First-Hand:Origins of Hewlett Packard 35 (HP-35)
|Line 1:||Line 1:|
'''Contributed by:''' Dave Cochran
'''Contributed by:''' Dave Cochran
|Line 35:||Line 37:|
Revision as of 17:49, 27 August 2012
Contributed by: Dave Cochran
After the development of the 9100 desktop scientific calculator in the mid 1960s, Bill Hewlett became obsessed with the idea that HP should develop the same capability to fit his shirt pocket. Every few months he would walk into the lab in building 1 Upper and ask how we were doing; he often approached me as I was investigating architectures amenable to the scientific algorithms I had used in the 9100.
Although semiconductor density was increasing yearly, bipolar technology was never going to be suitable, too large and power hungry. Metal Oxide Semiconductor (MOS) promised high density and low power, but was still in its infancy. However this didn’t stop Hewlett from getting the Industrial Design group of HP Labs to mock up some ideas of shape, key layout, etc. and fit in his shirt pocket. The Solid-State Laboratory was also working on LED displays with minimal-power bipolar driver circuits.
I had been obtaining samples of various semiconductor architectures that performed simple four-function calculations from U.S. vendors as well as the Japanese. Most were bipolar but some MOS circuits with hundreds of transistors per chip were starting to be designed. In 1970 Fairchild Semiconductor showed Tom Whitney and I a PMOS architecture that looked promising as a candidate for scientific algorithms; a binary coded decimal (BCD) adder and up to 20 digit long multiple words in a circulating shift register (race track) arrangement that was very efficient of both chip size and power. They were intending to offer this chip set as the platform for fixed point four-function calculators
I spent about two weeks scoping out a modified architecture based on what I has seen at Fairchild and determined I would need only three 13 digit (56 bit) registers and a microcode word length of 11 bits; this was later shortened to 10 bits by using only an inferred conditional branch. A ten percent reduction in circuitry was very significant. Thirteen digits would be sufficient for ten digit accuracy with an overflow or carry digit and two guard digits. The word could be displayed as either a mantissa with two exponent digits or variable length fixed point. The product would have an arithmetic and register chip, control and timing circuit and several Read-Only-Memories. How often does someone have a chance to design a microinstruction set?
Fairchild decided it didn’t want to do a custom design for HP so together with the group manager, Tom Whitney and the lab director Paul Stoft went to Hewlett thinking he’d be overjoyed that we had a combination of technology and architecture that would fit in his pocket. We told him that we would have to contract out development of several new technology PMOS circuits.
The selling price with all those custom circuits would have to be certainly more than $100 that current models of four-function calculators were going for. Hewlett wasn’t sure that the million dollar development would have a positive pay-back, so we used the existing R&D budget to fund the project while Hewlett contacted the think tank SRI to do an independent market feasibility study. SRI came back many months later stating after studying the market through focus groups, etc.; what HP had in mind couldn’t be priced”.
As the program was moving along we had a “name the baby” contest with many entries such as the Math Marvel, Athena, etc., Hewlett came by and said, “Well, it should be called the HP-35”; why, “because it has 35 keys”.
It was late in 1971 when we put together the first HP-35 prototypes and handed out a few to distinguished scientists in the area. Stanford Engineering Dean Fred Terman, the person responsible for bringing Bill and Dave together to start HP was one of the first. He was overwhelmed, looking for an umbilical cord connected to a big computer doing the precision calculations. Wouldn’t you know he’d find the first bug, he entered 90 degrees and pushed the TAN key; the unit started blinking as the algorithm divided by zero. I had to put in a trap to make it show 10 to the 99th power to show infinity. Nobel Prize winner Dr. Charles Townes was so impressed he called it “the eighth wonder of the world”.
The HP-35 was introduced with little fanfare for $395 through the normal sales channels; but word spread and orders quickly backlogged. The first production run scheduled to carry us for six months was limited to 100K; a few months later it was doubled. When told about an RFQ we had received from GE for 20 thousand units, Hewlett couldn’t believe it; I suggested that maybe they want each of their engineers to have one. He said, “No they should just borrow from one another”.
The minute attention to detail in every aspect of the HP-35 from battery life, the shape of the seven-segment display to the position of each key paid off. The HP-35 was truly a product which you knew would be successful because the engineer at the next bench wanted it.