The Museum of HP Calculators
The classic series used a multichip CPU. The Control and Timing (C&T) chip performed all the major nonarithmetic functions. It polled the keyboard, tracked system status and generated instruction addresses sent to the ROM. The C&T chip provided a word select feature which allowed the A&R chip (below) to operate on partial words (for example the exponent field) separately.
The Arithmetic and Register (A&R) chip contained 7 56-bit (14 BCD digits) registers. Three of these registers were the X, Y, and Z registers of the four-level stack. Two more were working registers and one was the user-visible memory register (STO/RCL.) The final register did double duty as the topmost register in the stack (T) and as a working register for trigonometric functions. Thus, when a trig function was used, the topmost value in the stack was lost. The A&R chip also contained instruction decoding circuits, a decimal adder-subtractor and a display decoder.
The calculator used digit-serial, bit-serial organization to reduce size, cost and increase reliability. This serial design also made the word select feature easier to implement. The A&R chip responded to a word select signal from the C&T chip as the appropriate part of the number passed serially through it. HP designed the logic and most of the chips were manufactured by outside vendors such as AMI and Mostek.
Additional information on the register layout and programming model may be found on the CPU and programming page.
The classic series machines used ROM chips which contained 256 instructions of 10 bits each. The number of ROMs used varied by machine (for example, three in the HP-35 and seven in the HP-80.) ROM chips respond to addresses and word select signals from the C&T chip.
Special LED clusters were developed by HP because existing LEDs consumed too much power and were too expensive. The LEDs were grouped into 5 decimal digits into a single 14 pin DIP with magnifying lenses over each digit.
The classic series machines were built with two circuit cards supported by a plastic "backbone". The larger card contained the key contacts and the LEDs and LED drivers and the smaller card contained the (A&R), (C&T) ROMs, and other components.
HP took pride in using the latest industrial engineering research and their own ideas and studies. The spacing of the keys on the HP-35 was based on a simple test - they needed to work well with Dave Packard's large fingers. Dick Osgood also analyzed how people pressed calculator keys. Based on the way that most people poked at the keys he designed a hinged key and avoided the "oil can feel" of TI and other calculators. It took many design iterations to get the final design which used bent beryllium copper strips. The key bottoms were designed to be easy on the copper while still providing the right feel.
The keys were made with a state-of-the-art double shot injection molding technique. This was very difficult but ensured that key legends would never wear off. Since the keypads were subject to very little wear, they were printed until the HP-65. On the HP-65, the top legends had a magnetic card sliding over them and the metal oxide was abrasive, so these were also molded using a secret process. A difference between the key molding and the keypad molding is that there could be no enclosed spaces in the letters, so the R in the R↓ legend was not completely closed and other letters with closed spaces were avoided.
The original design for the HP-35 was a cigarette pack with a keyboard drawn on it. The final size requirement was that it fit Bill Hewlett's pocket (if only barely.) The requirement for the HP-65 was that it had to be the same size as the HP-35. The HP-65 required even more electronics, a motor, gears, a card reader channel, and a place for the card to label the keys. The team thought they were going to be fired because they just couldn't do it. They went to Bill Hewlett with their data - and their resumes updated - and asked Bill for an extra 1/8". To their surprise he just said "OK."
The power of Bill's comments also affected the color. When the HP-45 was shown to Bill, he said: "Black is a very unimaginative color. Are they all going to be black?" This casual comment was responsible for the greens, browns and beiges that followed.
Bob Taggart was in charge of the HP-65 card reader design. he originally planned to make the unit out of aluminum, believing that this was essential to achieve the stability needed for the data density of the cards. Bill Boller wanted to keep the calculator light so he convinced Bob to try a new material using polycarbonate, fiberglass and Teflon. To Bob's amazement, the lighter material worked.
In order to meet the design goals, the card reader needed to read/write at 300 BPI. It needed to do this consistently in a temperature range of 0-40°C and the cards needed to be interchangeable between calculators. If the card had been allowed to be just one inch longer, the job would have been far easier, but the decision had been made by Barney Oliver to use the card to also label the top row of keys. Adding a prerecorded clock track would have also helped but it would have made the cards more expensive. Originally the team didn't know how the cards would be propelled through the card reader. They considered a spring-driven music box type system, a constant force spring with an air cylinder, a gravity-fed system etc. A motor was originally resisted in part because it didn't seem particularly inventive (it had already been done in the HP-9100), as well as concerns about power consumption and the quality tiny motors.
Later, it appeared that the motor was the best option. The team started with simple motors of the type used in toys. These motors produced so much noise that they developed $6 of circuitry to fix the noise produced by a $0.12 motor. Eventually they found a precision Swiss motor that pulled the cards without producing the noise.
The next challenge was developing a tiny worm reduction gear. HP had some experience with this, but the screw companies didn't have the proper equipment to make a worm gear of the needed dimensions so HP purchased the machinery for them. When the first gears came in, they were examined under magnification and they looked gorgeous. They had a perfectly cut involute curve which was considered essential. However - they didn't work. Due to tiny mechanical variations, they produced so much noise that, at most, they might allow 10 programs steps to be recorded on a card. This was a major disaster - one of many on the HP-65 project. The engineers on this project continued to update their resumes thinking that the calculator would never work.
With less than 24 hours before a go/no go meeting with Bill Hewlett, the card reader still didn't work. At 8PM, backup worm gears arrived. Because the HP-65 was pushing the technology in so many ways, there were many backup technologies developed. The backup gears were made by a simple thread-rolling technique and didn't have the desirable involute curve. They were considered inferior to the gears that had already failed and no one expected them to work at all. Much to everyone's surprise, the night before the go/no-go meeting, the backup gears worked! At the time the reason wasn't understood and Bill Boller said: "Let TI try to reverse engineer that!" (Later, HP and TI cross-licensed many patents giving TI access to HP's card reader technology.)
Another challenge was designing a tiny battery powered device that could reliably pull a card even after it was handled and covered in oil from the user's skin. Bob Taggart developed a test that Bill thought was a little too tough: Bob grasped a card tightly in his hand, then dunked the card a can of STP motor treatment - the slipperiest stuff he could find - and then he fed the dripping card into the card reader while still holding on to it as tightly as he could! After testing many materials and techniques, only one stood up to this stress test - a machine-cut polyurethane wheel. Bill tried to convince Bob to ease the test a bit to allow something a little less expensive, like molded polyurethane or vinyl, but Bob insisted on card reader that would work under the worst conditions.
When TI came out with their first programmable scientific calculators, they were embarrassed that their engineer's desks were covered with HPs. TI announced that they would be replacing all the HPs with their own calculators ("Eat your own dog food.") Almost all of the HP calculators vanished off the desktops before TI could replace them. Most TI engineers continued to use their HP calculators but did so discreetly.
Producing accurate answers for transcendental functions on a very small/slow computer was no simple task. HP considered power series, Chebyshev polynomials, polynomial expansions, and continued fractions but all were too slow due to the number of divisions and multiplications needed. The algorithm chosen was an iterative pseudo-division/pseudo-multiplication method first described in 'Arithemtica Logarithma' in 1624 by Henry Briggs. Because of the complexity of the programs, a subroutine capability and a set of status flags was built into the chip set.
Picture of an HP-35 CPU board. (~60K)
Picture of an HP-45 CPU board. (~55K)
Picture of keyboard contact/LED board, front case, key bracket, membrane and some keys. (~42K)
Picture of an HP-55 CPU board. (~60K)
Picture of an HP-55 CPU board on "backbone". (~42K)
Picture of an HP-65 interior. (~52K)
Picture of an HP-65 CPU board. (~49K)
Picture of an HP-65 card reader from the back. (~43K)
Picture of the HP-65 card reader switches (start/stop, write protect etc). (~16K)
Picture of an HP-65 top case from inside showing how the on/off, and top-row legends are molded through. (~35K)
Picture of an HP-70 CPU board. (~37K)
Picture of an HP-80 CPU board. (~47K)