Some lab projects are endothermic: the desired reaction proceeds only with the application of considerable heat and pressure and stops the moment these are relaxed. Others are exothermic: when the proper ingredients are brought together the reaction starts automatically and it is only necessary to harness and control the power that is generated. The 9100A project was one of the most exothermic I have known.
The ingredients started coming together in the late summer of 1965 when we were shown a prototype of a calculator invented by Malcolm McMillan that had one interesting feature: it could calculate all the common transcendental functions. The machine operated in fixed point and took a few seconds to calculate a function, but it did demonstrate the feasibility of providing these functions in a small calculator, and the power of the algorithm used to compute them.
The second ingredient was Tom Osborne who came to see us carrying a little green balsa wood calculator, which he had built on his own to demonstrate the virtue of some of his design concepts. What impressed us was its millisecond speed and its ten digit floating point operation and display.
A third ingredient was the imagination of Paul Stoft and other engineers in his group. It took no genius to see the appeal of a calculator that combined the speed and dynamic range of Tom's machine with the transcendental computing ability of the other machine. But to combine them into a small machine faster than either prototype, to adapt the transcendental algorithm to floating point, to add programmability and magnetic card storage and entry of programs, to provide the flexible display with automatic roundup, and to design the whole assembly for automated production required not only imagination but engineering skill of the highest order.
Another ingredient, the read only memory, which stores all the calculating and display routines, was already under development by Arndt Bergh and Chuck Near before we began the 9100A project. By carrying printed circuit techniques beyond the existing state of the art, Chuck was able to compress the required 32,000 bits into an amazingly small space. While the 9100A uses only discrete diodes and transistors, it it fair to describe the read only memory as one large integrated circuit with extremely long life expectancy.
As soon as the development began, everyone, it seemed, had ideas for new features. Hardly a day went by without someone proposing a new keyboard with a new key arrangement or new functions. It was Dick Monnier's responsibility as section leader to steer the project through these conflicting currents of ideas. Although we went down a couple of blind alleys, for the most part the course was held true. That we arrived at such an elegant solution in a short time is a tribute to Dick's navigation.
Tom Osborne joined the development team as a consultant on the general architecture of the machine. His contributions include the basic logical design, and the details of the control logic, flip-flops, mother board gates, and the memory drive and sense circuitry. He also contributed a large measure of sound judgment that resulted in an economy of design, low power consumption, high performance, and the ability of the 9100A to interface with peripherals and systems.
Tony Lukes developed the display routines, which include the features of automatic roundup in fixed point display, choice of decimal places, suppression of insignificant zeros, and the display of program step addresses and key codes, as well as numerical data. Tony also developed the program storage, editing, and execution routines which, together with the display, make the 9100A easy to use and program.
I should at this point emphasize that the 9100A, while small, is, in a sense, much more complicated than many general purpose computers. Most general purpose computers have relatively large memories but can execute directly only certain elementary machine instructions. To compute complicated functions and indeed, in some cases, even to perform simple arithmetic, the computer must be externally programmed.By contrast the 9100A has a very sophisticated external instruction set: the entire keyboard. The 9100A is a small computer with a large amount of 'software' built in as hardware.

The task of compressing the floating point arithmetic operations and functional computations into the limited read only memory of the 9100A was accomplished by Dave Cochran. To make sure that all the calculations were accurate over the enormous range of arguments allowed by the floating point operation, to assure exact values at certain cardinal points, and above all to get so much in so little memory space was an enormous achievement.
I sometimes wonder if Dave realizes what a remarkable job he did. It took several passes. On the first pass it appeared hopeless to include all the functions. But by nesting routines and by inventing a number of space saving tricks he was able to save enough states to crowd them all in. Then various bugs were discovered and more states had to be freed to correct these. The 'battle of the states' continued for several months and the end result was one of the most efficient encoding jobs ever done.
The necessity for magnetic card program storage and entry became apparent as soon as we had an operating prototype. Don Miller, Dick Osgood and Bob Schweizer deserve praise for the speedy development of this unit, which adds so much to the convenience of the 9100A.
Clarence Studley supervised the overall mechanical design and assembly, reducing to manufacturing drawings and to the final metal the handsome cabinet styling of Roy Ozaki, Don Aupperle and others in the Industrial Design group, while Harold Rocklitz and Doug Wright handled much of the tooling.
Many other people contributed to the 9100A — too many to give proper credit to all; but I must mention the fine art work of Frank Lee on the read only memory and other printed circuit boards, Chung Tung's work on the core memory electronics and Bill Kruger's development of the short high brightness cathode ray tube. Chris Clare made many contributions to the project especially in the area of interfacing the calculator with printers and other peripherals. A special measure of recognition is due Ken Petersen whose expert technicianship and whose genius at trouble shooting saved us weeks of time and bailed us out of many tight spots. Ken also laid out the multilayer mother board with its thousand diode gates and interconnections.
The transfer of the 9100A from Hewlett-Packard Laboratories to the Loveland Division took place gradually rather than abruptly. As various portions reached the final prototype stages, responsibility for these was assumed by the Loveland group headed by Bob Watson in engineering and by Jack Anderson in production. Many visits both ways and some transfer of people to Loveland accomplished the transmission of much unwritten information. The Loveland team introduced several engineering improvements. Especially significant were the improved read only memory margins obtained by Rex James and the many contributions by Ed Olander, whose comprehensive understanding of the entire machine helped greatly. That we were able to go from an incomplete lab prototype stage in Palo Alto to a pilot run of final instruments in Loveland in only 10 months attests to the skill and dedication of the Loveland group and to the fine cooperation on both sides.
Finally, I must confess that very few projects receive as much direct attention from corporate management as this one did. Early in the spring of 1967 a skiing injury landed Bill Hewlett in the hospital. We learned about this right away when he called up to have some 9100A programming pads sent over. I found myself hypnotized by the project and unable to share my time equitably. Here was management in the unusual role of consumer, for if Bill and I did anything constructive it was mainly to assess and modify the developing product from the user's standpoint. I owe the 9100A group an apology for being constantly in their hair, and everyone else in HP Labs an apology for slighting their projects. Now that it's all over I find the 9100A as fascinating to use as it was to develop. Caveat emptor!