New Scientist 12 June 1986 33
decline of Uncle Clive
by Margaret Thatcher and widely considered as the most well-known scientist in
Britain, the chairman of Sinclair Research seemed unstoppable. What went wrong?
Ian Adamson and Richard Kennedy
The third and most intriguing
option - and the one which presents the most daunting technical challenges—is
wafer-scale integration. This approach
to the design of semiconductors offers financial savings by producing complete processing systems, laid down on a
single wafer of silicon. It could
also pave the way towards compact implementation of the new generation of processing
techniques currently under development.
The opening in 1983 of the prestigious Metalab research unit near Cambridge provided
a base for the realisation of Sir Clive's visions,
among them the much-publicised "Fifth
Generation" project to develop artificial intelligence. Sinclair made patriotic
noises about beating the Japanese
at their own game - whatever that might be, and to what end. One of the elements of this
fantasy was the investigation of wafer-scale integration.
Catt: The image of the inventor.
High-tech products for the high street - but how
useful are they? Ivor Catt (above) sold Sir Clive his
patents to wafer-scale technology. According to who you speak to in the
semiconductor industry, Catt is either a crank or a visionary
Sir Clive's initiation into the
world of the wafer took place in the summer of 1983, with the arrival of Ivor Catt who had answered Sinclair's advertisement
for people to work at Metalab. Depending on who you
talk to in the generally conservative semiconductor industry, Catt is either a
crank or a visionary. For 20 years, he had been refining the theoretical foundations
for a revolution in the semiconductor industry, and thus was tailor-made for
the Sinclair project. Sir Clive took on Catt as a consultant and bought up
Catt's patents to the wafer-scale process.
Catt himself has succinctly summarised the appeal of
the wafer against existing chips and methods of manufacture: "I
noticed that the silicon wafer was a hundredth of the cost of the total system,
so why not use that cheap commodity to build the system on the wafer
instead of sawing it up to form separate circuits?"
Currently, the computer
industry produces multiple chips on each wafer of silicon. The production
process involves chopping up the wafer, testing each chip and then
separating the working chips from a significant number of faulty chips. The
working chips, after mounting, wiring and packaging in plastic, become
part of a larger system mounted on a printed circuit board. Catt's
alternative method involves preserving the entire wafer (including the
faulty chips), which has internal connections between chips so as to eliminate
the printed circuit board. It also avoids the need to test and encapsulate each
chip. An electronic logic test built into the wafer circuitry allows each
chip to be tested. If functional, the chip becomes incorporated in the
circuit and then tests an adjacent chip. Faulty chips are bypassed as a spiral
sequence of working chips is established on the wafer. The simplest form would
be a memory wafer, but there is a potential to develop new, alternative
computer architectures on the wafer.
Throughout the 1970s, the attempt to realise
such a product dominated the R&D strategies of many of the semi-conductor
giants. ITT, Texas Instruments and Burroughs, among
others, sunk undisclosed fortunes into the dream. The kiss of death for the wafer as an
investment option was the debacle of Gene Amdahl, formerly a designer with IBM.
Amdahl's pursuit of a "supercomputer" based on the wafer-scale
attracted around $240 million in backing from heavy-weights that included
Sperry, Digital Equipment and the Bull Corporation
of France. By June 1984, Amdahl's company, Trilogy, had conceded that it could
not overcome the problems of implementing its version of wafer-scale
failure of the big boys came as no surprise to Ivor
Catt, whose approach had always
radically differed from those of his rivals. Axiomatic to Catt's technique was
a reduction in the number of connections made to the chip. In the latter stages of Amdahl's mega-wafer, the
doomed prototype had an astounding
1200 pins packed on to its 6.4-centimetre
design. Since, according to
Catt's theoretical design, communication with
the wafer passed through the first chip on the spiral, his chips were designed as bipolar
components, thus needing only two pins as connections.
in the wafer
years in the wilderness, the National Research Development Corporation
eventually funded Catt's theories in the late 1970s. This at least enabled him
to patent their implications. At Middlesex Polytechnic, Malcolm Wilkinson ran aresearch team which examined the problems of implementing
Catt's work. Wilkinson and his team went on to develop their research with
Burroughs, where they success-fully realised a
provisional "test structure". At this point, the project fell foul
of company politics. A new and predominantly American management, presumably
with the experience of Amdahl fresh in their minds, wanted nothing to do with
research into wafer-scale technology.
Clive's interest in the technology could hardly have come at a more opportune
moment. At the end of 1983, his relatively
small, if momentarily profitable, company was able to poach not only Catt, but Wilkinson
and a significant proportion of the team from Burroughs. In time, valuable additions
from research groups working in related technologies from Plessey, TI, STL and
DEC, would arrive.
association with wafer technology does nothing to enhance his self-styled
stance as inventor and innovator, Sir
Clive's support of these discredited research objectives was undoubtedly a canny move at a time
when Sinclair Research was in a position to fund such an enterprise. In
acquiring Catt, Wilkinson et al. and the wafer-scale patents, en masse and
cut-price, it is arguable that Sir Clive was making an acceptable high-risk
investment in the future. Sinclair's appropriation of Catt's work mirrors his advocation and adoption of Denis Gabor's work in the
development of flat-screen technology at Imperial College in the late 1950s.
a relatively short time it looked as if the investment would pay dividends. By
spring 1985, Wilkinson's research suggested that the company could economically
produce a wafer with a memory of
half a megabyte for Sinclair's ill-fated QL
microcomputer. Unfortunately, at the same time, the price of conventional memory chips
fell dramatically. A few weeks
later the financial crisis at Sinclair Research came to a head, precipitating the sequence of
events which ended in theabortive "rescue" by Robert Maxwell. It seems likely
that Sir Clive's preoccupation with the wafer-scale project exacerbated his
lack of interest in the computer division of Sinclair Research, hastening a
deterioration of the financial crisis to the point of no return. The fact that
Sir Clive later turned down an
offer that would have ensured the survival of the
computer products tends to support the impression that, as far as he was
concerned, home computers were history. However, while Sinclair may have been
intrigued by the "intellectual
challenge" of wafer-scale,
it is equally clear that his
much-lauded vision was decidedly myopic.
soon as it became apparent that wafers with memories were unlikely to provide the funding
for more sophisticated research,
Robb Wilmot, chairman of ICL, was recruited onto the research board as troubleshooter.
brief was to drum up investment for the wafer-scale project. He soon recognised a potential that had eluded Sir Clive. Up until Wilmot's
intervention, Sir Clive's exclusive direction for research into wafer-scales
was towards theenhancement and development of
Sinclair's existing technology and projects. Wilmot approached the problem of
investment with the conviction that a solution to the production of wafer-scale
chips could propel Sinclair Research into a position where the company would
challenge the leaders of the semiconductor industry.
to Wilmot, wafer-scale chips could revolutionise the design and production of all types
of computers, and play a major
role in communications products and defence systems
(particularly radar equipment). In other words, thedevelopment
of wafer-scale technology seemed poised to take Sinclair Research well out of its
depth. Ironically, the company's capacity to raise finance was in a sense
impeded by the exciting potential of its R&D resources. The public'srecognition of Sinclair Research's managerial,
marketing and financial
shortcomings called into question its corporate ability to exploit effectively
such an innovation. During the crisis
in 1985, the odds were stacked against even ICUs well-connected supremo, Wilmot, coming up with a result. Malcolm Wilkinson
sums up the difficulties facing the project, which are the same today as they
were six months ago: "It's
semiconductors, which are bad news to the City at the moment . . . It's
wafer-scale technology, which has had some
notable failures . . . and then there are the problems that Sinclair Research has got, and
questions about the viability of the
business side of it."
a broker commented when the price of shares in Amstrad fell following the
announcement of the deal with Sinclair,
"The City . . . gets wobbles in the stomach when the name of Sinclair is mentioned."
In the event, Wilmot failed to
find the backers. A fortuitous deal with the Dixon chain of shops enabled Sir Clive's company to
struggle on into the New Year until Alan Sugar came to the rescue in April.
the Amstrad deal came the announcement that two separate companies would continue the
projects on the radio telephone
and wafer-scale technology. Sir Clive made it clear that he would have no part in the
day-to-day running of either corporation. Barclays. the
company's bankers, agreed to a limited
investment package for wafer-scale technology with Sir Clive retaining a majority interest
in the company, and the bank
having an option to take up minority holdings. Desperately under-capitalised, it is hardly surprising that the team researching into wafer-scale
technology is directing its attention towards distinctly unspectacular goals.
The only project announced by the
company is a wafer with a memory of 5 megabytes. It remains to be seen whether
the experimental pilot production achieved in September 1985 can be
sufficiently improved to create a product that can compete with conventional memory
components in 1987.
Catt has always insisted that memory products are merely an incidental spin-off
from the main work of wafer-scale development. The main purpose of wafer-scale
technology, he believes, is to assist in the design of systems that will revolutionise
computer architecture. A growing number of
computer theorists are inclined to view these developments with interest, but Sinclair's company
is hardly in a position to fund
such ambitious research programes. So while wafers may yet hold a hope for the future, it
seems unlikely that they hold out
much hope for Sir Clive.