Fullscreen HTML5Med Res (??MB)HTML4

A short history of... by Chris Morris SincLAiR cALcuLAToRs Sinclair’s innovative scientific and programmable calculators were all designed and assembled at Enderby’s Mill, St. Ives, Huntingdon by Sinclair Radionics. C onsultant Nigel Searle developed the pioneering firmware for the first scientific calculator in Texas in 1973. In March 1974, the first Sinclair Scientific went into full production. This calculator used the diminutive but elegant 1973 Cambridge case, in white with just 18 keys. It was nicely colour coded with no key lettering to wear off, other than on the arithmetic operators. Selling in large numbers, it put Sinclair on a roller-coaster ride at a time when the new scientific calculator market was very dynamic. The internal microcontroller was based on the pioneering Texas Instruments ‘calculator-on-a-chip’ (also used as the TMS0803 in the four-function TI Datamath). That sleek Datamath case, styled by Fred Gore & Associates of Carrolton, Texas, was meant to compete with the popular Bowmar ‘Brain’ - which had the largest sales in the world at the time for four-function calculators. Incorporating compact mathematical algorithms suggested by Clive Sinclair (and probably Marvin Minsky at MIT), Nigel Searle reprogrammed the TI chip firmware in late 1973 to cram a dozen scientific and arithmetic functions into the same Datamath four-function ROM, using the sparse 320 micro-instructions available. Reverse Polish Notation (RPN) was chosen for optimum efficiency, along with a two-level stack and the bare three registers on-chip. With RPN, rather than entering, say, 3 + 5 × 7 = to get 38, you would input 5, E, 7, ×, 3, +. The operators (× and + here) work on the last two entered or calculated values. Ken Shirriff reverse-engineered that Sinclair/TI chip in 2013 and produced an ingenious simulator of a very clever calculator. The Scientific could be slow and was accurate to only three decimal places at the margins, but it did the job at a very good price for the time (£50 on introduction, later falling rapidly in mass-production). The reprogrammed 28-pin PDIP PMOS chip (running from -10V and -16V supply rails) was named the TMC0805 by TI for Sinclair. Power came from four AAA cells that had a 25 hour operating life. It used an excellent Bowmar Optostic nine-digit miniature LED display Originally, Bowmar had hoped to develop the Scientific chip for Sinclair as well, but they were pre-empted by Texas Instruments. Bowmar, started by Ed White in Boston, was the first manufacturer to The 1854 Corn Mill on the River Ouse, home of the Sinclair Radionics Factory. 4 successfully mass-produce the pocket calculator in North America. The sturdy Bowmar 901B Brain used the TI TMS0103 ‘calculator-on-a-chip’ in September 1971. Their LED/calculator factory, in Acton, Massachusetts, was later supplemented by Bowmar factories in Ottawa, Canada and Nogales, Mexico. Demand for both pocket calculators and LED displays grew at a dizzying rate between 1972 and 1974. Clive and Ed were good friends, and Clive bought hundreds of thousands of Bowmar LED displays. As it turned out, their calculators both shared the same pioneering LSI chip supplier, Texas Instruments. This became a mixed blessing later when, impressed by Bowmar’s success, TI started making its own Datamath calculators using the same chip series from their own factory. The 12 functions found on the Sinclair Scientific were the usual four arithmetic ones (+, -, × & ÷) plus sine, cosine, tangent, arcsine, arccosine, arctangent, log and antilog (exponentiation). The calculator worked in radians, and scientific notation (eg, 1.23E4 = 12,340) was used to conserve instructions, with six-digit precision. It became a best-seller in its time (1974-76). Most ingeniously, the key constants for deriving yet more functions were The World’s first £50 scientific calculator, released in March 1974. It has 18 keys and a small, attractive Bowmar LED display. Source: https://pemag.au/link/ac4r Practical Electronics | May | 2025 printed on the case below the display, to six figures, as there was no room left in the hard-pressed ROM. This enabled the user to also derive xy, ex, √x and π. This was well-explained in the comprehensive 32-page manual. More recently, the original Sinclair Scientific’s later ‘super-hacking’ popularity has resulted in several modern ‘emulated’ versions being produced in both the USA and Canada. They use Ken Shirriff’s Sinclair retro firmware that includes the original 320 word microcode, though it has been moved to a modern Atmel AVR microcontroller. The next big thing The next development at Sinclair came in a larger and restyled case. This was developed on contract for Gillette as a simple four-function calculator called the GPA. Sinclair reused the case for their new Oxford series, while Gillette decided to stick to razors. This initially resulted in the Oxford 300 scientific, introduced a year after the pioneering Scientific, in March 1975. However, it used a single scientific chip (the CF5966) that General Instrument Corporation developed later. The chip was not unique to Sinclair and, in fact, was widely used by other competitors. Trig and log results were slightly more accurate than before, to the fourth decimal place instead of the third. The Oxford case, much bulkier and heavier than the slim Sinclair Scientific, had just 19 large keys. It was powered by a 9V PP3 battery, driving a green vacuum fluorescent display (VFD) from Noritake Itron, powered by -24V derived using an Astec DC-DC converter module (which also provided the -16V for the GIC PMOS chip). VFDs had been developed in the early 1970s by the Japanese calculator industry, as they were initially barred by tariffs and patents from obtaining LED displays from the USA. LEDs used simpler lowvoltage circuitry and drew less current than VFDs. The Oxford case snapped together with plastic rivets; no screws were used. The thin main circuit board required unexpected bracing inside in some cases (with glued wood offcuts on my original factory example!), indicating a rushed, post-Gillette GPA development of the Oxford design. Despite the heavy-duty outer case, the ‘bendy’ main circuit board inside was only 0.7mm thick; that’s precariously thin, around half the ‘standard’ PCB thickness of 1.5-1.6mm. By comparison, the pioneering Sinclair Scientific had a sturdy 1.6mm-thick single circuit board. Radians or degrees were selected by a slide switch on the case’s upper end. The Oxford does not seem to have been as popular as the Cambridge series, and was never available as a kit (it was probably too awkward to assemble). Cambridge Scientific mail-order kits were very much in favour amongst Heathkit enthusiasts in the USA, according to Nigel Searle in a 2021 YouTube interview. In retrospect, the Oxford looks large and clunky even now, when compared to the best-selling Hewlett Packard HP-25 programmable of that year, never mind the The contract Gillette GPA calculator. The same case was used in the first Sinclair 24-step Programmable calculator. The very first US pocket calculator, the Bowmar ‘Brain’ 901C, priced at $240 in September 1971. Mine was made in Ottawa, Canada. Usage instructions were on a label on the underside. Image source: National Museum of American History. The stylish but bulky Datamath II (top) compared to the diminutive Scientific calculator below. Both were 1974 designs that used the same chip with differing firmware. Practical Electronics | May | 2025 5 The Sinclair 24-step Programmable calculator reused the Gillette model’s case. Source: https:// pemag. au/ link/ ac4s sleek Cambridge or later clever Sinclair Enterprise. Inside, the main circuit board/ display and the keyboard PCB are rather crudely interconnected and supported. Having only 19 keys in such a large case seems like an opportunity missed. Further improvements In the following year (1976), the follow­-up Sinclair Oxford Scientific came out, in a white case. This used the GIC CF596 chip instead, adding yx, log(x), x2, ± and double brackets. There were not enough keys left to use the 10x and % functions also available on the chip. For the same reason, three of the five previous memory options on the Oxford 300 keyboard had to be dropped. In August 1975, five months after the popular HP-25 programmable launched at US$200, the Sinclair Scientific Programmable was produced for £30, in the same black Oxford case as the 300, with a green VFD display but with RPN operation, using custom Sinclair firmware. This was a failure as a product, despite being only 40% of the HP cost. With no branching, no step/debug, a single memory and 24 steps, it could not be considered a real programmable calculator, even at the time. It was more like the equally unsuccessful National Semiconductor “Novus Programmables” of 1975 in function, just a learn-mode machine. Even as a ‘scientific calculator’ it didn’t 6 match the original 1974 Scientific, inexplicably lacking tan, arc sin and arc cos. The firmware was likely again programmed by Nigel Searle, who had been assigned by Clive Sinclair to rapidly develop a programmable calculator after the success of the Scientific, though undoubtedly with a too-small ROM and RAM. The Sinclair Cambridge Scientific was released in March 1976, using the small white Cambridge case with 19 keys. This time, they used the same standard scientific chip as in the Oxford 300, the GIC CF596 in 28-pin PDIP, despite the absence of a fluorescent display (the ‘F’ option in the chip code). As mentioned earlier, that 1975 chip was used in numerous rival makes of calculators at the time; hence, Sinclair only had styling and price to differentiate themselves from a raft of competitors. It eschewed RPN for algebraic notation. Interestingly, only two AAA cells were used instead of four, saving weight and space, although reducing the running time with the heavier current draw. A radian/degree slide switch was fitted in the unused half of the battery compartment, requiring the battery lid removed in operation to change this trigonometric option. Compared to the original Scientific, natural log (ln)/ex was substituted for log(x)/10x, with 1/x, √x, π and a single memory slot added. By early 1977, Sinclair was fitting the more advanced GIC CF5986 chip to this model instead, though there were no spare keys available to use the new yx or log(x)/10x functions. As with the original Scientific of 1974, the two Texas Instruments custom support chips (inverter and LED digit drive), IFC1 and IFC2, were also fitted. One wonders why the fluorescent drive version of the chip was used when it was only driving low-voltage LED segments. Only those at the Mill knew. The Sinclair Cambridge Programmable was introduced in September 1976. This was the first serious programmable calculator from Sinclair, although it came quite late during the fast­-moving mid-1970s world of programmable calculator development. It came in the white Cambridge mini case. It had just 36 steps and many of the 19 keys had to perform triple duty. With only a single memory register, it was not very effective, although it did have go-to and go-if-negative branching, plus a crude step/debug display. Algebraic logic was used, based on custom The 36-step Sinclair Cambridge Programmable calculator, introduced in September 1976. It wasn’t especially successful. Image source: Kevan Dickin (https://pemag.au/link/ac4t). Practical Electronics | May | 2025 Sinclair programming of a National Semiconductor MM5799 COPS microcontroller in a 28-pin PDIP package. The MM5799 chip had 1536 microinstructions in its ROM, plus 384 bits of RAM for the 36-step program and a single memory slot. A DS8874 shift register was used to drive the nine-digit LED display. This display was also from National Semiconductor. Bowmar (the biggest calculator manufacturer by volume in the world in 1973) had gone bankrupt by February 1975, at a time when calculator prices had fallen to 10% of their 1972 levels. Radio Shack sold a re-badged version of this same Sinclair Cambridge Programmable called the EC-4001. Both machines had an unsightly ‘camel hump’ 9V battery compartment at the back. This spoiled the sleek Cambridge look and prevented the calculator from sitting flat on a desk. It could no longer fit in the neat hardshell carrying case that had helped to define the Cambridge calculator family. Only 16 op codes were displayed; many were used ambiguously for three different functions at a time. No fewer than seven steps were needed to halt an accidental endless loop in a program; an example of the awkward programming model and limited number of keys. The operating manual was carelessly written (Sinclair manuals were usually very good) – it didn’t even include an illustration of the keyboard! The programming section occupied only a third of the slim 25-page manual, compared to half of the later 45-page Enterprise manual. As with the Cambridge Scientific, the lettering rapidly wore off the number keys (the original Scientific only needed lettered keys for the four arithmetic functions). Ironically, this was the one area where the Oxford series were superior; they used the Hewlett Packard method of double injection moulded keys, with two plastic colours in each key, so the lettering could never fade. However, this more expensive feature was dropped by 1976, as is apparent from the faded keys on the later Sinclair Oxford Scientifics from that year. Finally, in July 1978, as Sinclair Radionics was in its final phase, the most successful Sinclair programmable calculator yet was produced. Available for the very reasonable price of £25, this was the Sinclair Enterprise Programmable. “Enterprise” referred to the new owners of last resort, the National Enterprise Board. Housed in an elegant new slimline case, parting easily to access the 9V PP3 battery without a separate cover, the Enterprise had the go-to and go-if-negative branching instructions of the 1976 Cambridge Programmable, plus a respectable 7-entry storage register and memory arithmetic. Practical Electronics | May | 2025 A much more usable op code display for debugging was provided (with 42 separate op codes). A substantial three-volume program library was available. Precision was excellent at eight digits. Despite a very capable program length of 79 semi-merged steps, subroutine calls were sadly not included. Nevertheless, it was a good, reliable and practical machine intelligently laid out, showing the same innovative spirit as the 1974 Scientific. As with the earlier 1976 programmable, a National Semiconductor 4-bit COPS PMOS family chip was again used (this time, the MM57146 with the MM57126 serial RAM for storing the 79 program steps) with custom Sinclair firmware, but to much better effect. The keypad had a good ‘click’ feel (as indeed did most Sinclair calculators with their Klixon-style action). The expanded set of 24 keys (versus 18 or 19 keys in all previous Sinclair machines) was much easier to use and reduced the separate steps needed in a program. A more durable key printing process ensured the key lettering did not rapidly fade, as had happened with the earlier Cambridge models (which had used poor contrast grey keys with fragile white lettering). In a sense, Sinclair’s scientific and programmable calculators came (1974) and went (1978) with a bang, incorporating excitement, innovative design and styling with good prices. The years in between, not so much. The world-wide evolution of the scientific calculator in the 1970s was unbelievably fast. Sinclair played a significant role during that energetic time, bringing both an elegant appearance to an everyday object and (sometimes) very clever firmware and operational design, with well-written manuals and program libraries. Their calculators were widely exported, often as kits and through mail order, particularly to the United States via their New York City office on East 57th Street. To this day, the 1970s Sinclair calculator family is regarded as iconic, both in England and North America. In July 1979, Clive Sinclair left the plush top floor of the Mill and Sinclair Radionics, never to return. His next successful endeavour, the ZX80/81 and Spectrum personal computers, began from Kings Parade, Cambridge with a new company, Sinclair Research. But that is another story… The third, successful Sinclair Programmable calculator had 24 keys in a slim, attractive case and easy battery access (by lifting off the top cover). Source: Kevan Dickin. References • http://files.righto.com/calculator/ sinclair_scientific_simulator.html • https://hackaday.io/project/91895sinclair-scientific-calculator-emulator • https://www.tindie.com/products/ arduinoenigma/sinclair-scientificcalculator-emulator/ • https://theretrohour.com/sinclair-inside-story-nigel-searle-ep256/ • https://www.wass.net/othermanuals/ (look for General Instruments MicroElectronics Calculator Chips Catalog datasheets 1978) • https://spectrum.ieee.org/theconsumer-electronics-hall-of-famebowmar-901b • https://www.rsp-italy.it/Electronics/ Databooks/National%20Semiconductor /index.htm (look for 1977 MOS LSI data book, pp9-27) PE 7