Stargate dreaded 1-3-1 error

[Mitchell Gant]

Board reset on me mid-game, annoying!

Showed static rug screen with no wipe across. LED showed 1-3-1 RAM error.

Checked 5V, -5V and 12V, all OK.

Checked/wiggled ribbon cables to ROM and IO board.

Swapped out a spare working CPU board, worked OK so ROM board and IO board must be OK. Therefore CPU board most likely has fault.

Swapped out all RAM, row of eight at a time, from the working board. No change from the 1-3-1, not that my gut feeling was telling me it was an actual RAM fault.

Swapped out decoder PROMs from working board. No change, so that's the obvious easy tests done then.

Next step it to try a bit of logic probing to see what the '153's around the RAM are doing. All seem to be active where expected.

Then I tried the '374's at 1H, 3H and 4H. All pins seemed active, except for pin 11 clock input on 4H which showed as stuck high, and it should be pulsing rapidly as it's the RAM to CPU data latch.

Hmmm, 4H clock pin 11 is connected in parallel to 1H and 3H clock pin 11's, which DO show as active on the logic probe. Strange, maybe a broken PCB track, but it hasn't physically moved so unlikely. Corroded maybe?

Switch off and check continuity. All checks as connected zero ohms from all three pin 11's and to AND gate pin 8 at 6J.

I try again with the logic probe and it shows 6J pin 8 and 4H pin 11 as stuck high, but 1H and 3H pin 11 as pulsing! Even though they are all connected!

I can only assume that the 7411 AND gate output has failed high impedance so showed as high on the chip and on the nearby 4H, but the logic probe picked up noise on 1H and 3H so showed INCORRECTLY as pulsing.

I had a spare 74HCT11 that I piggybacked on and bingo, all tests pass and game starts.

Well it showed '0' on the LED and made a startup sound, the picture had disappeared between me fetching the soldering iron and coming back while left switched on... and the monitor smelled faintly warm in a bad way. Yep it's died.

So one fault fixed and another appears... welcome to 30 year old electronics!

The lesson here is that logic probes are only a guide and cannot necessarily be trusted to be 100% depending on the failure mode of the circuit.

 

[Judder]

"Next step it to try a bit of logic probing to see what the '153's around the RAM are doing. All seem to be active where expected."

Ah... Now this is very interesting to me!

I have a Robotron CPU that I sent to Robotron2084 for testing and it came back with a note saying something like "RAM floating pins x and y - check 74153 at 4E"

I looked these up on the schematics last night and the 74153 at 4E provides select to the memory 4116 at 1R, 1Q, 1P etc. so the first bank of RAM

Do these fail often and would piggy backing a 74HC153B1 (which I do have) on top sound like a sensible way of testing this?

Would be awesome to be able to fix this

 

[Mitchell Gant]

I think that any of the logic chips directly connected to the Williams 4116 RAM banks are more susceptible to failure than any of the other logic on the board.

Here's my theory: The 4116 RAM needs -5V and +12V supplies as well as the normal 5V. Depending on how a RAM chip fails, or if it is loose or in a corroded socket so that it's ground is not connected, it may well pipe -5V or +12V through to those connected logic chips, stressing or destroying them.

Let's say one day you were stupid enough to plug in a chip that wasn't an actual 4116 RAM, but some other RAM chip that you *thought* was a 4116 clone. Well that could get very hot as it took +12V and -5V on pins that were never meant to take those voltages. As well as getting very hot as it destroyed itself it could also conduct excessive current to the other connected logic chips, like the '153's. No one would really be that daft to do that though, would they? (Looks down at feet sheepishly...)

Piggybacking an HC153 may help to prove failure, it's certainly something I'd try these days. But if it didn't help it doesn't mean that you've 'tested' the board's '153 as being good. Only a logic analyser or swapping the chip could truly prove that. Though if there's still excessive voltage going through from the RAM it would just get destroyed again!

Anyway that's my pet theory on why a lot of collectors seem to see 1-3-1 errors and it not simply be a RAM failure. If it is true then it's another good reason to upgrade to 4164 RAM and remove the +12V and -5V supplies.

 

[Judder]

Interesting theory about the -5V and +12V voltages being able to take out the other TTL chips around!

On this board I've upgraded the board to 4164 RAM and built an adapter so that should stop any non +5V voltages from flowing around the board

So I looked at the notes on Robotron board problems again, and it is labelled that pin 14 of the RAM is floating.

Pin 14 connects to the Data Bus Bank 1 on D14, which in turn connects to the 74LS374 at 1H on pin 14.

The 74LS374 at 1H is an Octal D-Type Flip Flop so perhaps that's the one I need to test instead.

I'm guessing it's not the 74153s, as these connect to the Memory Address Bus, and that only uses pins 5, 6, 7, 10, 11, 12 and 13 of the 4116 RAMs...

Thoughts welcome

 

[Mitchell Gant]

If pin 14 of a RAM chip is 'floating' then all that suggests is that the RAM output has either failed, or that it's output isn't being enabled.

I assume you've already changed the RAM chip for a known good one?

So that would suggest that the RAM output isn't being selected by the CAS (pin 15 of a 4116). Probe or 'scope pin 15 and it should be pulsing very fast. If it's stuck high then that's the problem and you can try to chase it back to the 74LS74 Q output on pin 9 at 4K (well, on a Stargate that is).

If 6K pin 9 is pulsing, then check that you have an actual good connection between the 74LS74 and the RAM.

Hope that helps!

Mitchell Gant2014-05-22 14:12:45

 

[Judder]

I assume you've already changed the RAM chip for a known good one?

They are new 4164 chips which are from a set proved working on another board I upgraded, so kind-of hoping that's not the issue - I haven't tested the board yet myself so I'm not sure if it is _all_ pin 14s floating high, or just one. Will check over the weekend and report back!

So that would suggest that the RAM output isn't being selected by the CAS (pin 15 of a 4116). Probe or 'scope pin 15 and it should be pulsing very fast. If it's stuck high then that's the problem and you can try to chase it back to the 74LS74 Q output on pin 9 at 4K (well, on a Stargate that is).

Cool advice.

On the Robotron drawing set 4K is the same (a 7474 dual flip-flop), and pin 9 is connected to the pin 15 of the 4116s the same (Chip Select) so the logic you suggest should follow

If 6K pin 9 is pulsing, then check that you have an actual good connection between the 74LS75 and the RAM.

6K pin 9 is connected to MUX 0 as far as I can tell which is pin 14 of the 74153s at 4F, 3E, 3D 4E ('A Select' on the 74153) - I'm guessing this should be pulsing?

Also (*just checking) do you mean a good connection between the 74LS74 and the RAM?

Checking the Bill of Material there are 6 7474s at 4D, 4K, 6K, 6C, 7F, 7G but can't see any mention of 7475...

Hope that helps!

It definitely will!

Logic probe and schematics with a strong cup of coffee on Sunday - thanks again!!

 

[Mitchell Gant]

6K pin 9 is connected to MUX 0 as far as I can tell which is pin 14 of the 74153s at 4F, 3E, 3D 4E ('A Select' on the 74153) - I'm guessing this should be pulsing?

Also (*just checking) do you mean a good connection between the 74LS74 and the RAM?

My typos are bad today... I meant pin 9 of that 74LS74 at 4K, sorry!

Yes I meant a good connection, to eliminate bad sockets or broken PCB tracks.

Checking the Bill of Material there are 6 7474s at 4D, 4K, 6K, 6C, 7F, 7G but can't see any mention of 7475...

My typo... sorry again!