Copies of Bell’s correspondence to Olsen courtesy Ed Millis. Click on these for full size images

But those Type A Bell point contact transistors were fragile devices that did not deliver the promise of virtually infinitely long lifetime. Haggerty recalled in 1980 “it was by no means obvious in 1948 and 1949 and 1950 that the transistor would be the kind of fundamental contribution it has become. But slowly, during 1949 and 1950, a portion of it at least became clear to me, and this was that the future of electronics would be profoundly influenced by the knowledge already attained and the knowledge to be gained about materials at the structure of matter level.” [Texas Instruments 1980]

Accordingly, Haggerty created a strategy to develop the new opportunities in semiconductors. The company would:

Get a license to the Bell transistor from Western Electric;

Create a project engineering group to create, make and market semiconductors; and

Establish a solid state physics research laboratory.

Accordingly Olson, Jonsson and Haggerty pestered Western Electric for a license through 1951 without success. Jonsson recalled that they bemused Western Electric who doubted GSI could succeed in that business saying “This business is not for you. We don’t think you can do it.” [Texas Instruments 1980]

Western Electric were confronted by diverging National interests: Due to the Korean War the US Military favoured secrecy around the transistor while the Department of Justice demanded commercial openness through an anti-trust action. Finally, the US Military agreed that Bell could issue transistor licenses to anyone in a NATO country. Following the issue of Shockley’s junction transistor patent on 25^th September 1951, Bell began offering non-exclusive licenses to anyone prepared to advance $25,000 up-front. GSI quickly signed up.

On January 1^st 1951 a new parent company was launched initially named General Instruments and GSI was relegated to a subsidiary. But the new name was already in use by an East Coast defence contractor and a year later the company was renamed Texas Instruments or TI for short. TI made its semiconductors venture public in March 1952, Jonsson advising his employees “Finally, we have purchased from the Western Electric Company a license under which we may manufacture transistors and related semiconductor devices. The transistor is a very new development, primarily of Bell Telephone Laboratories, which promises to revolutionise electronics. It is, in a general sort of way, a substitute for the vacuum tube.....there is little question but that transistors and related devices will play an exceedingly important part in our future.” [Jonsson cited Pirtle 2005]

In 1952 Bell ran a transistor Symposium for its licensees. Pat Haggerty, Bob Olson, Mark Shepherd and Boyd Cornelison represented TI. Some attendees complained that the Symposium was unhelpful. For example, Stuart Seeley, head of RCA’s Industrial Service Laboratory complained bitterly: “[They] brought the group in, they sat them down in some hard chairs, and began throwing theory at them until they were just too fatigued to listen to any more.” [cited by Choi 2007] But Mark Shepherd thought it successful: “They worked the dickens out of us. They did a very good job; it was very open and really very helpful.” [Cited in Riordan 1997]

The key to junction transistors was high quality single crystal germanium (and later silicon) first made on crystal pullers using the Czochralski process. After seeing the Bell design during the Symposium, Cornelison returned to his hotel room that night and made two sketches in the back of his spiral note book of what he had seen that day. [Millis 2006]

Sketch of the Bell Crystal Puller by Boyd Cornelison. Scan courtesy of Ed Millis

Shepherd was appointed project manager of the Semiconductor Project Engineering Group and established a group of 15 engineers. Despite its enthusiasm TI had no semiconductors experience and it trailed behind other mainstream vacuum tube producers such as General Electric, Raytheon, RCA, CBS, Sylvania and others such as Transistor Products and who were already producing point-contact diodes and General Electric which also had a junction diode. And these companies were making good progress with transistors.

Collection of early point-contact transistors in development in the period 1948-1952 courtesy Joe Knight. For full legend see here.

For example, by 1951 there were several point-contact transistors manufacturers in addition to Western Electric: Sylvania had already released the GT-372, Raytheon had released the CK703 and General Electric had released its SX-4A and Z2 transistors. General Electric had invented the alloy junction transistor and presented it at the Electron Devices June 1951 conference at Durham. At the same conference Bell discussed the development of their grown junction transistor; an approach that General Electric were also pursuing leading to their invention of the rate grown junction transistor.

Cornelison was given responsibility for building TI’s first crystal puller subsequently named “Old Betsy” based on what they learned at Bell and from its scientists such as Gordon Teal and Ernest Buehler on building crystal pullers and producing double doped grown junction transistors. TI successfully pulled its first crystal in June 1952 using the last of two seed crystals supplied by Bell. By August TI was producing its first point-contact transistors.Having succeeded in obtaining a license and setting up its semiconductor project engineering group, Haggerty addressed phase three of his strategy: A solid state physics laboratory.

Towards the end of 1952 Haggerty persuaded Gordon Teal to move from Bell Laboratories to TI and head up the research. Teal set out his agenda for a grown junction silicon transistor: “Shortly after arriving at TI, I set up a program on silicon crystal growing aimed at producing high-perfection silicon single-crystal p-n junction structures to facilitate our development of a silicon transistor with useful amplification properties. I persuaded Dr Willis A Adcock, an able young scientist, to leave his catalysis studies in one of the oil industry laboratories and join TI to undertake these crystal growing investigations, which I believed to be the key to the achievement of a silicon transistor. I reasoned that going the grown junction route would avoid the differential expansion difficulties between silicon and an alloying electrode inherent in use of alloy junctions.” [Teal 1976]

There is little doubt TI scored an enormous coup in securing Teal and significant

help from him even prior to his arrival. Others were less favoured. At General Electric the inventor of the alloy junction transistor, John Saby complained: “Gordon Teal wrote papers on crystal growing, but never disclosed a lot of the details of the process to get the crystals to grow.” [Morton 2000] Right: picture of Gordon Teal.It is scarcely surprising that the first junction transistors produced by TI belonged to the grown junction class and that TI invested a great deal of its research over the following two years into developing junction transistors for radio frequency germanium transistors and silicon transistors.

Other manufacturers concentrated on germanium alloy junction transistors which became the volume workhorse of the 1950s due to their better adaptability to mass production. But the grown junction was the only option in 1954 for credible RF germanium and silicon transistors and Teal’s dedication to this approach enabled TI to get a head start in the industry in two important domains.

First Transistors

In September 1953, TI introduced Types 102 and 103 in an improved hermetically sealed can. They were produced by D D “Mac” McBride who started at Texas Instruments in July 1953. Previously a watchmaker, he had mastery of the delicate touch needed to assemble these fragile devices. The entire process was carried out manually. For example to create the points McBride started with strips of phosphorous bronze and beryllium copper that were glued onto a fibre insulator. Then he would “use a microscope and cut, by hand, each point to the correct shape” and then to finish the transistor “mount (solder) the base and point contacts in place. Under a microscope, I’d adjust the points on top of the germanium wafer to about 5 mils apart. We’d use LePage cement to hold the points in place. Once dried, we canned the transistors in oil for heat sink purposes.” [Ward 2002 Picture courtesy Joe Knight] McBride reported to Ed Millis who recalls “McBride would build a batch of thirty units every two days, test them, and put the good ones into stock. Then someone noticed that no point contacts had been sold in months. My short and easy career of being the foreman of the point contact transistor assembly line came to an end, as we soon ceased production of this type.” [Millis 2000]

Germanium Grown Junction Transistors

The grown junction transistor was invented at Bell Laboratories in April 1950 by Morgan Sparks and was an important verification of Shockley’s theory of the junction transistor. It was made possible because Gordon Teal’s determination (while still at Bell) to champion the use of high purity single crystal germanium. The NPN transistors were grown by pulling crystalline germanium from a crucible of molten N-type germanium. At controlled intervals the germanium melt was changed to P then back to N creating an NPN block from which small bars could be sliced to make individual transistors.

The first grown junction transistors produced by TI were the types 200 and 201 introduced in November, 1953, and exhibited at the March, 1954, IRE show. They were originally made in the tall cans used in the 102 or 103 point-contact types before standardising on the lower profile cans. [Smithsonian G00043 Ward 2001] Elmer Wolff was on the development team and recalls “Of course, at this time, only germanium was used, and all these units were of that material. This transistor was much more rugged, could dissipate more power, was more reliable and easier to build. The combination of these attributes opened up a wider range of potential uses and applications.” [Ward 2001]

A transmitter was designed and built using TI germanium grown junction transistors for the ceremonial opening of the enlarged Lemmon Avenue Plant on November 18, 1953. It was assembled in a small case with a strap to resemble a wrist watch and produced a voice actuated signal. A conventional receiver several feet away picked up the signal and actuated a spark which cut the ribbon for the official opening. [Smithsonian G00314]

Picture of a Type 200 Courtesy Joe Knight

A germanium grown junction power transistor and phototransistor were also exhibited for the first time at the 1954 IRE show.

Advertisement from November 1953 Electronics Magazine courtesy Steve Reyer

The power type had a copper heat sink soldered to the can for heat dissipation and was rated for 1 Watt at 25 degrees C. The bar size was .40"x.190"x.190". [Smithsonian G00037 & G00038]

Wolff recalls how TI engineers had to work on new applications to create a market for their transistors. “TI added some applications engineers to the Semiconductor team. At first, this applications group consisted of two people – Jim Nygaard and Ed Millis (if my memory is correct). TI had to develop the application, take it to a potential customer, and work with the customer’s engineers to explain the circuits and how to use the devices.....The relatively small size and low power characteristics of the junction transistor led to the exploration of use in hearing aids with the Sonotone Corporation. At that time there was no experience in the electronics industry in the use and applications of transistors. We had the devices available, but no one outside the manufacturers of transistors knew how to use them.” [Ward 2001]

October 1953 TI received its first major order: Sonotone ordered 7500 transistors for its hearing aids. These transistors were probably intended for the Sonotone 1111 hearing aid which was its first fully transistorised model and for which examples survive with original Texas Instruments transistors. Left: Picture courtesy of Bob McGarrah of a Sonotone 1111 with two Texas Instruments transistors types 204 and 205 made in early 1954 and one Germanium Products TN-29/RDR 1 type.

Sonotone clearly had difficulty sourcing their transistors and examples exist with mixed line-ups of CBS-Hytron, RCA, Raytheon and Philco transistors.

Problems of quality control explains the lack of consistency in what could be used. Wolff recalls “In the early manufacturing stages, we did not have enough production controls to produce all the transistors with the same electrical characteristics, such as noise levels low enough to become the input device in a hearing aid amplifier. As a result, we had to select devices from a manufactured batch that met a combination of gain and low noise sufficient to support the hearing aid business. We used a color code to identify the performance level of specific devices. Can you imagine using colored paint (yellow, green and red) to identify transistors that could be used to make a complete set for a hearing aid? Well, that’s how we did it in the beginning, and we didn’t think it was such a bad idea at the time.”

Making a 200 Series NPN Grown Junction Transistor

A crucible the size of a teacup was loaded with pure germanium and heated to a red heat in order to melt it. Then a thin seed crystal was lowered into the molten germanium and spun while slowly withdrawn from the melt. By careful selection of the melt temperature and the rate of withdrawal a single crystal of very pure germanium could be pulled. [Picture courtesy Texas Instruments]

Grown junction transistors were made on crystal pullers The NPN structure was achieved by starting with N type germanium, changing it to P-Type then back to N-Type while pulling the crystal. This was done by adding just the right amount of impurities of the right kind and the right time.

Then the crystal was sliced into thin bars and mounted on a header each end providing the collector and emitter connections. Through a delicate operation a connection was made to the narrow base region and terminated to the base lead on the header. Picture: a 2N243 grown junction transistor. For more information

Early Germanium Power Transistor

The X-2 transistor was developed from the early germanium junction transistors in about 1954. It featured a substantial copper chassis mount for improved heat transfer giving a modest 330mw dissipation. Picture courtesy Joe Knight

Germanium Tetrode

Tetrode transistors were grown junction transistors that featured two base layer connections each made to opposite sides of the transistor bar.

The first tetrodes produced by TI were the grown junction 700 series sold for automatic gain control applications (AGC) in low level audio amplifiers. The 700 was part of the range by March 1954 as indicated in TI advertising. Tetrodes were more normally used in high frequency applications (see below for a more detailed description of the tetrode).

The gain of these transistors could be varied by up to 20dB with only 100μA applied to the second base electrode. [Electronics 1954]

These examples date from around 1953-54 as indicated by the use of the tall can style. Pictures courtesy of Joe Knight.

Type 800 Photo Transistor

TI promoted these transistors as having the same functions as phototubes “with the further advantages of smaller size, weight and power demand. They operate essentially as grounded emitter transistor amplifiers displaying great sensitivity to the direction as well as the amount of light received.”

Light was admitted through a glass lens mounted on top of the case and focussed on the base layer of the transistor which was hermetically sealed. They were part of the range by March 1954. (Photo courtesy Bob MacGarrah)

These were junction transistors but had no base connection. Elmer Wolff worked on example applications and built an auto-light dimmer for cars [Ward 2001]

Radio Frequency Transistors for Portable Radios

The stand-out potential application for transistors were clearly in portable radios. Tube portable radios required expensive high tension batteries that did not last long. A transistor radio offered the possibility of much greater battery economy

This work continued after the deal was done with IDEA and TI was supporting Dick Koch finalise the design: “While the final configuration of the radio was being designed, Frank [Horak] continued to make slight modifications in the structure of the RF transistors in order to improve even further the RF characteristics of the units to be produced in large quantities.” [Davis 1993]

Picture shows a line-up of crystal pullers used in production based on the original Boyd Cornelison design with electronic temperature controller designed by Art Evans. From Automation November 1955 courtesy Steve Reyer

Working against exceptionally tight deadlines the radio went into production in late October 1954. The production line was state of the art for the day featuring a solder bath. A remarkable video of the production has been posted here

Pictures from the Video show (a) hand assembly (b) soldering and (c) production output of one unit every 5 seconds!(wildly exaggerated production rate)

TI sold each set of four transistors for $10. As part of the deal IDEA agreed to assign its USA commercial rights in three circuit design patents to TI who were keen to make sure that any USA radio producer could enter the market. In a rationalisation of the businesses of each party and to assist IDEA’s cash flow, the Radell resistor manufacturing Division of IDEA was sold to TI. [Pies 1998]

These patents were

2820890 Radio Chassis Construction

2880312 Transistor oscillator-Mixer

2892931 Transistor Radio Apparatus [Pies 1991]

Transistors in the Regency TR-1

The colour coding is shown in the composite picture below:

Left: View of rear of the TR-1 and Right: Close up of the four transistors

Courtesy Steve Reyer

Producing 400,000 Transistors and the Ladder Process

TI had to meet the huge demand for grown junction transistors that would arise as a result of the success of the TR-1 but existing manufacturing processes were too slow as Ed Millis recalls: “This was a very labour intensive process, and during the development of the Regency radio, it became apparent that this wasn't going to hack it, cost-wise, and TI needed a faster way to build the NPNs.”

That better method was developed by Elmer Wolff and became known as the ladder process. It provided a faster and more efficient manual method for assembling germanium bars into transistors. In addition, by using all corrosion resistant metals such as gold and platinum, the final clean-up etch could be done without the masking step necessary previously.

Rather than manually solder the bars directly to the transistor header large numbers of transistors were given all their terminations in a single step. Firstly the germanium bars were cut. “When they were dicing up the germanium crystals, they would slice it open and then do a light etch which would make the base junction visible. It looked like a line across the surface. They set up the saws to center that on the bar length.”

Using a graphite boat with a cut outs in the shape of a ladder the individual germanium NPN grown diffused bars (rungs of the ladder) were placed over thin die cut squares of a gold-germanium solder positioned at the end of each rung. This would solder them to the sides of the ladder (thin strips of platinum positioned below the rungs). The base connection was made by positioning an indium dot over the base of the transistor and tensioning a platinum wire across the top of these. The boat was taken to an oven to fuse all these elements. “The indium dot was big enough to be somewhat forgiving on hitting the P layer.”

This sped up the processing at the expense of performance: “The main problem was the indium dots in the ladder process caused a huge base-collector capacitance, which necessitated a neutralizing capacitor, custom-fitted, for each HF stage” according to Ed Millis. Even so “this process provided improved productivity at the time and was used at TI from 1954 to 1956 for some production” recalls its inventor. [Cornelison 1955 Millis 2011 Smithsonian G00014]

Legacy of the TR-1

The TR-1 sold at a retail price of $49.90. In retrospect Haggerty and Harris rue the decision to sell it too cheaply Haggerty suggesting “it was a serious strategic mistake. The facts are that at $60 or $65 it wouldn’t have made an iota of difference.” But it would have made a huge difference in the profitability of the business. Haggerty noted that with more money to invest in further consumer products “I think the likelihood is very high that we would have been the Sony of the consumer business.”

Some 105,000 TR-1 sets were sold requiring 420,000 transistors representing significant mass production in the context of the mid 1950s. On the 25^th Anniversary celebration of the TR-1 and silicon transistor Pat Haggerty observed “One of the most important aspects of our strategy, as I commented, was proving that transistors could be made, that we could make them, that they would work and that they could be made in large quantities.”

Haggerty credits the TR-1 for landing them major supply contracts with IBM. He and Mark Shepherd had been trying unsuccessfully to gain this prime account. The breakthrough occurred in December 1957 when IBM finally signed with TI to supply a large percentage of IBM’s requirements for many years. IBM had been persuaded after “Tom Watson Jr had bought 100 or 200 of those pocket radios in about 1955 and scattered them among IBM’s key executives...he said that if that little outfit down in Texas can make these radios work for this kind of money, they can make transistors that will make our computers work too.” [Texas Instruments 1980]

A similar example was recalled by Roger Webster as to how the TR-1 opened the doors of other portable radio manufacturers: “There was a radio design section that contacted all of the radio manufacturers – Warwick, Emerson, Zenith, you name it. We talked to all of them.” [Wolff 1985]

Germanium Alloy Transistor

The alloy junction transistor was invented by John Saby at General Electric in March 1951. In addition, it was independently developed by Jacques Pankove at RCA. RCA were the first to file a patent on the method leading to litigation over which company had priority. Eventually this was settled in favour of General Electric.

The first alloy junction transistors made by TI were the 300 series. [Picture of an unmarked TI alloy junction transistor courtesy Joe Knight] Ed Millis started with the company in 1950 as an engineer on their military products range and transferred into semiconductors in June 1954. By this time the 300 range was in limited production. Millis was put in charge of four female workers building the 300 series transistors as part of his induction into semiconductors. He recalls that it was an entirely manual process. “Before long, the ladies had trained me to all the operations necessary to build transistors, although they did better with their delicate touch on certain of these process steps than I did. For example, holding a piece of microscopic wire with a pair of precision tweezers just so against the indium dot on the side of the germanium chip and spot welding the other end to a header post while peering through a microscope.”

“We made a pretty fine team if I may say so, and on a good day could crank out a hundred transistors, although sometimes I had to come back at night to finish canning them.” [Millis 2000]

While it was relatively easy to make an alloy junction transistor it was much harder to do this in any consistent manner. Starting with a thin wafer of N-Type germanium (the base) two indium dots were positioned on opposite sides. This assembly was heated in an oven causing the dots to fuse with the germanium creating islands of P-Type germanium (emitter and collector).

The alloy junction process did not permit good control of the transistor base width: the main determinant of gain (and also of high frequency performance). Production transistors were elusively variable despite the best efforts to standardise production.

Thus in common with other manufacturers of the period, transistors were not so much made as selected. The TI 300, 301 and 302 transistors were all made on the same line and selected according to their gain from low (the 300) to high (the 302).

Efforts were made to improve the productivity of the manual process. In common with other manufacturers TI used graphite boats in which large numbers of transistors could be assembled and alloyed in the production ovens. More information: Making a PNP Alloy Junction Transistor

The CAT Machine

Because of the need to select transistors from production to meet TI’s own specifications, testing each transistor was an additional manual step that need to be carried out post-production. Equally it was a target for automation.

Following the success of the TR-1, TI developed a range of circuits for low powered four transistor radios to more powerful six transistor designs. Transistor demand was high. The larger six transistor sets typically used two 2N185 transistors in a class B audio output and for reasonable audio quality each pair needed to be matched. Millis recounts the reality: “Since at that time, to quote good old Bob Chanslor, “We could hardly make one alike” matching up our shotgun distribution of transistor variations was an almighty tiresome chore.” Millis built a machine to automate the testing of the transistors which sorted them into any of 108 bins. Transistors in the same bin were considered “matched.”

The Centralised Automatic Tester or CAT for short became the first of many machines that TI developed to sort transistors. [Millis 2000]

Alloy Germanium Power Transistors

Unlike grown junction transistors, alloy transistors could be readily scaled up for high powered audio output. Typical applications were in servo control systems and automobile radios.

TI's first power transistors were the 356 and 356A rated at 12 watts which became the 2N250 and 2N251. [Knight 2007]

Tentative specifications for the 2N250 and 2N251 were issued in August 1956. It was designed for 12 volt audio auto applications with a collector dissipation of 12 watts. [tentative datasheet August 1956] although by March this had been revised to 25 watts [DL-S 726 March 1957]

TI continued their policy of producing designs and demonstration prototypes to encourage the use of their transistors. Here is the first transistorized auto radio prototype built at TI to demonstrate feasibility to Delco, the electronics manufacturing subsidiary of General Motors. It was built in 1955. [Smithsonian G00084]

This prototype contained the automobile radio built into a steel cabinet to hold the speaker and battery.

By 1956 the alloy junction range was extended using JEDEC type numbers and included:

The germanium alloy transistor operations were shut down on June 1, 1979. Over Two billion transistors were produced during the life of that operation.

The Silicon Transistor Breakthrough

In the history of transistors, Texas Instruments is best known as the first company to commence production of silicon transistors. This was in 1954, a blockbuster year for the company. Prior to its success with the transistor TI had, in March 1954, begun production of a silicon junction diode use for signal rectification. [Texas Instruments 1954]

Teal decided to follow the grown junction route using silicon noting that even then braver choices were available. “Some scientists at the time were advocating leapfrogging silicon and going to the III-V compounds to achieve an even higher temperature capability than potentially available in silicon. Certain companies followed this advice and concentrated on III-V intermetallics.” [Teal 1976]

By 1954 most companies were pursuing germanium alloy junction technology while Bell were developing germanium and silicon diffused transistors. The silicon alloy junction proved to be a dead end due to the failure of alloy-silicon interface due to differential expansion. Ultimately diffusion technology dominated transistor production and was essential for the integrated circuit era that began in the late 1950s.

The Silicon Transistor Team: left to right, are Willis Adcock, Mort Jones, Ed Jackson, and Jay Thornhill.

A team consisting of Willis Adcock, Ed Jackson, Mort Jones and Jay Thornhill succeeded in producing the first commercially feasible silicon transistor. The transistor was assembled by Mort Jones on April 14^th 1954 and the lab book pages and text narrative describing his experiments are reproduced here courtesy of the Smithsonian Museum.

These transistors were n-p-n silicon grown junction transistors with base layers only 0.5 mil in thickness. Jones began his entry noting “On April 14 1954, bars were cut from a silicon N-P-N grown junction crystal number A-130, grown the previous day. These bars were .040" x .080" x .200" with the junction perpendicular to the long dimension and approximately in the center of the bar as shown in sketch 1.”

The bars were etched and cleaned making them ready to make a base connection. Under a microscope the grown base was visible as a perpendicular line in the middle of the bar. An aluminium wire 5 mil thick was lowered to contact the base and electrically welded in place. Because the wire was thicker than the base region, the wire would be expected to bridge the NPN junction. By using aluminium this was solved since it created P-type doping and thereby an enlarged P-type zone around the weld.

After further cleaning each end of the bar was nickel plated and soldered to a standard transistor header. The base connection was then made. Jones concluded the four page note saying: “Other data on this transistor were measured and are in a folder on these units. Since the construction of this initial transistor about 150 more good units from about seven grown junction crystals have been constructed and their electrical characteristics measured.” This quantity justified the claim made at Dayton that the TI silicon transistor was in commercial production.

[Pictures courtesy Smithsonian Museum]

Teal recalls the dramatic circumstances of the public announcement of the silicon transistor at the IRE National Conference in Dayton, Ohio in 1954.

“During the morning sessions, the Speakers had unwittingly set the stage for us. One after another they had remarked about how hopeless it was to expect the development of a silicon transistor in less than several years. They advised the industry to be satisfied with germanium transistors for the present. We of TI listened with great respect-and mounting exultation-because I had a handful of excellent silicon transistors in my pocket.”

Knowing this Teal realised he had not done enough to capture the drama of the moment in his prepared paper he added an addendum to the statement that Texas Instruments was already in commercial production of the silicon transistor (really only on a pilot scale). “Contrary [to] the opinions expressed in this morning’s session this will begin immediately!”

Teal then brought out a record player and demonstrated the performance of its all transistor amplifier which had one of its transistors mounted on a paddle. With a germanium transistor on the paddle he dipped it in oil at 150^OC. The sound quickly died out. He then repeated the demonstration dipping one of the new silicon transistors in the hot oil: The sound was unaffected! (A fuller account of this demonstration is given here)

Those 150 silicon transistors secured the future of Texas Instruments and gave it a permanent place in the pioneering period of semiconductors. Teal quotes Fortune magazine for November 1961:

“The silicon transistor was a turning point in TI’s history, for with this advance it gained big head start over the competition in a critical electronic product; there was no effective competition in silicon transistors until 1958. TI’s sales rose almost vertically; the company was suddenly in the big leagues.” [Teal 1976]

Postscript: “They got the silicon transistor down in Texas!”

This phrase is part of the folklore of Texas Instruments. After hearing Gordon Teal’s presentation Frank Ducat of Raytheon called his boss Norman Krim to tell him that they had been beaten in the race to produce a silicon transistor saying “they got the silicon transistor down in Texas!”

In his interview with Andrew Goldstein Gordon Teal asserted that only TI were working in silicon: “Most of them didn't know anything about silicon, except sand. Pure silicon—they didn't know beans about pure silicon.” And on why other companies might have neglected silicon Teal suggested that this was a case of unconscious incompetence: “most people did not know enough to have any brilliant thoughts about the subject. Why should you make something out of silicon when you already have germanium?” [Goldstein 1991]

But several companies were working on silicon transistors. The following month at the Semiconductor Research Conference of the IRE at Minneapolis there was an entire session on silicon: RCA described work on their SX-152 silicon alloy junction transistor; Raytheon their silicon grown junction transistors; Bell Laboratories presented three papers on silicon rate grown transistors, silicon alloy transistors and on wide area junctions by diffusion from the gas phase and Hughes Aircraft on alloy junction transistors. [Kurshan 1954] In addition TI presented their silicon grown junction transistor disclosed the previous month at Dayton, Ohio.

While there is no doubt that Texas Instruments were the first to announce and commercialise silicon transistors Morris Tanenbaum of Bell Laboratories claims to have beaten Teal by a few months. At Bell Morris Tanebaum commenced a programme on silicon transistors in 1953 using at first the crystal pulling and double doping technique Bell had used in its early germanium junction transistors. This failed because the quality of the silicon used gave insufficient minority carrier lifetime even across the thinnest base widths they could create (as little as one mil). So Bell tried the rate growing technique invented by Robert Hall of General Electric [Hall 1952] which gave even thinner base widths eventually enabling a viable transistor to be produced early in 1954. [Tanebaum 2008] Other companies made earlier announcements: Philco made a silicon surface barrier transistor which was reported in Electronics for February 1954 and a Raytheon silicon transistor was reported in Electronics for March of the same year.

Grown transistors were too expensive to make and silicon alloy transistors too difficult to make reliably to be long term survivors. The most far sighted company of all in the field was Bell Laboratories which recognized that the only diffusion could meet future performance needs and single-handedly invented just about all the technologies needed for the production of high performance silicon transistors and integrated circuits. By 1955 Bell had produced the first silicon double diffused transistor. [Tanenbaum 2008]

Launch advertising in the June 1954 Electronics magazine of the Texas Instruments Type 900 Silicon Grown Junction Transistor (Courtesy Joe Knight)

First Commercial Silicon Transistors

The 900 was given its industry launch at Dayton Ohio by Gordon Teal and advertised the following month (for example in Electonics above). Picture of the 900 courtesy Joe Knight. This was followed by the 901 and 902. The can was soldered to header for hermetic seal. [Smithsonian TI G00006]

The type X-15 was an experimental device designed to increase the rate dissipation of the silicon grown junction transistor from 150 mw up to 1W. The first X-15s were filled with silicone oil and featured a strap style heat sink. In production the oil was changed to an alumina filled silicon compound and the X-15 became the 951 introduced late in 1954 and was the first TI medium power silicon transistor. Its structure was similar to the 901 transistor and the greater dissipation was achieved through the improved heat transfer. [Smithsonian G00127 G00128 Electronics 1954]

Late in 1954 TI replaced the 901 and 902 with the 903, 904 and 905 transistors. They retained a maximum power dissipation of 150 mw at 25C and were graded according to increasing gain (30.5 dB to 36.5 dB) and increasing frequency performance (frequency cut-off 4 Mhz to 6Mhz). [Smithsonian G00337 and TI bulletins DL-S 559-562 September 1955 1954, Electronic Design 1954]

The medium power range was extended with three transistors in the series 951-953 which had similar characteristics other than their maximum collector voltage being 50 volts, 80 volts and 120 volts respectively [Bulletins DL-S 563-565 August 1956] At the same time TI released the 2N243 and 2N244 with characteristics similar to the 951 but featuring a tighter spread of beta permitting closer control of circuit design. The 2N244 offered a higher gain. [Bulletins DL-S 639-640 August 1956]

Typical early silicon types (courtesy Joe Knight)

2N117 and 2N118

These transistors were released in June 1956 when provisional datasheets were published. They are very similar to the 903 and 904 and were sorted into groups according to gain with the 2N117 having a beta of 9-20 and the 2N118 having a beta of 18-40. Both transistors were commercial versions of the US Navy USN-2N117 and USN-2N118. [Texas Instruments draft data sheets June 1956 (no DL number)]. The can was a welded flanged can with glass lead to header seals.

Silicon Grown Junction Power Transistor: X-36 970 and 2N122 Series

Grown junction transistors were not good power transistor candidates because the short length of semiconductor between the collector-base junction and the collector termination to the header had a high resistance and associated ohmic heating. Thus heat transfer from the collector was inefficient.

The device was first developed in 1955 as the X-36 for a servo-amplifier by increasing the cross section of the bar to 125 mils square giving nearly 5 times the cross sectional area of the first silicon transistors and soldering the collector end to a copper header. This was developed into the 970 which was the same as the X-36 except for the package. The emitter was grounded through the case with the direct connection improving thermal cooling. The JEDEC registration of the 970 is 2N122. These transistors had a rating of 8.5W at room temperature [Smithsonian G00040 G00046]

Date Codes

The six letter date codes starting with “980” combine the TI Federal Stock Number (980) followed by the date code in the conventional format eg “715” means the 15th week of 1957 [Smithsonian G00339]

2N332

The 2N332 – 2N336 series date from 1957 and were the first in the industry standard TO-5 package [Bulletin DL-S891 march 1958] The first of these, the 2N332 was identical to the earlier 903 and 2N117 differing only in their packaging. The series 2N332-336 represents increasing beta.

[Smithsonian G00010 G00269]

2N1149 Series

In 1957 the 900 series were replaced by JEDEC types

Silicon Diffused Mesa Transistors

Bell Laboratories pioneered the development of silicon diffused technology from 1954 as part of a major programme of work to develop a PNPN switch for use in telephone exchanges. By 1956 it had developed most of the key technologies that would underpin the production of semiconductors for many years to come. In January 1956 it ran a major symposium for its licensees in order to transfer its diffusion technology.

At Bell Tanebaum made NPN silicon transistors by starting with N-Type silicon and simultaneously diffusing donors and acceptors to create a P-Type base layer and an N-Type emitter. [Tanenbau663m 1956]

Silicon Production

High purity silicon was needed for the production of silicon transistors. Dow supplied the industry early on but the purity of its silicon was problematic. Haggerty decided that TI would invest in its own silicon production as part of the company strategy to produce silicon transistors.

At the 25^th anniversary celebrations of the silicon transistor and Regency TR-1 radio, Ray Sangster recalled TI’s rationale for producing its own high-purity silicon: “The properties of semiconductor devices are ultimately completely limited by the chemical composition of the semiconductor crystals from which they are fabricated. Failure to control internally, within the company, our supply of semiconductor quality silicon would have meant that we were deliberately, as a company, leaving control of a vital aspect of our technology to outside suppliers.”

During World War II Gordon Teal produced pure silicon by high temperature reduction of silicon tetrachloride at Bell Laboratories and used this material to make microwave diodes. At TI Teal inititated work on high purity silicon in 1954. Process work was led by Willis Adcock and James Fischer who made high purity silicon tetrachloride and then reduced this to polycrystalline silicon. TI became a significant supplier of silicon to the semiconductor industry Sangster noting “we used to joke that our goal was to put DuPont out of business. We did put them out of that business.” [Teal 1976 Texas Instruments 1980]

First Silicon Diffused Transistor: the 2N389

Despite Bell’s leadership role TI was first to market with a diffused silicon transistor: the 2N389. Its development was led by Elmer Wolff collaborating with Robert Anderson and Boyd Cornelison. Wolff followed the method developed by Fuller and Tanenbaum of Bell.

The method exploited the fact that in silicon the diffusion rate of P-Type impurities is up to 100 times faster than N-Type impurities. Thus with the right choice of impurities and their concentrations a P-Type subsurface layer could be formed because:

The N-Type impurity had lower diffusion but the higher surface concentration and

The P-Type impurity diffused in faster but had the lower surface concentration.

Wolff’s work was presented by him at the August 1957 Western Electronic Show and Convention held that year in San Francisco in a paper entitled “Diffused Fifty-Watt Silicon Power Transistors” [Proc IRE 1957, Wolff 1957] His description of the process was quite sketchy saying only "For assembly of a transistor, we choose such a slice and alloy an aluminum ring into one surface, remove the back side of the wafer and make appropriate contacts to the emitter, base and collector. Then by selectively etching we have the structure of Figure 9." [Wolff 1957]. A summary of a similar kind is given in his lab book in entries for May and June 1957 (Courtesy E Wolff).

Wolff’s transistors were made by taking wafers of N-type silicon, dicing them and using simultaneous diffusion to produce an additional diffused NP layer on all sides of each dice. Then the die were etched on one side to remove the NP layer creating the NPN sandwich.

An aluminum base ring was alloyed into the silicon die. This followed the Tanenbaum method of “simply melting the aluminum wire directly through the heavily doped N-type layer that covered it.” The wire did not short the emitter-base junction because as the wire solidified, it created an isolating p-n junction with the emitter surface layer. [Tanenbaum 2008]

2N389 transistor courtesy Joe Knight

The rings were made from very soft aluminum wire by winding it on a mandrel then cutting through the coil lengthwise with a razor, making many C-shaped rings. These were placed on the transistor die around the emitter connection and furnace alloyed into the surface. This was not a pretty approach which Wolff describes as “a brute-force production technique.”

After the base ring was alloyed into the die, it was electroless nickel plated all over in order to create the required ohmic contacts for the emitter and the collector. The die was masked and the nickel etched off leaving only the collector connection on the back of the die and an emitter connection within the aluminum ring. Lastly the top of the transistor was masked out to the aluminium ring and the bottom similarly and the sides etched back to create the final transistor. [Millis 2011 Citing Wolff] Full details are given here.

2N389 transistors courtesy Joe Knight

Cross section of the 2N389 from Wolff 1957

Reflecting on the events at WESCON Wolff recalls: “When I presented the 2N389 at WESCON there was considerable response! As it happened I was preceded by a man from Bell Labs who presented a paper in which he calculated that a silicon device could not be built that would carry 10 amperes.

I did therefore get many questions and comments from the audience when I showed pictures of finished devices that were immediately available for sale.”

2N1047

Following the success of the 2N389 TI introduced a lower powered diffused alloy transistor, the 2N1047 early in 1958. This was a stud mounted transistor made by a similar process:

Slice N type wafers which become the collector

Create the base layer by diffusion

Dice (cavitron) into 150 mil wafers

Emitter formed by alloying on an aluminium ring

Create mesa and mount

[Smithsonian G00028]

Grown Junction Tetrode

The Tetrode was invented at Bell by Wallace, Schimpf and Dickten and a description of some of their early work published in 1952. These were grown junction transistors where a fourth electrode (b₂) was fixed to the base layer opposite the normal base contact (b₁) and given a negative bias (in the case of an NPN transistor). This creates a reverse bias with respect to the emitter over most of the emitter-base junction restricting normal transistor action to a region very close to b₁ reducing the effective cross section of the transistor and reducing input and output capacities and base resistance thereby increasing high frequency performance. [Wallace 1952]

TI first developed a germanium tetrode, the type 501 or 3N25 around 1956. It had a cut-off frequency of 200 Mhz and a maximum frequency of oscillation of 250 Mhz.

Silicon grown junction tetrodes were also developed, the first being the 3N28. Between 1954 and 1957 work on silicon tetrodes was sponsored by the US Army. In 1954 the tetrode was the only known approach to obtain VHF performance although over the period of the contract tetrodes became obsolete due to the development of diffused base transistors in germanium and silicon.

As the work proceeded TI released non-military specification silicon tetrodes (Types 924-927) in early 1956 with a specified power gain at 4.3, 12.5, 30 and 70 MHz.

The corresponding JEDEC type numbers for these transistors in a TO-12 package outline were issued in 1957 as the 2N32-5. TI registered these tetrodes but commercialized only the 3N34 and 3N35 completing their registration in 1958.

Germanium Tetrode 501 Unregistered types 3N26 and 3N27 Silicon Tetrode 926 Courtesy Terry Hosking

The contract also required some development of the package outline and in the course of this work TI settled on the TO-12 for tetrodes. This became an industry standard outline.

Below is example advertising for the 3N35 silicon tetrode in 1959.

Advertisement from Proc IRE from February 1959 Courtesy Joe Knight

Germanium PNP Mesa Transistors

The diffused base transistor was first developed by Lee of Bell Laboratories and featured a diffused base and an alloyed emitter. [Lee 1956]

As a licensee TI had access to this technology and Elmer Wolff, Boyd Cornelison and Mark Shepherd visited Bell Labs to learn the process for making them.

Thus TI made transistors of a similar kind to the Lee transistors fabricating multiple transistors on a single wafer. TI started with a wafer of P-type germanium that would become the collector. Then an N-type base layer was diffused in and lapped off from the collector side. Next multiple pairs of parallel metal stripes were evaporated onto the surface: N doped gold to make the base layer connection and aluminium that was alloyed into the base layer to convert it the P-type emitter region. The smallest area around the stripes was masked and the exposed N layer etched away to create a transistor of the smallest possible size: a strategy that favoured high frequency performance. Full details are given here.

The resulting area of the base collector junction dictated the capacitance that coupled the emitter, across the base junction, to the collector and therefore limiting the frequency capability of the device. Minimising the cross section of the mesa improved frequency performance.

The first germanium diffused mesa transistor developed at TI was the 2N623. It had an alpha cut-off of 90Mhz and was intended for television IF service (typically 43Mhz) as demonstrated in Webster’s TV prototype and other VHF applications. The 2N623 had a maximum frequency as an oscillator of 200Mhz and was packaged in a TO-5 can. It was advertised in the IRE journal for July 1958 as shown below. [Proc IRE 1958]

Advertising for the 2N623 courtesy Joe Knight

Within a year TI had improved the performance of their diffused base transistors and launched the 2N1142 series featuring an impressive alpha cut-off of 750Mhz and 750 mw power. The transistors in the series were selected for minimum current gains of 12dB, 10dB and 8dB at 100Mhz for the 2N1141, 2N1142 and 2N1143 respectively.

When these transistors were introduced the 2N623 was withdrawn. The 2N623 was never formally registered and is missing from the JEDEC registrations.

Advertising for the 2N1141 series courtesy Joe Knight

Germanium Diffused Base Switching Types

The following types were made for high speed switching:

2N705 June 1959

2N710 June 1959

2N725 Military TO-18 case

2N711

The major customer for the switching types was IBM who had their own specification that was difficult to produce. In 1960 while IBM pressed TI to provide such transistors in 100,000 lots yields of good transistors were just a few percent. TI could not meet demand without radically changing its production methods and it could not afford to let its prime account down. There was little choice and TI switched to a relatively unproven process. [Millis 2000]

Demonstration Portable Television

In 1957 Haggerty saw an opportunity for an all transistor TV based on advances by TI in high frequency transistors: the 3N25 could be used in the front end and the new 2N623 in the IF strip.He assigned his best engineer, Roger Webster to lead the development. Reminiscent of the approach taken three years earlier to develop the TR-1 Webster recalls he and Harry Cooke “bought a 9-inch vacuum tube set manufactured by GE, gutted it, took the picture tube to one of the local picture tube rebuilding houses and got them to put in a special 12-volt 150 mil filament gun on it. Now working with Boyd Cornelison and Elmer Wolff who were designing some germanium mesa transistors that had a much higher frequency capability than grown junctions did for these things.”

The set that Webster bought was a 9T002. The picture tube required 6500 volts and there was a major problem with the EHT rectifier: here a 1V2 vacuum tube was used.

Webster developed two tuners. One version used two 2N623 diffused base transistors for the oscillator and mixer. The second used three 3N25 tetrodes adding an RF amplifier stage giving 10dB more gain. The 44.5Mhz IF strip also used five neutralised 2N623 in common base mode. Webster says he carried the set all over the country in 1958 creating good publicity for TI (for example, in Electronics Design Strategy News for September 1958). But the idea was ahead of its time and makers of TV portables were reluctant to move to transistors.

[Wolff 1985, Radiomuseum, Electronics Design Strategy News 1958, Smithsonian G00086 picture of the General Electric 9T002 retrofitted with transistors courtesy Smithsonian Institute]

Texas Instruments Data and more Information

See the link to my data file. There are links in the left hand column that lead you to further published information.

References

Adcock W, Jones M, Thornhill J, Jackson E 1954 Silicon transistor Proc. IRE. 42. 1192

Choi H 2007 The Boundaries of Industrial Research Making Transistors at RCA 1948-1960 Technology and Culture 48 758-82

Cornelison B et al Materials and Methods of Manufacturing Solid State Devices US Patent 3076253 filed March 10^th 1955 Issued February 5^th 1963

Cornelison B Adcock W Transistors by Grown-Diffused Technique Proc 1957 IRE WESCON Convention Record Part 3 22-7

Davis P Kleiner D 1993 The Breakthrough Breadboard Feasibility Model: The Development of the First All-Transistor Radio The Southwestern Historical Quarterly

97 No. 1 (Jul 1993) 56-80

Electronic Design 1954 Silicon Transistors Offered in Five Types December 1954

http://people.msoe.edu/~reyer/regency/ti_silicon.jpg

Electronics 1954 August 302

Electronics Design Strategy News 1958 Self-contained TV receiver uses 24 transistors EDN September 1, 1958 http://www.edn.com/article/463621-Self_contained_TV_receiver_uses_24_transistors.php

Goldstein A 1984 “Finding the Best Material: Gordon Teal as Inventor and Manager,” in Sparks of Genius: Portraits of Electrical Engineering Excellence, ed. F. Nebeker, IEEE Press, 1984.

Goldstein 1991 Gordon K. Teal, Electrical Engineer, an oral history, IEEE History Center, Rutgers University, New Brunswick, NJ, USA.

Hall R 1952 P-N Junctions Produced by Growth Rate Variation Phys Rev 88 139

Kroemer H 2001 Quasi-Electric Fields and band Offsets: teaching Electrons New Tricks Rev Mod Phys 73 783 - 93

Knight J 2007 Texas Instruments 1950s/60s Germanium Power Transistors

http://www.semiconductormuseum.com/Transistors/LectureHall/JoeKnight/JoeKnight_EarlyPowerTransistorHistory_TI_Index.htm

Knight J 2007 Texas Instruments 1950s/60s Germanium Power Transistors

http://www.semiconductormuseum.com/Transistors/LectureHall/JoeKnight/JoeKnight_EarlyPowerTransistorHistory_TI_Index.htm

Kurshan J et al 1954 Report on the Conference of Semiconductor Devices Research Minneapolis MN June 28-30 1954 RCA Technical Report PEM-452

Lee C 1956 A High Frequency Diffused Base Germanium Transistor Bell Sys Tech 35 23-34

McGarrah R Sonotone Transistor Hearing Aids

http://www.transistorhistory.50webs.com/sono1111.html

Millis E 2000 TI, the Transistor and me Ed Millis Books, Dallas

Millis E 2006 Cornelison Crystal Puller Sketch 14^th March 2006 Unpublished

Millis E 2011 Personal communications May 2011

Morton D 2000 John Saby, Electrical Engineer, an oral history conducted by David Morton, IEEE History Center, Rutgers University New Brunswick, NJ, USA www.ieee.org/organizations/history_center/oral_histories.html

New York Times 1948 The News of Radio 1 July 1948

Pies D 1991 Regency's Development of the TR-1 Transistor Radio http://www.regencytr1.com/Regency_Early_Years.html

Proceedings of the IRE 1954 New PNP germanium 90 Mc alpha cut-off diffused base transistors July 1954 edition

Proceedings of the IRE August 1957 IRE News and Radio Notes 45 issue 8 1153-60

Pirtle C 2005 Engineering the World: Stories from the First 75 Years of Texas Instruments. Texas Instruments Incorporated 1ed 2005

Radio Museum http://www.radiomuseum.org/r/texasinstr_prototype_portable_tv_sol.html

Riordan 1997 Crystal Fire]

Tanebaum M Thomas D 1956 Diffused Emitter and Base Silicon Transistors Bell Sys Tech 35 1-22

Tanenbaum 1999 IEEE History Center: Morris Tanenbaum Abstract

Tanebaum M 2008 First-Hand: Beginning of the Silicon Age IEEE Global History Network August 2008

Teal G 1976 Single Crystals of Germanium and Silicon-Basic to the Transistor and Integrated Circuit IEEE Trans Electron Devices ED-28 7 621- 39

Texas Instruments 1954 Press release May 10^th 1954 New transistor will shrink electronic "brains"http://www.ti.com/corp/docs/company/history/timeline/semicon/1950/docs/54commercial-brain-vignette.htm

Texas Instruments 1980 Proc 25^th Anniversary Observance of the Transistor Radio and Silicon Transistor, March 17^th 1980

Wallace R Schimpf L Dickten E 1952 A Junction Transistor Tetrode for High Frequency Use Proc IRE 40 1395-1400

Ward 2001 Bill Brower The Art and Science of Building the First Commercial Silicon Grown Junction Transistors

http://www.ck722museum.com/history/Transistormuseum/LectureHall/Brower/Brower_Index.htm

Ward J 2002 Oral History of D D McBride taken January 2002 http://www.semiconductormuseum.com/Transistors/TexasInstruments/OralHistories/McBride/McBride_Index.htm

Ward J 2001 Oral History of Elmer Wolff August 2001

http://www.semiconductormuseum.com/Transistors/TexasInstruments/OralHistories/Wolff/Wolff_Index.htm

Wolff E and Anderson R 1957 Diffused Fifty-Watt Silicon Power Transistors Proc IRE 1957 WESCON Record Part 3 40-7

Wolff M 1984 Richard Koch: An Interview Conducted by Michael Wolff, IEEE History Center, 10 December 1984

Wolff M 1985 The Secret six-month project IEEE Spectrum December 1985 64-9

Wolff M 1985 Roger Webster, an oral history conducted in 1985 by Michael Wolff, IEEE History Center, New Brunswick, NJ, USA http://www.ieeeghn.org/wiki/index.php/Oral-History:Roger_Webster