Mannerisms

Dr Masatoshi Shima was one of the Busicom team which went over to Santa Clara in June 1969 to commission Intel to produce a set of chips for a calculator. Shima wrote the functional description of the 4004 chip-set's CPU, and designed its logic.

"I brought the idea for a ROM-based decimal computer architecture," says Shima, "in 1968, Busicom had already introduced the stored programming design instead of the hardwired logic design. Busicom had already become successful in the OEM business with a ROM-based calculator-printer, introduced in 1968, which had the programme stored on ROM."

"Dr Hoff's proposal," continues Shima, "was to use 4-bit binary micro-level instructions instead of of n-digit decimal macro-level instructions. Busicom realised that Hoff's idea was the next generation's logic design approach, but his proposal was only the concept. We brought the idea for the ROM-based architecture, and Hoff brought the idea for a simple architecture like the 4-bit binary architecture."

"Busicom's first idea was that all the peripherals (keyboard, display, printer etc) had to have hard-wired logic chips. Then we found out how to control the hardware by software," recounts Shima, "so Busicom added many more instructions. That was the basis for the success of the 4004 - finding out how to control the peripherals from software."

At first, however, things did not go smoothly. Busicom had chosen to go to Intel, says Shima, because of the reputation which Intel co-founder Dr Robert Noyce had in silicon-gate process technology. Silicon-gate technology allowed for higher levels of integration than the industry's then mainstream process technology - metal-gate.

However Intel had been founded to make memory and, as a memory house, had not envisaged using silicon gate technology to make logic. Consequently, Intel had not hired any logic designers. This may account for what Shima found on his second visit to Santa Clara.

"I had defined all the functional specifications for the 4004 by December 1969. I next visited Intel in April 1970. In those four months, nothing was done. There should have been some progress, but there had been no progress in the development. Intel did not have any logic designer at all who would be able to understand the proposed schematic," remembers Shima, "most of Intel's engineers were process people - they didn't have logic engineers or systems experience."

"Busicom expected the product to be out soon," recollects Shima, "my job in going to the US was to check on what Intel had done. But there had been no progress in four months! Everybody was upset. We had been planning to pay $100,000 for the development (and 60,000 chip-sets were envisaged) but no logic engineer had been assigned to the project. Intel did not assign any logic engineer to the 4004 logic until Federico (Dr Federico Faggin) came on-board."

Intel nearly lost the project, recalls Shima. "Busicom had a chance to go to Mostek to develop Busicom's original idea because Mostek also had silicon gate technology and had developed, in 1969, the first single chip calculator using a Busicom schematic." That chip had been made for use in a handheld calculator.

"Several USA semiconductor houses had said that Busicom's original idea could have been done," remembers Shima, adding, "if Busicom had not gone to Intel the microprocessor might have been born at TI (Texas Instruments)." However, on reflection, he reckons: "If Hoff, Faggin and I had not been there, the microprocessor would not have been born."

It was a close-run thing. "Without Bob Noyce, Intel wouldn't have had a contract with Busicom," recalls Shima, "only Bob Noyce had relations with systems customers. He listened to whatever I asked him. When Busicom asked Intel to develop a specialised ROM, RAM and shift register, some people at Intel didn't want to do that and said Busicom should use a standard RAM, ROM and flip-flop' - because all the people were processing engineers, not systems engineers. They didn't have a sense for marketing."

Noyce, however, saved the day. "Frankly he supported Busicom's project well", recalls Shima. So Shima got down to designing the 4004's logic and in doing so ruined the no-smoking status of Intel's layout room.

"In the layout room, whenever I went there for checking the 4004 layout, quite often I found Dr Noyce was checking the layout - always smoking lots - always putting two packs of cigarettes on the light table. Then I started smoking too. It was only later that I discovered that the layout checking room was a no-smoking room!" he recalls.

The budget was tight. Shima remembers being told that the logic simulation had to be stopped when it had exhausted its budget of $5,000.

"Federico and I worked so hard together", recalls Shima. Faggin had the overall responsibility for getting the chips designed and put into silicon; Shima, with his knowledge of calculator design, was responsible for the system logic; Hoff was still there to resolve architectural issues.

"I was not so excited when the 4004 became functional, but when the 4004 worked as the engine for a calculator with my programme, I was very excited and happy, because my responsibility was to develop a calculator," recalls Shima.

So, for Shima, the project worked out well, because Busicom sold 100,000 calculators based on the 4000 chip-set and later developed a cash register, a billing machine, and a teller machine for NCR, all using the 4000 family of chips.

In addition, Shima gained a formidable reputation as a logic engineer - there was not a single error in the 4004's logic.

NEXT FRIDAY: July 25th, Federico Faggin tells how the first microprocessor was made.