Logic Board reworking: A Tale of Two IIcx

[Addicted]

I got back to the IIcx's last week. Good news, I think, is that after they sad-chime, their modem ports do echo serial data back (9600 8N1). So either the serial controllers power-up in a loopback mode, or the ROM diagnostic is a real thing and is running.

Bad news, not too bad really, is that I misunderstood what application to run on the remote (host) debugger Mac. What I have is an app that Apple meant to run on the machine being debugged. I need to go back to the books and find the chapter on remote serial boot debug and determine what I need, then go find it on the net.

Once I get one of these systems up, I will have a small amount of wisdom to return to the community. Parts list for recapping a IIcx/ci PSU. A 3D print for the plastic clip that stabilizes the PSU output plug. Info on how to use this debug tool I seek. The Bomarc schematic nit I mentioned above.

And since the fan bracket is deteriorating on one of my IIcx PSUs, I may create a 3D print to replace it, and I'll include that here.

 

[Addicted]

The first IIcx has been checked out by AppleCAT. AppleCAT is finding an issue with SIMM Bank A, and the door remains open for SIMM Bank B - the tests end after the Bank A issue.

I have swapped the 4x4MB SIMMs in bank A out for 4x256KB SIMMs that I had on hand, but AppleCAT still fails Bank A. The 256KB SIMMs are not suspect but, I should verify them by other methods by other means to rule out two sets of bad SIMMs. Old hardware.. it can happen.

I did replace all the SIMM address bus multiplexors due to heavy corrosion. They were replaced with the PSU on/off NANDs, and the PSU wouldn't remain on prior to that. A bad solder joint, lifted pad, broken trace.. all possible.

I've probed the address multiplexors outputs to the resistor packs, and stopped there as the resistors and everything 'beyond' (CPU's perspective) appeared to be free from any corrosion. I do have a set of replacement resistor packs on hand, but my understanding is that they are a very simple passive part that is unlikely to fault. I will undertake the probing from the resistor packs all the way to the SIMM sockets next. Then, I'll check the multiplexor traces to the CPU, and the data lines for each SIMM socket to the CPU. <sigh> And yes, there are some traces between the multiplexors and GLU that might be involved. Looking like hundred(s) of traces to check, and a single break on a single one of them would explain everything. Even lint under a SIMM socket might be the cause.

No one said restoration would be easy

Another mystery remains: the system doesn't even get to the sad chimes unless I have a video card in a NuBus slot. Not focusing on that. One hurdle at a time.

 

[Addicted]

Summary

Macintosh IIcx #2 booted System 7.5 from internal HDD and from floppy today.

It displayed the UI in 256 colors over the RasterOps NuBus card and used MacWeb, Network Time, and MacTCP Ping to access the Internet via a NuBus 10Base-T Ethernet card. 16MB DRAM in Bank A. ADB power-on and OS soft power-off are working. Accessed an external Zip drive. No overclocking; stock 16MHz. It was solid and stable for the hour I was using it, and I may leave it running Speedometer or something overnight.

So:

• CPU, NuBus, Modem Port, ADB, Clock/PRAM, ROMs, PSU and on/off logic, SIMM Bank A, SCSI, speaker, front panel LEDs.
• FPU, Printer port, Sound in, Sound Out, External Floppy. (Test these later)
• SIMM Bank B

Details

Re-lifted and re-seated the six 74F258 RAM multiplexors. Some had dubious solder (my fault). Then, literally two hours of just really dousing the upper face of the logic board with 99% IPA and brushing, brushing, brushing. Found a tiny bit of something, thread? dried up old spider leg?, possiby bridging two pins of UG7. Really focused on GLU due to its proximity to big leaky caps (already recapped). Inspected the underside of the logic board for scratches, broken traces, popped-off SMDs, debris. Let the board air dry for a solid hour.

The RasterOps 8L is proving a little tricky; it insists on coming up in 1024x768, but my monitor is detecting 640x480. So things keep landing off-screen. The RasterOps Monitors Extension shows it has four "Extended" modes all 1024x768 or larger, but the RasterOps controls claim I have no RasterOps card capable of pan, zoom, etc. This has three DIPs on the edge that set the resolution; I'll review to see if I have them inverted.

The Apple Macintosh 8•24 NuBus card may not work. Or perhaps I have not discovered the right DIP settings for my VGA adapter, yet. Would the Apple graphics card sync differently than the RasterOps card?

The PSU now behaves even with no NuBus cards installed; Previously, it would not chime without the Apple NuBus display card installed. Something was pulling down #PFW. Never figured that one out.. but now, it can run headless (as a file server, say).

SO GLAD to have one of these systems booting now. I could live without SIMM Bank B, but, I may try to debug that later. I want to just get the monitor resolution figured out and then enjoy this system for a little bit before plunging in again. I'll run MacTest and AppleCAT again in a few days to try and figure out SIMM Bank B and/or the Apple video card.

Next

There is still work to do. Placing SIMMs in Bank B keeps the system from chiming at all. Possibly bad SIMMs (they came with a DOA system) but not sure how a SIMM could stop a sad chime. And for some reason, cards do not sit fully 'down' in the innermost NuBus slot. I'll look to see if I melted any plastic on its plug. But more worrisome, if a card is in that slot and I put the lid on the chassis, the system stops booting. Logic board may be prone to flexing, pushed down by the lid through the too-high card.

Fun things to do after: I have a box of RAM SIPs to fill the empty sockets on the RasterOps card, and maybe that will let it do higher color. The RasterOps ROM may dictate that, though. Worth a shot.

 

[Addicted]

To debug the SIMM Bank B issue, I placed a SIMM into the four slots in turn, power cycling between attempts, hoping to see a clue that might narrow the traces to check. But this experiment returned the IIcx to sad chimes.

I think I may give up on this hobby.

 

[Addicted]

Positive things to share:

Expanding the memory of the RasterOps 8L does not increase the pixel depth. As suspected, that requires replacing the onboard ROM too. But the RAM alone does enable larger virtual screens, panning, and zooming features. It was cool to have a Nubus video card that could 'do' an enormous pannable screen.

 

[falecore]

Did you try 2,3 and 6 for the dip settings on Apple video card.

 

[Addicted]

I dusted these two boxes off this week. The better one booted again, and after adding the RasterOps 8L and an Ethernet card was stable for a few days until it began to just "stop" erratically. Lost video (black monitor screen) and disk activity would cease, but the PSU fan and disk drive keep spinning. I imagine this is a thermal intermittent somewhere, possibly a cracked logic board trace since shifting the PSU or reseating the NuBus cards sometimes revives it, however it now rarely stays up for more than tens of minutes.

Or, it may be that the burn-in of several days drove some other component to the end of its life. I'd love to think it is anything other than a defective PCB. Finding intermittent traces on such a large logic board may be beyond my patience.

The second IIcx is still in on-off-on-off-repeat mode, even though I have replaced the entire set of gates in the startup circuit just as I did the first IIcx. It may be the PSU. I have recapped the PSU, but the relay logic on its daughterboard might be faulty..

Thanks falecore for the tip.. I did try some of the other likely modes on the VGA adapter on the Apple Video Card and I'm 99% certain 236 was one of them. 167 works for the RasterOps. 2367 works for that monitor on the LC Macs. I'll write the modes I try down next time I have a NuBus platform that boots.

 

[Addicted]

I think one of the two IIcx's may now be stable. Taking what I learned from others here in eliminating the video jitter from an SE, I resoldered about a quarter of the joints on the logic board. Things I thought "looked fine" before.

I focused my rework on the `030 because the logic board came with leaky caps, and some large caps were near it. Then resoldered the ROMs and all the bus support chips on the Nubus third of the board, because the board seemed sensitive to flex over there (inserting SIMMs and/or cards).

I have been able to fill SIMM Bank B, which previously would kill the chime, evidently from flexing the logic board during insertion. I also have been able to swap a few Nubus cards in and out, which would have certainly killed the chime for the same reasons.

I won't feel confident until the system burns in overnight, but it has me pysched...

 

[Addicted]

Got some disk corruption overnight. My guess is the RAM address lines. I replaced the RAM bank muxes UH1 - UJ3 due to corrosion when I replaced the PSU on-off logic, and so now I just need to probe the traces and be sure I didn't break any, or leave opens between the leads and the pads. Will check UG7, UB7 and the address lines resistor arrays too.

The MacTest "short" RAM test says it will take 10 minutes but never finishes, and also overwrites the graphic card's frame buffer, which I do not think is supposed to happen. The "long" RAM test prints two characters of garbage instead of the starting message, then immediately locks up.

 

[falecore]

Thanks for circling back. This is really good info, I havent yet had to go this far down the rabbit hole with a IIcx/IIci, but I am sure one day I will. I am going to break out my IIcx's this week and mess around a bit.

 

[Addicted]

@falecore - glad that this lengthy topic is encouraging someone out there.

I have a CayMac ROM SIMM.. anyone know how to force it to do the RAM test on power-on when there's 16MB or more on the board? Will it provide the sort of bank or byte fingerprint that is displayed with a Sad Mac? I figure, run multiple flavors of RAM tests and get all the info I can to narrow the debug ahead.

 

[falecore]

@falecore - glad that this lengthy topic is encouraging someone out there.

I have a CayMac ROM SIMM.. anyone know how to force it to do the RAM test on power-on when there's 16MB or more on the board? Will it provide the sort of bank or byte fingerprint that is displayed with a Sad Mac? I figure, run multiple flavors of RAM tests and get all the info I can to narrow the debug ahead.

I do not, I only have Rominator II's. What OS are you using? Perhaps you can use that key command in Memory control panel to overide the ram test skip? IIRC, That might only work in OS 8 and above

Also, not sure if I mentioned it, but I had a IIcx last year with a bad SIMM slot which did not respond to my meager efforts at fixing it. Got sad chimes without Rominator and could not boot into ROM OS.

I was able to narrow down the simm slot but wound up selling it at VCF swap before I got around to fix it.

My issues were compounded by the fact I was testing using CayMac 4MB LED simms, which despite working in my 950 (as long as all four are not installed in same bank.) I have occasionally had issues using more than two Caymac 4MB simms at a time in IIcx's and in SE/30's. Cant recall if I ever tested them in my IIci.

In the 950 I staggered two Caymac 4mb simms with 2 silicon insider Purpleram 4mb simms on two separate banks and sad chimes went away (see attached vid).

 

[Addicted]

As of Monday night, the system was reliably running 7.5 with 16MB using Bank A only. No new signs of disk corruption appeared. I booted off of an external disk and put a clean install on the internal drive without any mysteries. The new internal install boots and is reliable,

I still have nothing concrete about whether I have a RAM issue or not. It is only a theory. The disk corruption I saw was that apps would run, but post errors -38, -39, or -40 when quitting. Those are codes for things like unexpected EOF and file not open.. weird. I ran Norton Disk Doctor and it found only bad modification dates. If I have a RAM error, it's either only for Bank B or it's the SIMMs I put there. I use SIMMs from MemoryMasters, which has been trusty.

I'll swap the Bank B SIMMs into Bank A and see if problems return.

---

I also discovered that both video cards are working. So, no more reason to suspect the Nubus slots, yay. I did not learn until yesterday that if an Apple video card gets no sense code (111) or an unsupported sense code at init, it will disable the card. I have to have a monitor attached plus dial in the right magic code on the VGA adapter DIP switches prior to boot, and I had my only monitor on the other video card to see what I was doing.

I put VRAM SIMMs in the 4*8's expansion slots and was thrilled to get millions of colors in 480x640, and 256 in 1152x870. My monitor isn't perfect for 1152x870, though. I have not found a combination of switches on my VGA adapter to obtain 1024x768 in millions, which I'd like best. At first its video appeared to be unsteady, with evenly spaced vertical wiggles wiggling down the screen, but using the auto timing feature on my monitor eliminated them.

The RasterOps 8L always runs 1024x768 (DIP switches on the card select that) but is also limited to 256 colors. It loads its driver even if no monitor is attached, which is why I was confused by the 4*8. I'll keep the RasterOps for a while to test Nubus and provide video on the other IIcx I am fixing.

 

[Addicted]

Serious egg on face moment. I had MODE32 V1.2 installed on top of System 7.5 .

Explains everything. Corrupted files. Why it took a clean reinstall to get back to level ground.

I still had the no-MODE32 clean install of 7.5 on my internal disk. I installed the System 7.5 version of MODE32, populated SIMM Bank B, enabled 32-bit addressing and VM. Initial tests look good; Ethernet performance is as expected; Finder memory info looks sane; system is responsive.

Let the burn-in begin.

 

[falecore]

Serious egg on face moment. I had MODE32 V1.2 installed on top of System 7.5 .

Explains everything. Corrupted files. Why it took a clean reinstall to get back to level ground.

I still had the no-MODE32 clean install of 7.5 on my internal disk. I installed the System 7.5 version of MODE32, populated SIMM Bank B, enabled 32-bit addressing and VM. Initial tests look good; Ethernet performance is as expected; Finder memory info looks sane; system is responsive.

Let the burn-in begin.

Hahaha....been there, done that.

 

[Addicted]

Summary: decalring victory with the first IIcx. Second IIcx now boots, but there's a (possibly single) problem ...

---

The first IIcx appears to be 'done'. I took the RasterOps video out of it, leaving it with the Ethernet, 8*24 video and an empty slot; I hope someday to find an 8*24 GC to round it out - or settle for a third-party Nubus accelerated video card. I may stuff more RAM into it, if I find a good deal. This system is likely to become an AppleShare server on my LAN, conversation piece, and legacy Mac-game system.

---

I spent the afternoon applying what I learned from the first IIcx towards repairing the second IIcx. Checking traces, cleaning, cleaning, cleaning!, and re-soldering even decent-looking pads. Good things followed.

1. Its PSU won't stay on (cyclic clicking of its relay if I hold down the power button), but the other PSU I have works fine in it. ADB soft-on works. So, the logic-board PSU on-off gates I replaced are working. Time to open the PSU up and explore; I've already re-capped it.
2. Instant sad chimes with the Apple ROMs.
3. With the ROMinator, mod-chimes then proceeds to boot from ROM disk, SCSI HDD, HDSC (Tash20), or floppy.
4. Detects and properly sizes RAM in both banks.
5. Inits the RasterOps Nubus video and displays a good image.
6. ROMinator ROM disk image contents work well. I can access the PRAM, play TimeOut, etc. Some games hang the system - I hazard to guess that it's when they attempt to play audio.
7. If I boot MacTest IIcx from floppy, the logic board tests proceed until it plays the standard A chord. This triggers a bus error.
8. Boots of 6.0.8 (Tash20) or 7.5 start loading extensions and inits, then hangs. Will try 7.5 with shift-key boot tomorrow.

So I have to debug the PSU and also look for a problem with GLUE or SWIM, I think. Reading more on what each ASIC does before I go further. With so many subsystems operational, there should not be many things left to check.

Still - it had never been beyond sad chimes until today, and it was great to see the SCSI, floppy, sound, RAM, ROM, Nubus, ADB all behaving. Progress!

---

Humorous observation: if you're in the "not even sad chiming" phase of a restore: validate the speaker first. No joke: early today this IIcx stopped happy OR sad chiming, and I wondered how in the heck did I go backwards? Then I heard the speaker sort of 'rattle'. Yup, the contacts inside the speaker cable's header were oxidized and needed sanding and tinning. Chimes restored.

 

[Addicted]

Somewhere, I once found a detailed write-up of the logic gates in the Iici/IIcx logic board PSU control, the state machine they drive.. which helped me choose the probe points to debug this same issue. I've been searching and have not found it again. Anyone have a pointer?

 

[Addicted]

Inspected the suspect PSU. The only suspicious thing was the thermostat. It closes on high temperature (105dC), and at room temp should be open. It was showing 105kΩ. I do not know if that 'counts' as open. Still an active part and in stock, so I may order a new one, just as a hope.

Kemet OHD3-105M

 

[Addicted]

Cancel that. I measured it in circuit. Shortly afterwards, I realized my mistake, dismantled the PSU again, and measured the thermostat after removing one of its leads. It's fine.

And now I have somehow broken the PSU. Too many awkward dismantlings and reassemblies. It won't power up. More challenges...

 

[dougg3]

Somewhere, I once found a detailed write-up of the logic gates in the Iici/IIcx logic board PSU control, the state machine they drive.. which helped me choose the probe points to debug this same issue. I've been searching and have not found it again. Anyone have a pointer?

I wonder if you're thinking of my writeup of the II/IIx power circuit? The IIcx definitely isn't exactly the same but I think it's pretty close, the big difference being that the gates are powered by +5V trickle from the power supply instead of the batteries.