The HP 35 Handheld Calculator

In a real sense, the idea for the HP 35 handheld calculator came from a number of sources. Certainly the HP 9100A desktop computer, of 1968, showed the popularity and power of a desktop, technical (personal, not depending on a timeshared mainframe) computer in the hands of engineers. Transistorized desktop adding machines were beginning to take the place of the clacking mechanical "comptometers," built by business machine suppliers like Friden Corp., and others. Japan was building desktop calculators for business use, with Nixie tubes for display. Light-emitting-diode numeric displays had barely been invented.

Time for a handheld concept. At the HPA (components) Division, in early 1970, we had just concluded technical negotiations, for supplying keyboard and digital display lightemitting-diodes (LEDS), with a company called Unidynamics of Phoenix, AZ. And yet, after some months of research and production planning, the company, a division of Universal Match Corp., had decided not to proceed toward building a commercial pocket calculator. The main reason was that their core business was production of super-high-tech electronic sub-assemblies for atomic weapons. Their management had concluded that they should not move into the purely-commercial sector.

But, in retrospect, they had the hand calculator product concept, dead on. They had envisioned a handheld calculator, the size of a king-size pack of cigarettes. It had a 6-digit display, mounted horizontally, and four functions of add, subtract, multiply and divide (what we termed a four-banger). Further, they ingeniously had planned two or four more math functions, to be built into the ROM. These had keys that were to be covered up in the first models, but which could be uncovered, as needed, to confront competition, without having to re-design anything inside, merely put a new faceplate on the front.

The reason Unidynamics had come to HPA, was that we had a first-class ceramic facility for building photo-conductors. Photo-conductors were used for the electronic counters, built by the Santa Clara Division, to decode the BCD logic of counting circuits, into the 10-line code needed to drive NIXIE numeric display indicators. Unidynamics envisioned the need for a cheap keyboard, while at the same time, they needed long-life reliability. A photo-conductor keyboard, with no moving parts, seemed ideal.

Their idea was to use a photo-conductor resistive element at each key location. They had built prototypes, which demonstrated that when a finger was pressed into a shaped depression "key," the photo-resistance would change by about 1 million to 1. They had tested pulse-shaping processing circuitry, to eliminate false key inputs, caused, for example, by shadows of moving fingers across the panel, or variations in ambient lighting.

Unidynamics had done their homework. They had hired an experienced consumer-marketing man from 3M Corporation, who became their V.P. of Marketing. He had lined up international distribution for the product to the tune of 10,000 units per month. The price was to be $200, at the time that Japan's Sanyo Corp. was selling a calculator that was the size of a small cigar box for $400. Japan’s Canon was moving to a "pocket" calculator, but it had to be a coat pocket because it was the size of a small book. Their price was projected to be about $250.

In other ways, too, Unidynamics was ahead of their time. They envisioned 4 integrated circuits mounted on the rear of the ceramic “mother-board” which carried the key photoresistors and display on the front. Production would be so certain, that they planned no re-work, because of the expense, and therefore, would simply throw away bad units.

George Klock was project leader, and visited a number of times, along with Philip Tai, who was project engineer. Naturally, as soon as we in the LED group heard about this large project for keyboards, we immediately began our campaign to change their plan, for displays, to LEDs. We proposed either monolithic numeric chips (later used in the HP 35), or an abbreviated 3 x 5 dot matrix. That deal looked like a huge amount of business. They bought into the LED idea, and there were more visits with Tai, to work out the "LED strobing" technique, which we were just perfecting for use in HP 5300 portable counters.

Suddenly, and without warning, Unidynamics called us with the information that they were suspending their project. Amazing. They had the product concept in their grasp. They were about a year ahead of the market. They would have blown the market away. That year, 1970, Japan built about 150,000 total calculators, desktops, "portables" and "handhelds." The 10,000 per month sales estimate of Klock, for $200 units, was probably low by a factor of 5 or 10.

Once we had confirmed that the project was clearly cancelled, we felt we were released from our self-imposed HPA rules, about revealing any sales or technical details to other HP entities. We always held such information strictly confidential during our contract periods. In the components business, such technical and business details were sacrosanct. If another HP entity should discover any such details about a competitor who was our customer, we would lose all our credibility.

Then, and only then, it was HP’s turn. I then fed back to HP Labs, an overview of the proposed Unidynamics calculator. My memo below mentions that I had just read a speech transcript, by President John Young, that the present HP technology would now permit HP to put the ROM of the HP 9100A desktop calculator into four chips of integrated circuits. That was amazingly close to the Unidynamics four chips, although the HP chips were a full engineering algorithm set. Talk about serendipity.

*************************************************************************************************

January 27, 1970

From: John Minck

To:

• Marco Negrete
• Tom Kelley
• Milt Liebhaber
• John Nebozuk
• Volkmar Schaldach
• Paul Stoft

Subject: ELECTRONIC SLIDE RULE

Marco, I thought I would pass along a little information for what it's worth in considering a small, portable pocket slide rule or pocket calculator for HP.

We've had some recent negotiations with a U.S. company regarding both a digital display and a photoconductor keyboard application that looks like very large numbers. These people are talking about 10,000 units per month, minimum, with the possibilities for two or three times that in a few years.

While the particular company's name involved is not too significant, (in fact we probably shouldn't talk about them by name, since they have given us some information that was relatively private) I do think the overall market data and some of the particular components that we have are things that Hewlett Packard should consider. The particular product envisioned would be about the size of a pack of filter cigarettes, turned sideways, with an 8 or 10 digit light emitting diode cluster in the upper left hand corner. These would be about 0.1" in height and meant for viewing within arm's length. The keyboard would be

made up out of a photoconductor matrix for which HPA has already furnished samples. The computation will be done by a few, and I'm not sure how many, MOS integrated circuit chips done with LSI. I'm not sure who is doing their design but I do think that TI is involved. There are the normal numeral entry keys as well as a series of functional keys on the right hand side. We know nothing about the particular functions involved.

These people have already lined up an international distributor who has guaranteed to take 10,000 units per month and the proposed selling price is in the $300 range. Although these particular people are neophytes in the calculator business, they do plan to get knowledgeable people in their production and it appears that they will be going toward some state of the art interconnection techniques so that they can do assembly at very low cost. They would plan to have very few rejects at the end and would essentially do no testing, but instead would just scrap the completed package if it doesn't work. That's fairly revolutionary by itself. They might be able to make it work.

The main reason I'm writing however, is to inform you of several specific HPA components that might be well-suited to getting Hewlett Packard started on some kind of a project like this. The first of these is a multiple cluster, strobed array, built out of 7 segment numerals, which we plan to have ready in about August of 1970. The product ideal is to "monolithicly" produce the characters about .1" high and arrange them so that they can be strobed or scanned horizontally. We plan to produce these in various length, specialized clusters so they can be used for such things as pocket calculators, portable thermometers and a wide range of other numeric-only readouts. There are also some interesting possibilities of making characters even smaller than 0.1" and using plastic magnifiers of various types so we can get the power drain down even more.

Number two component is a photoconductor array that serves as a keyboard input. This is presently being made in John Nebozuk's area and has been specifically selected for the other company's calculator. They looked through a whole series of different keyboard entry ideas and rejected a lot of them because of contact bounce, reliability and just plain cost. It turns out that they have done a substantial amount of work in adapting the HPA photoconductor to use as a keyboard.

I think this is where Hewlett Packard needs to find some division with a keyboard need to pick up the present component. The other company puts a plastic mask over the top with holes cut in it about the size of thumbtacks that are appropriately shaped and spaced off of the photoconductor element. The finger that is tapped across the keyboard, much like an adding machine, will go through the pulse shaping circuitry in the right fashion to energize the integrated circuit drive. On the other hand, fingers just moved across the keyboard in front of a bright light or placed gently on the "key" do not have the right closing rate to cause false entries.

According to the engineers, one of the main reasons that photoconductors are quite good, is because they have a very wide range of resistance change, approximately one million. They have designed their key covers and shaping circuitry such that a substantially less change of light causes the triggering. That allows them to work from a very low level illuminated room out into direct sun light without having false triggers.

I'm particularly excited about this keyboard because it could represent an important reliable component for Hewlett Packard Company and all the variety of keyboard inputs that we are ultimately going to have to have. I would like to find some project that could develop some further human factors work on the key entry. It would also seem to be important to have some sort of audible mechanical feedback to tell the person that he has made an entry. In the case of a calculator, however, it appears that entering data onto the little light emitting display itself becomes the assurance element.

When John Young recently mentioned that you had been able to put the entire 9100 onto four MOS chips, it really got me to thinking that an exciting product could be put together from that technology and a couple of display and key entry technologies. A pocket calculator that can do slide rule functions and some additional arithmetic functions and ought to be worth $500 to many technical people.

I hope that sometime when you're in the area you might be able to stop by and perhaps we could discuss it in more detail. In any event, I'd appreciate knowing your thoughts on this general market area, and specifically whether light emitting displays and other keyboard elements might have some place in your product plans.

JLM:lb

CC:

• Barney Oliver
• Dave Weindorf
• John Young

*************************************************************************************************

Displays ideal for the HP-35. By that time, our HPA monolithic display chip technology could present a numeric digit, 1/10 inch high, and we could build them reliably. The HP 9100A project team had been pretty-well held together after transfer of the project to production to Loveland. Engineering for the HP 35 got rolling strongly in 1970, encouraged by both Bill Hewlett and Barney Oliver, who became the cheerleaders.

Think about the possibilities. With the ability to create dozens of calculation functions, the team leaned towards Electrical Engineering functions. 35 keys maximum resulted from a product design mandate, that the calculator must slide into Bill Hewlett's shirt pocket. But the power of the machine called for up to 15 digits of numeric display. Barney was leaning toward liquid crystal displays (LCDs), but we, in HPA’s LED department were pressing for LEDs, naturally. LCDs were dramatically cheaper, and only for the fact that LCD technology couldn't put 15 digits side by side, did we get the nod on the LED display decision.

In the meantime, we, at HPA, had perfected the use of a simple spherical plastic lens to magnify the 0.10-inch high character. It caused a minor reading drawback, which we called the egg-crate effect. This meant that the user had to place their eye more or less directly in front of the numbers. Otherwise the display would look like the numbers were down inside the partitions of an egg carton. But the fact that a 1/10th inch monolithic number only used about half the drive power as a 1/8th inch high unit, the designers preferred more battery life to larger numbers. It was still a hit, and never, ever, mentioned as a drawback.

In retrospect, the LED decision was very fortunate, for although the LCD technology had been around for 50 years, they were notoriously unreliable during those years. A LCD display set could develop an "infection," on one corner, and it would creep across the whole window, and you'd be out of the display business. Infection is the right word, since LCDs are organic materials. Now, 20-30 years later, things are different, and LCDs are everywhere.

Tom Whitney, HP 35 project leader, decided they could pay $1.05 for each monolithic digit. Our HPA production costs, at the time, showed that we should sell them internally for about $5.00 each. So we figured that with Bill Hewlett behind this project, so strongly, now was the time for a gigantic leap ahead in HPA's LED manufacturing automation.

Our team then developed a Gallium-Arsenide-Phosphide process lab expansion plan, which promised to be the biggest thing we had ever done. Until that time, we would slave over a $30K process machinery capital investment proposal. The proposal might run 20 pages, and still not be accepted. These equipments weren't cheap. Epitaxial reactors for growing layers of semiconductor atoms, diffusion furnaces, saws, polishers, dicers, thin-film deposition, photo-resist stations, automated probing stations, ran to tens of thousands of dollars. Our plan for ramping up calculator digits looked like about $550K.

Sales projections for the HP 35 were interesting. With the outside market, in 1972, expecting to feature four-function machines at around $200 to 250, no one really was confident about how many people would plunk down $495 (the first price proposed for the HP 35) for a pocket slide-rule. Even at $395, the final introduction price, I recall the sales estimate as 500 per month and going to 1000 per month, on the outside.

At the LED department, fortunately, Marketing Manager Milt Liebhaber and Rick Kniss, the LED Product Marketing Engineer, had big plans for outside sales of LED digits, that would leverage on the production of the HP 35. We were sitting around one evening with the thick equipment proposal in front of a group of us, presided over by Division Manager Dave Weindorf. After much discussion, Dave suddenly said, "What the hell, let's make it a proposal for a cool threequarters million".

So we corrected the equipment lists, to get up to $750K, and went up to see Bill Hewlett the next day. We had our arguments well honed and practiced, for proving the need for supporting the all-important HP 35, with incidental outside sales thrown in. After 5 minutes of preliminaries, presenting our executive summary, Bill stopped us, saying, "I’ve got another meeting, so is there anything else important, that I should know? If not, let’s go with it." We almost wanted to say, “But, Bill, we have all this backup argument material, don’t you want to hear it?” It was by far the easiest project I have ever seen sold. But it shows the magic of that calculator project in those days.

In retrospect, it was extremely lucky we sandbagged Bill on the materials processing lab for GaAsP. The 500-1000 per month forecast for the HP 35 was wildly wrong, on the low side. Sales went through 10,000 per month, climbing almost straight up. The reasons, of course, are clear with hindsight. The price was a stroke of genius, because the $395 was easily within the reach of every engineer, not just managers. It was not only a prestigious personal possession, but an amazing drudgery beater. It made better engineers, and it made them faster and more efficient.

Pick a keyboard layout. Meantime, Tom Whitney was having trouble freezing the keyboard design. With literally hundreds of different key functions one could choose, Barney and Bill were continually coming up with better ideas for key functions and layout. Tom had finally arrived at a crucial date, where the IC ROM masks had to be frozen, to be sent to the IC vendors, and be ready for production dates. Tom sent out a memo to all the algorithm people, Dave Cochran, Bill and Barney. The notice stated that he had reserved the meeting room for as long as it took, that day, to freeze the keyboard functions to everyone's satisfaction. At the end of the meeting, all would sign the final keyboard paper, and he would walk out to call the IC manufacturer with the final specs.

The meeting did take a long time, hours. Back and forth, since this was to be the way the calculator would not only look, but perform. To all those engineers, with those options, it must have been excruciating. At the end of a long period, they agreed, and all signed the penciled-in key nomenclature. Tom thankfully walked back to his office, ready to prepare the material to call the IC supplier the next morning, and the phone was ringing. It was Barney. "I've got another idea." "Too late," said Tom, and he called Bill to see if he could make it stick. Bill said to go with the agreed paper.

The HP 35 introduction, in 1972, couldn't have come at a better economic time for HP. The US business economy was in the doldrums, and this high-profit product energized our balance sheet. The thing we had failed to comprehend, was that engineers would find this personal calculation powerhouse indispensible. They would not rely on their companies buying them, instead, parting with their own money to posses one.

Personal power. Tom Osborne, the project consultant, told me of an occasion of his visiting the Smithsonian Institution in Washington, DC. He went to see the display of the original ENIAC computer, used in the WWII project for artillery-table computations. At the time, he was carrying the HP 35. He told me he was suddenly struck with the realization that, inside his coat pocket, in the HP 35, was more computing power than this entire roomful of equipment racks. The racks held 18,000 vacuum tubes, that burned out at the rate of about 5 per hour. Of course, it was an unfair technology comparison that spanned 2.5 dramatic technology decades. Tom was humbled to realize that his little baby computer in his pocket was more powerful, and certainly more reliable, than that monster system.

Field engineers found clever ways to sell HP 35s. Jim Bunn of the Las Cruces, NM sales office called on the U.S. Army’s White Sands Proving Ground. He discovered that, while military operations sometimes ran out of expense or capital budgets, there was often a lot of training budget left, near the end of their fiscal year. Jim cleverly arranged to put on a training seminar for “Reverse-Polish-Notation Scientific Computers.” It was an honest description for the HP 35. The special HP training course cost was set at $500, and included an HP 35, as a class "training tool," which the student was allowed to keep.

Dave Cochran remembers, “Soon after that (the success of the HP 9100A), Hewlett began pestering me about putting the 9100 in his shirt pocket. Even several years later the right architecture for the state of IC miniaturization just wasn't there to accommodate what we wanted to do. I was even thinking of getting hold of Hewlett's tailor to get his pockets made bigger.”

“I stumbled across an architecture during a visit to Fairchild that made it all gel. It contained a ‘racetrack’ register memory that greatly reduced the access and interconnection requirements and was excellently suited for the ‘CORDIC’ algorithm to calculate the transcendental functions.”

“Fairchild was trying to make a twenty-digit four-function calculator but hadn't applied for a patent. In the next few weeks I roughed out architecture for arithmetic and logic chip, control and timing chip, register chip and ROM structure. I also devised an instruction set and flow-charted a few functions including subroutine calls.”

“Excitedly, I went running off to Hewlett, ‘Look, I can do it now, if you give me the million dollars for development.’ Hewlett, who sometimes was unsure of himself without Packard who was at the Pentagon, said, ‘Wait a minute, I want SRI to do a market analysis on this.’ Four months later SRI said that if it was four-function it should be under $100 but they could not supply a recommendation for a pocket scientific calculator because there had never been one.”

“While waiting for Hewlett's decision we still went ahead on all fronts, I even devised a sampling scheme for the LED display that Minck mentioned that had the added benefit that it used only 10% of the calculated power due to the ‘super linearity’ of LEDs of the day. If you strobed each segment at a 1% duty cycle the current only went up by a factor of 10 for the same light output. I remember thinking that it may affect the reliability so I ran a sample at 0.1% duty cycle and compared the light output after six months, luckily it was the same.”

“I remember when Fred Terman got his sample HP 35, we made about six prototypes, he just couldn't believe it, kept looking for the cable. Louie Alvarez called it the ‘eighth wonder of the world.’ Having an HP 35 in the early days meant you could attract attention without having to know how to play the piano, but mostly you got all the men hanging around you at cocktail parties.”

The battle of the algorithm gurus. The inherent power of these hand calculators was contained in their algorithms. These abstract mathematical flowcharts are really the processes the machine goes through, to solve the transcendental and other complex equations. Some people's brains are born with the power to envision these algorithms, in their total flow pattern, and somehow see through their complexity. One of these men was Dave Cochran. He created the heart of the HP 35, and probably has not gotten near enough credit for it in the outside world.

Dave told me that as other competitors came into the scientific calculator market, notably Bowmar and Texas Instruments, that he came to meet the algorithm inventors of those companies. They tended to be the same kind of geniustype people, clever and crafty. In one sense, it was a reputation thing, with each conceding points, if a particularly clever algorithm was being tested and qualified.

But in other ways, these geniuses were playing other games. It became a kind of chess game. They might tell about a new algorithm idea, but might salt-in traps and false leads, to put the other company on the wrong track for a time. Naturally, there were advantages to be gained in talking together since, often, it established industry standards and processes, and it was such an arcane art. Yet, all information had to be taken with huge caution. Did we gain more than we lost? Knowing Cochran, I suspect we gained a lot more.

The HP 35 was probably HP's number one product winner of its 6 business decades. It captivated our technical customers. It raised HP's already high prestige. It brought untold computing power to the fingertips of the common engineer, and they worshiped HP for the innovation. Finally, it sparked a long succession of more and more powerful machines, the magnetic-strip programmable HP 65, financial versions, and many more.

I personally believe that the HP 35 changed our HP history. It most certainly launched HP into the "consumer" market, even though this particular product had a professional (engineering/consumer) focus. HP later introduced other powerful consumer machines, some years before competitive $10 units began showing up in drugstores. The line of successful pocket calculators also reinforced HP's dominance in desktop computer technologies, such as the follow-on HP 9810 and 9830, and later the highly-programmable HP 9826 and 9836.

In looking back on those heady days, I can recall one evening in the small office area we all shared, talking about how the power of personal calculators would change our world. We were then immersed in a manufacturing technology where semiconductor packaging still required careful handling. IC chips were contained in separate plastic in-line packages and those dozen components had to be inserted into printed circuit boards for all the hundreds of interconnecting wires.

Our conversation got around to real visionary stuff, like attaching the IC chips directly to the PC board and bonding to the board. Then you could just squirt out a drop of plastic silicone to protect the whole thing from humidity, and sell the whole calculator in a drug store for $10. Then we all laughed at our folly! However, knowing what I knew about the rapid decline in anything made of semiconductors, I did have the feeling that it just might happen someday.

Finally, there was a fascinating postscript to the Unidynamics story. In the 1980’s, long after the 1970 negotiations for LEDs. I was back at Bldg 5, working on MW products at SPD Marcom. In those days, all customer inquiries sent to corporate headquarters in Bldg 3, were parceled out and sent to the appropriate division marketing people to answer as they had time. It was probably the late 80's, when a letter arrived from Australia, a teacher in a tech school had written to HP for some technical or product posters for his classrooms. The teacher's name was George Klock.

I gathered together a bunch of HP product posters and banners plus a frequency allocation chart and some other generic posters and sent them along. I also sent a note telling that years before I had known a George Klock from Unidynamics, and wondered if that were he? It was. After the calculator project was shut down, he migrated to Australia to start life anew, and ended up teaching, which he loved. Now, imagine the tiny statistical chance that I would pick up that particular letter out of a pile that just happened to be sent to our division. Another stunning coincidence of my life.

John Minck