History of Philips’ Semiconductors in the 1950s
Copyright Mark P D Burgess 2009
Phiilips was founded in 1891 by Gerard Philips in Eindhoven, the Netherlands. Early on its products included incandescent bulbs which led it to begin producing vacuum tubes initially under contract in 1917 and then in its own name from 1919. [Tyne 1977] The company quickly became a major European tube and radio receiver manufacturer. Later it branched out into consumer electrical and electronic products.
From 1914 its research and development was carried out at the Eindhoven Natuurkundig Laboratorium usually abbreviated to Natlab. This is where the history of Philips’ semiconductor developments begins.
In the 1950s Natlab was comprised of three sections each with a Director:
Physics (vacuum tubes and microscopy) directed by Casimir
Chemistry (materials, semiconductors and transistors) directed by E Verweij
Engineering (applications) directed by Rinia
Philips had formed a dedicated solid state physics group within NatLab in the 1930s. It worked on the known semiconductors of the time: copper oxide and selenium and developed a selenium diode that went into production at the Electron Tubes product division. But it had no success with its research in solid state amplifiers and field effect devices. However, this group provided an important platform and the focus as Philips positioned itself for the semiconductor era.
Bell Laboratories announced its point-contact transistor in June 1948: “An amazingly simple device, capable of performing efficiently nearly all the functions of an ordinary vacuum tube, was demonstrated for the first time yesterday at Bell Telephone Laboratories where it was invented.” The press release was coupled with three short papers by Bardeen, Brattain, Shockley and Pearson in the letters to the Physical Review for June 1948 of which the first of these set out in the briefest terms the geometry, performance and theory of the device. [Bardeen 1948]
The importance of this breakthrough was not lost on Philips. For example, after reading the article, Verweij said to Hazeu, who was the commercial director of the Electron Tubes product division: “The content of this article will shake your department to its foundations!” [cited from Davids 2007]
As Bell Laboratories had a pioneering patent position on the point-contact transistor, Philips needed a license to the Bell patents if it wished to enter the transistor market. Short term it had a more pressing problem: it had neither the technologies nor the organisation to enable rapid development. The latter took four years of work. Philips had a point-contact transistor four years after Western Electric, the Bell Laboratories associate company and manufacturing arm of AT&T, made its A Type available in 1949. This reflects the disadvantage of European companies compared to their US counterparts. US government war time research contracts gave companies such as AT&T, General Electric and RCA a huge head-start post war while their European counterparts were recovering from the war time shut down.
Bell Laboratories’ License
In 1947 Philips signed a cross licensing deal with Bell Laboratories securing a royalty free non exclusive license to Bell’s patents in the fields of telephone, appliances, systems and tubes. In return, Bell obtained a license to Philips’ Ferroxcube ferrite material. [Davids 2007] But this neither anticipated nor covered the principal applications of the point contact transistor. Cross licensing arrangements of this kind by which rights to substantial patent estates were swapped for little or no royalty were relatively common, particularly with US electronics companies. [Tilton 1971]
Bell, through Western Electric, its commercialization associate, did not begin to issue licenses to its transistor patents until late 1951, after the Shockley junction transistor patent issued.
There were three imperatives that dictated Bell’s licensing strategy: Firstly Western Electric knew it could not exploit the transistor in all its potential applications. The rudimentary device badly needed diversified development by many companies in order to be successful. Western Electric was accustomed to a liberal licensing strategy: But the Military intervened. The transistor had such an extraordinary military potential that for a time the Military considered classifying it entirely. But comprehensive applications information was withheld until September 1951 when Bell ran a symposium for government and institutional engineers and scientists. After this event Western Electric was permitted to grant manufacturing licenses to NATO countries. Manufacturing information remained classified and restricted to licensees. Lastly the Bell Laboratories parent, AT&T, was subject to an anti-trust suit brought by the US Department of Justice which sought to break down its monopolies. This encouraged a liberal attitude to licensing issues.
Scant design details were disclosed in the Bell Laboratories 1948 publications in Physical Review: just two short paragraphs in the first of these [Bardeen 1948] which assumed manufacturing knowhow with respect to point-contact diodes, something Philips did not have at that time. The transistor was described as having two electrodes of tungsten or phosphor bronze “of the point-contact rectifier type” placed in close proximity (2-10 mils) on the upper surface of a block of germanium. A third electrode had a large area low resistance contact with the base.
The N-type germanium was “prepared in the same way as that used for high back-voltage rectifiers” and had a resistivity of 10 Ωcm. The germanium was “ground and etched in the usual way” and lastly “the collector point may be electrically formed by passing large currents in the reverse direction,” a process also used in the production of some radar diodes.
As other manufacturers such as RCA found, it was relatively easy to make a point-contact transistor by adding a second point to a commercial point-contact diode. But to produce transistors de novo required technologies for semiconductor purification, doping and surface treatment, point contact design and placement and experience in the black art of “forming.” To obtain a useful transistor for any specified application, the laboratory needed to understand how to optimize noise, gain, frequency response, stability and dissipation. And a design was needed that lent itself to reproducible manufacturing.
In 1948 no such assistance was available from Bell Laboratories. Philips had to find its own solutions. It responded by reorganizing and looking to the United States for know-how. Wisely it chose to emphasize work on semiconductor diodes which offered immediate markets and built on its experience in producing and selling selenium rectifiers.
The semiconductor group supervisor, P W Haaijman, was given more resources and initiated a germanium programme. At the end of 1948 a second group dedicated to transistor research was formed under the direction of F H Stieltjes. L J Tummers joined the transistor group in 1949 and played a significant role in later developments including making Philips’ first transistors.
Early on resources were modest: By 1952 the transistor group had 10 persons and by 1955 it had grown to 35. [Davids 2006]
Natlab began work on point-contact diodes as information on the World War II development of radar detectors, largely in the United States, came into the public domain. The United States radar programme was a major war time effort coordinated by the Radiation Laboratory at MIT on all aspects of radar of which one element was crystal detectors. Participating institutions working on radar detectors met every two months throughout the war. Regular participants were the Radiation Laboratory, the Universities of Pennsylvania and Purdue and General Electric with less frequent participation by the Naval Research Laboratory, Bell Laboratories, Westinghouse, Du Pont and Carnegie Tech. At the end of the war the results of the programme were published in 28 volumes including Crystal Rectifiers, a substantial treatise on radar diodes by Henry Torrey and Charles Whitmer. [Torrey 1948]
Initially Philips tried to develop these using silicon but this was unsuccessful. They may have been influenced by one of the earliest publications post war that describes work by Scaff and Ohl at Bell Laboratories on the use of silicon in radar detectors. [Scaff 1947]
E Verweij who was Director of Chemistry at Natlab visited the United States and returned recommending that the research should be redirected to germanium. [Davids 2007] This proved successful and the germanium point-contact diode was launched in 1950.
Philips used hermetically sealed construction. The pellets for each diode were molded by fusing purified germanium powder in multiple well quartz molds to produce pear-shaped polycrystalline balls. These were copper plated, soldered to a stud, ground down in lots of 50, etched, washed and mounted. Tungsten used for the whisker was formed and welded by machine.
“The Germanium itself is of extremely pure quality, but contains a very small and accurately controlled proportion-about one part in ten million-of impurity, the chief of which is arsenic, and in some cases antimony.
The metallic catswhisker is of tungsten wire, and crystal and catswisker are enclosed in a glass capsule, and after adjustment are firmly secured in position so that movement is impossible.” [Mullard 1952]
In 1953 two types were produced distinguished by the conductivity of the germanium used:
OA70 using higher conductivity germanium designed as a video detector
OA50, OA51, OA53, OA55, OA56 and OA61 selected and graded from production
Diodes were produced at the Electron Tubes division which in 1951 formed a dedicated unit for semiconductor production at Eindhoven headed by J van der Spek. In October 1953 the unit moved to Nijmegen where the development laboratory was located. At this time Philips were selling a million diodes annually and were the major supplier in Europe. [Herold 1953]
Picture left: two Diodes sold by Philip’s subsidiary company, Mullard showing the encaspsulation style used from 1952. [Mullard 1952 and Mullard 1954 courtesy Jon Evans. Below:click on these icons for full sized images]
Philips first goal was to recreate the Bell Laboratories point-contact transistor.
Natlab developed a copy of the Bell Laboratories A type and early in 1952 transferred development to the Electron Tubes division laboratory. Radiomuseum gives the release date as March 1952. In the UK the new transistors were publically announced much later being reported in Wireless World in February 1954 [Toute la Radio 1954]
The transistors were:
OC50 general purpose amplifier
OC51 switching [Photo courtesy Arnaud Cramwinckel]
Philips described its point-contact transistors in general terms as follows: “For the point-contact transistor a small slab of N-type germanium consisting of a single crystal and having a volume resistivity of 5 Ωcm is taken as the starting point. The dimensions may be 2mm x 2mm x 1mm. One of the larger surfaces of this germanium slab is soldered on to a metal carrier and thus forms the base. Two thin metal wires provided with sharp points are placed at a distance of approximately 0.2mm against the opposite surface, which has previously been smoothed and cleaned by polishing and etching. These “whiskers” must obviously be kept in position very firmly. The forming for obtaining the desired transistor action is then achieved by temporarily feeding electric current to the point contacts. The transistor is mounted in such a way as to protect it against vibration and atmospheric influences to which it is obviously very sensitive.” [Ploos van Amstel 1954]
The stated spacing of 0.2mm or 8 mils is very wide. Edward Herold, Director of the Radio Tube Laboratory at the RCA Princeton Laboratories, visited Natlab and the Electron Tubes Division Laboratory in 1953 and noted in his report: “The point-contact transistors have 2 mil point spacing and are close copies of the BTL type A; they had little of an original nature so far as I could see.” [Herold 1983]
By 1953 Philips was producing 1400 point-contact transistors per month. But these transistors shared the frailties common to those of all producers. They served to advance applications research, in computing for example, but were not suitable for use in consumer or industrial products. Richard Grimsdale worked on the first transistor-based computer at Manchester University and recalled issues in obtaining supplies of point-contact transistors: “There were other sources: the GET2 from GEC and the OC51 from Mullards. It was necessary to test the transistors on arrival: in the earlier batches, up to half did not work at all, while many of those that did work had very varied characteristics.” [Grimsdale 1995]
In 1954 Wireless World carried an article that exhibited their more robust performance. It described an experimental amateur transmitter using an OC51 transistor in negative-resistance mode at the limits of its powers: It reported an output of 100mw at 1.8MHz and that signals were received at distances of up to 30 miles. [Cockle 1954]
The Junction Transistor
The Junction transistor was invented and patented by Shockley at Bell Laboratories in 1948 but it took until 1951 for Bell to make a viable prototype: The grown junction transistor. It presented this work at the IRE Electron Devices closed conference at Durham NC in June 1951. At the same conference Saby of General Electric presented his alloy junction transistor. Junction transistors were quickly recognized as having more potential than point-contact transistors even though those presented at Durham were relatively undeveloped. For example, it took several years before production junction transistors could be made to perform well at radio frequencies.
Bell laboratories 1952 Symposium
Philips obtained a license to the Bell Laboratories transistor patent estate and accordingly was entitled to attend the 1952 Bell Licensees Symposium where they were introduced to the grown junction transistor as well as many of the key process technologies in relation to this and the point-contact transistor.
Philips sent a team of four: F H Steiltjes, P W Haaijman and J S van Wieringen attended from Natlab, with the Director of the Electron Tubes development laboratory, J C van Vessem. [Davids 2006]
Other licensees attending from Europe included GEC, Telefunken, Siemens, Ericsson and BTH. [Lojek 2007]
The Bell Symposium covered semiconductor materials science, transistor fabrication and testing. Participants received a two volume reference set entitled “Transistor Technology” that was classified and disclosed production knowhow not previously available. Its contents were: Technology of Germanium Materials; Preparation of Single Crystals; Principles of Device Fabrication; Principles of Device Characterization; Design for Manufacture; A Manufacturing Procedure. [Bell 1952]
Bell produced their new junction transistor by double doping a germanium melt while pulling a single crystal: hence described as the double doping method.
To create an NPN transistor, for example, this method involved pulling a single crystal from an N germanium melt. Midway, gallium doped pellets were added to the melt converting it to P-type overall and a P-type layer was grown. A few seconds later arsenic flakes were dropped into the melt converting it back to N-type and the rest of the crystal was grown creating a large scale NPN structure. Individual transistors were cut as tiny bars from the crystal. Transistors made this way were called “grown junction transistors.” Given tight production control the width of the base layer was determined by the doping interval. [Bridgers 1958] The double doping method was the first and crudest of the grown junction form which was further developed by General Electric and Texas Instruments.
Philips Grown Junction Transistor
Work at Natlab on grown junction transistors began after the Bell Laboratories Symposium in 1952 exploring Bell’s double doping technology. Philips quickly acquired the resources needed for single crystal preparation and the double doping process and by the end of 1952 had made 100 transistors. Philips then sought to transfer development responsibility to the Electron Tubes division laboratory but this proved more problematic than expected.
The junction transistor brought challenges: new technologies needed to be implemented at Eindhoven for both the laboratory and to facilitate production. These problems “illustrate how different preferences and lack of coordination between the product division and Natlab researchers slowed innovation at Philips.” [Davids 2006]
In addition the Electron Tubes Laboratory Director, Van Vessen was not convinced of the viability of the double doping approach. Controlling the base width, detecting the position of the base layer in the crystal, cutting out transistors and making a connection to the base layer were demanding and largely manual processes. Performance was poor: such grown junction transistors were only viable as AF amplifiers due to their relatively wide base region. Subsequently, improved methodologies by other companies such as General Electric enabled grown junction transistors to become the first viable radio frequency transistors but this was not apparent in 1952.
Philips Alloy Junction Transistor
Van Vessem through his relationship with RCA knew that RCA was developing an alloy junction transistor. [Davids 2006] RCA had decided this after Edward Herold, Director of the Radio Tube Laboratory at the RCA Princeton Laboratories, attended the IRE Electron Devices closed conference at Durham where General Electric described their alloy junction approach. By mid 1952 RCA had developed its first viable prototype alloy junction device, the TA-153 which was marketed as the 2N34 from May 1953.
The TA-153 was made from a wafer of N-type germanium 6 mils thick. Indium dots that would create the emitter and collector were fixed to either side of the wafer and alloyed by heating the assembly in an oven under strict temperature control. On heating about 2 mils of the germanium was dissolved by the molten indium on each side and on cooling the germanium recrystallised as P-type creating a PNP structure between the indium dots. The residual N-type layer now 2 mils thick became the base.
Van Vessem agreed with the assessment by RCA that the alloy junction approach was the best option and history proved this to be the case: alloy junction transistors dominated World transistor production through the 1950s.
Thus Eindhoven decided to adopt alloy junction technology and sought a license from RCA. Philips wanted to extend the exclusive license it had from RCA to include transistors but anti-trust concerns at RCA inclined it to reject exclusive relationships. Negotiations were complicated by Philips demanding its standard terms from RCA for a Ferroxcube license. Negotiations dragged on past November 1952 when RCA held a Symposium on transistor production and applications for its licensees. [Davids 2007]
Philips did not obtain the benefits of attending the RCA symposium but had the right to receive RCA confidential Licensee Bulletins. [Choi 2007] They also received a visit from Edward Herold in October 1953.
Herold was in Eindhoven because after its successful symposium in November 1952, “RCA decided to impress its foreign licensees by sending over a team of four scientists for an extensive tour of European laboratories.” Herold represented the transistor work, Humboldt Leverenz the work on electronically active materials, such as phosphors and ferrites; Albert Rose the work on photoconductivity; and Robert Janes (of RCA’s Lancaster operation) represented camera tubes and photo devices. [Herold 1983]
He reported some observations on Philips’ developments: “Alloy junction transistors are being made in some quantity. Slicing and dicing is done by a set of parallel 5-mil wires which move back and forth using carborundum abrasive. The indium is cut and weighed and, before using, melted into tiny spherical balls. Although all laboratory indium is zone purified, I am not sure if this is done at the factory as well, but believe it likely…The Philips manufacturing group expect to emphasize junction transistors at the expense of point-contact transistors because they believe the former have much greater utility.”
OC10 OC11 and OC12 Transistors
The OC10-12 series were the first junction transistors released by the Philips Group for which the earliest documented release date is March 1953 [Radiomuseum]
A press release by Mullard in March 1954 gives informative details:
Three types of junction transistor, the Mullard OC10, OC11 and OC12 are now available for circuit experiments. In the past, the lack of supplies has prevented circuit designers in this country from gaining direct experience of junction transistors in their own laboratories. Now, however, the availability of the first junction types invites practical investigation into their many possible applications. As junction transistors provide no current gain when connected with grounded base, they are more usually employed in grounded emitter circuits, where they function well as AF amplifiers. In both amplifier and oscillator circuits these transistors will operate with supply voltages as low as 1.5 V and with current consumptions of the same remarkably low order. The OC11 is a general-purpose amplifier, while the OC12 is intended for operation in an output stage, although it can, of course, be used otherwise. A low-noise version of the OC11 is provided by the OC10, a special transistor for early stages in high-gain amplifiers. The OC10, OC11 and OC12 are readily available for experimental purposes at a price comparable with that of mains subminiature valves. [Cited by Harrison 1998-2008. Photo courtesy Arnaud Cramwinckel]
These were PNP alloy junction types:
OC70 OC71 and OC72
The OC10-12 series were made “available for circuit experiments” and quickly became obsolete. These transistors were never made in quantities that could sustain the production volumes needed in any successful consumer product. In 1954 they were succeeded by the OC70 and OC71 which were made for many years: the OC70 replaced the OC10 as a low noise pre-amplifier, the OC71 replaced the OC11 as a general purpose audio frequency transistor. [Kendall 1954] The OC72 was introduced in 1955 as an output transistor shown here in the original long glass case compared to the OC71 below it. [Photo courtesy Arnaud Cramwinckel]
While Philips depended on its RCA license for the core alloy junction
technology, it introduced innovations of its own: It developed an all glass hermetic seal using silicone grease to protect the active elements and improve heat transfer. While the principal alloy dopant was indium advances were made in the use of co-dopants such as gallium for improved power handling. Both these improvements were patented. In this way the capacity of the OC72 to handle greater currents was improved. Producing the OC72 was problematic and its formal release followed the OC70 and OC71. [Davids 2006]
Early versions of the OC72 were glass encapsulated in cases longer than the OC70 or OC71 (see figure) but subsequently power dissipation was improved by using a metal sheath which gave better heat transfer via a metal fin to a heat sink (such as the chassis).
The specifications for these transistors migrated over time towards higher ratings (presumably reflecting improved production quality and consistency). Early specifications were conservative:
The first power transistor produced in the Group was the OC15 released by
Valvo in October 1954 [Herzog 2001] Early versions were known as the 100 O.C. Both versions are shown. [Photo courtesy Arnaud Cramwinckel] It was rated at 3 watts dissipation in class A and at 5 watts class B output. [Knight 2007]
Despite the apparent fragility of small signal devices, power transistors of respectable dissipation were made by several manufacturers relatively early. Key features in the approach to these were to have “large area” junctions and good heat transfer to the case which were designed to fit to heat sinks. The penalty for large junction areas was control of the base width and consequently consistency and high frequency performance.
The OC15 had an elaborate stud mounting design: “three layers of steel housing, three glass relief insulators, a large insulated bottom layer, and the top-hat heat sink, which is made with resin inside with a copper heat tab.” [Knight 2007] Part of the complexity was due to the electrical isolation of the transistor from the case; a practice not followed in subsequent designs such as the OC16 where the collector is connected via the case with consequent improvement in heat transfer.
The OC16 was available from 1956 and is seen in two forms: and earlier version on the left and a later more compact form on the right. [Photo courtesy Arnaud Cramwinckel]
It was intended as a general purpose power transistor for use in audio output or driver stages, switching and DC converters.
At around this time the OC22 and OC30 were in development. The OC22 is another stud mounted design for high speed switching with a surprisingly high alpha cut-off of 2.5 Mhz. The OC30 intended for medium power output in class A and class B was notable also for its first use of a TO3 style outline. (The TO3 has a width of 25mm whereas the OC30 was 18mm wide.)
Data for the early power transistors is shown in the following table:
In a subsequent development a TO3 case was used for transistors from the OC22 onwards.
Hobby Transistors: the OC13 and OC14
Many semiconductor manufacturers branded low specification transistors for the enthusiast market which was profitable since otherwise these transistors would be scrap. Philips rebranded its OC71 and OC72 as OC13 and OC14 and announced these as part of its new product releases for the Firato over 8-15th October 1956:
“The radio amateur world reveals a growing interest in using transistors. No wonder, since these semiconductors have minimum dimensions offering special properties related to the possibility of very interesting experiments with the construction of very small devices, while the lack of a filiament is also very attractive. The price of transistors for professional purposes is usually beyond the reach of the amateur and therefore we have decided to make available two types of experimental transistors OC13 and OC14 available at prices far below those of the professional types. The technical details of these types and an application sheet some interesting transistor circuits will be available at Firato.” [Philips 1956 courtesy Arnaud Cramwinckel]
Both were audio frequency for pre-amplifier and low output power applications. No gain was specified. The outlines were the familiar glass used for the OC70 or OC71 (OC13) and metal/glass OC72 (OC14).
First Radio Frequency Transistors
The first radio frequency transistors produced by the Philips Group were the OC45 and the OC44 both announced by Philips for the Firato 8-15th October 1956 [Philips 1956 courtesy Arnaud Cramwinckel]
These were made by similar alloy junction processes in use for the audio frequency transistors but optimized for higher frequency service. This meant thinner wafers in order to obtain thin base widths and an annular base connection to reduce the base lead resistance; two important determinants of improved high frequency performance.
Although Philips was a licensee of RCA which developed similar radio frequency transistors at the same time, there were significant differences in their design: The base wafers were thin at 4 mils but this was twice the wafer thickness used by RCA for a similar transistor. The inference is that Philips used deeper alloying depths to achieve a viable base width; something that RCA found problematic due to difficulties in controlling depth and achieving planar junctions. Both companies used ring base tabs.
Whereas other Philips transistors were made from rectangular dices the OC44 and OC45 made from round wafers 1.45mm in diameter cut with an ultrasonic drill. [Mullard 1961]
In the 1970s some production was selected for specified duty: The orange spot identified a first stage IF amplifier and the blue spot the second IF stage.
Data for the new transistors is given below:
Bradley (1959) states that the characteristics for both transistors are identical with specified exceptions. This suggests that the transistors may have been made on the same line and then selected according to measured performance.
The advent of the new radio frequency transistors completed the line-up for a six transistor radio:
OC44 Oscillator mixer
OC45 Intermediate frequency amplifier (2)
OC71 Audio amplifier (2)
OC72 Class B Output (2)
Investment in research and development and production increased rapidly over the early years. By 1955 the Nijmegen semiconductor plant employed 1100 workers of which 122 were employed in the laboratory.
Resources employed at Philips Group plants are shown in the following table.
Philips’ Subsidiary Companies
“Philips diffused production of germanium alloy transistors and signal diodes throughout its European organisation. While specialising production for specific markets, Philips subsidiaries in Great Britain, Germany France and Switzerland shared a common technological base in semiconductors. In fact all the European subsidiaries were producing signal diodes. In addition, Mullard (Great Britain) and Nijmegen produced general purpose germanium alloy transistors for radios while the Swiss subsidiary produced semiconductors for watches.” [Malerba 1985]
Philips’ major subsidiaries with common semiconductor production in the 1950s were:
Valvo, West Germany
Mullard, Great Britain
Radiotechnique Compelec, France
All these companies were acquired because of their tube businesses. In the following sections some information is given on the transistors they either made or imported from other manufacturers in the Group and re-branded. It is not surprising that geographically close Group companies might have all been independently manufacturing: Production of transistors was an intensely manual process and in the early years were always in short supply. Philips also produced transistors in Australia from 1959 along with three other manufacturers (Amalgamated Wireless, STC and Ducon) illustrating the viability of production in even quite small markets.
Valvo GmbH began as a transmitting tube company in 1924. They were acquired by Philips in 1932.
They announced their first transistors in 1954: the OC50 and OC51 well after Philips had prototyped them in 1952.
1954 was the year of the junction transistor in Germany. Valvo announced the OC70 and OC71 at the Hannover Industrial Fair in April. Technologically they were similarly placed to Telefunken who exhibited their OC601 and OC602 PNP types. Siemens announced their TF70 and TF71 NPN audio frequency types.
In October 1954 Valvo introduced the OC15 power transistor. At the Hannover Fair of 1955 the OC72 was on show for the first time. While Telefunken exhibited a transistor radio with selected OC602 AF transistors pressed into front end service there were no radio frequency junction transistors on show.
By 1956 Valvo celebrated producing a million OC70 transistors and the first RF junction transistor: the OC45 announced in the autumn of 1956, matched by Telefunken who brought out the OC612, both transistors being for intermediate frequency use. In January 1957 Valvo released the OC44, their first transistor suitable as an oscillator/mixer at broadcast band frequencies. This was matched again by Telefunken with their OC613. [Herzog 2001]
Mullard was founded in 1920 by Stanley Mullard and began making receiving and transmitting tubes. Philips acquired 50% of the company in 1925 in return for granting Mullard a license to their vacuum tube patents and took over the remainder of the company in 1927.
Mullard began semiconductor research and development in 1954 at Redhill, Surrey with assistance from Philips and a year later in 1955 began limited production of germanium diodes and alloy junction transistors at Mitcham. In 1957 production was transferred to a purpose designed facility at Millbrook. Mullard quickly dominated the British market with a 55% market share within a year. [Morris 1990]
In 1961 Philips and General Electric Company formed the Associated Semiconductor Manufacturers Ltd, owned two thirds by Philips and one third by GEC. Their semiconductors were marketed by Mullard. [Clayton 1989]
The case study of the development of the alloy diffused transistor recorded in the following section illustrates the loose nature of the relationship between Philips and their key subsidiary companies. Both J Beale of Mullard at the Salfords laboratory and a Natlab team led by L Tummers independently worked on the same concept for a radically improved radio frequency transistor. By
the time Beale was ready to publish his work he was aware and acknowledged the work at Natlab but the publications of both teams showed that they did develop independent solutions to the alloy diffusion device.
Mullard made transistors for Amperex as illustrated in a rare case of an OC16 also marked with the JEDEC equivalent: the 2N115. [Photo courtesy Joe Knight]
This group of OC73 transistors, supplied in a single three-pack, illustrates Group sourcing from Philips and Mullard
Amperex is a famous name in vacuum tubes and was acquired by Philips in about 1955 in order to help Philips expand its share of the US tube market. This was at a time when the market for tubes in consumer electronics and hi-fi was still growing strongly.
Amperex began producing semiconductors at their factory in Hicksville NY that was opened in 1953. This location was also the headquarters for North American Philips.
Initially Amperex marketed transistors using the European numbering but moved to the conventional “2N” numbering commonly used in the USA and JEDEC registration from 1956. This example of a AC107 is branded "Amperex" but made in Holland.
Initially the 2N115 was reserved for the OC15 (photo right from a product announcement in Radio Electronics December 1955). However, in the final JEDEC release documentation Amperex give specifications for the OC16. [JEDEC 1958]“Amperex registered their transistors with JETEC and the 2-OC72 became the 2N282, a matched pair of individual 2N281's, the OC-72 equivalent. The 2N282 came standard with the copper cooling fin. The whole Amperex 2N279-2N284 line was short lived apparently because by 1956 most other U.S. manufacturers had moved far beyond the power and frequency range of these types.” [Knight 2007. Photo courtesy Joe Knight]
In the table above the comparable Philips type has been inferred from close examination of the published specifications. [JEDEC 1957, JEDEC 1958, JEDEC 1962, JEDEC 1963, Bradley 1959, Philips 1959]
In 1959 a new plant at Slatersville, Rhode Island was built to produce gold bonded germanium diodes for the computer industry. As demand for product swung to silicon Philips used its spare capacity at Slatersville to produce alloy diffused transistors. [Narragansett Imaging 2004]
In 1962 Amperex introduced the “Universal” communications transistor in two packages: the 2N987 in a TO-18 envelope and the 2N2084 in a TO-33 envelope. Amperex published a design for a 100mw Citizen’s Band transceiver using this transistor in all the RF stages. [Rudich 1962] This transistor appears to be very similar to the Philips AF124 which was produced in a TO-72 outline having dimensions very close to the TO-18 (2N987). At 100Mhz the 2N987 provided a gain of 12.5dB compared to 8dB for the AF124.
La Radiotechnique was formed in the wake of World War I in 1919 to produce vacuum tubes: one of four companies making these in France. They were bought out by Philips in 1931 in order for Philips to avoid paying damages to La Radiotechnique when Philips lost a court battle over patent infringements. [Tyne 1977]
In 1955 they were producing the:
OC15 power transistor
OC70 first stage amplifier
OC71 audio amplifier
OC72 output transistor [Toute la Radio 1955]
This data shows that La Radiotechnique were releasing their transistors in step with Valvo in Germany
The outline data published illustrates the OC72 in a longer glass encapsulation than the shorter kind familiar to the OC71 and other low power devices. Later versions of the OC72 had a tightly fitted metal shell to improve heat transfer to a fin heat sink.
Alloy Diffused Transistor
Alloy transistors such as the OC44 and OC45 suffered from three issues that seriously reduced their high frequency performance:
Slow transit time of minority carriers across the base: a function of carrier diffusion rate and base width;
Base resistance; and
In the field, early high frequency transistors were produced by reducing the base thickness by using the thinnest viable pellet width (about 2 mil) while controlling alloying depth from the collector and emitter dots. In these structures the base width was determined by three independent production variables: initial pellet thickness and the depth of alloying of the collector and emitter. This resulted in variability in the base width and a practical limit to the base width of about 0.5 mils. [Rudich 1962]
Philips’ alloy diffused transistor was developed from:
(1) The RCA diffusion transistor;
(2) A conceptual breakthrough at Bell Laboratories whereby the pellet is not the base but the collector; and
(3) The use of a double doped dot developed at Westinghouse from which two junctions could be formed
Improving Minority Carrier Diffusion Rate
Between 1954 and 1956 RCA developed the first diffusion transistor. Its improved performance was due to the increased minority carrier diffusion rate across the base due to the base field created by a graded doping profile. This transistor (the 2N247) was made by diffusing N-type dopant into the base pellet from the emitter side and then creating the emitter and collector by alloying. This was a significant advance but the design offered no improvement to base width.
Thin Diffused Base
Bell Laboratories designed a transistor in which the pellet was the collector, not the base, and used diffusion to create a thin diffused base layer on one side. [Lee 1956] This was a conceptual breakthrough that eliminated the issue of variability of collector depth alloying. Diffusion is slow and the base width made from this process could be accurately controlled. However, their transistor had an alloyed emitter and alloying had to be exceptionally shallow to avoid shorting the base.
Longini Hook Transistor
A hook transistor is a four layer device with a common base alpha of over one that can act as a switch. The first examples were formed point contact transistors where the Schottky-barrier collector point contact was modified by a high current pulse.
Working at Westinghouse, Longini developed a hook collector transistor made in the conventional alloy junction manner with the exception that the collector region was made from a dot of indium-arsenic alloy. Arsenic (N-type) has a fast rate of diffusion and indium (P-type) has a slower rate of diffusion. To produce an NPNP hook transistor he started from a P-type pellet, alloying the indium-arsenic dot at 570 C and then dropping the temperature to 550 C for 30-60 minutes during which time both dopants diffused beyond the alloy area. Since arsenic diffused faster an N-type region (the collector) was formed beyond the surface P-type region. The transistor was completed with an alloyed N-type emitter on the opposite face. Longini pointed out that many combinations of dopants were possible. [Longini 1953]
Longini’s post alloy diffusion step was adopted by Mullard and Philips to produce a thin diffused base from the emitter alloy region.
Alloy Diffused Transistor
The alloy diffused transistor was developed independently at Philips’ Mullard subsidiary company at its Salfords research laboratory in Britain and by Philips at the Natlab in Eindhoven. Its genius was to eliminate the problem of emitter depth by diffusing the base from the emitter.
The possibility of independent development reflected on the loose technology transfer between Philips and its subsidiaries. “There were no joint projects and no extensive exchange of information. Van der Spek, for example, commented that “the contributions from these foreign labs were only minimal,” a statement that clearly illustrates the company’s close-minded attitude toward the foreign laboratories.” [Davids 2006]
The Mullard development was led by J Beale and the Philips development was led by P Jochems, O Memelink and L Tummers. Both acknowledge the work of the other laboratory, the antecedent work noted above and that the concept arose from Longini’s hook collector transistors. [Beale 1957, Jochems 1958]
Beale noted in regard to the genesis of his approach “the development of this process was prompted by an idea contained in a lecture on an N-P hook collector by Benjamin and Longini in Philadelphia in 1955.”
Beale produced PNP transistors by starting from a P-type pellet that would become the collector. A dot of 1% antimony and 2% gallium in lead was alloyed onto one side then held at a controlled high temperature for a precise period. Antimony diffuses about 100 times faster than gallium but gallium is more soluble in germanium than antimony. Thus after the diffusion step a PNP structure was formed with gallium forming the P type emitter and the diffused antimony forming a thin base under the emitter dot. An ohmic connection to the base layer was made by diffusing antimony from the surface around the emitter down to the base and making the connection through a antimony alloy dot positioned very close the emitter dot.
Beale tablulated some typical data for a group of transistors made over a range of conditions [Beale 1957]:
Philips published more details of their approach which used a bismuth dot containing arsenic and aluminum. The bismuth formed the emitter and the fast diffusing arsenic formed the base. Aluminium was used to compensate the remaining arsenic in the recrystalised P-type region (the emitter).
The process was carried out in an atmosphere of arsenic vapour arising from
the arsenic in the emitter dot or from a separate arsenic source. This created an N-type layer all over the surface of the wafer as well as under the emitter providing a method of making an ohmic connection to the base. Rather than using a ring base connection as employed in the OC44 and OC45 which created a high base collector capacitance, a second dot was used of about the same dimensions as the emitter dot but made only of bismuth and arsenic. Then the
surface was etched with the exception of the bridge between the emitter and base dots. [Jochems 1958]Data released for this transistor showed it was very similar to the Mullard device:
The OC170 was the first alloy diffusion transistor released by the Philips’ group. Preliminary data released Mullard indicates a lower specification than that set out in Beale’s paper [Bradley 1959]
The OC169-171 family were produced because production variations resulted in a spread of characteristics due to the large variation in base resistance. They were graded by selecting for gain at 100 Mhz.
Subsequently this class of transistors was improved by utilizing gaseous diffusion prior to the alloying process which gave transistors of lower base resistance and high yield of high frequency transistors. [Rudich 1962]
Postscript: Silicon into the 1960s
The first silicon transistors released by Mullard were the OC201-OC206 series [Wylie 2008] of which the first two are reported in Bradley (1959):
They are described as alloy junction types for audio and general industrial applications having identical specifications other than gain, alpha cut-off and with minor variations in their noise figure and collector-emitter saturation voltage. In 1959 these transistors were not generally available being “for manufacturers’ use only.”
Alloy junction transistors were difficult to make in silicon due to stress fracturing at the alloy junctions caused by differential expansion between the alloy and the silicon. Producers needed to minimize this by careful selection of alloy composition. RCA would have made its data available to Philips: it abandoned silicon alloy transistor development in 1954 and put comprehensive details of its programme into the public domain. [Nelson 1954]
Due to its success with the germanium alloy diffused transistor, Philips was slow to move to silicon in any significant way. Van Vessem, who had introduced the alloy junction transistor to Natlab, urged the company to take silicon much more seriously but was ignored. At the end of the decade Natlab reviewed the use of silicon but concluded that its prospects were limited, particularly in the mainstream consumer markets of most interest to Philips. This echos the RCA position when it concluded that silicon had markets mainly for military use; a declining market following the end of the Korean war. [Sadler 1990]
In 1958 Natlab sanctioned a programme on a silicon alloy diffused transistor which was transferred to Mullard for manufacture in 1960 but which proved difficult to produce.
Despite the growing importance of silicon planar technologies in the USA, Philips did not undertake any significant development in this field in the 1950s. In 1962 when it had became clear that silicon was set to be an important semiconductor, Philips signed a license from Texas Instruments to its planar technology. But this did not include a technology transfer package and in 1963 Philips was forced to implement its own planar research and development programme. The Philips factory at Faselec developed a silicon planar transistor which was transferred to Natlab and subsequently to Nijmegen for production. This was problematic since silicon and planar production technologies were much more challenging than germanium alloy transistor production.
In parallel Philips tried to keep abreast of US developments in integrated circuits invented by Texas Instruments in 1958. In 1959 a group from Philips visited Texas Instruments to evaluate their integrated circuit technology but concluded it was too under-developed. Texas was still to implement planar integrated circuits at that time and it was not clear what technology would prevail.
Texas overcame its early production problems and was producing integrated circuits successfully by 1963 and Philips knew it had to enter this field. It was assisted by its takeover of the General Electric Company Wembley laboratory which strengthened its science but not its production capability. Rather than commence an in-house development programme, Philips looked for a technology partner and after evaluating several US companies signed with Westinghouse in 1965. [Davids 2007]
References and Acknowledgements
The author acknowledges the kind permission of Arnaud Cramwinckel, Philips Historical Products, Eindhoven (www.philipsmuseumeindhoven.nl) and Joe Knight to reproduce their photographs of early transistors in this article. The author is indebted to Joe Knight for his review of the original manuscript and helpful suggestions.
JEDEC data has been made available courtesy of extraordinary efforts by volunteers of the Tube Collectors Association who scanned and catalogued many thousands of registration records submitted by manufacturers to the Council.
Bardeen J Brattain W 1948 The Transistor, A Semiconductor Triode Phys Review 74 230-31
Beale J Alloy-diffusion: A Process for Making Diffused-base Junction Transistors Proc Phys SocB70 1087-89
Bell 1952 Transistor Technology Volume 1 July 1952
Bradley E 1959 British Transistor Manual Norman Price (Publishers) ltd
Bridgers H 1958 Transistor Technology Volume One Van Nostrand Company Inc
Choi H 2007 The Boundaries of Industrial Research Making Transistors at RCA 1948-1960 Technology and Culture 48 758-82
Clayton R, Algar J 1989 GEC Research Laboratories 1919-1984 Institution of Engineering and Technology
Cockle A 1954 160-Metre Transistor Transmitter Wireless World 60 May 217
Cramwinckel A 2009 Transistorontwikkeling Bij Philips Radio Historisch Tijdschrift 128 1 28-30
Davids M Verbong G 2006 Intraorganizational Alignment and Innovation Process: Philips and Transistor Technology Business History Review 80 4 657-88
Davids M Verbong G 2007 Absorptive Capacity in Solid-State Technology and International Knowledge Transfer The Case of Philips Comparative Technology Transfer and Society 5 1 1–31
De Transistor in theorie en praktijk. De Muiderkring N.V. Januari 1958
Grimsdale R 1995 The Transition From Valves To Transistors Bulletin of the Computer Conservation Society Autumn 1995 http://www.cs.man.ac.uk/CCS/res/res13.htm#c
Harrison M 1998-2008 http://www.electricstuff.co.uk/oldadverts.html
Herold E 1953 Herold’s Visit to Philips Research Laboratory at Eindhoven Oct 8 and 9 1953 RCA Report PEM 364
Herold E 1983 Focus on a Career Engineer: Excerpts from Bygone Days: An Autobiography (David Sarnoff Library http://www.davidsarnoff.org/ewh.html)
Herzog R 2001 The Early History of Transistors in Germany at the Semiconductor Museum http://www.ck722museum.com/history/Transistormuseum/LectureHall/RHerzog/RHerzog_Index.htm
Hilberink H 2004 http://www.hilberink.nl/techinfo2.htm
JEDEC 1957 Joint Electron Device Engineering Council Release 2063 December 23 1957
JEDEC 1958 Joint Electron Device Engineering Council Release 2183 April 28 1958
JEDEC 1962 Joint Electron Device Engineering Council Release 3987 November 19 1962
JEDEC 1963 Joint Electron Device Engineering Council Release 4501 November 18 1963
Jochems P Memelink O Tummers L Construction and Electrical Properties of a Germanium Alloy-Diffused Transistor Proc IRE 46 1161-65
Kendall J 1954 Practical Transistors & Transistor Circuits Bernards (Publishers) Ltd London
Knight J 2007 A Survey of Early Power Transistors: Amperex 1950s Germanium Power Transistors
Lee C 1956 A High Frequency Diffused Base Germanium Transistor Bell Sys Tech 35 23-34
Lojek B 2007 History of Semiconductor Engineering Springer
Longini R 1953 US Patent 2836521 Hook Collector and Method of Producing Same Filed September 4th 1953 Issued May 27th 1958
Malerba F 1985 The Semiconductor Business: The Economics of Rapid Growth and Decline University of Wisconsin Pres, 1985
Morris P 1990 History of the World Semiconductor Industry Institution of Engineering and Technology
Mullard 1952 Germanium Crystals Outlook October 1952
Mullard 1954 Mullard Germanium Diodes Outlook January 1954
Mullard 1956 Transistors for the Experimenter August 1956 Technical Service Department Mullard Ltd.
Mullard 1961 Manufacture of Junction Transistors Outlook January 1961
Narragansett Imaging 2004 http://www.nimaging.com/about/history.html
Nelson H 1954 A Silicon NPN Junction Transistor by the Alloy Process Transistors 1 Radio Corporation of America 1956 172-81
Philips 1956 Philips op de Firato Philips Nederland N.V. Eindhoven August 1956
Philips 1959 Transistors for Radio Receivers and AF Amplifiers
Shockley W 1948 U S Patent 2,569,347 Circuit Element Utilizing Semiconductive Material Filed June 1948 Granted September 1951
Ploos van Amstel J 1954 Transistors reprinted in Miniwatt Transistors Germanium and Silicon Diodes 4th Edition 119-31 Philips Electrical Industries Pty Ltd
Rudich I 1962 The Universal Communications Transistor IRE Transactions on Product Engineering and Production 6 36-43
Saddler J 1990 The Human Side of Early Electronics and Semiconductors SMEC Vintage Electronics 2 1990
Scaff J Ohl R 1947 Development of Silicon Crystal Rectifiers for Microwave Radar Receivers Bell Sys Tech J 26 1-30
Shockley W 1949 The Theory of P-N Junctions in Semiconductors and P-N Junction Transistors Jnl Bell Sys Tech 28 435-89 (reprinted in Sze S M 1991 Semiconductor Devices: Pioneering Papers World Scientific Publishing)
Tilton J 1971 International diffusion of technology: The case of semiconductors. Washington, DC: Brookings Institution.
Toute la Radio 1954 Transistors Anglais March-April 184 137
Toute la Radio 1955 Guide des Transistors 198 September 1955 284-5
Torrey H Whitmer C Crystal Rectifiers 15 MIT Radiation Laboratory Series. McGraw-Hill, New York, 1948
Tyne G 1977 Saga of the Vacuum Tube Howard W Sams & Co
Wylie A 2008 Mullard Semiconductors