SE/30 with Micron Xceed Grayscale & Socketed Daystar Acceler

[JDW]

I just made a 720p YouTube video that shows the vertical line problem:

 

[JDW]

Still not sure what edition video card this is. The evidence I see points to it being a Micron Xceed MacroColor 30HR instead of just a normal "MacroColor 30." The reason I think this is because it came with a floppy disk and on the floppy it says "30HR." The README inside that floppy goes on to say that the manuals that ship with the MacroColor 30 are the same manuals that ship with the 30HR version too.

Of course my card looks different from the photo shown here:

http://sites.google.com/site/micronxceed/

But if you scroll down on that page it says that V3.5 comes with the 30HR, and the floppy I have shows V3.5.

I simply don't have a VGA monitor to test if it will get 16 million colors or a max resolution of 1024x768, which is described here:

http://home.earthlink.net/~gamba2/microngray.html

Sure would like to fix that vertical line problem though. Strange it never shows when in B&W mode -- only in the grayscale modes.

(See my previous post for the YouTube video I made.)

 

[trag]

Sure would like to fix that vertical line problem though. Strange it never shows when in B&W mode -- only in the grayscale modes.

In black and white mode, only 1 bit of memory is used per pixel. So your VRAM is only actively using 512 X 384 = 196, 608 bits or 24KB of memory.

When you go to 8 bit greyscale, it's 192KB of memory. In 4 bit greyscale, it uses 96 KB of memory.

So the problem row in your VRAM memory is somewhere above the 24K mark, and somewhere below the 96K mark, it appears.

 

[Gorgonops]

A sort of dumb question... when an SE/30 with one of those boards is set for 1 bit (B&W) mode is the video generated by the Micron card, or is it generated by the onboard video?

This is a wild guess, but the way the line marches across the screen when color depth is changed makes me think what's going on is there's a bad bit in the SRAM output cache of one of the RAM chips, not a bad bit in the main memory. I forget the precise details from the datasheet for those dual-ported RAMs but I vaguely recall that they had an SRAM buffer with dedicated clocked (auto-increment?) data output pins. (And that the RAM chips responded to commands to selectively load that output buffer with pages from the bulk DRAM storage cells.) The distance of the line from the left side of the screen doubles when color depth is dropped from 8 bit to 4 bit, and doubles again when it goes to 2 bit. This is complete and utter speculation, but I suspect what's going on is that during the horizontal refresh interval the SRAM cache on the dual-ported RAM is loaded with the memory contents necessary to render a line of pixels, which is then clocked out. This is why the bad bit appears on each affected horizontal line in the same vertical location (at a given bit depth.) This theory also explains why the line is missing whether the card handles 1-bit video modes or not. If the card *does* handle 1-bit the bad bit location is too far to the right to show up on a 512 pixel wide monitor. If you used a higher-res external monitor you might see it.

If I'm correct, which is a big if, the fact that the line is dashed vertically tells us what the memory organization of the card is. It's interleavened with each bank of RAM serving every other scanline. One bank works perfectly.

 

[trag]

A sort of dumb question... when an SE/30 with one of those boards is set for 1 bit (B&W) mode is the video generated by the Micron card, or is it generated by the onboard video?

If the gray scale adapter is installed, and no external monitor is connected, then the video is generated by the Micron card, unless the (thin) documentation we have is inaccurate. One bit color depth is mentioned in the Micron card concept description. And the description of the logic for switching what drives the internal monitor only discusses whether an external monitor is attached, not whether 1 bit color is selected. Of course, it's possible I overlooked something.

This is a wild guess, but the way the line marches across the screen when color depth is changed makes me think what's going on is there's a bad bit in the SRAM output cache of one of the RAM chips, not a bad bit in the main memory.

That's entirely possible. The difficulty is that I think we could construct two or three or four possible scenarios depending on how we assume the memory is organized.

I don't have time to work through all the possibilities at the moment. I will leave with the quick thought exercise. The SRAM is 512 X 4 per chip. So eight chips per bank gives one a 32 bit bank, with 512 words stored in it.

At one bit color, sixteen of the 32 bit words are used per 512 pixel line. If there's one bad bit in one of the chips, then the bad bit would appear once every 512/16 lines = one pixel per 32 lines. Probably not visible without very careful examination.

At four bit color, the bad pixel should appear once every eight lines. At 8 bit color the bad pixel should appear in every fourth line.

I haven't looked at the videos, as I usually have trouble with them. Is that what the data shows?

 

[Gorgonops]

That's entirely possible. The difficulty is that I think we could construct two or three or four possible scenarios depending on how we assume the memory is organized.

I don't have time to work through all the possibilities at the moment. I will leave with the quick thought exercise. The SRAM is 512 X 4 per chip. So eight chips per bank gives one a 32 bit bank, with 512 words stored in it.

At one bit color, sixteen of the 32 bit words are used per 512 pixel line. If there's one bad bit in one of the chips, then the bad bit would appear once every 512/16 lines = one pixel per 32 lines. Probably not visible without very careful examination.

At four bit color, the bad pixel should appear once every eight lines. At 8 bit color the bad pixel should appear in every fourth line.

I haven't looked at the videos, as I usually have trouble with them. Is that what the data shows?

Wait. I don't follow your math... no wait, I get it. Your math makes sense if you assume "bad bit" means a bad row or column. What I meant was a "bad bit" in the sense that there's a single stuck bit in an entire 512x4 bitfield. (IE, one bad memory cell out of 2048.)

For cheap thrills I fired up BasiliskII with a 512 pixel wide screen, overlaid a window from a graphics editor, and used a crop tool to roughly measure the distance between the left edge of the screen to roughly where the line appears in 8 bit graphics mode. It seems to be somewhere around pixel 96. (Over/under about 5 pixels either way. Let's call it 96 because it's a more convenient number than some others.) If we assume that on a given line the memory layout is entirely linear then pixel 96 in an 8-bit color mode would be inside the 24th 32-bit word. (We get 4 pixels per word, so 96x4=24.) That would mean then that in one of the 8 4-bit wide chips making up the bank there's a bad bit in the 24th word of the 512 word array. Such a problem would create the following symptoms at these color depths:

32 bit: wrong color pixel at location 24

8 bit: Wrong shade pixel around* 96

4 bit: wrong shade pixel around* 192

2 bit: Wrong shade pixel around *384

1 bit: wrong shade pixel around* 768.

*(You have to say "around" because the exact location would depend on which SRAM chip had the bad bit.)

The reason I said the memory arrangement is probably 32-bit line-interleaved is because the line is apparently dotted vertically. However; an alternative is that the memory arrangement is 64-bit linear, in which case the bad bit gets moved to around position 12 in one of the 16 chips. If that's the case for the artifact to appear in the same place every other horizontal line the card would have to be set up so when it does video modes requiring less than 512 words per line it uses a single buffer refresh to handle two (but no more than two) scanlines at once. (If it simply marched the buffer through an in-memory framebuffer in a completely linear fashion then the bad pixel would be staggered across vertical lines, which it is not.) Other arrangements are conceivable, of course. It would be semi-instructive to connect the card to an external monitor and see if/where the bad pixel shows up in video modes that can't be rendered internally, such as a wide 1-bit and a 32 bit color mode.

(This has me wondering actually if all resolutions supported by the card are set up in the same "grid" on a per-colordepth basis, IE, a 512x342 SE-size screen has the same memory layout as, say, a 1024-pixel wide grayscale screen. The system merely displays the first 512 pixels of any scanline and just throws away the rest. That would actually simplify a lot in the driver. The same calculations for drawing slanted/vertical lines would apply at any resolution, you'd just need Quickdraw to understand that there's "dead space" at the end of each line.

Say, that gives me an idea! The Micron Xceed supports some sort of "virtual desktop" thing, right? Enable that and try dragging the desktop around. It would be interesting to see if the line moves. Whether it does or not would depend on how the output buffer is used.)

 

[tt]

JDW, it might be easier to explain if you say draw a black or plain white background in hypercard and make lines in the exact locations and mark them. Then you could count pixels. Although it might be tricky to do fullscreen (maybe a really old HC version) since the micron xceed adds an extra pixel in one of the directions.

 

[JDW]

With a spray can and the little tube it comes with, you should be able to squirt solvent in the hard to reach spots... You could also try the dishwasher method. A soak in hot water will probably get the hidden spots. But....it's also possible that the goo ate a via.

Thank you for the advice. Since I don't have a dish washer (such are uncommon in Japan), I just spayed an enormous amount of Ethanol on the board while it was at an angle to let it drain. At the same time I used a can of compressed air to blast the Ethanol off the board quickly and under chips and connectors. I can truly say the underside of the PDS connector is now very clean! I cannot see the underside of every soldered in chip though. And I cannot help but feel some deep sense of reservation about dunking a PCB in water!

I also pulled the socketed chips as you suggested and I reseated them carefully. I then too the pains to use a continuity checker, touching the top of each pin on the chip with one problem and then touching the other probe to the opposite site of the logic board, proving that each pin was perfectly connected to the board.

And yet, when I booted I again got the sad Mac "chimes" and the horizontal lines, as shown in my new 720p video here (please take time to watch it):

I need to get one of my SE/30s in running order and put back together with my MacCon Ethernet card so I can get files out of the machine. I cannot physically insert my MacCon in the Xceed machine, and I don't have a USB floppy drive, which would be the easiest means of getting files off the SE/30 and onto my Intel iMac so I can upload them to the web for all to see. I will work on this over the next 48 hours.

 

[bigmessowires]

OK, I'm wading into a thread without having really read the whole history, but here are a couple of thoughts:

1. If you suspect a power supply problem, then you might want to monitor the 5v output with a scope, and set it to trigger on anything below 4.8v or so. By measuring with a multi-meter, you won't catch brief power supply glitches. I think a bad power supply is likely to suffer brief voltage droop in an instant of high current demand, rather than a sustained period of low voltage that could be measure with a multi-meter.

2. If you suspect a DRAM or SRAM chip has gone bad, can you try poking at the chips with a probe while the computer is running, and see if you can affect the display in a way that identifies which chips contribute to drawing which parts of the display? Then you might be able to determine which chip is bad. Offhand I'm not positive how you'd do that, but you might be able to see a visible result if you use a resistor of around 1K ohm to temporarily connect a chip's data pin to power or ground. On the other hand, there's a chance you might damage something that way.

 

[PowerPup]

Found this applefritter topic where someone posted the rom dump and some assembler files. He also shared the schematics here.

I have re-uploaded the rom dump file with the proper file extension here: http://www.mediafire.com/?84wasyd26m74y6l

Not sure if there's anything that will help, but it may end up useful somewhere along the road.

 

[JDW]

...you might want to monitor the 5v output with a scope, and set it to trigger on anything below 4.8v or so. By measuring with a multi-meter, you won't catch brief power supply glitches.

Good idea. That will have to wait another week though, since I would need to bring a scope home from the office and have time to do the tests. But what I can say for now is, my PSUs work perfectly when powering my main SE/30 machine, which has 128MB RAM, a DiiMO 50MHz 68030 PDS accelerator and a PDS MacCon Ethernet card.

So I cannot say yet if the PSU is the problem. All I can say is that when I install the socketed Daystar 50MHz 68030 accelerator onto the SE/30 logic board, the machine locks up after a while. If the accelerator were bad, I would expect it to not boot at all or lock up right away, but that is not the case. So I then was suspecting it could be that "socketed" accelerators perhaps require a much more stable 5.00v source. It's just a guess. But to test that theory I would need a totally different PSU that could supply 5.0v at say 10A or more continuously, so there are no voltage dips.

So this is why I am currently using the Xceed setup on a non-accelerated SE/30. Having grown accustomed to my 50MHz DiiMO, I must say that going back to 16MHz is quite painful indeed. It's also annoying because I want to see this socketed accelerator work without lockups. When the socketed accelerator does work though, it's just as fast as my 50MHz DiiMO (obviously, since both are 50MHz 030's).

If you suspect a DRAM or SRAM chip has gone bad...

In terms of main system RAM on the SE/30 logic board, there are no worries whatsoever it is bad. It is known good RAM. Indeed, I have numerous known-good SIMMs and I have tried them all. In good logic boards, they work fine. But in my socketed logic board that I recapped yet still gives me simasimac at cold boot, no RAM or ROM works. So I know something is awry with that logic board. I just don't know what. I really want to know though because it's my only other socketed logic board. And I want to resurrect it so I can try the socketed Daystar upgrade on it to see if the logic board could have anything to do with the lockups.

Now if your comment was referring to the RAM on the Xceed card, I'm not that far yet. But hopefully the information I am providing below will assist those smarter than I to direct me to the specific area to test/fix.

...draw a black or plain white background in hypercard and make lines in the exact locations and mark them. Then you could count pixels.

I've done better. I've taken screen shots on the SE/30 and at the same time I've shot photos. The screen shots don't show the dotted line, but I put both in Photoshop, resized the photo to overlap the screen shot, then I simulated the dotted line with perfect pixel precision.

My full set of photos and associated screen shots are found here:

https://docs.google.com/leaf?id=0B3mQLg8d1xThZDhhYWE3N2MtMzkzNy00NDBmLTg3NDUtM2MwNzQ3ODUwZDEy&hl=en_US

Measuring from the left side of the screen, we have the following data for the vertical line anomaly:

Gray Depth     Dotted Line Location

256                96th pixel from the left

16                  192nd pixel from the left

4                    383rd pixel from the left

If you doubt me, download the GIF files and examine them yourselves. I increased the size of the screen shots by 400% in Photoshop using "Nearest Neighbor" so it is a perfect 4x magnification without any fuzziness, making it easier to pick out individual pixels and count them.

The only odd thing I noticed is that the distance from the left (to the dotted line) for 16 grays is exactly twice that of 256 grays. So I thought 4 grays should be twice that of 16 grays, but I am 1 pixel short for some reason (393px instead of the expected 384px).

Also note that despite the fact I had the Monitors control panel set to grays, two of my screen shots came out in 256 colors for some reason. I was switching back and forth between CANVAS 3.5.6, and CANVAS was doing strange this to the grays in the Monitors control panel, so perhaps that contributed to 2 of the shots being in color and the other being gray. Not sure. Anyway, you'll notice that I have 2 photos labeled "256." One is labeled "smooth." You can clearly see the difference between then when looking at the bottom of the Monitors Control panel. The smooth one was created in the Finder, when I switch from another depth to 256. But the other non-smooth 256 file is showing choppy grays at the bottom of the Monitors Control Panel. This happens after I switch from the Finder to Canvas. I think Canvas is doing this, switching around the grays.

But something else to consider...

Have a close look at the photo (not the "smooth" one) and screen shot for 256 grays. Note how there is a strange set if 9 stray pixels appearing toward the right side of the screen, between "PM" and the "CC" menubar icon, running in a vertical fashion from the top to bottom of the screen. They appear to be shades of red and dark blue only. What's interesting about this is that those pixels even appear in the screen shot! So what do you make of this? Clearly, Canvas caused this because if you look at my other 256 "smooth" photo you will not see those artifacts at all (when the grays at the bottom of the Monitors Control Panel are smooth).

Lastly...

For those of you just entering this thread, please be sure to see the two videos I've made and my photos, which visually show some of the things I've been talking about in my recent posts:

https://picasaweb.google.com/103365672326265854011/SE30MicronXceedGrayscaleVideoWithDaystarSnapOnAccelerator#

 

[bigmessowires]

OK, I think I understand now. You have three unrelated problems being discussed concurrently.

1. SE/30 logic board with all upgrades removed shows horizontal lines and won't boot (SimasiMac), even after re-capping

2. Daystar 50 MHz accelerator kind of works, but caused occasional lockups.

3. Xceed grayscale card works, but shows a vertical line whose position changes with the monitor color depth.

Does that sound right?

For 1, there are a couple of other suggested remedies for SimasiMac (other than re-capping) mentioned here: http://www.biwa.ne.jp/~shamada/fullmac/repairEng.html#SimasiMac I know nothing about it, but maybe it helps? For all I know, you may have written that page yourself.

For 2, it could be anything really. Power supply problems is one possibility, but I'd guess it's more likely a borderline component on the Daystar card, maybe triggered by temperature.

For 3 I agree with trag's earlier statement that it's very likely a VRAM problem. The way the line moves about with color depth makes that pretty clear. His suggestion of blowing cold air on the chips sounds much better than my idea of probing with a resistor, although the end goal of identifying the faulty VRAM chip is the same. Assuming you could find the faulty chip, do you have any source for replacements?

 

[JDW]

Does that sound right?

100% correct.

For 1, there are a couple of other suggested remedies for SimasiMac (other than re-capping) mentioned here: http://www.biwa.ne.jp/~shamada/fullmac/ ... #SimasiMac

I have been performing SE/30 logic board recaps for several years now. And until now I have had one one other board (the standard, non-socketed type) which continued to show horizontal stripes at cold boot despite the new capacitors. And at that time, I tried the 1k-ohm pull-up resistor across UB11's pins 7 & 15, but to no avail. But I've never been able to figure out that UB11 part logically, since UB11 is a "sound" chip and as my video shows I am clearly getting sound -- indeed robust, high volume sound! So if UB11 really did not a pull-up, would it not adversely affect sound? And I don't see why that sound chip would result in horizontal lines on the screen. And lastly, since I've never had access to a Bourns filter (for RP2), I never have been able to rest that.

For 2, it could be anything really. Power supply problems is one possibility, but I'd guess it's more likely a borderline component on the Daystar card, maybe triggered by temperature.

The only way for me to test if a rock solid 5.00v power source would solve the problem is to have an ATX or similar high-output PSU directly connected to the SE/30, which I currently do not have.

The underside of the Daystar board contains only ceramic SMD caps, SMD tantalum caps, and SMD resistors -- none of which are prone to failure even after many years of constant use, nor are any of these physically marred/burned/cracked:

https://picasaweb.google.com/103365672326265854011/SE30MicronXceedGrayscaleVideoWithDaystarSnapOnAccelerator#5641720098640690242

Therefore, if indeed there is a problem on this board, one would logically suspect the chip-side (which is not something to troubleshoot for the faint of heart):

https://picasaweb.google.com/103365672326265854011/SE30MicronXceedGrayscaleVideoWithDaystarSnapOnAccelerator#5641720071352856226

For 3 I agree with trag's earlier statement that it's very likely a VRAM problem. The way the line moves about with color depth makes that pretty clear. His suggestion of blowing cold air on the chips sounds much better than my idea of probing with a resistor, although the end goal of identifying the faulty VRAM chip is the same. Assuming you could find the faulty chip, do you have any source for replacements?

Well, not having a can of that cold air, I've not yet tried it. But as you properly point out, even if I do try it and identify a chip with it, I have no chip replacements nor access to any.

So for now, I look at those 3 problems in the following order in terms of repairability (with 1 being the easiest to repair and 3 being the hardest):

1. Simasimac board

2. Daystar accelerator

3. Xceed PDS video card (vertical line problem)

This by no means even 1 is "easy," as I would have solved it by now had such been the case. Further technical thoughts would be appreciated. Thanks to all for your input.

 

[JDW]

With regard to my Simasimac SE/30 logic board, I have made some headway. Aided by my Apple and BOMARC schematics, I used a DMM's Continuity Check feature (beeps when both probes are shorted) to painstakingly check the board, pin by pin. I started off in the most logical manner, assuming the problem may center around the ROM. I felt that most logical in light of past experience with the ROM showing unusual artifacts on the screen. Furthermore, the capacitor that leaked the worst was SMD cap C12 (47µF). Even though I cleaned that area thoroughly more than once with high grade dehydrated Ethanol, the problem remained. So I decided to investigate using Trag's advice about thru holes.

I checked the ROM pins one-by-one, with one probe on a ROM pin and the other probe testing various legs of different chips to which that pin should lead (accordingly to my schematics). All the "Address" pins of the ROM SIMM checked out OK except one, pin 45 (A22). Pin 45 of the ROM should lead to pins 3 & 6 of UI4, as well as to the Glue Chip (IU8) and the CPU. I found A22 connecting the IU8 to the CPU, and the CPU to IU4 and IU4 back to IU8, but none of those points led back to pin 45 of the ROM (A22). I then desoldered C12 and found the problem.

Pin 45 of the ROM goes to a via, then on top of the board nearest C12 and the battery, then to another via about 1.5cm away. It's that 1.5cm long trace that is broken (on the top of the board). And it appears the break is at the top of the thru hole via, on the top of the board. Have a look at my photo:

https://picasaweb.google.com/103365672326265854011/SE30MicronXceedGrayscaleVideoWithDaystarSnapOnAccelerator#5642529984101043730

Note the white silked text "C12" and "D1" and "D2". Now note the "+" sign above C12, and that cap's corresponding via. Just 2mm above that is another via. That leads to pin45. Follow that via straight down to just 2mm below C2, and you can see the other via. That's the broken trace, broken at the bias. It makes logical sense too, as that is where the majority of the spilled fluid was, sitting atop those two vias for a few years I would expect. The via above the "+" in the photo is hard, but the other one is soft when I press on it with a metal pin.

I could just run a wire to attach those two points, but I am not sure if that is best. This is a multi-layered board. And unfortunately, I do not know if there are more than 3 layers. Further, I have schematics but I don't have PBC layout files which would show how many wires splice off the same via. Hence, if I've got a bad via, it's easy to fix that via by running a wire on the top and bottom of the board, but that won't help me if there are breaks to that point in different layers WITHIN the board.

Since I am not experienced fixing vias of multi-layered boards, I would appreciate your advice. (And if anyone happens to know how many layers the SE/30 board is and/or have access to PCB layout files, I would certainly appreciate getting that information from you.)

PM me if you wish to have access to my BOMARC and Apple SE/30 schematics.

I look forward to your advice. Thanks.

 

[trag]

And yet, when I booted I again got the sad Mac "chimes" and the horizontal lines, as shown in my new 720p video here (please take time to watch it):

It seems unlikely after the thorough cleaning you gave it. There could be that hidden nugget of conductive detritus... I'd be more inclined to believe an eaten via which doesn't display it's destruction to the eye alone.

 

[trag]

I could just run a wire to attach those two points, but I am not sure if that is best. This is a multi-layered board. And unfortunately, I do not know if there are more than 3 layers. Further, I have schematics but I don't have PBC layout files which would show how many wires splice off the same via. Hence, if I've got a bad via, it's easy to fix that via by running a wire on the top and bottom of the board, but that won't help me if there are breaks to that point in different layers WITHIN the board.

Ah, now I'm caught up with your latest post. I would run the bypass wire. I'm pretty sure these are four layer boards, with internal ground and power planes and signal layers on the top and bottom. More importantly, I can't imagine how running a bypass wire would do any harm, even if the board has more than four layers. At worst (I think) it just won't work.

You might also probe A22 in the PDS slot and see if it connects to either side of your broken connection. If it does connect that would seem to reduce the likelihood that there are additional broken connections inside the board.

Oh, and regarding #2, the Daystar Accelerator. I agree that there are a lot of variables, so try altering the ones which you easily can. The 68030 is socketed. Try swapping it with one from your reliable PowerCache board. Note down the date codes on the two chips first. It's surprisingly easy to get them confused in the heat of testing and not know which came from where.

Similarly for the FPUs. Those chips, at least, are easy to swap for a quick reliability check.

Finally, in my post above, it looks like I was wrong about the SE/30 on-board VRAM chips being socketed. All the chips (PALs) below them in your photo are socketed, but it looks like the VRAM is soldered down. With your discovery of a broken address line to the ROM, it's probably irrelevant, but I thought I should mention it.

 

[JDW]

I'm pretty sure these are four layer boards, with internal ground and power planes and signal layers on the top and bottom. More importantly, I can't imagine how running a bypass wire would do any harm, even if the board has more than four layers. At worst (I think) it just won't work.

First of all, thank you for the reply. And thanks for viewing my videos. I talk a lot in those videos, so you may wish to find a way to get the sound to work!

In my examination of several thru-hole vias last night, I noticed that some wires from connectors will route from a connector pin to a via, but then you cannot physically see any wires on the top or bottom of the board leading from that via to another place on the board. Yet, if you hold the board up to the light so you can see light shining through the back of the board, you can then see a tiny little wire inside the board leading away from that via somewhere else. And this holds true for "address" lines that are certainly not "ground or power planes." Because of that, I was thinking there is a "signal" plane (perhaps more than one) inside the board. And that's where a PCB layout diagram for the SE/30 motherboard would come in handy. If I know all the lines that lead to/from a particular via, I then don't need to worry about resurrecting the via itself, as I could run direct wires to all the points.

I agree with your last sentence that running a wire from the bad via on the top and bottom of the board would not harm anything, it may not solve the problem either. But then we don't know if the problem is not solved because there are other wires connecting to that via inside the board (which my new wires on top/bottom do not re-attach), or if there are other eaten traces. Only by truly fixing each problem can I then rule out that fixed section as possibly contributing to additional problems I find.

 

[bigmessowires]

Hmm, whether or not there are broken signal traces in the inner layers, can't you proceed as you have been, checking out the board with a continuity meter and comparing to the schematics? Once you add this first wire, you can confirm that the ROM pin 45 (A22) connection is working now. If that doesn't get the Mac running, then you can keep checking for more continuity problems and fix them with additional wires, verifying each fix as you go.

 

[JDW]

can't you proceed as you have been, checking out the board with a continuity meter and comparing to the schematics?

Of course. But that is more time consuming and error prone as a result. A PCB diagram would show me all I need to know about each and every via at a glance.

It's not a matter of sheer laziness. I really try to make the most of each free moment because I don't have that many of them.

Furthermore, I have found some minor inconsistencies between the BOMARC and Apple schematics, which lead me to believe there could be an error or two in there. I do keep notes though, so if I do find an error, it will be documented and I will perhaps even change the schematic at some point (to provide to others who ask for them).

So later this evening I will carefully examine the schematics again, then decide if the two bad bias truly can be restored simply by soldering in an external wire between them.

 

[trag]

In my examination of several thru-hole vias last night, I noticed that some wires from connectors will route from a connector pin to a via, but then you cannot physically see any wires on the top or bottom of the board leading from that via to another place on the board. Yet, if you hold the board up to the light so you can see light shining through the back of the board, you can then see a tiny little wire inside the board leading away from that via somewhere else.

That does sound like more than four layers. Perhaps six layers. I think the chances of a complete PCB diagram turning up is low. Unless someone made one by examining the SE/30 board. Apple certainly isn't going to have released it. Unfortunately.

You might apply a large dollop of solder flux and apply heat to the hole. That should remove the oxides in the hole. Perhaps use some desolder braid, or maybe blow the hole out while it's hot. I'm not certain of the best way to clean it out, but I'd work toward cleaning out the oxides and getting fresh solder adhered in there.

Once it's cleaned out, insert a bit of stripped wire wrap and solder in place. There's a pretty good chance that if there are additional layer connections in there, the solder and wire will make a connection to them.