| +Home | Museum | Wanted | Specs | Previous | Next |
Sharp Compet 20 Electronic Desktop Calculator
Updated 1/10/2006
This poor machine sadly suffered at the hands of careless shippers, receiving some damage on its way to the museum. The damage to the machine was unfortunately functional as well as cosmetic. While the cosmetic damage is somewhat repairable, the real difficulty is in determining what damage was caused to the functionality of the machine. When received, one Nixie tube was shattered, three others would not light up, and the machine was non-responsive to keyboard input. The shattered Nixie tube was replaced with an exact replacement, courtesy of a fellow calculator enthusiast, Brent Hilpert (see his wonderful calculator technology Web site by clicking HERE). Over time, I've found that a number of the circuit boards had small fractures in them (due to the rough handling in shipment), causing some circuit traces to be broken. Upon carefully inspecting each board with a laser backlight and repairing the fractures found, the three initially non-working Nixie tubes are working again. After quite a bit of electronic detective work, the machine is now nearly 100% functional, with only some quirks (which may have been part of the original design), and a problem with the error/overflow detection circuitry.
Profile View of Sharp Compet 20 In the process of working through the
problems with the machine, I've also been doing some reverse-engineering and
data gathering. Even though I've been quite successful in
resurrecting the machine, I'd love to find more information on it.
If anyone out there has service experience or documentation for
this Sharp (or Facit equivalent) machine (e.g. Service Manual or other
service material, or an Operator's Guide), or if you have one of these
machines in operating condition, I'd greatly appreciate hearing from you.
Click HERE to send me
an EMail message.
The ID Tag on the Compet 20 The Compet 20 is the second model of
electronic calculator that Sharp produced for consumer use. According
to the Company History section of the
Sharp USA
Web page, Sharp's first electronic calculator, the Compet 10 (Model CS-10A),
introduced in June of 1964, is claimed as the world's first all-transistor
electronic calculator.
The Compet 20's Predecessor, the Sharp CS-10A While it is true that Sharp's Compet 10
was indeed among the earliest of the all-transistor electronic calculators,
it is generally recognized that the Friden 130,
introduced in late 1963, was the first all-transistor electronic calculator,
though the 130 used a magnetostrictive delay line for its register storage,
which, technically is not a "solid state" device. The Compet 20 was both a redesign and improvement on the Compet 10,
relying on advances in transistor technology to reduce the size and
increase the reliability of the machine. The Compet 20 was introduced
in September of 1965, about a year and a half after
the introduction of the Compet 10. Just a month after the introduction of
the Compet 20, Sharp introduced a higher-end
version of the Compet 20, called the Compet 21 (Model CS-21A) that added
automatic calculation of square roots. The Compet 20 and 21 are
visually identical, the difference between them being additional circuitry
in the Compet 21 to provide the square root function, activated by
pressing the ÷ key, immediately followed by the "=" key.
Near the end of the Sharp Compet assembly line, circa 1967. The Compet 10 used all Germanium transistors, which suffered from
reliability and longevity problems. The Compet 20 used more reliable
Silicon transistors, which were smaller, faster, and consumed less
power than the older Germanium transistors used in the first Sharp
machine. The Compet 20 also improved upon the Compet 10 with a
'10-key' keyboard versus the 'columns of digits' keyboard arrangement
that was used on the Compet 10. As early as the mid-60's, the Swedish company Facit, a powerful force in
the European mechanical calculator market, had an OEM relationship with Sharp
to provide designs for (if not manufacturing of) electronic calculators
for sale into the European marketplace.
Based on date codes on some of the parts in this Compet 20, I would target
the exhibited machine as being produced sometime in mid-to-late 1966.
The Compet 20 is a fairly basic
four-function electronic calculator. It has a 14-digit (plus a 15th sign digit)
Nixie-tube display. The digit Nixie tubes indicate 0 through 9, and also
include a right-hand decimal point. There is no leading zero suppression,
but the machine does have floating decimal point (and a 'fixed' point mode
determined by the setting of the "M" key, explained later). Add and subtract
operations work 'adding-machine' style, with the keys actually performing
a "+=" or "-=" operation, even though the keycap nomenclature doesn't
indicate this.
Compet 20's Unusual Multiplication Display (1234 X 5678 entered) The multiply operation on the
Compet 20 works rather differently than many calculators. The multiplicand
is first entered, then the 'X' key is pressed. A small indicator lights
up on the 'X' key to indicate a multiplication is in progress, and the Nixie
display after the last digit of the multiplicand lights up both the 7 and 9
digits at the same time (see photo above). Then, the multiplier is entered,
and the display shifts all of the digits to the left for each digit of
the multiplier entered. The resulting display looks rather odd, with the
multiplicand and multiplier all on the display at the same time, separated
by this odd looking digit with both 7 and 9 lit simultaneously. The "=" key
is then pressed, and the product is calculated and displayed. This method of
putting both the multiplicand and multiplier on the display at the same
was used on early transistorized Canon calculators (
Canon 161 and
Canon 130S), and
and on the early Small-Scale Integration (SSI)
Integrated Circuit-based Brother Calther 412. Division on the Compet 20 operates much more conventionally, with the
added twist of an indicator light in the keycap like on the multiply key
that lights up when the dividend has been entered and the machine
is waiting for the divisor to be entered, followed by the
"=" key to obtain the result. The Nixie tubes are left active during
calculations, and during some of the longer calculations (though the
machine is quite fast) the numbers in the display dance quite a jig while
such operations are in progress. The "M" key is a "push to latch, push again to unlatch" switch that
changes the way that decimal point positioning is handled. When the "M"
key is not activated, the calculator behaves in full floating decimal mode.
When the "M" key is locked down, the decimal point position is determined
by where the user enters the decimal point in the first number of a
calculation. For example, with the "M" key depressed, entering 1 divided
by 2 will result in 0. However, performing 1.00 divided by 2 will result in 0.50; and performing 1.00000 divided by 2 will result in 0.50000.
The "X-" key (with the - located under the X) appears to work the same as
the "X" key, except that the multiplier is negated. The "RC" key recalls
the last number 'entered', for example, If you do a '12 + 13 RC, the
12 will return to the display. The machine sports a backspace key,
which was a thoughtful addition which many early calculators (and many
even today) lack.
The Compet 20 with Case Removed As part of the effort to resurrect
the machine, I have been working on instrumenting the electronics as well
as reverse engineering partial schematics of the machine to try to better
understand how it works. As part of this project, I've come to some
assumptions about the general technology and operating principles of the
machine. All of the technical information includeded here is based on these
observations and assumptions. The Compet 20 is built entirely with transistor technology. There are
no integrated circuits anywhere in the machine. Transistors are used for
all active circuitry, including flip-flops, level shifters, gating, and
display drivers. An assortment of different types of transistors are used,
however the majority are packaged in small ceramic-cased 'pancakes', that
show up on the photos as white 'dots', some with green or yellow markings
on them. Diode/resistor arrays are used for logic gating functions. The
machine appears to have been designed using standardized circuits, with the
same basic core of parts for flip-flops and gates, with component variances
occurring in cases where different logic levels are needed or fan-out
requirements differ. Most, if not all, of the flip-flops used seem to be of the
RS style. The machine appears to use two different sets of logic
levels...one set of levels for use 'on board', and another set of voltages
used for communications across the backplane. For on-board logic,
10V represents a 1, and near ground represents a 0. For backplane levels,
4.5 to 5 volts represents a 1, and ground represents a 0. It isn't clear
to me why the backplane logic levels are different than the 'on board'
logic levels. One of the 'glue' circuit boards The brains of the machine are made up of 20 circuit boards which
plug into a hand-wired backplane. Two large boards are located at
the rear of the machine, spanning the width of the
cabinet. These boards, based on nomenclature found on one of them, form
the "program" for the machine. One of these program boards is populated
primarily with a large number of diodes, while the other has a lot of logic
and active circuitry. My assumption here is that this machine, in contrast
to the asynchronous bit-serial architectures of other early electronic
calculators, is actually microcoded, with the board full of diodes
serving as a ROM (Read Only Memory) containing the instructions
governing the operation of the machine, and the other board as the
sequencer which steps the machine through the microcoded instructions.
If this assumption is correct, Sharp probably can lay claim to the
first all-transistor microcoded electronic calculator. A microcoded
architecture for such calculating equipment makes for a lot of flexibility.
If the microinstruction set and its execution engine are properly designed,
it becomes a fairly simple matter to make modifications to the behavior
or functions of the machine without having the 'rewire' the whole thing.
There are also other hints in the way that this machine is put together
that bolster my assertion that this is a microcoded machine. A 'fully populated' digit board (left) versus a digit board with only register, decode and driver logic (right) The remaining 18 boards are smaller than the program boards, and are
plugged into the backplane perpendicular to the program boards. Fourteen of
these boards are what I call 'digit' boards. Each digit board contains four
flip-flops connected in FIFO (First In, First Out) shift register fashion
which serve to hold the BCD (binary-coded decimal) digit to be displayed
at that digit position; an array of diode/resistor gates that serve to
decode the BCD number stored in the 4-bit register into a 1 of 10 selector;
and finally, driver transistors (2SC287) which switch on the appropriate
digit in the Nixie tube based on the outputs of the decoder logic.
The Nixie tube for the particular digit is mounted directly to the board,
situated so that it is properly positioned to shine through the display
window in the keyboard bezel. This parallel digit design, where each Nixie
tube has its own display decoding and driving circuitry, is much more
straightforward to design and troubleshoot than multiplexed displays,
but costs significantly more to manufacture because many more parts are
required. The remainder of the real-estate on the digit boards is dedicated
to almost what I call a "breadboard" area, where up to 8 flip flops and
various gating functions can be built up. This breadboard (though it is
not really a breadboard in the true sense of the word, as there are pre-etched
traces on the boards for some of the components, and there are definitely
etch differences between the different digit boards.) area by my guess is
where the working registers, data path, and switching circuitry resides,
being distributed throughout all of the digit boards, with interconnections
through the backplane. A special 15th digit board has a unique Nixie tube
which contains only a "+" and a "-", used for indicating the sign of the
number in the display. The hand-wired Backplane The rightmost card in in the cage is the keyboard interface. This card
has an edge connector at its top edge that provides a connection for
the wiring harness from the keyboard. This card provides keyboard encoding
and signal conditioning from the switch contacts that make up the keyboard.
The two left-most cards in the cage are pretty tightly packed with parts,
and appear to be glue boards which contain numerous functions, including
decimal point logic and drivers and error/overflow detection and lockout. Compet 20 factory assembly line power supply adjustment, circa 1967 The calculator uses a conventional linear power supply. Voltages produced
are +12 and -12 volts, +5.25V, and +185V (used for driving the Nixie tubes).
The 5 Volt and +12 Volt supplies are transistor regulated. The Compet 20 is actually a fairly fast calculator for its time.
Compared to the Friden 130, the Compet 20
is a speed-demon. Addition and subtraction are return virtually
instantaneous results. The worst-case multiplication seems to take
about 1/4 second (999999 X 999999), and thirteen 9's divided by 1
takes about 1/3 second. Note that in the 'all nines' calculations, the full
capacity of the calculator is not exercised. As with many early
electronic calculators, the Compet 20 gives incorrect answers when the
most-significant (14th) digit of an operand is non-zero.
Typically this behavior is related to the fact that one of the digits in
the accumulator is used as a single-digit counter for counting up or down
(up for division and down for multiplication) as each digit of the multiplier
or quotient are used/generated during multiplication and division operations.
Image Courtesy Nigel Tout
Photo Taken at Science Museum, London
Note finished Sharp Compet 30 calculators in foreground
