.

That should give you a stable clock for the SA to work from.

If you still have numbers changing rapidly then look at the LS08 at D4 which in my case was the other problem chip for this one.

After that I was able to probe all the proms properly and found one bad 2901 on the board and bam the L error was gone.