Reverse engineering vintage PCBs (wrt: 3B2 main board schematics)
alan at alanlee.org
Fri Jan 9 10:12:34 CST 2015
There are actually twelve Signetics '153 PLAs, two registered 16R4s, and
a custom 18-pin ASIC that doesn't have a part number at all. So some
info will need to be derived through inference. Just the 12
combinatorial parts alone will need to be de-soldered to pull either
fuses or an observed truth table. So much programmable logic on the
board and the high density complicates functional level tracing. Memory
is actually on the left and the expansion bus is on the bottom. Power
comes down through the bottom back-plane from the PSU.
Manual is doable. It's just time consuming. The nice thing is all
components are layed out on a perfect 100 mil grid. I already have a
perfect board layout in eCAD with correct offsets for 100% of the 4,009
holes, including 7 mounting holes, just by looking at the board. I was
hoping to find someone on cctalk with a short-cut, beyond the
traditional methods I had already considered and you have echoed, for
doing the rest. I'm still looking at using a friend's home brew CNC step
table and a couple pogo pins to automate net-list creation. But we'll
see. I have a few more boards beyond the main board that I will trace
in-circuit by hand before I do the main board; for practice. Most option
cards are essentially complete 80186 PCs on each card! (PORTS, EPORTS,
QIC TAPE, StarLAN NAU, etc). AT&T must have reused the 80186 design on
everything for NRE cost reasons.
There doesn't seem to be any archived copy of the AT&T 305-490 "3B2
Computer Technical Reference Manual" anywhere nor any indication if it
included board schematics. Beyond that, I cannot find a hint of anything
previously archived that would have schematic level information on any
3B2 computer. That's why I'm doing this.
I know of Seth's work. I pulled ROM images for him which started me down
this rabbit hole! Otherwise my 3B2s would be quietly rusting in the
Thanks for the feedback.
On 2015-01-09 09:53, Al Kossow wrote:
> On 1/8/15 6:35 PM, Alan Hightower wrote:
>> I appreciate all the input from folks. With 3600 though holes and at least 4 layers, I'm afraid I will make mistakes using any manual method like ohm tracing without a check or balance. With so many permutations and little personal knowledge about the WE chipset, I'm sure I will miss an occasional trace that shoots to the other side of the board.
> Looking at the picture, it isn't that hard to decompose what's there into
> functional blocks. At that density, the PC layout guy will have kept the
> parts that are closely related together. You have an expansion bus interface
> on the left with bus buffering, HD controller, UART, floppy interfaces and a
> DMA controller for it on the top, processor upper right, four eproms, so there
> is a 32 bit data bus, and some sort of memory interface coming out the bottom.
> Get the SYS V driver sources and headers from Seth, that will tell you what the
> interrupts and DMA channels are hooked to and where the peripherals are mapped.
> Start with the CPU and trace the address and data bus paths. You don't need to
> buzz out every pin, just identify what buffer is connected to what and start
> ticking off the ICs from the top board layout out through the peripherals,
> expansion bus, and memory bus. Once that's done, you'll have decoders, state
> machines and glue logic which should be functionally associated with one of
> the big blocks. It doesn't look like there are many bipolar proms or PLA/PALs
> I see one 82S153 sequencer near the CPU, but Signetics parts (82S100, 105, 153)
> don't have protection fuses.
> I have one document that I haven't put on line yet that I think describes the
> expansion bus, I'll try to get that uploaded today.
> Is there a description somewhere of what the various 3B2 CPU boards had on them
> or what the memory board looks like?
> It's a drag the guy on eBay wants so much for the docs they've listed.
More information about the cctalk