Moving to Semiconductors
To maintain their competitive edge, Honeywell shifted their focus away from the development of minicomputers to the advancement of mainframe technology. As a result, the company needed to reorganize and staff were informed in a meeting that Honeywell was forming a new semiconductor memory group.
During that meeting, Regitz recalled sitting near his colleague, Bill Jordan, who worked in the same department. Neither of them knew about Honeywell's plans for this new unit. Honeywell assigned Regitz and Jordan to the new semiconductor memory division and provided them the budget and resources to investigate new technologies in semiconductor memory.
Although the semiconductor memory division was newly established, it was not haphazardly developed. Born out of his work with Texas Instruments in developing integrated drivers for core memories, Bill Jordan had advocated for the division. He believed semiconductors were the future of technology. Jordan proposed two semiconductor programs to focus on: the bipolar junction transistor (BJT) and the metal-oxide-semiconductor field effect transistor system (MOSFET).
Early bipolar semiconductors at this time used transistors made from germanium. However, MOS transistors primarily used silicon. The three layers within the MOS transistor included a metal conductor, an insulating silicon layer, and a semiconductor silicon layer. When applying an electrical voltage, the balance of the electrical charge between the metal conductor and the semiconductor layer was altered, allowing for the storage of binary information.
Regitz chose to concentrate on the newer MOS technology. To get the foundation he needed, Regitz carried out research and read papers about MOS transistors. John Defalco, another Honeywell engineer, worked in a semiconductor group at 3C that looked at simple MOS design. In collaboration with Defalco, Regitz built a circuit simulation program that analyzed models and tested how MOS worked within an integrated circuit.
Based on his research, Bill requested funding from Honeywell to produce a Three-transistor cell 512-bit MOS memory device, referred to as the F30. Honeywell provided him a list of approved manufacturing companies to work with, which included Motorola, Fairchild, Intel, and Honeywell's own internal manufacturing. However, 512-bits was a modest density for a chip. Intel wanted to push the envelope a bit further and aim for 1024-bits, arguing that the 512-bits' die size, which is the size of the silicon wafer used to make the chip, remained too small to keep production costs low. As a newly established company, Intel needed a major project to establish itself as a cutting edge competitor in the computer manufacturing industry. Producing a cost competitive alternative to core memory would give Intel that leading advantage, so they showed keen interest in the F30.
Each company was given 4 possibilities to test:
-
A 512-bit chip in a p-channel process circuit (Honeywell's standard)
-
A 1024-bit chip in a self aligned gate circuit (Honeywell's standard)
-
A 1024-bit chip in a self aligned gate circuit with a double x-line
-
A 512-bit chip in a self aligned gate circuit with a double x-line
Motorola and Fairchild produced modest results that met expectations, but Intel tested the 1024-bit chip using Honeywell's process, quickly releasing their results and exceeding the product expectations. As a result, Regitz and his team chose Intel.
At Intel, designers referred to the chip as the i-1102 since the process followed production of the i-1101. From the start, the i-1102 posed some challenges, namely inefficient connection methods between components. The chip also demanded increased usage of its peripheral circuitry, the parts of a RAM cell that connect it to the rest of the board, than deemed preferrable.
Jordan recalled a conversation with Regitz while in the thick of working on the chip, as the two were sitting in a cable car. Regitz wondered aloud: “Can you imagine, one hundred years ago, some guy told his boss about an idea to build a transportation system with five mile long cables? If he could get that running, we ought to be able to get the 1102 to work!”
Once the 1102 proved successful, Bill, alongside Joel Karp who was one of the team's collaborators at Intel, co-authored a paper addressing their research on the 1102 and submitted it to the Institute of Electrical Engineering (IEEE). The IEEE accepted their paper, and invited Regitz and Karp to give a presentation of the paper at the February 1970 International Solid State Circuit Convention (ISSCC) in Philadelphia. On October 31st of that same year, "Three-transistor-cell 1024-bit 500-ns MOS RAM" by Bill Regitz and Joel Karp was published in volume 5 of IEEE's Journal of Solid-State Circuits.
Presenting at the ISSCC was a long-term professional goal for Bill, yet despite this, preparation for his presentation was still nerve-wracking in the lead up to the convention. Later, he described the intimidation he felt arriving in Philadelphia “at the convention center. It [was] huge. For a little guy from a little town…as a coal cracker, it’s a scary thought!”
Still, setting aside his nerves, Bill enjoyed the overall presentation experiencem, encouraging him to seek out similar opportunities to establish himself as an innovator within the electronics industry. Regitz and Karp continued working together after the convention, eventually writing another article on computer memory that was accepted by the IEEE in 1972 for presentation and then published in 1973.
Occurring around the same time, and unbeknownst to Regitz and the Honeywell team, while they worked on Honeywell's chip, Intel also quietly started a separate design team of Ted Hoff, Joel Karp, and Leslie Vadasz to develop an alternative draft of the 1102.
Shortly before the conference, Intel's marketing team decided they favored their new design, pushing for the company to drop the original 1102 design and to put the modified chip into production under a new name: 1103. Though Regitz was disapointed about the change at first, he understood that the larger goal was to “set an industry standard, not develop a custom part for the sole use of Honeywell. [The Honeywell team] truly understood the power of buying standard parts and the economic impact on pricing.”
Regardless of Intel's rejection of the 1102 in favor of the 1103, Intel impressed Bill over the course of their collaboration with Honeywell. The young start-up offered opportunities for innovation and a more trustworthy environment for visionaries like Bill. He interviewed with Intel and they offered him a job. Although tempted to stay at Honeywell, he accepted the position to work more intimately within the semiconductor industry and chip design process.
On the plane leaving the conference, Jordan tried to convince Regitz to stay at Honeywell. However, Regitz had made up his mind about joining Intel's team. In a twist of fate, Regitz ended up convincing Jordan to join him at Intel instead. They “agreed to put a business plan together to convince Intel to go into the memory systems business.”
Regitz sent his job acceptance to Intel, along with the proposal he and Bill Jordan drafted that argued for Intel to join the memory systems business.
The Three Transistor Cell 1024-bit MOS memory device
