2 Apr
2010
2 Apr
'10
3:20 p.m.
From: M H Stein <dm561 at torfree.net>
From: "Chuck Guzis" <cclist at sydex.com>
As a matter of fact reading the card 'sideways' was a basic principle
of the electro-mechanical punched card systems that (along with
card sales) were IBM's bread and butter until the mid-sixties.
With the exception of keypunches, paper tape converters etc. these
machines were 'clocked' through the 12 states (12,11,0-9); the (firmly
clamped) cards and various gears, relays, punches, type bars, etc.
were cycled through these 12 states in synchronized parallel unison,
much like most of the manual 'posting machines' with full row and
column keyboards that preceded and coexisted with them.
Timing diagrams looked much like the diagrams we're familiar with,
but they were usually calibrated in degrees of rotation of the master
camshaft instead of time.
The punch of my ICT1301 is an IBM design, originally built under licence by BTM (British
Tabulating Machine co), which merged to form ICT. It clocks through the 12 states and a
few more while the card is being fed. It mechanically pauses in the middle of each state
by means of a Geneva mechanism. It has two camshafts with bakelite cams which operate
contacts which controls the punch and signals the CPU when data is ready to be punched and
check read. When one card is being punched the previous card is at the check reading
station with 80 wire brush contacts. It does indeed have a timing diagram as you
describe.
The main mechanism sits on a steel casting, and I'm told its the top part of an
original Holerith design which had integral cast ball and claw feet. This plate is mounted
on rubber bushes to a massive angle iron frame with the widest bit of copper earth braid
I've ever seen. It operates at 100 cards per minute and the CPU has to feed it 80 bits
of data for each row and reads 80 bits back from the previous card. Yes the interface is
80 bit parallel in each direction plus lots of control and status signals, 200 wires in
total (8 cables of 25 cores each).
The manual tells programmers to avoid punching more than 60 columns in any row. How any
programmer was supposed to avoid punching a card full of zeroes I don't know. I have
punched fully laced cards with it no problem, and I have a complete spare mechanism
anyway, which I'm thinking of linking to my Mac, though its a big project.
I've never been told that the card reader or line printer have any connection with IBM
but I suppose it is possible. The reader is 600 cards a minute (80 column first) and the
printer is a 600 lines per minute, 120 column drum printer (all the similar characters
aligned not staggered) with a sprag mechanism which advances the paper. The card reader
has a reject hopper and a main stacker. The main stacker turns the card around a big
rubber drum and feeds it into slots in a pair or rollers which then turn and flip it out
under the stack of cards already read.
Roger Holmes