Last Update: January 22, 2003 -- THE HP REFERENCE
An inexpensive, pocket-sized machine revolutionizes calculator design, and we did it right here.
"They got the jump on us with the transistor radio and kept the ball rolling with tape recorders and TV sets. Then they started out on a simple pocket calculator in the same way, using American manufacturing know-how and circuits. Hell, we were exporting not only the necessary components, but also the technology itself. This company finally decided that it had to beat the Japanese at their own game."
The "game" is the business/scientific calculator market, and the speaker is Dave Cochran. Cochran is a project leader for Hewlett-Packard's HP-35 a pocket-sized calculator recently announced and is chief algorithm-brewer for HP's calculator line.
Why is it that the Japanese -- those masters of the miniature haven't yet achieved what this American electronic instrument firm decided to do, and did? Tom Osborne, technical advisor to HP's Data Products group, answers this way: "Look, suppose I just wanted to make a simple, low-cost adding machine. Just to add two numbers up to, say, a million. I'd take two counters and store one number in each. To sum the two, just count one counter down to zero, and let the other increment up to the sum of the two."
"It's a very simple algorithm, and doesn't take much hardware. If I counted at a megahertz rate, it'd take a second to get the sum. With a small refinement you can get a very acceptable speed for addition, subtraction, multiplication, and division. But if you want to take on more complex things, you've got to let go of this type of simple technology, and use more finesse."
Cochran sheds some light on the finesse. "We designed this thing (the HP-35) to do transcendental functions -- logs, exponentials, and trig. We decided on the minimum we would need not how we could add-on to a one-chip design or a four-function machine to make it do these other things. And this is an approach that the Japanese have certainly not taken yet."
Tom Whitney, HP-35 project manager, agrees: "The Japanese were committed to an early MOS design, and wanted to use those same circuits in newer machines. Well, those chips just don't have the horsepower needed. And to start all over again, from scratch, would be, in expensive proposition for them."
"We got it all together, right here in one company," says Cochran, "while the Japanese have been buying LEDs outside, MOS outside, and so forth. For instance, with MOS, they've had to sail across the water; the Sharp calculator needed several years of Pacific crossing to really get going. The Japanese just don't have the necessary skills in-house as yet, although," he cautions, "they're coming on-line now."
The boss wanted it
These three men Tom Osborne, Dave Cochran, and Tom Whitney were key figures in the development of the HP-35, a development that took less than half of the usual two-and-a-half years product development time. Each understates his own role, but there's no doubt that their talents and personal commitments to the project paved the way for its successful conclusion in record time.
And there's a fourth man involved. The boss, William R. Hewlett, president of Hewlett-Packard, also understates his role in the development of the HP-35. But without his foresight, personal involvement, and willingness to fully commit the corporation to the goal, the U.S. still wouldn't have achieved this feat.
According to Osborne. "About 15 minutes-or-so after we finished the 9100 (HP's first programmable calculator, introduced in l968 at $4900), Bill Hewlett said we should have one in a tenth the volume, ten times as fast, and at a tenth the price, Later he added that he wanted a shirt-pocket machine."
"I think," says Hewlett, "it was pretty evident that there was a gap in the small-calculator market. The Japanese had concentrated on the very large, consumer-oriented market--for four-function machines--a market Hewlett-Packard doesn't really touch."
"But on the other hand, we have a very great background in dealing with professional people.. We understand what their problems are. And this encouraged us to move into the area of a very small, business/scientific-based calculator that fitted into both our technology and our concepts of logic structure. These (concepts) were not really being developed by the Japanese, who are concentrating on a much simpler problem."
"But we were already in the sophisticated-calculator business, and it wasn't much of a problem for us to winnow these down to where you could put them in a hand calculator. On the other hand, if you're only doing four-function machines, it is much more difficult to escalate your thinking and your technology into this area."
When Hewlett finally said "I want one,' the ball really started to roll. He forced a definite company commitment: this schedule was not to be slackened; it was to get the support it needed from any of the operating divisions; and it was to get the highest priority needed."
So the check was written. People were relieved from other duties to help on the project. Many things had to be investigated in parallel; cluster LEDs with magnifying lenses; flexible PC boards; MOS logic; etc. etc. Even production and tooling were cranked in from the very start. Normal cycles had to be short-circuited. Responsibility and decision-making were at the level of whomever was doing the job: there was no central legislating negotiator it was a team effort from the very start.
Package design to fit a shirt pocket
One of the important inputs in the development of the HP-35 was the industrial design. Informally, you take the breadboard and see what kind of box it will fit into. For the HP-35, the reverse was true it was designed from the outside in.
Tom Osborne drew up a set of rough sketches for a package the size of' a Benson and Hedges' cigarette pack showing a first idea of the calculator keyboard, display, etc. He left this packet of drawings with the industrial design group and went away for a month. When he returned, he found a plastic model based on his sketches. He showed it to Hewlett, who simply put it into his shirt pocket. It fit, and the HP-35 was underway.
"That was one of the objectives on this thing," Hewlett comments. "You ought to be able to put it in your pocket and attend a meeting without having to haul out some damned thing that isn't convenient. The form factor is very important in a portable machine, How do you design something to be functional? Our industrial design group did it, and came up with a helluva package, in my view. It's really a very practical little device."
A keyboard like a thumb piano
That "helluva" package posed one helluva set of problems. Take a look at the keyboard, for example. Very early in the game. the question came up: should this calculator have the same key spacing as all the other calculators in the world -- 0.75-in. center-to-center? Since this was to be a palm-sized machine with many keys, the answer just had to be "no!" Besides, on most calculators, the keys are large, and fill the key-to-key spacing. Some companies had gone to small keys, but provided a stylus for keyboard operation.
Two factors led the designers to believe that they could use small keys and spacing, and not need a stylus. One factor was the accordion, an instrument with a large mass of small, densely-packed keys. The other factor was, interestingly, a calculator. About 30 years ago, Monroe made a calculator with small keys. It was a small machine, but a very large success.
Further, people around HP, who were shown the prototypes were asked for their impressions of the small keys. The answer: small keys are okay for small machines. And as it turned out, the small keys and spacing are right on; some users, in fact, operate the machine by holding it in both hands and using their thumbs, playing it as though it were an African thumb-piano.
The keyboard functions much like those found on a log-log decitrig slide rule- are the result both of experience with other calculators and of the successful resolution of personal preferences among the designers (log10 x instead of x^2, etc.). The final layout of the keyboard is a model of functionality, with similar functions grouped together: log and exponentials, trig functions, power of x, and data manipulation. On the basis that most users are right-handed, the +, --, X, \ keys are at the left of the arithmetic keys, and not hidden when you enter a number.
Designing the keyboard switch mechanism itself was far from a trivial problem. The keyboard had to interface directly with MOS circuitry (for economy), it had to have a low silhouette, and had to give tactile feedback to show that the key was properly entered. All in a small size, at low cost, and with high reliability.
The solution? A buckled piece of metal that operates in much the same way as bending a metal tape-measure the wrong way -- the oil-can principle. The result'? A 3-x-5-in. keyboard in which each of the 35 keys has a touch similar to that of a high-quality electric typewriter.
Five chips do the algebra
November, 1970, saw the start of the HP-35's logic design. Hewlett-Packard knew how to do the logic algebraically, but not how to draw logic diagrams that MOS vendors like to see, nor how to physically implement the dynamic MOS/LSI circuits. The HP people started catching up. They talked to everybody they could--including MOS manufacturers. They studied texts on MOS LSI design.
A desired battery life of three to five hours, plus a knowledge of what it takes to do the transcendental functions, set the logic power requirements. Further, the clock rate needed was about 200 kHz--an order of magnitude greater than that found in four-function machines. Dave Cochran "kind-of-knew" -- through experience with similar, but slower, algorithms in the 9100 -- what clock and bit rates he'd need to keep the transcendental operations to less than a second. (Typical trigonometric-operating time turned out to be 500 ms.)
In February 1971, HP went to seven vendors who were in production with the latest technology. The specs called complex MOS to be built with the vendor's most advanced technology. Complementary-MOS was ruled out. It was just becoming available, and though it was a very-low-power technology, it needed too much chip area.
There were significant differences in the vendors' replies. The partitioning was to be for a three-chip set. There are five chips in the machine, but only three different designs: three ROMs; a control and timing chip; and a register and arithmetic chip. Some vendors said it couldn't be done, that the design had to be partitioned into more chips, and quoted on that basis. Some even felt that the chips would prove too large to build -- 300-mils square.
Finally, the job went to AMI and Mostek, although HP notes that two other vendors were close and probably could have done the job. Two vendors were insurance for HP that chips will always be available. Though the innards are somewhat different, the chip sets are completely compatible, and pin-for-pin interchangeable. Some HP-35 users, in fact, will find AMI chips inside, and others will find Mostek chips. Both manufacturers are using ion-implant, depletion-load technology. These circuits, by the way, are believed to be the largest presently in volume production in the world. Each is equivalent to 6000 transistors -- for a total of 30,000 on five chips.
A magnified display
The display had to be low power and low cost. Liquid crystals were ruled out because of an uncertain technology, low reliability, high voltage, and low contrast. And so it was decided to use a small-digit LED) display, magnified for ease of viewing. Magnification reduces the area needed for a digit, thus significantly reducing material cost and power.
Although outside vendors were considered, a corporate decision was handed down that said HPA shall make the needed display, in time, and at the right price (or else!). Needless to say, HPA did come up with the display; it is now available separately as a standard product. The actual digit height is 80 mils; but the visual height, with the magnifier, is effectively 112 mils.
Originally, the designers worked with a manufacturer of fresnel lenses -- plastic, stampable, and flat planning to attach a lens over each digit. HPA then proposed molding a plastic lens into the package, thus saving on material and parts handling. The resulting display is beautifully clear, easy to read, and without a trace of the magnifier. It's as though you were looking at the display itself, without a magnifying lens between your eye and the digit.
A further advantage of the plastic lens is that it reduces the mismatch of the indices of refraction of the GaAsP and air, and thus reduces light loss.
Another trick used to increase brightness and efficiency is the strobing and multiplexing of the display. The duty cycle is low, and refresh occurs every 300 us. Bipolar display-drivers are used to handle the high peak currents.
Even the display package is new for LEDs. It uses a straightforward, old-fashioned lead frame. The display is a hybrid on a metal leadframe, encapsulated in plastic. There's no way to test it until the package is completed.
You can replace these batteries
In the HP-35, even the battery pack is unusual. In most rechargeable-battery calculators, the batteries are not replaceable by the user. The HP-35 designers felt all along that the batteries might prove to be one of the less reliable parts of the system. So the NiCad batteries from GE and Union Carbide -- are replaceable, just in case.
Rough sketch to masterpiece
And so an American company finally got it all together and outperformed the Japanese -- and everybody else -- in a market that most of us thought was lost to this country. Do you see how it was done how a team worked together?
o Bill Hewlett -- he wanted an inexpensive shirt-pocket calculator, and continually made sure that's what he was going to get;
o Tom Osborne -- he knew what he wanted the calculator to look like, what it had to do, and what the users wanted;
o Dave Cochran -- he made the machine do what the keyboard said it could do;
o Tom Whitney -- he got it built by coordinating all the forces at work within the company, and channeling them into an end product.
This is a simplistic view, of course. Many things had to come together at the same time. A machine of this complexity and in such a small size needs a very competent logic-design team familiar with state-of-the-art techniques and calculator system design. It needs a capability in algorithms for transcendental functions. It needs a corporate commitment and allocation of all resources. And the timing must be right. This is critical; the MOS technology had to be at a certain level and, even at that, the vendors had to be pushed to their limit and beyond, Hewlett-Packard had these things come together for them.
Bill Hewlett put it this way: "We had inherently a critical mass for this type of operation."
Tom Osborne put it this way: "I brought some paint and brushes to some people, sketched out something, and then got to look at pictures much more beautiful than I ever imagined."
Dave Cochran, Model 35 project leader
Tom Whitney, project manager for the Model 35
Arithmetic: Add, subtract (60 ms); multiply, divide (100 ms): square root (100 ms).Display --
Trigonometric: sin x, cos x, tan x, arc sin x, arc cos x, arc tan x (500 ms).
Logarithmic: log10 x, ln x, e x (200 ms).
Other: x y (400 ms); 1/x: PI; plus data storage and positioning keys.
(There is an operational stack of four registers, and a memory register for constants. The stack holds intermediate results and, at the appropriate time, automatically recalls them for further processing. This eliminates scratch-pad notes and reentry of intermediate answers. Any register can be shifted to the display for review.)
Ten significant digits with the decimal point automatically positioned. Numbers larger than 10 ^ -2 and smaller than 10 ^ 10 are automatically displayed in floating point. Outside this range, numbers are expressed in scientific notation, with two-digit powers of ten shown at the extreme right. Trailing zeros are automatically blanked. A flashing display indicates an improper operation.Power source --
Rechargeable NiCad batteries, 500 mW; or 115/230 Vac, +/- 10%, 50-60 Hz, 5 W.Operating temperature range -- 0 to 40C. Size --
3.2 x 5.8 x 1.3 (max.) inches.Weight --
9 oz. (calculator with batteries).Price --
$395, including ac adapter, battery recharger, soft leather case, travel case, operating manual, and owner name-tags.
For more information, contact Hewlett-Packard Advanced Products, 10900 Wolfe Road, Cupertino, Calif. 95014 (408) 257-7000.
HP-35, Circle Reader Service 210
LED display only, Circle Reader Service 211