Tempest PCB Watchdogging in-game

EclipseEye

Well-known member
Joined
Oct 18, 2010
Messages
1,594
Reaction score
102
Location
WARRINGTON, Warrington, United Kingdom
I recently took ownership of a Tempest PCB set, I have no information on it's history. This is the first time I have looked at a Tempest, although I do have some experience repairing Asteroids PCBs.

The PCB was initially dead, and I have now got it to the point where it will display on TEST, which is almost working 100%.

In-game forces the PCB to reset constantly.

So far, I have done the following;

  • Confirmed my test rig is fine with a working Tempest PCB set.
  • Connected a proven working AUX board (no change).
  • Used a new (Bob Roberts) Interconnect cabe (no change).
  • Reflowed all Interconnect header pins.
  • Verified all ROMs, and even used a proven working set to eliminate (no change).
  • Verified all 8 Vector RAM 2114 ICs as good (no RAM check sound in TEST either).
  • Checked all voltages on the MAIN board and AUX board.
I have also gone over much of the circuit, looking at buffers and counters, but have still not found where the problem is.

One thing to note, when I flick to TEST, I do not get the 1st screen which shows the number of seconds played and so on. Instead, TEST goes straight to screen 2, not sure why, although it did display screen 1 the one time fine);

picture.php


picture.php


picture.php


The following pic is supposed to be a square being scaled down (with the audible tones dropping with it). I am getting the tones, but the screen is completely static;

picture.php


picture.php


picture.php


Note, the grid is missing in the next pic;

picture.php


Does anyone have any clues as to where to concentrate to find the problem?

Is there a way to disable DMAG0, as per Asteroids to determine if the problem is with VSM?

Here is a video of what I am seeing in-game, occasionally you can get a glimpse of the Tempest game for a split-second before it resets (not shown in video though);

 
Last edited:
Id guess its a rom socket? You eliminated the Mathbox, The vector engine is drawing the test screens, so it appears to be running. It looks like its crashing when its reading the attract screen code.
 
Id guess its a rom socket? You eliminated the Mathbox, The vector engine is drawing the test screens, so it appears to be running. It looks like its crashing when its reading the attract screen code.

Thanks for the reply, yes I did think it may be sockets but have eliminated them for the time being. Reason being is that I have found that bad sockets usually produce intermittent results. Remember, the TEST does not work 100%, the scaling square test does bot work. I have also probed all the pins of the ROMs and found no floating lines.

The sockets are gold plated dual swipe and look in good order.

Something tells me this is a logic problem or bad trace somewhere.
 
Ok, I'd made some assumptions here, however, these were made by comparing the SELF TEST in MAME to what I had on a real PCB.

Screen 1 of the SELF TEST will only show if the SELF TEST is activated during in-game. Otherwise, if the PCB is powered from SELF TEST active, it will jump straight to Screen 2.

Screen 4 - In MAME this shows a square being scaled down, however, on a real PCB all you see is a static cross. EDIT: I've just realised that I am running Rev.1 ROMs on my PCB. Checking MAME using Rev.1 ROMs indeed shows a static cross on SELF TEST. Only Rev.3 ROMs show a square scaling down in size.

Therefore, it would seem SELF TEST is working 100%.

The only thing to add, is the on the top left hand corner of the screen on Screen 2 SELF TEST, I can see some digits which suggest ROM issues? The thing is, the re-trace lines on my scope make it hard to make out?

picture.php


I'm beginning to think this is a ROM addressing issue, I know the ROMs are good, and I doubt the sockets are the problem. I noticed that this PCB is populated with 8304 logic for addressing, rather than using LS245. My experience with Asteroids has taught me that these 8304 logic have a high failure rate - I will start with these.
 
Last edited:
Screen 1 of the SELF TEST will only show if the SELF TEST is activated during in-game. Otherwise, if the PCB is powered from SELF TEST active, it will jump straight to Screen 2.
...
Therefore, it would seem SELF TEST is working 100%.

The only thing to add, is the on the top left hand corner of the screen on Screen 2 SELF TEST, I can see some digits which suggest ROM issues? The thing is, the re-trace lines on my scope make it hard to make out?

picture.php


I'm beginning to think this is a ROM addressing issue, I know the ROMs are good, and I doubt the sockets are the problem. I noticed that this PCB is populated with 8304 logic for addressing, rather than using LS245. My experience with Asteroids has taught me that these 8304 logic have a high failure rate - I will start with these.

I agree re: 8304s. The error code at the top I'd say is a 2 which means ROM D1 (ignore second line). But your other problem is shown by the M in the middle of screen #2, indication a Mathbox failure
 
Just a small update, I checked the 8304 bi-directionals and they were fine. Still searching for the fault - sheesh!

I may be able to get hold of a Fluke 6502 pod which will eliminate most things.
 
You've got a Rom problem, and a Mathbox problem.

I would work on the Rom problem first. Since none of the others are showing errors, it's most likey a bad rom, or socket.

Then you need to figure out why the aux board isn't playing nice with the main board.

That's what is causing your watchdog to bark
 
You've got a Rom problem, and a Mathbox problem.

I would work on the Rom problem first. Since none of the others are showing errors, it's most likey a bad rom, or socket.

Then you need to figure out why the aux board isn't playing nice with the main board.

That's what is causing your watchdog to bark

Thanks Talon - I've tried a known working AUX board with the CPU board in question and it still behaves the same way. So, I know the problem which is preventing the PCB set from running game is on the CPU board.

Is it worth swapping out all the ROM sockets at this stage? I did probe the pins on the ROMs and they were all active (no floating lines). Also, I did swap over a known working set of ROMs, also verified existing ROMs.
 
Gold plated sockets are generally trouble waitin' to happen when used with tin plate chips. The electrolysis due to the dissimalar metals creates a corrosion that causes bad contact. Keep gold with gold, tin with tin.

Somethin' like Deoxit may be a temporary fix, but it'll likely be troublesome again down the road. And....Atari used some shitty sockets.
 
While it's cheap to replace them all, without the right tools to do the job you may wind up causing more damage. I've seen plenty of lifted traces, or cut traces from people just trying to pull eproms.

I would pull the offending rom, check the CRC if you are able. If you have single wipe sockets they most likely are contacting on the inside of the eprom legs so make sure they are clean.

Then have a look at the socket itself. Sometimes just pulling and re-seating a couple if times will work (at least for a while).

If I'm looking at it right, it's 2 that's showing, which is D1, which is a vector eprom. So that infact could be causing the aux board problems as well.

Use a logic probe to make sure the rom select is working. Pin 20 on D1, and pin 6 C1 ls139

All Data lines, and address lines are shared between both vector eproms D1 and E1 so those are most likely good.

Once they rom error clears up, I bet the rest of your problems will go away.
 
I went ahead and replaced the socket at D1 anyway, and saw no change, therefore knew it must be a logic problem as the ROM was verified as ok via ROMident.

As per Talon's suggestion, I went back to the multiplexor LS139 at C1. I had previously probed this IC, but had made the fatal mistake of leaving the PCB in TEST mode where the pin would not be active anyway. Hence, I never spotted that pin 6 was stuck HI, and therefore it could never select the ROM during game mode.

The Maths box problem was due to a bad ALU and PROM - all working 100% now.

Many thanks to all.

IMAG0566.jpg
 
I went ahead and replaced the socket at D1 anyway, and saw no change, therefore knew it must be a logic problem as the ROM was verified as ok via ROMident.

As per Talon's suggestion, I went back to the multiplexor LS139 at C1. I had previously probed this IC, but had made the fatal mistake of leaving the PCB in TEST mode where the pin would not be active anyway. Hence, I never spotted that pin 6 was stuck HI, and therefore it could never select the ROM during game mode.

The Maths box problem was due to a bad ALU and PROM - all working 100% now.

Many thanks to all.

IMAG0566.jpg

I want to get a tempest board to test my scopes with :).
 
Back
Top Bottom