I once used an old computer made by 3M that was transistorized and used
a magnetic
drum as main storage. The machine was actually two processors, each of
which fit in
12U rackmount drawers. Along with the two processors was another 19" 6'
tall rack
full of acquisition electronics...DACs, comparators, and counters, which
were
peripherals to the CPUs, which had I/O instructions to talk to the
stuff.
Included in the rack full of electronics was a real-time clock that kept
track of
the time in HH:MM:SS format. It did not keep track of the date, only
the time.
The clock was transistorized, using ring counters. The clock also put
out 1Hz
pulse that set a flip flop that could be polled, and reset by the CPU.
I used
this 1Hz signal along with the time from the clock to maintain date/time
in a small
operating system that I wrote for the machine as a class project. As I
recall,
there was no interrupt capability on the machine. The clock had a front
panel
display that showed the current time, as well as controls to halt the
clock and set
the time (advance hours, minutes or seconds). As I recall, the clock
derived its
timing from the 60Hz AC power line.
The clock could he read by either CPU by accessing a read-only
peripheral register.
The time was returned in BCD format.
My understanding is that the machine was designed in the early 1960's.
It was designed
specifically for data acquisition processes. I've written to 3M to try
to find out
any information they may have, but alas, none survives. Unfortunately,
I can't
remember the model number of the machine. The machine I used was
donated to the high school
I attended sometime in the early '70's, after being used for quite a
long time as a natural
gas pipeline monitoring system at Northwest Natural Gas in Portland, OR.
The machine
used a 24-bit word, with two-address (operand & next instruction)
format. Opcodes were
5 bits in length. It had a hard-wired "read in mode" where it would
accept input from a teletype
(ASR-33) in the form of octal address followed by a space, followed by
the data to be
written to that address. Addressing was stated in block/track/sector
format.
The instruction set included information for optimal programming to
allow for the rotational
latencies of the drum. The machine was rather slow, even when
programmed optimally.
I recall that I had a hard time writing code that could send output to
the teletype at
full speed.
Rick Bensene
The Old Calculator Web Museum
http://oldcalculatormuseum.com
Received on Thu Jul 29 2004 - 09:01:43 BST