| +Home | Museum | Wanted | Specs | Previous | Next |
NCR 18-2 Electronic Desktop Calculator
Updated 6/23/2008
Just about everyone knows that NCR (National Cash Register) made its fame by making cash registers, but few know that, for a time, they also sold electronic calculators. Note that I write "sold" rather than "made". The NCR 18-2 exhibited here is not designed nor manufactured by NCR. NCR partnered with Japanese calculating machine manufacturer Nippon Computing Machine Corp. (NCM, also known as Busicom) beginning sometime in the late 1960's, selling calculators designed and manufatured in Japan by NCM. NCR, like many other American makers of mechanical and electro-mechanical calculating machines (a cash-register is a specialized calculating machine), was caught without electronic design expertise when electronic calculators began to erode market share from earlier mechanical calculators in the mid-1960's. The corporate management at NCR wanted to get in on this lucrative market, but didn't have the local talent to do so. Like other major players in the mechanical calculating machine marketplace, NCR needed to find digital electronic engineering talent. The best place to look: Japan. At the time, Japan's Hitachi and NEC had some of the lowest-cost integrated circuit manufacturing capabilities, with lines of small to medium-scale bipolar and new Metal Oxide Semiconductor (MOS) devices that would serve well for electronic calculators (and eventually cash registers). Japan also ad unmatched manufacturing capacity, with comparatively inexpensive labor rates. Busicom had managed to make the transition from mechanical to electronic calculators by buying up and analyzing the design of pioneering machines in the calculator business (the early calculators from Sumlock/Anita in the UK, and the advanced electronic calculators from IME in Italy), and then crafting their own designs based on what they learned. Busicom's first electronic calculator, the Busicom 161, was essentially a re-engineered "copy" of the IME 84, IME's first electronic calculator, and one of an early group of electronic calculators that utilized coincident current magnetic core memory as the primary storage element in the calculator. Busicom successfully marketed their electronic calculators in Japan, but was looking for ways to grow their market. NCR had a great opportunity for Busicom to get into the very lucrative U.S. marketplace, with a very well-established sales and service network. An alliance was formed, and Busicom's early electronic calculators started making their way across the ocean with NCR logos proudly adorning their cabinets.Like so many other tidbits of history that are being lost every day, the details behind the story of NCR's foray into the electronic calculator market may be lost. My hope is that someone who was connected with NCR back in those days will read this and get in touch with me, and help add to the story of the relationship between Busicom and NCR. NCR's own web site has no information at all about them even having been in the calcluator business. There is a large following of collectors that preserve and document NCR's wonderful cash registers, but there's virtually no information anywhere about NCR's calculator history. If you know anything more about the history of NCR's calculator business, or the company's relationship with NCM/Busicom, please write me.
Another View of the NCR 18-2 The 18-2 is the entry-level machine
amidst the two Busicom-made machines that brought with them NCR's debut into
the electronic calculator marketplace sometime in mid-1968. Along with
the $1095 Model 18-2 exhibited here, the Model
18-3
stablemate was the flagship machine, with all of the features of
the 18-2 along with a one-key automatic square root function.
for a suggested retail price of $1,275. Both the NCR 18-2 and 18-3 are
identical (other than minor cosmetic differences) to the Model 162-C and
162 calculators sold by NCM under the Busicom brand. The Busicom
machines preceded the NCR versions of the machines to market by a period of
about four to five months.
The 18-2 was quite a capable machine
for the time. It performs the four basic math functions, offers a
full sixteen digits of capacity and two independent memory accumulators.
The calculator uses mixed floating/fixed decimal point logic, with a
thumbwheel selecting the fixed decimal point location (at 0, 1, 2, 3, 6, or
9 digits behind the decimal point) for addition and subtraction (as well as
recall of items from memory registers), and automatic decimal placement for
multiplication and division operations. This mixed-mode decimal
point logic is somewhat unusual, and takes a little
getting used to, but is flexible and affords the greatest accuracy
when multiplying and dividing fractional numbers. Other features include
a convenient "back arrow" key that deletes the last digit entered for
easy correction of mis-entered numbers, and a "clear indicator" (CI) key for
clearing the display without disrupting calculations in progress, also useful
for correcting input errors. For a retail price of just under $1100, the
NCR 18-2 was quite competitively priced against competitors like the
Monroe 990, a machine with similar capabilities
priced at $1250.
The cabinet of the 18-2 is all-metal, with the only plastic
part of the cabinetry being the smoked plastic display window that the
Nixie tubes shine through. The base of the machine is a thick metal
casting, with nicely machined bosses for the electronics of the calculator to
mount to. The upper half of the case, again all metal, is a thinner-gauge
casting, but still quite hefty. The upper half of the case fastens to the
base via four screws that are easy to access, which made it nice for
service personnel.
Overall Internal View of the 18-2 The chassis of the machine is a
combination of stamped metal parts and plastic. The backplane, situated along
the bottom of the chassis, is built upon a rather heavy metal casting
that makes up the base of the chassis, mounting on the bottom half of
the cabinet. The circuit boards plug into a plastic frame that provides
card edge guides that align the cards and keep them from moving around
during transport. A stamped metal panel covers the top of the card
cage to keep the circuit boards from working themselves out of the connectors
they plug into.
The Keyboard Connector The keyboard assembly, a rather massive
item by itself, is built on a stamped metal frame that sets into plastic
section of the chassis. A large multi-pin connector provides the electrical
connection between the keyboard and the backplane. The design is such that
getting the keyboard disconnected from the connector is difficult, because
the cables are short enough that the keyboard assembly can't be lifted too
far, meaning that it takes small hands to get under the keyboard while
holding it up. Plugging the keyboard connector back in after the keyboard
has been removed is even more challenging.
The "Business" side of the keyboard The keyboard design is quite unusual,
using leaf-type contacts actuated by the plunger of the keystalk. Most
calculators of the time used magnet-actuated reed switches, which provide
a much cleaner switching action (less contact bounce) and offer
longer service life. This design was likely used to
help reduce the cost of the calculator, however, this
seems somewhat contradictory, though, given that the rest of the machine seems
to be so overbuilt.
A close-up view of the keyswitch modules in the NCR 18-2 Each switch has a fixed set of contacts
and another contact that moves. When the key is depressed, the plunger
pushes the movable contact aside as it moves downward, causing it to come
into contact with the fixed set of contacts. A capacitor mounted directly
to the leads of the switch helps absorb some of the switching transients
caused by the mechanical action of the switch. Other circuitry helps clean
up and shape the switch closure waveform so that reliable keypresses are
detected. Each switch is encased in a snap-on clear plastic housing that
keeps out dust, but allows removal for adjustment and cleaning of the switch
contacts by service personnel.
The Backplane of the NCR 18-2 The backplane that provides interconnection
between the circuit boards, keyboard, power supply, and display subsystem
is hand-wired, with point-to-point connections painstakingly put in place
by workers that had to have immense patience to be able to do this type
of wiring day-in and day-out. Each of the ten circuit boards that make
up the logic of the machine plugs into four edge-connectors in the backplane.
Each connector has 20 gold-plated contacts, for a total of 80 possible
connections for each circuit board. One extra set of two edge connectors
provide an eleventh slot for the display subsystem to plug into.
The Power Supply of the NCR 18-2 The power supply for the machine takes
up the back portion of the chassis, and, like most of the other aspects
of this calculator, seems quite overbuilt. It is a complex power supply,
with it appearing that all voltages (with the exception of the Nixie
tube drive voltage) being transistor-regulated, and heavily filtered.
Detail of the 18-2's Nixie Display (note lit "Overflow" Indicator) The display subsystem of the 18-2 is
a modular assembly that plugs into the backplane. The subsystem consists
of a circuit board to which the Nixies are soldered, with a fairly elaborate
metal frame to align and secure the display tubes. A metal bezel with stamped
digit designator nomenclature presents the faces of the Nixie tubes to
the user. Along with the sixteen Nixie tubes, which have 5/8-inch tall
digits and a right-hand decimal point, are two incandescent lamps situated
behind jewels at the left end of the display which light to indicate a negative
number and/or overflow of the calculator.
A typical logic board from the NCR 18-2 The 18-2 uses small-scale integrated
circuits for its logic. A total of 134 IC devices combine forces with a
small core memory array, and an assortment of discrete components to make up the
logic that runs the machine. The IC's used in the 18-2 are a bit of an
enigma. I had expected, given that the machine was clearly manufactured in
Japan, that it would use Japanese-made ntegrated circuits.
This machine is an exception, as it uses IC's from the American IC manufacturer
Signetics.
A close-up view of one of the IC's
At this point, my guess would be that the IC's are
DTL (Diode-Transistor Logic) or TTL (Transistor-Transistor Logic) small-scale
devices, but the particular part numbers of these devices are unknown to me,
so I can't be sure. The part numbers used may be what are called "house"
numbers - custom part-numbers placed on the IC's by the manufacturer at the
request of the customer to allow the customer to use their own internal
part numbers to avoid giving away design secrets to cometititors, as well
as restricting service of the machines to "factory authorized" repair
facilities. A total of four different part
numbers are all that are used. My guess is that one of the parts contains
a flip-flop or two, and that the other parts contain various types and
combinations of logic gate elements.
Each circuit board is approximately
11 inches wide by 4 1/2 inches tall, and is made of a phenolic circuit
board material. Edge connector fingers are gold-plated, which makes for a
very reliable and corrosion-resistant connection. Components are mounted
only on the front of the boards, with interconnection traces etched on both
sides of the board. Feedthroughs connecting the two sides of the board are
implemented by placing a pad on each side of the board, with a hole drilled
through. A piece of wire is inserted in the hole and soldered on each side of
the board. Jumper wires are occasionally used on either side of the board
to provide interconnections when etched traces couldn't be routed.
A metal stiffening bar is riveted across the top edge of the circuit board
to provide some structural ridgidity for the boards.
The Core Memory Board (Core Array at Center) At the time this machine was designed,
during the late part of the 1960's, the levels of integration possible
in IC technology were such that it wasn't practical in most cases
to implement the operating registers of the machine using flip flops.
Some of the Japanese MOS (Metal-Oxide Semiconductor) IC's could implement
a 10 or 12-bit shift register in a single chip, but even at that level of
integration, it would require quite a number of IC's to implement all of
the bits of register needed for a typical four-function calculator. A number
of schemes were used for register storage during this period, including
acoustic delay lines (see the Monroe 990 exhibit
for an example of this technology) and magnetic core memory. The NCR 18-2
uses a small magnetic core memory array, manufactured by Japanese
electronics component and equipment manufacturer Hitachi, as its register
storage means. The core array consists of four 8x8 core planes for a total
of 256 bits of storage. Magnetic core memory uses small doughnut-shaped
rings of a special magnetic material, all woven into a grid of wiring that
allows each individual core to be magnetized in one direction or another,
as well as allowing the magnetic state of each core to be individually detected.
With such an arrangement, it is possible to store the working registers
of the calculator as a series of magnetic ones and zeroes within the
core array. Core memory technology evolved out of work by Dr. An Wang,
a pioneer in the field of magnetic core memory technology, and later,
the head of Wang Laboratories, a company that made a fortune in the
high-end electronic calculator business from the mid-1960's through the
late 1970's.
Keyboard Detail on the NCR 18-2 The color scheme of the NCR 18-2's keyboard
is somewhat odd...leading me to believe that perhaps some keycaps were replaced
over the life of the machine. Keycaps are colored white, ivory, beige, or gray,
in a mixture that doesn't make a whole lot of sense. The '4' and '6' keys
are white, while the rest of the digit keys are ivory in color. I believe
that perhaps the '4' and '6' keycaps were replaced, and perhaps there
were differences in the color of the plastic, or simply the fact that the
'4' and '6' are newer and haven't aged as much is why they keys have the
lighter color. Another example is the 'X' key, which is gray, while the
other math function keys are beige. Perhaps the 'X' key was also replaced
and somewhere in the process, the colors were changed. In any case, keycap
coloring aside, the layout of the NCR 18-2's keyboard is logically thought-out
and attractive from an aesthetic point of view. There are four groups
of keys, with the left-most group providing miscellaneous functions, including
the backspace key, memory register recall and clear keys, along with the
clear all and clear indicator (CI) keys. The next group of keys is the
traditional numeric keypad, with oversized '0' key, and a raised area
on the '5' key to allow the user to orient their fingers on the keyboard
by touch. The math functions make up the next group of keys, with the
memory accumulation function keys rounding out the groupings.
At the left end of the keyboard assembly
is a thumbwheel switch that is used to select the decimal point position
for addition, subtraction, and memory recall operations. The decimal point
selection thumbwheel provides selections for 0, 1, 2, 3, 6, and 9 digits
behind the decimal point. Above the decimal point selection thumbwheel
is the push-on/push-off power switch. At the right end of the keyboard,
a three-position slide switch selects the rounding mode of the calculator.
The switch has no detectable nomenclature on the keyboard panel, either
it was never there, or the nomenclature was printed on the keyboard panel
in such a way that it wore off over time and usage. The rounding mode
of the machine operates only with addition and subtraction operations, where
the fixed decimal point setting of the machine comes into play.
The uppermost position of this switch forces the calculator to round the
least-significant digit up in all cases. The center position of the
rounding switch causes the calculator to round up if the next less-significant
digit is 5 or greater, and to truncate if the digit is 4 or less.
The lower-most position of the rounding switch causes the calculator to
ignore the next less-significant digit, leaving the least significant
digit alone in all cases. As an example, performing 1 divided by 3, followed
by pressing the "+" key (thus forcing the machine to fix the decimal
at the selected position), with the decimal position set to '2' would
result in: 00000000000000.34, 00000000000000.33, and 00000000000000.33
with the rounding switch in the upper, middle, and lower positions.
Performing 2 divided by 3 with the same settings would result in
00000000000000.67, 00000000000000.67, and 00000000000000.66.
The Model/Serial Number Tag on the NCR 18-2 The 18-2 uses a curious mix of
algebraic and arithmetic logic. Addition and subtraction operate
arithmetically, with the function entered after the operand, like an
adding machine. For example, to subtract 15 from 30, the problem would
be entered as '30', '+', '15', '-', with the result of 15 showing in the
display after the '-' key is pressed. Multiplication and division
are entered algebraically, with the '=' key completing the operation.
As mentioned before, the decimal point location in multiplication and
division is fully floating, with the calculator placing the decimal point
wherever needed to wring the maximum accuracy out of the result. However,
as soon as a result of a multiplication or division is submitted for
addition or subtraction, or stored in a memory register, it is forced
to the number of digits behind the decimal point as selected by the
decimal point selection thumbwheel (and rounded according to the setting
of the rounding mode switch).
Internal QA and Modification Record Tag (located under the keyboard, glued to the cabinet base) The memory capabilities of the 18-2 are
quite handy, with two completely independent accumulating memory registers.
Each memory register has keys to clear, recall to display, add to memory,
subtract from memory, and add product/quotient (acting as an "=", followed
by add to memory function) to memory. The memory registers adhere to the
decimal point location selected by the thumbwheel, with rounding occurring
as defined by the position of the rounding mode switch. The memory registers,
although residing in non-volatile core memory, are automatically cleared
when the calculator is powered up, so it isn't possible to recall a number
in a memory register through a power off period, as it is on some other
calculators that don't clear the core memory at power-on time.
The NCR 18-2 does have a few quirks
in its operation. Division with dividend larger than 14 significant
digits causes incorrect results. This is a common limitation on many
early electronic calculators, mainly due to the fact that some digits of
the working register of the calculator are used as temporary storage during
the accumulation of the quotient. The overflow detection on the machine is
a bit dicey, sometimes failing to indicate an overflow (mainly in
large multiplication operations) when one should occur. When an overflow does
occur, the keyboard is not locked out, meaning operations can continue,
although the accuracy of results obtained when the calculator overflows
is poor. Division by zero results in the machine getting quite confused.
Pressing the "C", or "CI" key will stop the futility of the effort, but
pressing "CI" leaves the machine in a strange state, requiring a full
clear (using the "C" key) to restore it to normal operation.
The 18-2 is not a speed demon, with
some division operations taking nearly a second to complete. The "all nines"
(which in this case, is 14 nines due to the limitation mentioned above) divided
by 1 division problem takes just a shade under 1 second to perform. During
the calculation, the Nixie tubes put on quite a light show, as the displays
are left active as the calculator subtracts, shifts, and accumulates the
quotient. Addition and subtraction operations complete in a blink of an
eye, with just a slight flicker of the display indiciating the machine
is working.
