Williams Bubbles PCB repair

[saltbreez]

OK. Well let me know.

Whenever I've had one of these, it has usually come down to some component that would normally get overlooked.

Have you tried playing 24 RAM pickup? There was a thread a while ago about somebody having a dead board problem and they started randomly changing the RAM chips, found a dead one and the board came back to life. You could just pop the RAMs mix them up and reinsert at random. If it is an addressing problem, then it may be a bad RAM chip where the stack first gets asserted. Or it could be a bad RAM chip hosing the bus.

ken

Ken,

It is interesting that you say this; I have never observed these exact symptoms you describe. This board had 3 RAM sockets on the top row that were melted and had to be replaced. I was meticulous about the work and double checked myself but still this is a repair. I wonder if these might have something to do with this boards problems....

Saltbreez

 

[YellowDog]

Ken,

It is interesting that you say this; I have never observed these exact symptoms you describe. This board had 3 RAM sockets on the top row that were melted and had to be replaced. I was meticulous about the work and double checked myself but still this is a repair. I wonder if these might have something to do with this boards problems....

Saltbreez

There is only one way to tell. Break out your meter and do a continuity check on the sockets. In the past I have found about 1/4 of the sockets have a bad connection when I am done soldering them in. Cold solder joints, broken traces, accidental solder bridges, etc. Those traces are very small and easily damaged. Some of them rely on solder flowing through the hole and making the connection to the trace on the top of the board as well and that isn't always easy to do.

I usually repeat the continuity tests after I have inserted the chips, just in case inserting the chips flexed the board and broke a connection.

At this point, it is worth a try. You have looked under all the common rocks, it is time to look under some of the weird ones now.

ken

 

[saltbreez]

There is only one way to tell. Break out your meter and do a continuity check on the sockets. In the past I have found about 1/4 of the sockets have a bad connection when I am done soldering them in. Cold solder joints, broken traces, accidental solder bridges, etc. Those traces are very small and easily damaged. Some of them rely on solder flowing through the hole and making the connection to the trace on the top of the board as well and that isn't always easy to do.

I usually repeat the continuity tests after I have inserted the chips, just in case inserting the chips flexed the board and broke a connection.

At this point, it is worth a try. You have looked under all the common rocks, it is time to look under some of the weird ones now.

ken

OK, I have removed ALL the RAM chips, double checked the 3 new sockets, exhaustively checking for continuity and crossed lines. I replaced ALL the RAM chips and repeated the exhaustive checking for continuity and crossed lines [not that I really thought the inserting a RAM chip would cause a crossed line, but hey...]
Thanks Ken, no luck found under this rock.

Any more suggestions?

How about it Mike R?

Saltbreez

 

[saltbreez]

Alright Guys,

...

I have a working board for comparison [thanks Ken], my thinking is: If I could force the watchdog on the working board to repeatedly reset similar to a locked up CPU, I could scope out the differences in the startup of the 2 boards. Sound like an idea?
...

Saltbreez

OK so to force the watchdog to issue a periodic reset on the good board, you cut the watchdog trace and ground the right hand side.

What I see is that the R/~W line of the good board has very regular low pulses throughout the 133 ms time frame. On the bad board the pulses are almost random and only occur in a very short time frame.

The good board will write about 2" of rug pattern on startup. The bad board does not write any rug pattern.

Still looking for suggestions on this one

Saltbreez

 

[YellowDog]

OK so to force the watchdog to issue a periodic reset on the good board, you cut the watchdog trace and ground the right hand side.

What I see is that the R/~W line of the good board has very regular low pulses throughout the 133 ms time frame. On the bad board the pulses are almost random and only occur in a very short time frame.

The good board will write about 2" of rug pattern on startup. The bad board does not write any rug pattern.

Still looking for suggestions on this one

Saltbreez

Some of this makes sense.

On the good board, grounding pin 2 of the 74LS393 defeats the pullup resistor which would ordinarily keep the counter reset enabled. This allows the counter to free count until Q4 goes high, which cascades through chip 5H & 6H resetting the CPU.

This is ordinarily accomplished by the CPU periodically writing a specific value (it looks like $39) to location $CBFF before the counter trips the reset line (known as "petting the dog").

The only things left that I can see that could hose the address bus are chips 1D & 2D the hex buffers.

Another dumb question, do you have the widget board plugged in to the bad board? It sounds crazy, but I had a bad widget board stop the CPU from booting once. One of the address lines was being held from the widget board and since they are the low lines the CPU couldn't address anything correctly. Just another weird rock to look under.

ken

 

[armis2000]

Just a completely random suggestion, but have you replaced the multiplexers? I have seen a problem like this on a defender, and the signals looked good, but unless you have a logic analyzer, you cant see the write-thru, and the other input ports will make the outputs look "live"...

 

[saltbreez]

Some of this makes sense.

On the good board, grounding pin 2 of the 74LS393 defeats the pullup resistor which would ordinarily keep the counter reset enabled. This allows the counter to free count until Q4 goes high, which cascades through chip 5H & 6H resetting the CPU.

This is ordinarily accomplished by the CPU periodically writing a specific value (it looks like $39) to location $CBFF before the counter trips the reset line (known as "petting the dog").

The only things left that I can see that could hose the address bus are chips 1D & 2D the hex buffers.

Another dumb question, do you have the widget board plugged in to the bad board? It sounds crazy, but I had a bad widget board stop the CPU from booting once. One of the address lines was being held from the widget board and since they are the low lines the CPU couldn't address anything correctly. Just another weird rock to look under.

ken

Hey Ken,

Well I got the Bubbles ROMs in from Steph today, and I have a fully functional Bubbles ROM board, per testing on your Bubbles MPU.

I have not had the widget board plugged in for some time, but thanks for asking.

I see Armis asking about the MUX chips, 1 of 4 of the 74153's is socketed and new... the other 3 are on my list: next-in-line-for-the-total-shotgun-of-this-board.

Dave [Dokert] asks about the CMOS RAM. I know these chips can cause many types of problems. Before I go and socket out and use the last of my stash of 5114 chips (2 and Bubbles requires 2) I would like to at least hear a second opinion: Is this step remotely worth pursuit? I trust Dave, but he cannot say for certian this is the problem.

Still looking for feedback/suggestions

Steph at Hobbyroms.com, Thanks for the Bubbles ROMs!

Saltbreez

 

[YellowDog]

I'd do the 74153's. They are fairly cheap and have been known to cause addressing issues.

I've got to get back working on my second Bubbles ROM board. I got my Bubbles cabinet last week and I am just about done with the Karnov. I decided to leave it as a JAMMA cabinet (it's a Defender but has nice wood grain laminate sides with only a few nicks and scratches). So of course when I looked at the wiring, it was a POS and I had to completely gut it and start from scratch. The Karnov boardset that had been in it wouldn't come up, so I plugged my Yanton JAMMA tester and the bastard op had wired it for -5 & +12. Karnov needs +5 and +12. So two days of rewiring and it is up and working, except that the boards suck so much juice, it blows a 7A fuse on the Defender power supply every once in a while on start up. I guess I will have to put the switcher back in...

On the other hand, the rebuilt 4900 chassis I bought from Pacray works like a charm. I swapped it into one of my 4900's that wouldn't adjust the width down small enough. Just needed a little tweekinag and a good degaussing. It's almost too good to waste on a JAMMA cab. I also picked up a WG7400 from a local 8Liner shop for $20 today and that thing is awesome. It has some nice even burn lightly across the screen but you can't tell when it is running. I think I have the monitor for the Bubbles restore.

Sorry didn't mean to hijack the thread. I'm tired and rambling....

ken

 

[MikeRanger]

I'm sorry, but at this stage I think you have to shotgun some of the data bus chips, including ram access. Sometimes you can examine data lines here, and there, and figure it out, other times not. I happen to have a 9010, and a z80 pod (no 6809) and what I like about it is it can really help in these situations, and you can tell if it can access roms, or ram. It will test the bus, and in essence it will tell you what the cpu is seeing, and give you clues. Write FF to a single ram location, read it back and ger FE, would indicata a problem with data line 0, for intance. It is to expensive to justify in this case, so get out the shotgun and start swappin 1G 1D 2D 2G 3G. Eventually, you may have to pull the cmos ram, but no need to replace, it can run without them, and a 2114 (if I remember correctly) will work in there for testing.

Mike

 

[saltbreez]

I'm sorry, but at this stage I think you have to shotgun some of the data bus chips, including ram access. Sometimes you can examine data lines here, and there, and figure it out, other times not. I happen to have a 9010, and a z80 pod (no 6809) and what I like about it is it can really help in these situations, and you can tell if it can access roms, or ram. It will test the bus, and in essence it will tell you what the cpu is seeing, and give you clues. Write FF to a single ram location, read it back and ger FE, would indicata a problem with data line 0, for intance. It is to expensive to justify in this case, so get out the shotgun and start swappin 1G 1D 2D 2G 3G. Eventually, you may have to pull the cmos ram, but no need to replace, it can run without them, and a 2114 (if I remember correctly) will work in there for testing.

Mike

Mike,

All of 1G, 2G, 1D, 2D are socketed already. 3G is the decoder ROM 6, it has been swapped with a know good ROM chip. So... is the CMOS next in line, and if so, which of the 2 is most suspect?

Thanks,
Saltbreez

 

[mspaeth]

All of those suggestions are probably just a giant waste of time at this point, since he doesn't have code running well enough to start the RAM test.

Who cares if the RAM/RAM Addressing/CMOS/etc is bad if it's not even getting to those tests right now?!

 

[saltbreez]

Got rug!!!

Mike,

All of 1G, 2G, 1D, 2D are socketed already. 3G is the decoder ROM 6, it has been swapped with a know good ROM chip. So... is the CMOS next in line, and if so, which of the 2 is most suspect?

Thanks,
Saltbreez

OK, so my reasoning was simple: the lower 8 bits of data would be the most suspect to stop the CPU since they are the 8 bits most used in vintage Williams Bubbles games; therefore the CMOS RAM at the same location as on a Robotron would be the first to go. I cut this CMOS RAM out. Bummer. No help.

So... I cut the oddly placed CMOS RAM out, ... booted her ... and ...


GOT RUG!!!!!!!!!

Saltbreez


Still has a ways to go... like socket and replace 2x5114 chips...

 

[saltbreez]

Bubbles

Bubbles is running...

Saltbreez

 

[Dokert]

OK, so my reasoning was simple: the lower 8 bits of data would be the most suspect to stop the CPU since they are the 8 bits most used in vintage Williams Bubbles games; therefore the CMOS RAM at the same location as on a Robotron would be the first to go. I cut this CMOS RAM out. Bummer. No help.

So... I cut the oddly placed CMOS RAM out, ... booted her ... and ...


GOT RUG!!!!!!!!!

Saltbreez


Still has a ways to go... like socket and replace 2x5114 chips...

Who was that Dumb ASS that suggested this was your problem?

Spaeth 0
Dokert 1

Again Mark, this is getting to be old habit with you being WRONG.

 

[saltbreez]

Thank you

Wow, what a marathon debugging session... Now I need a Bubbles cab!

[Hopefully last] I had to swap out the lithium battery.

I would like to thank everybody that help me out on this one with hardware, technical info, suggestions and just plain encouragement. This really is a great group.

In particular [Ken] Yellowdog for sending me his working boardset and support...

[Dave] Dokert for solving the root cause...

Mike Ranger for his technical suggestions...

James, Chris, thank you...

Even that colorful fellow Mark what's-his-name was helpful on this one.

Thank you!
Saltbreez

 

[YellowDog]

You're welcome. I am just glad I was able to help a little to bring back a classic from the brink.

As far as getting a cabinet, just whine a lot. Eventually someone with a Bubbles cabinet rotting in storage will hear you. At least it worked for me.... (Thanks to Komodo (Pat) and Noel Johnson for transporting it here)

ken

 

[mspaeth]

Who was that Dumb ASS that suggested this was your problem?

Spaeth 0
Dokert 1

Again Mark, this is getting to be old habit with you being WRONG.

Occasional your blind shotgun techniques will be work.
It doesn't mean that's the correct approach to repairs.

Funny how you don't count all your previous WRONG suggestions before finally coming down to the 5114s.

 

[Dokert]

Funny how you don't count all your previous WRONG suggestions before finally coming down to the 5114s.

Spaeth 0
Dokert 2

Wrong again Mark, learn to read. Check the thread for any wrong suggestions by me.

Looks like you are just having a Spaeth Day.

 

[MikeRanger]

Woo hoo, another classic roars to life!!!!

I was suggesting that in a shot gun approach, when you cannot be sure a signal is bad, and with your symptoms, everything on the data and address bus were a suspect. I did not really thing the cmos would be your issue, but not too long ago, John Robertson has a cmos ram stop a board from working, I think it was an omega race, however.

Cutting chips out???? Argh! Start desoldering, will ya!

Mike - Former williams guru

 

[saltbreez]

Woo hoo, another classic roars to life!!!!

I was suggesting that in a shot gun approach, when you cannot be sure a signal is bad, and with your symptoms, everything on the data and address bus were a suspect. I did not really thing the cmos would be your issue, but not too long ago, John Robertson has a cmos ram stop a board from working, I think it was an omega race, however.

Cutting chips out???? Argh! Start desoldering, will ya!

Mike - Former williams guru

Just have not had that much luck with desoldering individual chips for removal; I always seem to lift traces when I attempt this. Cutting legs and then removing individual legs from the PCB has been, by far, the cleanest, most reliable method of removing IC's from these vintage Williams boards that I have found. Sure, I would like to have saved that $5 5114 CMOS RAM that probably was good, but I not gonna risk ruining the PCB it is on in an attempt to save it.

Hey, I hooked up a joystick to my test rig [looks like hell] and I'm playing Bubbles! I have on occasion hooked up a Defender CP to test the widget boards, but with Bubbles you really can play the game from the test rig: I'm even on the high score!

Saltbreez