Development of the HP-35 calculator: how innovation was created

The HP-35 is an example of an extremely attractive, miniature and revolutionary product. Four-function handheld calculators were already on sale [formerly “handheld” or “handheld”, or “portable”, called electronics small enough to fit in a hand and be portable - before all the equipment such / approx. trans.]. Few people could imagine a machine capable of producing scientific calculations and fit in a shirt pocket, but many were already beginning to dream of such. The HP-35 was developed by Hewlett-Packard, based in Palo Alto, California, at 1501, Page Mill Road, and launched onto the market in 1972. It was the first full-featured scientific calculator the size of a shirt pocket. This invention revolutionized the engineering profession, allowing him to perform almost instantaneous and extremely accurate scientific calculations at home, in the office or "in the field." The HP-35 was the innovative culmination of mechanical design, advanced technology, the development of algorithms and applications - all this was unique at that time.

Many of us consider ourselves to be inventors, but really we need to focus on “innovations” - making things that other people need, the kind that they want to buy. Research must support the development of the final product. If the result is something new, then perhaps a worthy patent invention will appear. Many engineers begin to develop from the inside, from the iron, from the engine, and then equip it with a body. However, an analysis of the most successful products in recent decades shows that they began to be developed outside; Appearance and sensations from use are prioritized over detailed engineering developments.

During the development of the HP-9100 desktop calculator, HP-35's older brother, I was responsible for developing algorithms that fit the architecture proposed by freelance inventor Tom Osborn. Tom brought HP with a four-function and floating-point scheme that became the basis of the 9100 architecture. The proposed algorithm methodology was taken from the Athena calculator, developed by Malcolm Macmillan, also responsible for the fixed-point calculator and transcendental functions . I had to read a lot of literature in order to understand the technique of various types of calculations, many of which are already over 1000 years old. Although Wang Laboratories used similar methods of calculation, in my research I found historical examples, the oldest of which dates back to the year 1624, which revokes the patents of Wang Laboratories.

The study helped to master transcendental functions (exponential function, logarithm and trigonometric functions) by using algorithms that meet the needs of the user and fit into the limitations of iron. Such an approach proved invaluable during the development of the HP-35, up to the use of 12-digit constants for generating functions, in order to reduce deviations to one in the 11th decimal place.

We have thought over the choice of algorithms for the HP-35 very carefully. Power series, decomposition of polynomials, continued fractions , Chebyshev polynomials - all this was considered for use in the calculations of transcendental functions. And all this was too slow due to the number of multiplications and divisions required to maintain accuracy up to the tenth digit in the proposed interval of two hundred degrees ten. A generalized algorithm, best suited to the requirements of the speed and efficiency of programming for the HP-35, was the method of iterative pseudo-sharing and provisioning, first described in 1624 by Henry Briggs in the work of Arithmetica Logarithmica, and later Volder and Meggit. The algorithm of the same type was previously used in HP desktop calculators.

In addition, as a result of the study, it was proposed to use the reverse Polish notation as the basis for the development of the HP-35 for developing the desktop calculator HP-9100. This affected all aspects of the development - from the number of keys to the architecture of the internal logic. Reverse Polish notation requires that the operator be entered after the operands, as a result of which brackets are no longer needed. This made it possible to enter data using fewer keystrokes, as well as to simplify the hardware.

The calculator project began literally as an attempt to create a scientific calculator that would fit in the pocket of William "Bill" Hewlett's shirt. After developing the HP-9100 desktop scientific calculator in the mid-1960s, Bill became obsessed with the idea that HP should develop a calculator with the same capabilities that fits in his shirt pocket. Every few months, he appeared in the laboratory of the 1U building and asked how his favorite project was going. He often turned to me personally, because I was researching architectures suitable for the scientific algorithms I used in HP-9100.

Although the density of semiconductors increased every year, bipolar transistors would not be suitable for our project - they were too large and consumed too much. MOS structures (metal oxide semiconductor; metal-oxide semiconductor, or MOS) have promised high density and low consumption, but were still in the early stages of development. But this did not stop Hewlett from instructing HP's industrial design team to jot down a few ideas, key layout points, etc., that could fit in a shirt pocket. The laboratory of solid-state electronics also worked on low-power LED displays based on bipolar transistors. From various manufacturers in the USA and Japan, I have assembled a large collection of semiconductor architectures that performed simple calculations with four functions. Most of them were bipolar, but some manufacturers have already tried to develop MOS circuits with several hundred transistors on a chip. At the end of 1970, everything changed dramatically when Fairchild Semiconductor showed HP’s department manager Tom Whitney and me the pMOS architecture, which seemed very suitable for working with scientific algorithms. Binary-decimal adder (binary-coded decimal, BCD) and support for several 20-bit words in shift registers with information circulation were very effective in terms of chip size and power consumption. Fairchild did not have a patent for this architecture, since they allegedly took it from Sweda, the manufacturer of electronic cash registers. They were going to offer this chipset as a platform for calculators with four functions and a fixed point.

For about two weeks, I studied a modified architecture based on what I saw at Fairchild, and decided that I would only need registers for 13 bits (56 bits) and words with a length of 11 bits; they were later reduced to 10 bits using an imaginary conditional branch. The reduction of the scheme by 10% was quite significant. Thirteen discharges should have been enough for an accuracy of 10 discharges, with one discharge for overflow or carry and two protection discharges. The word could be displayed either as a mantissa with two digits of the exponent, or as a result of variable length with a fixed point. The product had to have an arithmetic chip and a register chip, control circuits and a timer, and several ROM chips. How often does a person have a chance to develop a set of microinstructions?

Fairchild decided that they would not make a special custom scheme for HP, so we, along with the department manager Tom Whitney and laboratory director Paul Stoft, went to Bill Hewlett, hoping that he would be pleased with our combination of technology and architecture that could fit in his pocket. We told him that we would need to order the development of several new pMOS chips. And the final cost of the product will definitely be much greater than $ 100, for which then calculators with four functions were sold. Hewlett was not sure that he would get a good response from a million-dollar development, so we used the existing research and development budget, and Hewlett contacted SRI's analytical research center to entrust them with independent market research. SRI worked for many months, studied various focus groups, etc., and gave the answer: what HP conceived was “it is not possible to evaluate”.

The primary goals for creating the NR-35 were:

What should the HP-35 look like? It was supposed to be the size of a pocket - and therefore light and easy to carry. What buttons did he need, how would they all fit under this size limit? Will users accept the prefix and suffix buttons? How to place the buttons so that they are convenient to use? Can I avoid accidentally pressing the adjacent buttons? The battery had to work for several hours without recharging. The display was supposed to be readable at arm's length and in bright sunlight.

The industrial development of the HP-35 was a novelty not only for Hewlett Packard, but for the entire electronics industry as a whole. Usually, the mechanical and electrical components of a product were determined before its appearance was developed; NR-35 went the opposite way.

Since the calculator had to fit in a shirt pocket, size was the dominant limitation of design. Immediately, several more parameters were set. The calculator will require three batteries to achieve the stated operating time, with a highly efficient DC converter on the expected voltage of the semiconductors and with the appropriate power requirements. Based on the development of previous desktop calculators, the HP-35 decided to make 35 keys (obviously, didn't it, that the name was invented after the development?), As well as a fifteen-digit LED display with exponential notation, decimal point and signs for mantissa and exponents .

Industrial design began with the study of the keyboard, housing, the overall concept of the form. With the help of sketches and three-dimensional models, several basic form factors were studied, which made it possible to evaluate well the shapes and sizes under consideration. From the point of view of engineering psychology, the keyboard was the most critical moment. The problem was how to place the 35 keys on a 6.5 cm x 11.5 cm square, while still being able to work with the keys without pressing more than one at a time. It became obvious that the industrial standard of 19 mm between the centers of the keys would have to be abandoned.

A successful compromise was the use of a distance of 17 mm between the centers of the numeric keys, and 13 mm for the rest. This became possible after reducing the size of the keys, which increased the distance between them. The keys are divided into groups according to functions. Groups are divided by size, contrast, color and location. The number keys, as most frequently used, are made larger and have the greatest contrast. Their designations are applied directly on themselves. The next group of keys in frequency of use highlighted in blue. The enter key and the arithmetic keys are highlighted in this group in that their designations are applied to the keys themselves. The least frequently used keys have the least contrast, and their designations are printed on the panels above the keys.

The keyboard requirements for the HP-35 were particularly challenging. It was supposed to be reliable, inexpensive, with low keys, pleasant to the touch. The decision was based on the fact that curved metal strips attached to the ends may have two stable states. When you press a key, tactile feedback was released, similar to the one that the cricket's toy had [apparently, some famous toy among children in the USA of that time / approx. trans.]. The HP has developed a special spring contact with a height of 3 mm. The tactile sensation of the keys gave a clear understanding of the moment at which the contact occurred.

The case of HP-35 was developed taking into account engineering psychology and the importance of appearance. The edges of the special-form calculator make it convenient to hold it in one hand. They also allow him to easily enter the pocket. The keyboard and display are deflected upward so that they are easier to see when using the desktop. The upper part of the body is lighter than the lower - because of this, the product seems thinner than it really is. It seems as if he is soaring when you look at it with normal desktop use. The use of complementary textures greatly influenced its overall elegant appearance. The texture of the case gives a non-slip surface, which is important when held in the hand. An industrial design team led by Ed Lillenvola did an outstanding job without knowing anything about the product stuffing.

At that time, only general information existed about the calculator's electronic stuffing. The development and packaging of all necessary electrical and mechanical components in a tiny product has become a titanic task for electronics and mechanical parts developers and industrial designers. The NR-35 would not have come into being without a terrific working relationship between a development lab, industrial design, manufacturing, and tooling teams. Everyone who worked on the project had a common goal to keep the original size and shape, as a result of which many engineering innovations were invented. Many of the problems encountered during development could have been easily solved in the usual ways, but then the main goals would not have been achieved, and the product would have been less attractive.

NR-35 Development Schedule

In the early planning stages of the HP-35, it was obvious that it would require new display technologies. The existing LEDs consumed too much energy and cost too much. HP developed a five-digit scoreboard that saves energy and cost thanks to the integrated plastic spherical lenses in front of each digit. The efficiency of the LEDs is enhanced by using a low fill factor instead of direct current. In the NR-35 energy is stored in inductors and fed to the LEDs. Such technology allowed to actively use multiplexing; digits were scanned one by one, one segment after another. Extensive reliability tests showed a slight change in intensity after several years of current ripple with a fill factor of 0.1%. The readability of the display even in the bright sun was so important that the individual segments changed a little by adding small tick marks from the left edge of the upper and lower bands. Each segment was also modified so that the perimeter was comparable to the ratio of the areas in order to achieve a uniform visual intensity.

The display of the HP-35 was arranged similarly to the ten-digit displays of the HP desktop calculators. It consisted of 15 seven-segment signs and decimal points. Results between 10 10 and 10 -2 were always shown as floating point numbers, which were appropriately placed on the display, and the power field remained empty. Outside this gap, the HP-35 showed the result in an exponential notation with a decimal point to the right of the first significant digit and a corresponding power of 10 located on the right edge of the display. To improve readability, the decimal point has been allocated its own segment.

The HP-35 had five MOS / LSI (metal-oxide semiconductor / large-scale integration; MOS on a large integrated circuit) circuit: ROM, arithmetic circuit, register circuit (A & R), control circuit and timer (C & T). The logic circuit was developed by France Road and Chang Tan from HP Laboratories, and the electronic circuits were developed and manufactured by two third-party manufacturers. Three custom-made bipolar circuits were also developed at HP Laboratories and manufactured by the Santa Clara business unit — a two-phase clock driver, a clock and anode voltage generator for LEDs, and a cathode LED driver. HP-35 was going on two printed circuit boards. The top contained a display, drivers and keyboard. The lower, smaller one contained all the MOS logic, clock driver and power source.

System architecture NR-35

Battery selection and power source design were non-trivial tasks; To increase the efficiency of the DC / DC converter, a stack of three batteries was used. The efficiency of the single-transistor converter was above 80%, which ensured the task of battery operation for four hours. , ; , , HP Laboratories, , (+7,5 , +6 -12 , ). / .

. , , . 14 - , 56 . 14 , — , , — . .

-. (SYNC), , 56 , . (Is) ROM . , (WS), , C&T ROM. , , , . C&T . — , , , .

. C&T . , ROM. — , ROM, . .

, , . -35 12 , C&T. , , . ROM C&T ROM. — — , .

, . , -35 . , , .

. , «», . , , , . , , , , .

ROM, 256 10 . , .

. . , , , C&T. - , A&R, . A&R : , , 56- , /, . — . . , . , , , , , ..

, . / / . . , , , . / . , , .

. , 1001 (), 0110 (). , . . , .

, C&T, A&R ROM, : , . . — , — .

- -35 . , , 2-3 , , , .

. , , , , , . , . . , , , . .

- , , , , -35. , / . , . - , , . , — , .

. , , . , , , , . , , .

, , , , , , . , .

. , , , . -35 , . , .

, , arctanh(x) — — . , - . (Algorithmic State Machine, ASM). , . , , ROM- 750 .

. / , , .

-35 . 11- . , , , 10- . . , , , , . , .

. , 1,0 0,9999999999 . , , , 90°. 89,9° , 0,1°. 10 10 , .

-35 «». — , HP-9100, — . . « », pMOS, , . Fairchild Semiconductor . , . Fairchild , , , AMI Mostek. .

, (reduced instruction-set computer, RISC), 1970- . - , , . , , . , — , , , — . ROM, , , . ROM 10%.

, , , SRI . $1 .

« », «-35» . 1971 . , , , . , , , . — . 90 TAN. , . , 10 99 , . , « ».

NR-35 was presented without much hype and sold for $ 395 through regular sales channels. But they began to talk about him, and orders quickly exceeded the supply. They said that some buyers were ready to throw $ 100, just to speed up the execution of their order. Other sales channels have also opened - HP products, which were usually sold by technical representatives, began to be distributed through department stores. It was incredibly strange to see the HP-35 lined up on Macy's store counter. The first batch of 100,000 pieces was to last for half a year; after a few months, the plan was doubled. But even after production began, Bill Hewlett was not sure that the project would be successful. Once at dinner I mentioned that we received a request for a price of 100,000 from General Electric. He said: "This is probably a mistake, why are they so much?" I replied: "Maybethey buy pieces for each of their engineers. " Bill replied, "They just need to buy a few pieces, and let their engineers borrow calculators from each other."

-35, , , . , -35 , . Dietzgen, , . , -35 ? — , $395, ? ? , . , « » — , .



All Articles