Re: PDP-8 building instructions in your kitchen...

From: "Charles N. Choukalos" <chuckc@btv.ibm.com>

> Alan (I think was the one who posted this originally)....

Yup.

> Dude, please email/point me to the directions for this.  I'd love to
hack
> around and build a pdp-8.

OK, I'll help however I can. It was in 1983 that I last did this, BTW,
helping
a complete nutter who I worked with, just before the Falklands War. I
haven't done this for 20 years, so availability etc of chemicals is
unknown.

This is what I would advise:

First, get a good book on Digital Machinery. You're going to need this
in order to design the circuitry for a PDP-8. Fortunately, the PDP-8's
12-bit architecture is so mind-numbingly simple that it's no big deal.
FWIW my 3rd year DM course final exam was simple: "Here is the
instruction set of a CPU. Using the components you have built over
the course, design and build one. (2 hours)". Of course it was a PDP-8,
but I didn't know it at the time.

Anyway, I'd talk to your local education people about this, they may
be able to lend a hand here. Even supply some equipment. This is a
suitable Undergraduate or even senior High School project (with
supervision, see note below)

Second, breadboard your CPU using discrete components. In other
words, debug your design before even thinking about manufacture.
If you can pick up a Heathkit H-8, even better, as this has the design
work done for you, it's just a matter of soldering things together.

Thirdly, get a photoengravers kit. I'd recommend doing an Adult
Education
course (I didn't do this bit, my nutter friend did). Be warned, the
chemicals
involved are toxic, you'll need good ventilation ( a kitchen fan is
ideal).

Fourth, once you've got some amorphous silicon (dunno where you'd get it
if you're in the states, but shouldn't be too hard), use your Boys Own
Photoengraving Kit to make some simple components on a large scale.
We're talking about single transistors and diodes here. Get a good
multimeter and test, test, test.

Fifth, design and build some simple flip-flops using the same
techniques.
At this point, try to shrink the component size while maintaining
quality.
There'll be some trial and error here, it won't be a quick process.

Sixth, make a number of flip-flops on the same plate. Connect them to
become an ALU. Do the same so you have a working design for 12 bits
of memory (ie a single word).

Seventh, shrink them down as much as you can. You should be able to make
at least 1k of memory on one plate, with yield >30%  before going any
further.
This will be a painful process. It will take a long time. There will be
batches
with yields of 0%. Don't give up, back up one step and see what went
wrong.

Now try to implement your design on a large plate , just the CPU. Don't
get
discouraged, our yield on first attempt was less than 10% IIRC.

Shrink further, until either you have it as small as you want it, or
your
yield
becomes unacceptably low. We got a basic CPU in about a 1" square with
good yields, less than that we got one or two out of a hundred. But then
again,
we were amateurs.

The good thing is that once you've got it right, mass-production is as
easy
as
producing 1.

------IMPORTANT!!!!!------

Again, a stern warning: the chemicals are DANGEROUS. Don't just have
adequate ventilation and protective gear, have a plan on what to do if
the unexpected occurs. Have plenty of distilled water and various
neutralising
solutions handy, and a phone to call the local medics. ALWAYS work in
pairs.
I'm not joking. If one guy can't see to dial the phone, the other must
wash
out his eyes and call for him. If you have any children nearby DON'T
EVEN
THINK ABOUT IT. Really. This means you.

If you're in the states, I'd recommend a mutual waver-signing too. So
no-one
involved will sue the others.