Reverse Engineering the Macintosh LC Logicboard

[max1zzz]

So, pepperoni, mushrooms, bell peppers, and onions? Um... no, that's a pizza. Nevermind.

I'm with you up until the onion!

It occurs to me, that it might be helpful to people who are restoring original PCBs that got capacitor/battery bombed to have the raw scans of the various layers, so as to see what traces went exactly where...

The original scans aren't always the clearest, (Apparently my scanner won't scan at higher than 600dpi despite canon clearly claiming it is a 2400dpi scanner.....) I'll export the top and bottom copper from sprint to images as those are much clearer (I did just do this, then I remembered I still need to fix a couple of layout issues, so I'll upload them once I have done this)

 

[bdurbrow]

I'm with you up until the onion!

WHAT!?!?! No onion! What kind of uncouth, plebeian hooligan are you?!?!

---

2400 "digitally interpolated" DPI, perhaps?

Yes; but the original scans are of the original PCB; and your redraw will have small variations (those layout issues you mention)... so I would suggest posting both.

 

[Daniël]

Apparently my scanner won't scan at higher than 600dpi despite canon clearly claiming it is a 2400dpi scanner.....

As someone shooting film photography, (film) scanners lying about their true DPI is all too familiar. Sure, the software can save 2400DPI pictures, but the optics aren't going to push beyond 600DPI. It gets worse with my Epson V600, claims a grandiose 6400DPI despite it optically really not being able to go beyond 1600DPI or so, it's terrible.

 

[mg.man]

Well... as it happens, I do have a Creo IQSmart 3 - which is an A3, 5500 (true optical) flatbed scanner. Probably too late for the LC 'board (I assume it got sanded?), but if the LC III has not met this fate, I'd be happy to scan that...

 

[max1zzz]

I didn't actually sand the LC, didn't need to in the end (I managed to figure out all the splits in and connections to the planes with just a bright tourch) Also managed to get the III board scanned to a acceptable enough quality for reverse engineering

However I might send a stripped Classic II board and maybe a stripped LC II and 475 board to you for scanning

 

[mg.man]

However I might send a stripped Classic II board and maybe a stripped LC II and 475 board to you for scanning

Sure! I also need to send some bits your way - like those 30 pin angled sockets... I'll PM you on our other thread.

 

[max1zzz]

So, finally got around to doing some testing and everything seems to be working perfectly
Audio in and out both work fine, Ext scsi works fine, PDS works fine (At least a PDS Ethernet card works and I was successfully able to connect to my network), I connected up a real floppy drive and verified it formats, reads, writes and ejects the disk fine, serial works (at least on a basic level, I was able to send and receive data on both ports using a serial to usb adapter connected to a modern machine)

As I am now confident these are fully working I'll correct the errors in the design and get another batch assembled

Is there anyone wanting on of these boards? (They'll be £35 each excluding shipping, Assembly will be available for a additional cost) If not I'll just order the minimum quantity. This run of boards will be in matte black, any future runs will be in blue

 

[cheesestraws]

I'd love one with assembly if I can find where my donor board went

 

[max1zzz]

I'd love one with assembly if I can find where my donor board went

Great I'll PM you in a little bit

So, Just when I thought it was all working I have found another issue.
NVRAM data is not being kept when the board is powered off, this appears to be because when you power off the voltage on the EGRET's power supply line dips while it is transitioning from AC power to backup battery power (This is done by pulling the base of Q2 low through a 3.3M resistor, when on AC power this is pulled up to +12V by a 470R resistor. Q2 is a PNP transistor that connects the backup battery to EGRET's power supply line).

On a stock board the voltage never dips below that of the backup battery. (At least my multimeter doesn't see it, so if it dose dip it happens faster than the refresh rate of my multimeter)

If I connect the battery to EGRET's power supply line through just a diode NVRAM data is retained just fine, also if I tack a higher value cap over C9 (Like on some early / pre-release boards) NVRAM data also survives just fine so it appears the transition is just not happening quick enough

I have verified the pull up / pull down resistors are the correct value and have tried swapping Q2 and EGRET from the known good board and the issue still persists. I'm now at a bit of a loss as to what the issue is....

 

[max1zzz]

Swapping the 3.3M for a 100K also fixes the issue, I'm OK with ths fix as it dosen't require modifying the board or and bodges, but it still bugs me that the original board seems to work with a 3.3M in that position!

 

[max1zzz]

V1.1 PCB's are here!



Unfortunatly I forgot to connect /OE and /CE on the one chip I missed the connection on the V1.0 PCB so these will require a jumper wire....

 

[max1zzz]

After much confused probing of the board the mystery NVRAM issue has been solved!

So what was the issue? something bizarre and abstract? Not really....

One single via which should be connecting pin 4 of the MC34064 at UD9 to GND was instead connected to nothing

One small piece of bodge wire later (Connecting pin 4 of UD9 to a very conveniently placed ground pin on the Egret)

Testing the board with this fix in place reveals that the NVRAM data now survives power cycles just fine with no component substitutions

So what's going on here? UD9 is a MC34064 undervoltage sensing IC, it's output is connected to pin 12 of the Egret. What exactly pin 12 of the egret dose is not completely clear, it's not the reset pin of the MCU and the BOMARC schematic of the LCII mark it as "reset in" so perhaps once the voltage of the +5V rail starts dropping this instructs the Egret to yank /RST low and stop the CPU etc running and this slows down the rate at which the voltage on +5v drops, giving the power supply circuit for the Egret a little more time to switch over to the battery? Or maybe this simply instructs the Egret to go into a low power standby mode? Whatever is supposed to happen UD9 isn't going to be able to tell the Egret do it with no ground reference!

Either way with pin 4 of UD9 grounded the EGRET's supply voltage drops slowly rather than the sharp drop and bounce back I was seeing before and NVRAM data is not lost

 

[cheesestraws]

Nicely found!

 

[max1zzz]

Final boards are in, and one has been assembled and tested

This revision works fine with no issues or patches and NVRAM is preserved correctly when powered off
Many thanks to @djhaloeight for providing the donor board used for this one

 

[cheesestraws]

Beautiful. Very well done. Really pleased

 

[demik]

Nice job ! Is the small internal SCSI thing a terminator ?

 

[max1zzz]

Beautiful. Very well done. Really pleased

Thanks, I'm pretty pleased with how they came out myself

Nice job ! Is the small internal SCSI thing a terminator ?

Thanks and that little scs device is one of my 50pin to 80pin scsi adapters, I was originally going to boot form a scsi HDD bur then remmebered the drive was in a LC, somewhere in a massive pile of LC's sitting in the corner of my room so I elected to boot form my external SyQuest drive instead

 

[djhaloeight]

Sweet!

 

[demik]

I did, got chimes of death. Also tried dropping the roms form a LC II onto the board but got the same result, I guess the memmory controller just won't see more than 2MB of ram on the motherboard, Presumably the extra positions are to allow the use of lower density ram (where 512Kx4 IC's ever a thing?)

On a slightly amuseing note, the LCI (with a stock RAM config) will boot just fine with the LCII roms, and even identifys in Mac OS as a LC II I'm not sure if this gives any real advantage but it was amusing to see

Just noticed this on the LC II Developer Note:

• Changes to RAM sizing code to support 4 MB of permanent RAM on the main logic board and up to 6 MB of expansion RAM.

So yeah, no more than 2MB on the motherboard with stock roms at least.

 

[djhaloeight]

I'll be interested to see the result of your work. The other challenge with the LC is that the cases are quite afflicted with brittle plastic, so we may need @maceffects to make new cases for them.

- Alex

I'd love to see that myself! I've got one of each of the LC family, (LC, LCII, LCIII+, and Performa 475) and the plastics are definitely brittle. New cases would be sweet!