vintage computers in active use

[Jon Elson]

On 05/27/2016 01:30 AM, Eric Smith wrote:
> I wrote:
>> Another CHM volunteer (from the PDP-1 Restoration Project) and I
>> pushed for an IBM 360/30 Restoration Project, and the ability to build
>> replacements for failed SLT modules was part of our plan.
> On Thu, May 26, 2016 at 9:40 PM, William Donzelli <wdonzelli at gmail.com> wrote:
>> I am still trying to figure in which universe are SLT modules so rare
>> that one needs to fabricate replacements.
> As far as I can tell, if I suddenly need a specific SLT module, the
> odds of finding that specific module at any given time on eBay is
> essentially zero.
>
> Obviously if we could find an authentic replacement, we'd prefer that.
> We didn't want the entire restoration to be dependent on having to
> find authentic replacements.
>
> Some SLT modules are far more common than others. I don't know how
> many different SLT modules are used in the 360/30, nor what percentage
> of the SLT modules in that machine are common ones.  When we restored
> the DEC PDP-1, it contained quite a few DEC System Modules from DEC's
> standard catalog and that we had spares of, but it also contained a
> non-trivial number of specialized modules that are much harder to
> find, if not impossible. I'm guessing that IBM machines were probably
> similar.
>
OK, there are hundreds of different SLT "cards", ie. the PC 
boards. But, reading some FE docs on bitsavers, it seems 
that all SLT 360's were built with 95+ % of the SLT 
"modules" consisting of only 6 types.  (I've never 
understood why the SLT modules had to have 8-digit or 
something part numbers if there were only 6 types.)

So, when I was talking about making replacement SLT MODULES, 
I meant to fabricate tiny 1/2" x 1/2" PC boards with 12 
leads, that could be mounted where the failed small SLT 
module had been.

In IBM terminology, they have SLT modules, which are 1/2" 
square ceramic substrates with thin-film resistors fired 
onto the ceramic, then silver-bearing conductors are printed 
on and fired, and finally bump-bonded transistors and diodes 
are soldered onto the conductors.  A drawn aluminum cover is 
epoxied to the ceramic.

Then, these modules are installed on boards, which can be 
single-side or double wide.  They have connectors on one 
edge, which plugs into a "BOARD".

The boards were roughly 9 x 12", I think, and had rows of 
pins on both sides.  The SLT cards plugged into one side of 
the board.  The boards were multilayer, and the inner layers 
distributed power and ground.  Surface layers had 
application-specific wiring.  Then, more wiring was added to 
the back side with wire-wrap wire.  Also, around the edges 
of the board, there were places where 18-signal transmission 
line cables could be plugged.
These had the same layout as the SLT card connectors, 24 
contacts on .125" spacing.

The transmission line cables were clear flat ribbons with 3 
wires per signal.  So, they went 
ground-signal-ground-ground-signal-ground, etc.  The 
neighboring ground wires were much closer together than the 
ground-signal spacing.  These had a characteristic impedance 
of 93 Ohms, I think.

On IBM 370's they split up these cables into individual 
signals to reduce crosstalk.  They retained the 
3-wire/signal scheme, and called them tri-lead.

Jon