On 05/26/2011 10:05 AM, Rick Bensene wrote:
The disk drive was actually used as memory of the
D-17B (and also as
the storage for the registers I believe, but I can't remember, and
BitSavers has no docs for the D17). But it essence, the D17's disk is
similar to the disk of the Autonetics RECOMP II (which DOES have
documents on BitSavers), so for a better general idea of what the disk
was used for, go check those documents.
I would expect that the disk drive also has at least one pre-recorded timing track that
provided some clocking signal(s) for the machine. It was very common in the bit serial
disk/drum-based main memory computers to use one or more tracks with pre-recorded timing
information on them to generate some or all of the clocking for the rest of the logic.
This was a convenient source for clock pulses that were synchronized with the various
other data recorded on the disk/drum, and also helped correct for the small rotational
speed variations of the motor that spins the disk/drum, since the clock pulses governing
the operation of the machine skewed along with the data.
As such, there probably is some "content" on the disk that is required to get
the machine running, in the form of clock pulses.
Yes, there is. the disk provides clock, several register loops, a few
"fast register" multi head loops and the main program store.
Of course, if the characteristics of the clock
signal(s) are known, it's not a complex matter to reproduce them with external logic.
If the external logic duplicated the various serial loops as shift
register than the clock for the registers would
also be the system clock.
The high school I went to had a computer in the
computer lab made by 3M (the same people that make Scotch Tape and tons of other stuff).
It was a process control machine that was donated to the school by Natural Gas utility
that used it to monitor and manage gas pipelines. It was a 24-bit, bit-serial
architecture, with two main registers (A and B), and main memory was a magnetic drum that
held 8K words. Opcode was 5 bits, and each instruction had both an operand and next
instruction address. Disk was organized as block/track/sector, so addresses were formed
this way. The drum provided a master clock signal that was used by the rest of the
machine for timing. The motor that drove the drum was an inexpensive motor much like that
that drives a sewing machine. There was some phase-lock circuitry on the clock signal to
let the machine know when the drum was up to speed, which would then make the machine go
"ready", but there was no active feedback to control the speed of the motor
beyond that. If the clock rate from the drum fell outside a window of tolerance, then the
machine would lose ready state, and freeze where it was at. The drum was belt driven and
one time the belt broke, and the machine froze. I was able to remove power from the
motor, put on a spare belt (looked a lot like a vacuum cleaner belt), re-apply power to
the motor, and once the drum came up to speed, the machine resumed what it was doing
without a hitch.
There were a bunch of smaller Drum/disk based machines when transistors
wer cheap enough but ICs
were not yet available. That was a very narrow window in time as by the
mid 60s RTL and DRTL were
fairly common for gates and basic flops. The bridge to parallel
machines was integrated (board level or chip)
flipflops that were "cheap" enough to allow for having a lot of them as
registers.
Allison
-Rick