-------------Original Message:
Date: Fri, 30 Nov 2007 09:52:16 +0100
From: "Arno Kletzander" <Arno_1983 at gmx.de>
Subject: Mechanical calculators (was: Re: *updating* 8088's)
"Roy J. Tellason" <rtellason at verizon.net> wrote:
On Friday 23 November 2007 22:23, Chuck Guzis wrote:
There are many kinds of mechanical memory. In
particular, I recall
an early TTY device that used a large rotating drum with cams
embedded in the surface. One could flip a cam one way or the other
and then read them out. I'm trying to remember what sort of machine
this was used on and its application, but my memory sadly fails me.
This stirs a vague recollection of an old mechanical adding machine I
once had, the kind that had a big rectangular array of buttons instead
of just a "10-key" set of numbers. I have *no* idea how it stored a
number in there, though.
Hmm, like, say, an old cash register? This design is called
"Volltastaturmaschine" in German (would translate to "complete keyboard
machine"). They have a latching mechanism that holds in one button per column of keys
(0-9, representing one digit of a number). When a calculation is initiated, the protruding
shaft of the latched key acts as a stop for a toothed rack or similar device which is used
to advance the wheels of the accumulator register by as many teeth as the corresponding
digit says. There are also designs which involve an arrangement of levers positioning an
intermediate gear along the axis of a stepped drum, for example the Badenia
VA-17/VARE-17.
The big advantage is that zeros need not be entered (to enter 100.00, you just press the
"1" button in the fifth column from the right) and that operators could learn to
"touch-type" on these keyboards, effectively entering all digits of a number in
parallel and greatly reducing cycle times.
<snip>
--
Arno Kletzander
www.iser.uni-erlangen.de
-----------Reply:
The IBM unit record or Tabulating machines that we've been discussing
had their roots in the same principles as these calculators, adding machines
and cash registers, but instead of a depressed key stopping a moving rack in
a certain position (a certain point of time in the machine's "cycle"), a set
of
wire brushes (later replaced by photocells) would make contact through
the holes in the card as it moved vertically through the different digit positions,
and pick a relay, latch a rotating selector (and lock a synchronized moving
typebar if it was a printer) to represent that digit electrically. Like the manual
full keyboards, all 80 digits/characters were processed simultaneously.
Timing diagrams looked similar to the modern square waves we're all
familiar with (although they were sometimes arranged in a set of concentric
circles to represent a complete cycle), but instead of a continuous linear
time line they had a definite beginning and end and were divided into
equal-length "digit" times.
When electronics started to replace more and more of the mechanical gears
and relays there was a paradigm shift and everything switched to reading and
processing cards "serially" ("parallel" from our modern perspective
;-), i.e.
lengthwise by character, instead of all characters in "parallel" by digit.
The "Accounting" or "Posting" machines _also_ known as EAMs
(Electr(on)ic
Accounting Machines) made by Burroughs, NCR etc. and used in almost every
bank also used the same basic principle. Because a good operator could enter
an entire multi-digit amount in one "keystroke" and data entry speed was more
important than processing, they retained full keyboards for quite a while even after
the gears and cams had been replaced with electronic equivalents. The moving
racks and cams were simply replaced by a set of rotating 10-position selector
"switches" and connected to the electronic "adjunct" with thick
multi-conductor
cables. The processors in turn operated on "words" consisting of 15 or more
10-"bit" "bytes;" guess you could call it parallel processing...
m