| +Home | Museum | Wanted | Specs | Previous | Next |
SCM Marchant Cogito 240SR Electronic Desktop Calculator
Updated 2/24/2009
The Old Calculator Web Museum is very happy to add this amazing machine to its growing archives of early electronic calculator technology. This calculator has been one of the most challenging machines to be documented to date. The search for information on the history of this machine has been extremely circuitous, but the results of the research reveal a history to the Cogito 240 (SCM's initial entry into the electronic calculator market) that is truly fascinating. While the Cogito 240SR (a slightly later introduction that added square root and grand total multiplication) is most certainly not the most capable, fast, elegant, or easy-to-use calculator of its time, the story behind its development and, most specifically, the man behind its design, is one of the most fascinating stories in the history of desktop electronic calculating machines.Stanley P. Frankel, 1959
Photo from IRE Transactions on Electronic Computers, Sept. 1959
To begin the story of the development of the SCM Cogito 240, let us first start with a little about the background of the company that ended up marketing the calculator.
In 1910, two brothers, Rodney and Alfred Marchant, began manufacturing mechanical calculators with the intent of selling them to a marketplace that was hungry for better ways of performing mathematical calculations. Incorporating in 1911, Marchant Calculating Machine Co. of Oakland, California, made a very successful business making and selling desktop mechanical and later, electro-mechanical calculators, that used innovative mechanical engineering and design to make the machines fast, reliable, and reasonably-priced. It is interesting to note that a luminary in the world of mechanical calculators, Carl Friden, worked as a principal designer for Marchant before leaving to start his own calculating machine company, Friden.
Model Identification on Front Panel
In 1958, Smith Corona Co., the well-established typewriter manufacturer,
acquired Marchant Calculating Machine Corp. Smith Corona's management
felt that diversifying the company from just a typewriter manufacturer
to other office machine product lines was a good growth strategy.
Once the acquisition was completed, Smith Corona continued to market
Marchant's famous rotary calculators and adding machines under the Marchant
brand name, using the name "Smith-Corona Marchant". In 1962, Smith Corona
made this name it's new corporate name, becoming Smith-Corona Marchant,
or SCM for short. By this time, SCM was into numerous areas of business
machines, not just calculating machines and typewriters. SCM kept the
"Marchant" name on their calculating machines, leveraging the market
recognition of the Marchant name.
After World War II ended, and the
nuclear weapons projects at Los Alamos started to slow down,
Stan Frankel did various things, including a stint
at the University of Chicago's Institute for Nuclear Studies; forming
a mathematics consulting company with his colleague from the Manhattan
Project, Eldred Nelson; and in 1949, being appointed
the head of a newly-formed Computation Laboratory at Cal Tech.
During his time at Cal Tech, Frankel lost his security clearance
due to questions (thought baseless) about some of his past acquaintances
during the period of anti-communism spurred by Joseph McCarthy.
Though it was a blow
to him to be blocked from the inner circle of nuclear physics, Frankel's
interests had shifted somewhat away from nuclear physics, more toward
electronic devices that could perform complex calculations. His work
with early IBM Punched Card-based calculating
equipment during the Manhattan Project, as well as performing complex
programming on some of the earliest electronic computers had
convinced Frankel that electronic computing was the only way man was
going to be able to understand the myriad complexities of nature.
Unfortunately, this distraction led Frankel a bit astray as far as his
duties as the head of the Computation Lab were concerned. He became so
consumed with the design of a personal computer that he never was able
to gain tenure at Cal Tech, and left (with persuasion) in 1954. During
his time at Cal Tech,
he did consulting work on the side, and ended up developing the design
for a small computer he called MINAC, that was eventually licensed from him
to become the
Royal McBee/Librascope/General Precision LGP-30,
which became a reasonably successful vacuum-tube-based small computer used
in educational, military, and business environments. He also developed
(under contract) the design for another computer called CONAC that was
developed for Consolidated Oil Company. He also contracted with General
Electric doing design work for some of their early computers.
Sometime in the late 1950's, Frankel got interested
in the motion of building an electronic version of the desktop
electro-mechanical calculating machines that had been used by the
"human computers" in the early days of the Los Alamos neutron
diffusion calculations.
Enter a company called Electrosolids, Inc., in Sylmar, California.
The company, founded in 1956, had gained a solid foothold in a niche of
the fiercely competitive electronics marketplace of the time building
high-efficiency solid-state power inverters. The demand for such devices
was high in the aerospace industry, and Electrosolids was able to build
a flourishing business designing, building, and marketing various types of
power supplies. One of the founders of Electrosolids, Joseph Strick,
had come to know Stan Frankel through Eckert-Mauchly Computer Company (EMCC),
Strick's brother-in-law knew Frankel, who was consulting for
EMCC at the time, and introduced the two. Strick kept in touch with
Frankel, and in time the two become close
friends. Frankel had often chatted with Strick about the merits of electronic
computing, and mentioned his ideas about a desktop-sized electronic calculator.
Frankel was quite passionate that such a machine could be built, and didn't
take long to convince Strick of the market opportunity such a machine would create.
By 1958, Electrosolids was embarking on diversifying its business. The company
had developed a some solid-state walkie-talkies, a talking doll, and some
other electronic devices to help minimize the company's dependence on the
power supply business. Strick believed that Frankel's calculator idea might
be a great way to augment Electrosolids' market. Frankel had just finished
the design of a computer for Packard Bell Electronics, that ended up being
introduced in 1961 as the Packard Bell PB-250, another historical
"mini computer"
before the term actually existed, and was free to pursue a new project.
It was decided that a subsidiary company
within Electrosolids would be created, called Computron Corporation, with
Stan Frankel as the President. Frankel was given budget to build a
prototype of the calculator, and hire a technician to assist.
Together, Frankel and his technician would build the prototype calculator. Strick was convinced
that a personal electronic calculator would have market potential far
beyond the niche markets that Electrosolids had up to this point.
Frankel and his technician set up shop in a small area set aside for them
in the Electrosolids plant. Immediately
they set about building the prototype machine. The machine was a
fully-transistorized design, using a magnetostrictive delay line for the
working register storage. Numeric entry was through a 10-key keyboard
arrangement, and the display was presented as digits on the face of an
oscilloscope CRT. It took nearly a year for the prototype to be
built and made fully operational. By late 1959, it had taken the form of
number of circuit breadboards spread out over a large workbench.
It surely wasn't a desktop machine at this point, but in theory, based on
component counts, it could be built into a desktop-sized package.
The plan was to build and market the calculator was the "CC 1200". The "CC"
stood for "Computerized Calculator", and the 1200 represented the expected
selling price of the machine, of $1200.
"Leak" article announcing Electrosolids' Calculator
At one point, Electrosolids realized that Integrated Circuit technology
was going to become a force to deal with, allowing electronic calculators
to be built less expensively. Discussions were carried out about reducing
the price of the machine when it went to market to grab as much of the market
possible before IC-based machines came around. Prices of around $500 or so
were bandied about. Somehow this word leaked out to the press, which resulted
in the short article shown above. Clearly, this revelation did not transpire
until much later,
but the article did grab the attention of a number of folks who were also
in the market to produce an electronic calculator, and lit various fires
around the industry.
Unfortunately, during the time that the prototype was being built,
Electrosolids had embarked on an aggressive project to build a high-power
inverter for a customer who had very tight technical requirements. A tremendous amount
of money was invested in the development of this product. As it turned out,
the final product did not meet the specifications of the customer, and
it seemed no amount of rework would get the device to meet the specifications.
The customer finally backed out of the deal, leaving Electrosolids in a real
bind. The company needed some money. Coincidentally, the calculator prototype
was pretty well perfected, and running reliably. Funds were such that
there just wasn't a way to turn the prototype into a production reality,
so it was decided to try to sell the calculator prototype and design to an
established mechanical calculator company, in hopes that enough money
could be raised to help save the company.
Sometime in mid-1961, a press release offering Electrosolids' prototype
electronic calculator for sale was published in a number of technical
publications, soliciting a buyer. With a working prototype to demonstrate, it
should have been a pretty easy sell, as many mechanical calculator companies
were just beginning to think about the possibilities of electronics eclipsing
mechanical devices as far as calculating machines were concerned.
The response was quite immediate, with a number of the key mechanical calculator
manufacturers responding that they wanted to see the Computron prototype.
The first company to come look at the prototype was Friden Calculating
Machine Co., a major player
in the mechanical calculator market. Senior members of management and
engineering for Friden came to see the prototype. The prototype worked
flawlessly for about ten minutes, then suddenly quit working when a
transistor failed. When this happened, the Friden crew was clearly amused.
The Friden entourage left uninterested. Interestingly enough, at the time,
Friden was in the process of coming up with their own ideas for an electronic
calculator. While it will likely never be known for sure, it is interesting
to contemplate the motivation behind Friden coming to see the
Electrosolids/Computron prototype machine. Perhaps the visit by Friden wasn't
so much out of interest in acquiring the design, but more one of seeing what
potential competition Friden may have in the newborn electronic calculator
marketplace. The breakdown of the prototype machine may have been viewed
by Friden as an indication that the design was not yet ready for "prime time",
leading them to believe that anyone who acquired the design had a lot
of work to do to make it a production reality. One can wonder, though, if
the folks from Friden may have got some ideas that were incorporated into
thier eventual EC-130 calculator. The Computron prototype used a CRT display
with a seven-segment digit representation.
Recently information has surfaced that in late 1961, Friden
submitted a request for proposal to Stanford Research Institute's Computer
Lab, indicating a desire for SRI to develop a prototype CRT-based display
system for use in an electronic calculator. Coincidence?
Marchant Calculating Machine Co. sent a number of representatives to
view the prototype, and while they were quite impressed with the machine,
they weren't convinced that enough attention inside Marchant could be
generated for the
company to make an offer. Later, the CEO of Monroe's parent company showed up,
thinking that a move to electronics may make sense for Monroe, but he
too went away without making an offer.
It was late in 1961, and the very beginnings of electronic calculating were
starting to move ahead. The British firm Sumlock had collaborated
with another British company, Bell Punch, to develop an all-electronic
calculator, the Anita Mk8.
While not a solid-state machine (it used gas-filled, cold-cathode
Thyratron tubes), and quite simplistic in its capabilities, it clearly
demonstrated the ability of electronics to dramatically improve the
speed and accuracy over electromechanical calculators.
Hopes of a deal to sell the Computron calculator protytpe were starting to
fall a bit, when a European connection came
along. Ian Rose, the head of the calculator division of Diehl, manufacturer
of brilliantly-designed complex electromechanical printing calculators
(whom Stan Frankel would later end up designing the concepts behind
Diehl's first electronic calculator, the
Combitron),
got in touch. Diehl had a remarketing agreement with Marchant, by which
Marchant could market Diehl-made mechanical and electromechanical calculators
under the Marchant brand name
in North America. Mr. Rose managed to get the right people inside Marchant
to listen to Computron's electronic calculator ideas, and this time, a
deal was struck. Sometime in 1962, Computron Corp. was acquired by
SCM, and Stan Frankel and his technician began work as SCM
employees to make the calculator a production reality.
Conceptual drawing of what became the Cogito 240SR from US Patent 3518629 The work at SCM to get Frankel's Computron
prototype design packaged into a desktop unit moved along at a rather slow
pace. The exact date of introduction of the machine, which became
known as the SCM Cogito 240, is not known, but the best estimate that can
be made is that it was in production sometime in 1965. It seems likely that the
pre-sales announcement of the machine likely occurred sometime
in late '64 or early '65. The first Cogito introduced did not have
the square root or accumulative multiplication functions of the machine
exhibited here. The originally introduced machine was dubbed the Cogito 240
(without the SR). It appears that the addition of the square root and
accumulative multiplication function were added into the design at SCM
(either independent of Computron, or perhaps with input from Frankel,
although Frankel is not listed in the patent for the square root version
of the calculator), and introduced as a follow-on machine called the
Cogito 240SR. To the author's knowledge, no examples of the original
Cogito 240 are known to exist, but it is known for sure that it was
a true product, based on market data from 1965.
There's an amusing twist to the story
of the Cogito 240. SCM was in need of electronics expertise to help
build the machine that Frankel had designed. Ads were placed for a number
of high-level electrical engineers, as well as a bunch of technicians
and other support staff that would be needed to build the desktop
version of the calculator. One of the engineers that responded was a
very bright Electrical Engineering/Computer Science graduate from UC Berkeley,
named Thomas Osborne. Osborne was hired on at SCM to help with the
development of the Cogito calculator. After becoming familiar with the
architecture of the Cogito, he became disenchanted with the design.
He thought that the machine was too much like an electronic implementation
of a mechanical calculator. Osborne's background in computer
science told him that an architecture more like that of a small computer
was a much more efficient way to build a personal calculating machine.
Osborne tried to influence SCM management to scrap the architecture that
Frankel had developed, but his arguments fell on deaf ears. After trying
for a few months to influence change, Osborne went to his management saying
that he wished to resign, but was willing to work on his own, unpaid, to
develop his concepts for calculator design into a working prototype, which
he would offer SCM the first opportunity to buy once it was
completed. SCM wished Osborne luck, but made no committment to consider
any developments that he made. With that, he left SCM and began work on his
own to develop a calculator design that used computer science concepts as its
basis.
Osborne went on to develop a wonderful
little calculator prototype (called the "Green Machine", after the metallic-
Green color that the cabinet of the machine was painted) that provided full
floating point math to ten
digits of precision, was very fast, and quite compact. To learn more
about Tom Osborne and the development of the Green Machine and HP
calculators, please read Steve Liebson's
Reprint of some materials written by Thomas Osborne. In early 1964,
Osborne started showing his machine to many potential buyers, (including
IBM, Hewlett Packard, and Monroe to name a few), but got no real interest
in anyone buying his design.
This was in spite of the fact that the calculator was light-years ahead of
anything else on the market at the time, and would remain so until the concepts
behind it became a production calculator in 1968.
Osborne didn't bother asking SCM if they might be interested, as
they had already made significant progress on turning the Computron design
into a production calculator, and clearly felt that their machine
was all they needed to make their stand in the electronic calculator
marketplace. Word of Osborne's calculator eventually worked its way through
Hewlett Packard, to the ears of a group of folks who were involved in developing
an electronic calculator that would blow the doors off of anything else
out there. Osborne was contacted by these folks, and he was asked to bring
his prototype to Hewlett Packard to demonstrate to them. Once the right
people at HP saw Osbrone's machine, they were hooked. HP offered to give
Osborne a job on the spot, but Osborne was more interested in being a
consultant. Osborne went on for a time as an independent contractor to HP,
then later hired on as an employee, becoming a key contributor to HP's
revolutionary calculator developments, including the groundbreaking HP 9100A,
that incorporated many aspects of Osborne's little prototype calculator.
SCM's loss was Hewlett Packard's gain. This single decision on the part
of SCM's management likely contributes strongly to why HP is still very
successful in the calculator market today, while SCM left the electornic
calculator business in the 1970's without much success.
Architecturally, the Cogito 240 is a curious
mix of mechanical calculator concepts, merged with emergent computer technology.
The machine implements a three register architecture, similar to many
mechanical calculators. One register (called K) is where keyboard entries
are placed, another is an accumulator (P) that accumulates sums and products,
and the third register is essentially a counter used for forming
quotients and (later) square roots (Q). The innovation in the implementation of
the Cogito 240 is that numeric representation and math processing means of the
calculator is much more computer-like. Numbers are represented in
Binary-Coded Decimal (BCD) form, and all arithmetic is done in BCD.
Some early electronic calculators
utilized electronic equivalents of the ten-step counters used in
mechanical calculators. At the time the Cogito 240 was designed, the use of
binary-based math in electronic calculating was a concept applied typically
only to electronic computers. However, by the time that the
Cogito 240 hit the market, binary-based (rather than decimal-based)
architectures for calculators had pretty much taken over because
of the increased efficiency and flexibility of the use of the binary number
system.
In order to make the Cogito 240 more
usable for complex math, a twist was added to the three-register architecture,
in that each of the three main registers has what is called a surrogate
register. The surrogate registers serve as scratch pad memory for each of
the main registers. This feature allows more complex math to be performed,
as the surrogate registers can contain intermediate results, mostly
eliminating the need for manual recording and re-entry of intermediate
answers. The surrogate registers are called K', P', and Q'. The
surrogate registers are not displayed on the CRT, while the main registers
are always displayed. While the surrogate registers were helpful in
performing more complex math, their use was not inherently obvious. Keyboard
functions are provided to store the content of a primary (visible on the
display) register in its surrogate, to recall the surrogate to its primary
register, and to exchange the content of the primary register and its
surrogate. This method of providing intermediate result storage was rather
complicated and non-intuitive, and while unique, could not compare to Friden's
elegant Reverse Polish Notation (RPN) stack architecture.
The use of the surrogate registers of the Cogito 240 will be explained in more
detail in the operational section of the exhibit.
SCM Marchant Cogito 240SR With Clamshell Opened From a technology perspective, the 240
is reasonably conventional for its time. The machine is implemented with
all discrete transistorized circuitry -- there are no integrated circuits
to be found in the machine. A total of six circuit boards, each measuring
about 9" x 10", contain the main logic of the machine, with a few other boards
performing the various analog functions (deflection amplifiers, scanning,
and beam blanking) needed to drive the CRT display. Logic gating is done
with tried-and-true diode-resistor gates (AND and OR), and transistors are
used for level shifting (buffers and inverters) and various types of
flip-flops. A magnetostrictive delay line serves as the working
storage for the calculator. The delay line has a capacity of 480 bits which
provides all of the capacity needed for the registers of the machine.
All of the bits making up the six working registers circulate
in the delay line continuously, and are gated to various sections of the
logic as they are needed.
The Delay Line (in the bronze colored enclosure underneath the "kludge" card with the silver cover) It appears that somewhere in the design
process, there may have been some issues with the delay line circuitry, as
in the production implementation of the Cogito 240SR there is a "kludge board"
mounted to the
top of the delay line case that both of the paths into and out of the delay
line connect to. This board has a number of transistors and passive components
that appear to be related to additional conditioning for the signals that go in
and out of the delay line. A grounded metal cover protects this circuitry
from noise created by the CRT drive circuits, which are in close proximity
to the delay line when the clamshell is closed.
The manufacturer's label on the bottom of the delay line (click image for larger view)
The delay line was made for Marchant by a company called Digital Devices, Inc.
of Syosset, New York, a company that no longer appears to exist. (If anyone
out there knows anything about this company, I'd love to hear from you).
The delay line, model 414-1002, has a nominal delay of 460 microseconds,
and a maximum bit rate of 1.1 MHz. At the maxiumum bandwidth, the delay
line could hold 506 bits of data. In practice, the delay line is run at
a slightly lower speed to reliably allow it to store the 480 bits of data needed
by the machine.
Delay Line Kludge Board The calculator utilizes a
bit-serial architecture, with all operations occurring on data one bit at a
time. The data in the delay line is interleaved so as to present the bits
of each register such that only a couple of flip flops are required as
temporary single-bit storage registers to store the bits needed to perform
a binary operation on two bits. The timing of the calculator is generated
by a master clock generator that runs at approximately 1 MHz (1,000,000 cycles
per second), which is divided down by a chain of flip-flops to generate the
major and minor states of the control logic. The state control flip flops
are connected together with complex chains of gating that guide the machine
through the various steps of performing a function. The patent information on
the machine indicates that the machine has a total of 64 different states that
are generated by the state logic, with each state carrying out a particular
portion of the operational steps involved with running the machine. Originally,
the Cogito 240 (note the lack of SR) was designed as a four-function machine
only (no square root function). After the Cogito 240 was on the market, some
design work was done that added some extra logic states (or used unused states
of the original design), in order to provide the square root
function while the original design did not have this ability. The calculator's
logic is hard coded by the wiring of the various flip-flops and gates, as
opposed to later calculators that used a generalized microcoded architecture.
Mechanically the Cogito 240SR is a monster.
The machine weighs 39 pounds, and chews up a sizable chunk of desktop space,
with a footprint of 13-1/2" wide by 19" deep. The machine is not as large
as the Friden 130 in terms of footprint, mainly because the Cogito 240SR is more
aggressive than the Friden in terms of packaging density. The cabinet is
arranged in a clamshell, with the top part of the cabinet folding up, exposing
the main logic, keyboard, and power supply in the lower half of the clamshell,
and the CRT tube its associated drivers and high voltage power-supply circuity
in the upper part. The cabinetry is all thick wall aluminum castings or heavy
gauge stamped sheetmetal. The construction is of high quality, with many
machined surfaces and high quality connecting hardware. The overall design
of the machine shows a good deal of care in the mechanical design aspects.
A large and heavy heatsink takes up the entirety of rear panel of the machine,
dissipating heat from the power transistors and rectifiers that make up the
power supply. A small synchronous motor drives a fan that forces air drawn
through cooling vents in the lower part of the cabinet across the logic
circuit boards to cool the electronics. The fan is quiet enough such that
it is not obtrusive in an office environment.
Cogito 240SR Keyboard Mechanism The keyboard of the 240SR is a mechanical
marvel, similar in design to the mechanically encoded keyboard of the
Friden 130. The keyboard uses a complex arrangement of levers that
activate individual micro-switches for each key to identify
the key that is pressed. The levers also serve to provide lockout functionality
to prevent multiple keys from being pressed at the same time. Unlike
the Friden 130/132, the keyboard is encoded into a form the electronics
can use by electronic means (diode arrays and gating). The Friden calculator
encodes keypresses by entirely mechanical means. Like the Friden 130,
the keyboard mechanism of the 240SR is designed to lock the math function key
down during the time that math operations are taking place. A small solenoid
fires to release the key when the operation is complete. Given the
length of time that the machine can take to perform some operations, this
key locking prevents a fast operator from causing errors by getting ahead
of the calculator. Keyboard operation is rather noisy, but overall, the machine
is certainly much more quiet than the elcetromechanical machines it was
designed to replace.
Cogito 240SR CRT Display The CRT tube uses a medium-persistence
phosphor, as the display is scanned quickly enough that a long-persistance
phosphor isn't required in order to avoid flickering of the display. The tube
uses a blue-green phosphor, with a blue filter positioned in front of the
faceplate to provide a pleasing blueish color to the display.
The CRT display is organized as four rows of numbers, with the top row being
the K register, the next to the top row being the Q register, and the
bottom two rows making up the P register. The capacity of the K and Q
registers is 12 decimal digits. You may note on the display that 16 digits
are actually displayed for these registers. The first four digits of the
K and Q registers are always zero, and are used as place holders for decimal
point positioning, as explained the the operator's manual. The P register
consists of a total of 28 digits, 24 of which are useful, with the leading
four digits again reserved for decimal point placement. The P register
has double the capacity of the K and Q registers as it serves as the main
accumulator for the machine, and must have enough digits to hold the product
of two twelve-digit numbers. The digits are formed using the familiar
'pieces of eight' 7-segment rendition, with rather peculiar half-sized
zeroes. The decimal point lights to the right of the digit position it
is being displayed in. The sign of the number in the register is displayed
as a '-' before the most significant digit of the register. Two other
annotations show up on the display, in the form of two "streaks"
which can appear at the left, right, or at both ends, just below the digits
of a register display. These notations show up when the number contained
in the register needs its decimal point repositioned manually to properly
express the magnitude of the number being displayed. The rendition of
digits on the display is controlled by a series of diode-resistor gates which
essentially form a ROM that defines the sequences of strokes of the electron
beam to form the display features. The display cycle continues during
calculation, allowing the user to observe the steps of the calculation as
it proceeds.
Cogito 240SR Power Supply Detail The power supply of the 240SR is rather
complex, though it is of conventional design. The main logic supplies are
+20V and -20V DC, with various other voltages used in the analog portions of
the circuitry that drive the CRT. The logic voltages are electronically
regulated by use of zener diodes as references, and high current
pass-transistors for regulation. A bank of sizable computer-grade electrolytic
capacitors provides filtering to assure minimal AC ripple on the logic supplies. The
high-voltage power supply for the CRT is an epoxy encapsulated module that
provides the approximately 3000 Volt potential required by the CRT.
The cooling for the dissipation devices in the power supply comes via
convection from the large heatsink at the back of the machine.
Circuit Board Mounting and Interconnect Detail The electronics of the machine were
clearly not designed for ease of service. There is no backplane to speak
of. All of the circuit boards are individually wired together with plug-in
harmonica-style connectors, and a maze of wire between the connectors.
The circuit boards have arrays of connector pins on one end onto which the
harmonica connectors plug in to provide a connection. The connection pins
are tin plated, as are the sockets in the harmonica connectors.
The circuit boards are made of fiberglass, with traces on both
sides of the boards, and plated-through feedthroughs connecting traces
on each side of the board together. Components are mounted only on one side of
each circuit board.
Cogito 240SR Circuit Board Detail
The boards are connected together in a stack, with
insulating sheets of a stiff fiberboard-type material between the boards
to keep components on the top of one board from shorting on the traces of the
board above it. The boards are secured in the stack by a series of
six plastic spacers that hold the boards seperated from each other. This
means that in order to replace a given circuit board, all of the boards
above it in the stack must be removed to get to the board that is to be
replaced. Keeping track of the maze of harmonica connectors that had to be
unplugged when replacing a board had to be one of a service technician's most
difficult things to have to do. The other difficulty with this design is that
it makes it very difficult to troubleshoot in the field, as gaining access
to circuit board components and traces while the machine is running
is next to impossible. The upside to this design approach is that it is quite
efficient in terms of space, allowing the machine to be somewhat smaller than
it would have been had a more serviceable design been used.
Cogito 240SR Keyboard Layout Due to the rather unusual architecture
of the machine, operating the 240SR is different than just about any other
calculator ever made. The architecture of the machine
is not quite intuitive insofar as how to determine the input for a
sequence of calculations in order to arrive at the right answer. While
the Friden 130 required folks to learn the Reverse Polish stack method that
it uses, once the basics were understood, it became quite easy to
perform complex chains of operations with little trouble. The arithmetic
methods of the 240SR aren't quite as easily understood.
Numeric entry is pretty straightforward,
using the numeric keys and the [DECIMAL] key to enter a decimal point.
Negative numbers are entered by pressing the [NEG] key prior to entering
the number. An unusual inclusion is the [POS] key, which forces the number
to be positive, useful to correct the inadvertent entry of a negative number.
The decimal point may be entered anywhere in the number, and the machine
automatically adjusts the position of the decimal point on the display.
By default, numeric entries are entered into the K register, beginning
at the left-most (not including the place holder first four zeroes)
digit, and advancing digit at a time to the right as digits are entered.
It is possible to directly enter numbers into the P and Q registers
by pressing the [Q] or [P] key (located in the left bank of keyboard keys)
prior to making digit entry. The [CLEAR K] key clears the K register, and
is used mainly to correct erroneous entries.
Moving register content between the registers is performed by a group of
four keys; [Q->K], [Q->P], [P->K], and [P->Q]. The transfer from one
register to another results in the source register being cleared, and
the destination register receiving the former content of the source
register.
The [EXCHANGE], [ENTER], and [RECALL] keys control the transfer of data
between the main set of registers, and the surrogate
registers. By default, the [EXCHANGE] key exchanges the content of the
K register with the K' register. The [ENTER] key takes the content
of the K register, and stores in into the K' register, leaving the K
register clear. Lastly, the [RECALL] key moves the content of the K'
register into the K register, clearing the K' register. The [Q] and
[P] keys can be pressed before an [EXCHANGE], [ENTER], or [RECALL] key
to perform the function on the designated register. For example, to
exchange the Q and Q' registers, the [Q] key would be pressed, followed
by the [EXCHANGE] key.
Cover of Cogito 240SR User's Guide
The arithmetic function keys, grouped at the right end of the keyboard,
are a little unusual in that the multiplication, division, and square
root functions each have their own individual "=" key that causes the
result to be calculated. Addition and subtraction operate arithmetically,
with the [+] key causing the content of the K register to be added
to the P register, and the [-] key subtracting the content of the K
register from the P register. After addition or subtraction, the K
register is left undistrubed, allowing for simple constant addition or
subtraction.
Multiplication and division operate by entering the first number
then pressing the [X] or [÷] key. In the case of multiplication, the
content of the K register is copied to the Q register, and the K register
is cleared for entry of the multiplicand. In the case of division,
the K register is copied into the P register, and the K register
cleared. After this, the second number is then entered,
and then the [X=] or [÷=] key is pressed to calculate
the result. In the case of multiplication, the product is left in the P
register, and the multiplicand is left in the K register, and the multiplier
is left in the Q register. In the case of division, the quotient is
stored in the Q register, the remainer is left in the P register, and the
divisor is left in the K register.
The square root function of the 240SR is most interesting -- it is a
two step operation. Most calculators that perform square root operate
simply by typing in the number to calculate the square root of, then
pressing the square root function key. The answer is calculated and presented
to the user. In the case of the 240SR, the user must supply an estimated value
as a second operand to help the machine in the calculation. To calculate
a square root, the number to have it's root extracted is entered, then the
key with the square root symbol is pressed, which moves the number to the
P register, clearing the K register. Next, the guess value is entered into
the K register, and the [square root =] key is pressed to begin the
calculation. The Q register and P' registers are cleared, and the extraction
of the square root begins, using the guess value as a starting point. The
square root is presented in the Q register, and the radicand remains available
in the P register (as well as the P' register). The square root calculation
is extremely interesting to watch, with numbers zipping all about the
display (see the VIDEO for an example). The
method used for calculating square roots is based on the
Newton-Raphson method of successive approximation. The
root extraction calculation progresses slowly enough to observe the trial
roots converging on each other, until they are close enough together that
the calculation stops. Sadly, this method for calculating square roots, on
a calculator that is already somewhat so, makes long calculating times for
square roots, especially if the estimate provided is inaccurate. Some square
root operations can take nearly a half-minute to complete.
At the right end of the keyboard are three switches labeled "÷= ACCUM",
"X ACCUM", and "X= ACCUM". These switches are used to accumulate sums of
quotients, multipliers, and products. These functions are useful for
performing operations such as averages, sums of products, and other types
of statistical operations. When the "÷= ACCUM" switch is on, the
calculator accumulates results of division operations in the Q register.
The "X ACCUM" switch enables the accumulation of multiplicands in the P'
memory register. The "X= ACCUM" switch enables the accumulation of products
in the P register. In addition to these accumulation modes, the [GRAND TOTAL =]
key provides an additional means for accumulating sums of products. This
key functions similarly to the [X=] key, except that an ongoing sum of products
is accumulated in the P' register. The [GRAND TOTAL] key is used to recall
the accumulated sum of products to the P register, by transferring the P'
register to the P register, clearing the P' register.
Lastly, a unique feature, an addition to the original design of
the machine as designed by Frankel, is designed to help with performing
statistical correllation functions. This feature was added by SCM engineer
Jorge Hernandez, and patented seperately under US Patent #3553445. The switch,
located to the lower right of the numeric keypad, selects what is called
"Dual Multiplication Control". When this switch is on, the function of the
[DECIMAL] key is changed, such that five zeroes are indexed when it is pressed.
These five zeroes serve as a seperator to allow the K register to contain
two numbers, seperated by five zeroes. For example, to simultaneously
calculate the following summations, x2, 2xy, y2, x,
and y, the "X ACCUM" and "X= ACCUM" switches would be turned on, along with
the "DUAL MULTIPLICATION CONTROL". For the two groups of numbers X and Y,
let us use (40, 98, 81, and 40) for X, and (21, 57, 45, 26) for Y.
To begin, 40 is entered, followed by the [DECIMAL] key (to seperate X from
Y), then 21 is entered (with the K register display showing
"oooo40.ooooo21ooo"), the [X] key is pressed, follwed by the [X=] key
(which calculates the square of X and Y and presents the result of 1600 (for
x2) and 441 (for y2), and 168 for 2xy as
"oooo16oo.ooo168ooooo441oooooo" (using the 'o' to represent the half-sized
zeroes rendered on the display) in the P register. Continuing by
entering the remaining numbers the X and Y lists using the same method
results in a final answer in the P register of "oooo19365.oo22222ooo6391ooooo",
which indicates the sum of x2 of 19365, the sum of 2xy as 22222,
and the sum of y2 of 6391. Pressing the [GRAND TOTAL] key
provides the summation of the X and Y lists of numbers, displaying
"oooo259.oooo149oooooooooooooo" in the P register, indicating the sum of
the X list is 259, and the sum of the Y list is 149. The results take some
intepretation, but via this functionality, a lot of simultaneous calculations
can be done in a comparatively simple manner.
Overflow is handled rather uniquely
on the 240SR. Unlike other early calculators that simply roll over (e.g.,
on a 10-digit machine, 999999999 + 1 = 0000000000),
or stop the calculation, the 240SR always tries to perform the calculation
as best as it can. For results that overflow the capacity of the register
that holds the answer (the Q or P registers), a system utilizing 'streaks'
displayed below and to the right or left end of the register is used.
In the case of large numbers (> 1011), a streak at the right end
of a register indicates that the decimal point should be positioned 16 places
to the right of the displayed decimal point position to provide an approximation
of the result. If the streak is at the left end of the display, the decimal
point would be positioned 48 places to the right of its displayed position.
If both left and right streaks are displayed, the decimal point in the
approximate answer is located 32 places to the right of the displayed decimal
point location. In the case of small numbers (less than .000001), a left
streak indicates that the number to be read should be preceded with four zeros
plus one zero for each place to the right of the decimal point. A right
streak indicates that the number to be read should be preceeded by 36 zeroes
and one zero for each place to the right of the decimal. If both left and
right streaks are visible, the number should be read with 20 zeroes and one
zero for each place to the right of the decimal. The system can be a bit
tedious, but it does allow the machine to provide approximations of very large
or very small numbers.
Profile View of the Cogito 240SR
The 240SR is most definitly not a speed demon. A general weakness
of the design is the main reason why the machine isn't
particularly fast. When decimal point positioning must occur in the P
register, the only way that the shifting of the number a single place
to the left or right can occur is to circulate the content of the register
completely through the register, which takes a significant amount
of time. An example of this can be seen in the video. A typical addition that
does not require decimal repositioning occurs quite quickly. However, that
changes quickly when a decimal repositioning has to occur, with the shifting
process taking an operation that is almost instantaneous to one that can
take almost a full second to complete. The same issue plagues multiplication
and division. Simple multiplications take around 3/4-second, and more complex
multiplications can take up to two seconds. Division operations complete
in a somewhat slower times, with the most complex division of 999999999999 ÷ 1
taking just over four seconds to complete. The square root operation is by far
the slowest, with some calculations requiring nearly 25 seconds to complete.
The reality of the final product is that, in some cases, the electromechanical
machines that preceded the 240SR may actually be faster!
While the Cogito 240SR was not particularly
stunning in terms of its performance, it's legacy as the
brilliance of Stanley Frankel still lives on to this day.
Electronics Magazine, April 1961
Article courtesy Nigel Tout
Click Image to view a video of display in operation.
