Who's rewired their house for this hobby?

[Jon Elson]

On 11/26/2014 01:47 AM, Eric Smith wrote:
> On Tue, Nov 25, 2014 at 7:42 PM, Jon Elson <elson at pico-systems.com> wrote:
>> On 11/25/2014 02:18 PM, Eric Smith wrote:
>>> Actually to 13 massive linear regulators on multiple large heat sinks.
>> There may have been several versions.  The KL10B in a Decsystem 2020 we had
> A DECSYSTEM-2020 uses a KS10, which has no ECL, and uses a big
> switching power supply made by third parties. The KL10B was used in
> the DECsystem-1090, but not in any DECSYSTEM-20 models.
Not true!  The CPU was, absolutely, a KL10B.  It was in a 
big orange cabinet that was
fairly similar in size and shape to a VAX 11/780.  It had a 
PDP 11/40 (I think) in the
next bay over as the I/O processor.  I am not so clear about 
the exact "DECSystem"
designation, but it VERY much was a DECSystem 20<something>. 
During a major
upgrade, the FE had the print set out, plus all the boards 
that were being
replaced with upgraded boards.  I got a VERY good look at 
the print set and the
boards.  Very typical MECL 10K logic, a million terminating 
and pull-down
resistors all over the place.  I was most interested in the 
clock alignment system.
They had zig-zag traces with a piece of wire-wrap wire that 
were hand-soldered
to tap onto the zig-zag at the right point to set the clock 
delay. The label on
the front of the cabinet said "DECSystem-20" without 
specifically giving the
model number.
>    The KL10BC
> was used in the DECsystem-1090T, which was ran TOPS-20 but was not
> marked as a DECSYSTEM-20.  The other DECSYSTEM-20 models (2040, 2050,
> 2060, 2065) all use the KL10-C, KL10-E, or KL10-E/R processors.
>
>> had one big regulator that was a ring of "christmas tree" segments with a fan
>> about 10" in diameter.
> They wouldn't have used one big regulator because the remote sensing
> wouldn't have been able to get a sufficiently uniform voltage on all
> backplane segments, and that definitely *will* cause problems for ECL.
OK, the thing was all on one huge fan, but there were dozens 
of power transistors on
individual heat sink segments, all arranged in a circle 
around the fan.  So, yes,
there were several regulators, electrically, but the main 
pass transistors were all
mounted in a single unit.

Yes, if you don't go the
Seymour Cray route and run EVERY DAMN SIGNAL differential 
from one board to the
next, then you do have to be careful that the power supplies 
are fairly well matched.
>
> I don't see a print set for the KL10B online, but I'm confident that
> the H761 Regulator Assembly was the same.
>
No, I think there was an "early" version and a "late" 
version.  The system was not
greatly different, but they reduced the number of separate 
regulators and I think
cut one whole regulator board out of the system that way.  
This late model might have
been the KL10-BC you mention, although I'm PRETTY sure the 
print set did not carry
that ending C, at least on the front page.  And, maybe the 
system was a 2040.
This is all from almost 25 years ago, and I was not a user 
on that machine, we took
over the machine room after the 20xx was moved to another 
building.

Jon