Recalling 74181 history during the ICs Marketing War for TTL leadership in the 1960s and 1970s

David Laws, Semiconductor Curator for the Computer History Museum who worked in Silicon Valley semiconductor companies including Fairchild Semiconductor for more than 40 years, illuminated me about Steve Allen's photo and the origin of the 9341 ALU. 

In 1968 both Fairchild and Texas Instruments had bipolar variable array programs in operation to provide quickly custom designed circuits. Fairchild's program was called Micromatrix, while TI's was called Discretionary Routed Arrays. Fairchild had available 4600 (48 Gates/Array) and 4700 (96 Gates/Array) TTL Micromatrix arrays, which contained six and twelve cells of TTL logic respectively; each cell consisted of four TTL AND-OR-INVERT elements, counted at two-gate-per-element complexity [20]. The cells were interconnected by two layers of custom metallization to produce the customer's desired function.

A 4-bit arithmetic logic unit was initially designed by Fairchild a year or two before the 9341 or 74181 as the MSI 2-layer 4711 (Steve Allen's photo with Accession Number 102710038, 30 Sept 1968), an example of the capability of Fairchild's custom Micromatrix product line. The 4711 chip was a functional 4-bit ALU, but not optimized for yield, as were the successor devices [23].

The Fairchild 9341 MSI function was a derived function from the 9340, a TTL MSI 4-bit arithmetic logic unit with internal Carry Lookahead, capable of two arithmetic operations and six logic functions, which Fairchild designed in a 24 pin DIP package: it was conceived in the applications group under Bob Ulrickson by either or both Clive Ghest and John Nichols. Robert Ulrickson was a supervisor of Systems Engineering Group in the Applications Department at Fairchild (where he worked from 1966 to 1973) where his team of engineers invented the first dozen TTL MSI devices which were initially introduced by Fairchild Semiconductor as the 9300 series.

The 9340 die was very big, that made it expensive. Also not every application needed the carry look ahead built-in feature. Realizing that this would limit sales, they divided the function into two chips, the 9341 and 9342. To test out the idea rapidly the first version of the 9341 was implemented using Fairchild's computer-aided design system: this is the 3-layer chip shown in Steve Allen's photo.

Once the functionality was verified, the function was redesigned onto a smaller more economical chip suitable for high volume production: the 9341 ALU chip.

Fairchild did not market a TTL family until 1967. When Fairchild entered the market, during a period in which total industry sales of integrated circuits almost doubled, a battle for market leadership in TTL was already on between Sylvania (the first commercial manufacturer of TTL) and Texas Instruments. At the same time, National, which under license from Texas Instruments acted as a second source supplier of its 54/74 TTL family assuring a secure supply, began an aggressive cutting price campaign that helped increase the market share of Texas Instruments design [18].

By 1968, improvement in lithography significantly increased the number of transistors that could be integrated on a chip. Desirous to gain share in the TTL business, Fairchild (9300 Series) and Signetics (8200 Series) pioneered the design of TTL/MSI functions (Medium Scale Integration - up to 100 logic gates per chip) such as counters, shift registers and arithmetic logic units [19]. For a long time Fairchild supplied the more advanced MSI chips, including the 9300 4-bit universal shift register and the 9316 4-bit binary counter, while Texas Instruments was more volume-focused on TTL/SSI chips (Small Scale Integration - up to 10 logic gates per chip) such as simple gates and flip flops.

As computer market was opening up, demand for TTL/MSI grew explosively: allowing a superior way to assemble a minicomputer, TTL/MSI caught on quickly, particularly in the 74xxx numbering system originated by Texas Instruments, who soon grabbed the lead during the early Seventies.

Fairchild was actually the sole source on the proprietary TTL/MSI series 9300, while there were a lot of sources on the series 7400. This represented a serious limit for Fairchild's sales, since early computer makers, like Digital Equipment Corporation (DEC), avoided to design with components available exclusively from a single source in order to reduce the risk of not being able to sell their final products due to delivery problems with the sole supplier.

As a result, Fairchild had to adopt the 54/74 numbering scheme, becoming just an alternate source to Texas Instruments who was already a giant manufacturer [21]. In effect, the marketing departments of the two companies knew that standard products would be purchased by customers in higher sales volume than custom products or products which differed slightly. So, they collaborated to achieve higher sales for both through 2nd sourcing: this was a break from prior customer practice of specifying custom logic designs (at the gate and flip-flop level) to be "written" on Silicon by semiconductor companies [source: Mr Robert Ulrickson, personal email exchange]. 

Thus, of these products, the 7400 series from Texas Instruments became de facto standard, with many similar products being produced: 74195 was the direct replacement of the Fairchild 9300 function, the 74161 of the 9316, the 74181 of the 9341, and the 74182 of the 9342.

By 1970, after having introduced a much faster technology called "Schottky TTL", the design of Texas Instruments had become the industry reference, and Sylvania had effectively withdrawn from the semiconductor industry. The 74S181 was successively implemented by Texas Instruments in Schottky S/TTL technology in mid 1971: in that year Texas Instruments had 41% market share in TTL, and it would remain the market leader in bipolar logic until after 1980.

As per above historic reconstruction, the 9341/74181 was not the first conceived integrated ALU: a 4-bit ALU pioneered and implemented by Fairchild on custom computer-aided design system seems to have preceded it by some considerable time.

Furthermore, the Signetics 8260, TTL/MSI Arithmetic Logic Element with minimalist functionality (4-bit adder, XNOR, AND), was really the first integrated ALU to be marketed, at least in 1969 as per “6947” data code I recognised in the below picture. The 8260, which was employed in DATA GENERAL SUPERNOVA, is de facto the sole device that Texas Instruments cross-referenced to the 74181 (as a recommendation for new design) on its first data book, in March 1970. 

At the time of their inventions, Fairchild did not patent the logic functions or circuit designs and their patent filings were much more focused on semiconductor device process technology. Actually, there was no specific pattern of licensing in the early semiconductor industry and patents and intellectual property rights were the subject of frequent never-ending costly litigation. Prior to “The Semiconductor Chip Protection Act” of 1984, the application of copyright law to integrated circuits was not clear and any form of intellectual property to adequately cover a chip did not exist [22]. Also, the practice of "second sourcing" to provide customers with the assurance of multiple suppliers for popular logic functions as an incentive to buy their standard products, pushed manufacturers to cross-license their products. Additionally, the recurrent talent mobility due to the large number of spin-off firms in the early semiconductor industry, gave companies an easier access to one another's technology.  

On the Web (mainly at the Computer History Museum) you can find interesting documents and interviews about the lack of patents filed by the early semiconductor companies in Silicon Valley. Interesting are the conversations of Robert Wayne Ulrickson about the alleged paternity of the MSI design of either Fairchild's 9300 product line or TI's early 74/54 series. In particular it is mentioned the 9341/74181 ALU: you are invited to read the outcome of such discussion in this inteview [17].