The Siege Blog

Shocker Shocker, a System 7 power supply with a fried GI section.   To be quite honest, this is probably the most poorly designed section of a pinball machine fully of poorly designed sections.  If you ever find a System 7 machine that doesn't have this area roasted to a crisp.... nevermind that thought, they are ALL like this.  From System 7 all the way through the end of WPC-95 production,  Williams constantly had issues with overloading their GI lighting circuits,  System 7 just happened to be the worst.

.t.

The problem is under-rating of the connectors.   My Black Knight, for example, has 65 GI lights.  With the original #44's, at .25A each, that's 16.25 amps JUST for the bulbs.  That doesn't include additional amperage draw created by the sockets, connectors and wiring.  The input connector is only rated for 12A.  So that's why the input side always fails. 

 The output side though, the one that's the worst in this case, fails for a variety of different reasons.  Technically, the connectors here SHOULD be up to the task.  They're rated at 7A per pin, and average 4A per pin of power.  However, that's assuming perfect conditions.  Throw in the notoriously cracked solder connections, some oxidation, heat, and humidity,  and the actual rating starts to drop pretty fast.  This is made worse by the small contact area of the IDC connectors that the wire attaches to, then the flat connector onto a round pin.  This is, of course, assuming they used the phosphor-bronze connectors, not the plain brass which is only good for 5A when new.

As you can see here,  a poor attempt to re-solder the header pins was already tried.  But it's a useless piece of work, as the pins themselves have already been compromised.  Tarnished, burned up and oxidized pins like those on this machine need to be replaced.  A common fix is to bypass them by soldering the wire directly to the big trace on the board, but that's a poor hackjob repair.  The fuse clip also has to go, as that's been compromised too. 

..

Now, once everything is off, we see more of the issue here.  The fuse clip and connector have been overheating so badly, that it's delaminated the trace from the board on both sides. This is probably worse than most others that I have repaired,  but it IS repairable.   It's just a matter of getting down to good clean metal, re-establishing the thru-board continuity, and getting better quality parts on there.

...

Now, I'm going to do my best on this one.  Of course, this needs eyelets,  as so many of these burned up traces tend to need. As you can see, I'm getting much better with them as I get more experience using them.  I don't know how I did all these repairs without them. The front side of that fuse clip doesn't look as nice as I would like, after soldering between the eyelet and the trace,  but you must do that before installing the components to maintain good continuity.  Otherwise you're just going to end up repeating the repair eventually. The backside will get soldered together when I install the fuse clip. I also needed one in the input connector too, which came out very nice, eh?

....

Now I can go ahead and put in new fuse clips and new connectors.  I always replace the fuse clips in pairs, as you can see.  The .156 header is a square pin, which maintains much better surface area, and thus maintains the 7A amperage rating much nicer.  If you use the newer molex high current connector, with the box style crimp connector, these can be good for up to 13A per pin!   At the very least though, use trifurcon pins here.  They have the same amperage rating, but again have increased surface area to maintain that rating over a wider range of conditions.

..

The power input connector will be supplied with new pieces for the wiring side too,  to make sure all new and clean metal is in there.  Another option is to convert to a System 11 style pigtail, but in this case it was requested to keep it with the original style.  Not an issue if you can reduce the amperage,  but if you insist on running all #44 bulbs this connector will fail again eventually. The owner said he is planning on swapping all new #47 bulbs into the machine, which add up to a 9.75A draw, and thus within specs for the connector.  LED's are even better for this, as it gives a massive buffer of available capacity. 

As a historical note, Williams eventually abandoned this whole area of the board in later designs, and many early System 11 machines still used this basic power supply.  System 9 saw remote fuse cards introduced,  early System 11 saw the pinput connector remoted to a pigtail,  and the last versions of System 11 even remoted the relay.... using one each for the playfield and the backbox.

WPC-89 saw the GI lighting controls moved back onto the power-driver board,  but a drastically redesigned system that could dim the individual sections of GI lighting via a pulsewidth modulated control system.  Yet they STILL had output connectors burning up all the way to the end by putting too many bulbs per pin on the connectors. 

I recently ran out of 2532 Eproms after a couple of repairs.  I normally don't go through too many of these, but things happen, and I needed more.  But they are proving difficult to find these days.  I eventually came across a link for a place called epromman.com that claimed to have them.  There were some folks recommending them on a couple websites, so I opted to take a chance, particularly since the prices were fantastic.  2716's were $0.50 each, and 2532's were $1.00 each.  I half expected them to never show up, but they shipped fast and arrived in just a couple of days.

They arrived, like this in their anti-static bag.....

.

Taking them out, the 2532's were put on a sheet of black chip foam, but many of the 2716's looked like this......

..

Now, I didn't really expect them to all be erased, tested and verified as good. I don't mind having to clean off old labels and put them in my chip eraser, but banged up all to hell like that?  I eventually got the legs straight enough to at least fit them into tubes for long term storage.....

. .

Needless to say, I am not happy with the packaging that they arrived in.  Chip tubes only cost like $0.25 or so from Jameco that hold a dozen or more chips each.  So there's really no excuse for packaging like this if you're running a website that sells Eproms in large numbers.  I will not be placing any repeat orders from this source.  

Well, I'm going to let a bit out of the bag on Secret Project #1 here,  but that's how things go, as I'm pretty darned excited on this one.  Secret Project #1 is going to be a highly modified variation of the speech/sound boards used through the System 6/7 era in a single integrated board, as opposed to separate speech/sound boards.  Ok, no so secret anymore.   But that's ok, I have a LOT more behind the scenes on this project that you're not going to see for quite some time.  Lets just say that I am not making a simple reproduction on this one by any means, so this should build up a bit of anticipation. I am looking to make an overall better peforming board.

Part of the problem I was having was the voice decoder chip on the speech board.  There are two different speech boards out there, which are interchangeable, but have a few differences between them.  Mostly though the differences are centered around the speech decoder chip used,  one of them uses a Harris 55516, the other a Motorola 3417, and their difference in ancillary components.   These chips are known as CVSD decoders.  For a good idea how they work there is a great YouTube video by a woman named Jerri Ellsworth that explains it very well.

The CVSD chips used on the original speech boards haven't been made in quite a long time.  Typical, eh?  And this was a huge stumbling block as I worked on my design.  If I can't reliably get a chip to decode the speech,  I can't do a speech board, and it becomes a dead project.  It also doesn't help that there wasn't that much info out there about CVSD speech encoding and decoding.  In fact, it was the only part that really kept me from seriously proceeding with the project. Though to be honest I haven't looked much into the 1408 DAC chip yet.  So who knows what's going to come up when I get to that part of the circuit.

But lets get back on track here. I finally found a viable option for this. Jerri, in her video, used an FPGA which sounded great, and is an awesome bit of engineering. However I am looking for an easier design for production purposes.  Well, I finally found one in a current manufacture CVSD chip that can use the 3-bit algoritim used in the older 55516 chip. The 55516's CAN Be found, but at $25 per chip they just aren't cost effective and its an uncertain future supply as it's mostly just old stock that happens to be sitting in inventories.  As far as I can tell, the other chip I found is still in manufacture and is about 50% cheaper.  

What I'll probably end up doing though is starting off with just a speech board prototype that is compatible with the original sound boards, but using the alternate chip.  After all, I need to make sure to get the op-amp circuits built up properly as well.   But it's exciting stuff knowing that I can move along some more now.  If it works, I may offer it as a stand-alone product,  but it's not the main goal of this design exercise.

Man, I am glad I don't do pre-orders, as I'm sure I'd have some upset folks by now with these Firepower eprom adapters.  The good news is they will be shipping in about a week from today.  The last details have been worked out, and I am just waiting on a package of hologram stickers that is enroute to me already, then I can start kicking them out the door about a month or so later than I had originally scheduled.   Welcome to business, I guess, eh?  I'm going to be contacting the pre-order list again on Monday, and finally will start accepting payments and preparing them for shipment.  The way things are progressing, I'm going to be needing to do a second run in the very near future,  but I want to get this one shipping first before I worry about that.

The Gottlieb eprom adapters will probably be ready about the same time as well, they've already shipped from my board supplier and will be here in a few days.  Errrrr.... I should probably order the 24 pin sockets for them, eh?

I'm also going to be putting up a few e-bay auctions in the near future for surplus components that I no longer want/need.  Connectors that I ordered the wrong size,  machined pin sockets that I no longer use, stuff like that.  Nothing big really, but still dead inventory as far as I'm concerned.

Lastly, progress is going very well on the board layout for "Secret Project #1".  It's far more complicated than anything else I've done, so it's going to take a lot longer as well.  But I think people are really going to like it.

I just placed another order from my board supplier, and the next generation "quickfit" solenoid testers are on the way.  These are going to be modeled more like the diagnostic switch input boards, meaning that they will be press-fit straight onto the 9-pin molex connector, and will NOT require a wiring harness to function.  They will also be much easier for me to assemble.  As with the original design, it's made to be assembled as either an LED tester output, or a switch input for the special solenoid circuits. 

Realistically, the original model boards have been priced at a clearance price for a while now, and if I were to do another run of them I'd probably be charging about $10 more per kit.  The new ones are going to be pricing pretty much equivilent to the original style testers at their clearance price.  Expected street date is approximately March 15th.

Oh, and this is NOT one of the 'secret' projects.  I'm still slowly plugging along with them, but they are still a long time off.

Sorry, folks, but I have to turn the comments off again for a short time.  I've been getting hammered by nonsense, and rather large, spam comments the past couple of weeks.  While they were just going into the spam folder, it was no big deal, but they started sneaking into the 'approved' section today as well.

So, no comments for a couple of weeks, turning the feature off for a while.  Normally this makes all the trollers think the sight is 'dead' to comments, and they go away.  I hope so, because they're pretty darned annoying. 

Interesting repair, not one you see too often.  This is a sound board for Firepower that came in for repair. The owner had a Firepower machine with no sound boards, bought these used out of a Black Knight. The initial complaint was that it was getting no sounds using a Firepower rom set, but worked fine with a Black Knight rom set.  Sorry for the shadowy picture,  bad lighting in my workbench when I do these late at night.  The better photos are during the day.  But lets get to the repair work, eh?

..

Started off with Leon's test rom.  Yes, he has one for these sound boards too,  ties an LED into an address line for the characteristic blinking.  This board would blink for a few seconds, then lock up.  Other times it would just keep blinking,  but if I pressed on the chips in their sockets, it would lock up. 

..

The problem, Scanbe sockets of course.  Thankfully this is a very simple board in comparison to the others in a machine, only two sockets on the sound board for the CPU chip and the rom chip. I kinda knew this was going to be an issue when I installed the test eprom,  as there was very little tension in the socket.  You can really feel it when a socket goes bad sometimes.

...

Easy fix so far, install a pair of good stamped sockets, and the test rom works perfectly.  It also has a test of the 6810 ram chip, which also checked out ok. 

..

While I was at it, I also replaced the power connector.  Again, typical Williams.  Pins trimmed too short, no mechanical strength, and the solder joints crack up and loose conductivity.  This wasn't causing problems yet, but would be eventually.

..

So, I throw it into my Black Knight for a live sound test.  Sound portion works just fine,  great!  Attach the speech board, rejumper it accordingly, and zero speech response.  Sound working fine, but no speech.  This narrows it down a lot actually.  It means the mixer op-amp at IC2 is working fine.  Very little hardware on the speech board,  so first thing to check is the eproms.  On a hunch I pulled the labels.  AM2732 eproms.   Whoops!   Williams speech boards use 2532 style eproms.  Same capacity, but incompatible pinout. 

..

Though they are scanbe sockets, they still had good tension, so I burned a set of the correct 2532 chips, and installed them.  Speech kicks in just fine,  not a hitch to it. 

..

Just a bit of interesting thoughts about this repair.  On the one hand it's the usual Williams issues of sockets and connectors, followed with incorrect parts in a previous bit of rework, when the roms were swapped.  Par for the course really,  it's almost invariably a socket, connector or previous work that causes the problem.

But then again, you almost never see these sound boards go bad.  Driver boards have their 6821 PIA chips get fried all the time, but you never see them go bad in sound boards.  Why is that? I have my theory, and you've heard me say it before.  Heat and battery corrosion.  Driver board PIA chips are right below the CPU board batteries, and right above the lamp matrix transistors/resistors.  I've seen driver board after driver board come through here, and they all have some kind of heat damage coming off those lamp resistors.  CPU board is immediately above the driver board, connected to it in fact. Heat rises, stressing the interboard connector. 

Sound board is all the way on the other side of the backbox, and not thermally coupled to any real heat source.  Sure, the 7805 voltage regulator gets hot at times, but nowhere near as badly and it's got a lot of heatsink attached to it.  It's also right next to the backbox vents, so it's going to have more cool air available too.

Why am I rambling about this?  Simple.  If you want a long-term reliable machine, you need to cut down on all that backbox heat.  It's killing your boards, killing your displays, and killing your backglasses.   The two biggest sources are the lamps and the driver board transistors.  On my Black Knight pages I have a couple of ways I've dealt with this.  Biggest one is swapping to LED's,  the other is a MOSFET swap on the driver board.  Doing these cut the power usage down to less than that of your average home light bulb.