Hardware
and OMI
In the late 1980s, there was
a continuing decline in the European share of semiconductors,
and microprocessors in particular. This was despite considerable
innovation and technical developments. JESSI (Joint European
Sub micron Silicon Initiative, now MEDEA- (Micro-Electronic Development
for European Applications) was tackling the process and manufacturing
technology, with companies such as SGS-Thomson, Philips and Alcatel
collaborating on base technology e.g.HCM0S5.Technology is no
good without products, and 0MI was to form the bridge.
Although aspirations still
existed for the European computer industry, the real opportunity
was foreseen in embedded systems. Europe had a strong presence
in the emerging RISC (Reduced Instruction Set Computer) microprocessors
with the indigenous ARM and transputer, and the licensed SPARC
and MIPS architectures.
OMI's Strategy
It was clear that densities
i.e. the number of transistors in. a given area, would increase.
Moore's Law predicting a doubling every 18 months or so had held
for 20 Years, but with hindsight, speeds probably grew faster
than many of us realistically expected.
In 1990,with 1.2 micron technology,
a 32 bit transputer was a 10 x10 mm-chip, but actual CPU was
only a small portion - on chip memory, communications and other
functions accounted for over 60% of the chip's 300k transistors.
In a few years, with millions of transistors, the overall die
size would be unable to be reduced because of the number of bond
pads (the parts around the edge of the chip where the wires connecting
the outside world are attached) would not be decreasing. The
big question by the late 1990's would be, "what to do with
the silicon real estate?"
For desktop computers, the
answer was obvious-the CPU would be improved, and any extra space
left would increasingly be used for cache memory. For embedded
systems, the only obvious strategy was to put more of the system
onto the chip. So reducing external components and cost and increasing
reliability.(Modem chips have a reliability that is roughly proportional
to the number of pins they have .Reducing from 10 components
with 100 pins each to 1 component with 100 pins can give an order
of magnitude improvement in reliability).
This is the system-on-a-chip that is common today, but at the
time, this was thought to be somewhat visionary and impractical.
System on
a chip
And this is where 0MI came
in-if a large pad of the system is going to be on one chip, who
will design and make it? Previously, the systems integrator would
have designed the system from standard chips, and perhaps a few
ASICs (Application Specific Integrated Circuits). We now faced
a fundamental business, not technical, problem. Europe's strength
lies in embedded systems, with a large number of smaller companies
involved. The scenario above leads to only large companies being
able to participate and compete!
There is a solution: companies
can work in partnership with each contributing their expertise.
For this to work, designs could no longer be unique every time.
Much as we build houses from standard components e.g. bricks
and when we need to do this more quickly, from prefabricated
panels. Transistors are the electronic equivalent of bricks;
now we needed higher level functions AND a common framework so
that the components would together.
Applying divide and conquer
to the problem, it breaks down into issues of methodology and
standard interfaces.
Standards
0MI was launched with a standards
project to address these issues, and a series of demonstrator
projects in which the techniques would be incrementally applied
(businesses could not wait for the perfect solution, or they
would miss the market opportunities).
So how did the companies respond
to the challenge? Together, in the OMI Standards project, Siemens,
Philips, ARM, Temic (Matra) and SGS-Thomson (Inmos) developed
a common basis for an on-chip bus, for connecting peripherals.
This provided a focus for other companies to develop modules,
and encouraged more entrants, such as Nordic VLSI, Hyperstone,
Mitel (GPS), SDSA etc.
Products
0MI Bulletins are occasional
leaflets published by the OMI PROMISE project.
0MI,The Open Microprocessor
systems Initiative, is a programme set up by the European Information
Technology industry and the European Commission in the framework
of Esprit the European strategic Programme for Research and Development
in Information Technology.
Let's look at some of the products that have been developed.
Accepting that most of these are not directly recognised, as
they are buried within something else i.e. the application.
Working in partnership with
car manufacturers, SGS-Thomson took the transputer and turned
it into a modular 32 bit core, and for the first instantiation,
added the necessary functionality to create a cost effective
system for GPS (Global Positioning by Satellite) to create in-car
navigation systems. SGS-Thomson is purely a semiconductor company,
but with a full range of facilities from design through manufacturing
to packaging and test.
ARM developed the ARM family
of cores. ARM is a small company, founded at the start of 0MI
as a spin-off from Acorn, who had developed a 32 bit RISC processor
for their own computers. ARM was thus one of the early fabless
design companies. Partnerships with semiconductor foundries,
as licensees, were an essential part of the business plan. An
early example brought ARM and Mitel (formerly GEC Plessey Semiconductors-GPS)
together with one of the resulting chips being used in Apple's
Newton. Newton was the first hand held PDA (Personal Digital
Assistant) to use handwriting recognition, a feature previously
impossible without the computing power at low power consumption
introduced by the ARM processor. Another example was the partnership
with Hagenuk to produce a cheaper ISDN telephone supporting digital
services. Yet another was the use of the ARM microprocessor by
OptionExist to create the portable FAX, marketed by Fujitsu.
Philips is a large vertically
integrated company, including design and manufacturing not only
of semiconductors, but a wide range of electrical and electronic
products. They took a licensed processor, MIPS, and modularised
it for use with the 0MI peripheral bus, allowing system chips
to be very quickly developed through a very high degree of re-use.
A side effect of high re-use is a lower level of design problems-most
of the design is already well proven. With a standard set of
modules, it becomes feasible to make progress. n the related
area of hardware/software codesign - techniques to assess which
parts of the design should be realised with dedicated hardware,
and which are more cost effectively implemented in software.
This is a good example of the virtuous circle 0MI set out to
create-improved design m-use increases productivity, increases
reliability and reduces time to market and costs. Some of the
resulting chips can be found in mobile phones and PDAs.
Siemens is also a large vertically
integrated company, but the semiconductor division operates autonomously.
Work done in 0MI led to the 80C167, where a fuzzy logic co-processor
was added to an existing 16bit controller and used in control
systems for electric vehicles, where there is not a straightforward
relationship between the measurable parameters and the system
to be controlled. 0ther work has led to the development of Tri-core-
a 32 bit processor with some DSP (Digital Signal Processing)
characteristics, targeted initially at the automotive market.
Temic (Matra) took the licensed
SPARC architecture and developed a family of communication controllers.
This has now been spun out into a separate company, Tsqware.
These products are typically used in data concentrators ISPs
(Internet Service Providers). This highlights a feature of the
emerging IP (Intellectual Property - in this case, semiconductor
design) business. There are a limited number of people with a
deep understanding of a particular type of system. The world
market will buy the IP, rather than develop and maintain staff
with that expertise.
Integrating processor cores
and developing a range of peripherals is not enough to be successful
in the silicon systems on a chip. The business case requires
them to be cost effective. Generally, for the lower cost mass
market, this requires that the chips are put into cheap packages
i.e. plastic, which have limited capacity to dissipate heat.
Secondly, many of the embedded market opportunities are for new
products, especially portable new products. Silicon technology
development will evolve ways of reducing the power, but
OMI has pioneered a radical new approach - application of asynchronous
logic to real systems. This computing-without-clocks makes use
of the fact that the CMOS technology in common use for the last
20 years only consumes power when switching. Most of the time,
clock signals are fed to every circuit, but other than the clock
signal itself, there are no changes: many signals have low rates
of real change. Thus if signals only change when it is meaningful
for them to do so, considerable power can be saved, Within OMI,
Philips have developed this technique and applied it into a portable
CD player and Manchester University are now on their 3rd generation
implementation of an ARM microprocessor with this technique.
Conclusions
Some of these are examples
of what ESPRIT has helped to achieve - new ideas and businesses
grow up around a capability, as in Silicon Valley, California.
In Europe, 0MI has, through partnerships, and successes of the
companies involved, helped change the thinking of engineers to
a belief that new ideas and innovation can succeed here in Europe.
This has come about through the transition from a Euro-centric
view of business to a global view of the business. Ironically,
a number of companies have not survived in the form that they
started in, having either ceased to trade, or been taken over.
This morality suggests that the market in Europe is now much
more driven by the facts of business life, and that ESPRIT is
helping to overcome the lack of critical mass in any one country,
by reducing the barriers, financial and otherwise, to creating
a distributed critical mass. This is essential for a true single
market to develop, as it must encompass design and development,
not just sales of products.
In this analysis, we have been
looking at a few of the hardware successes- however the silicon
hardware is just one part of a three, way partnership:
Silicon
Software
Application
0MI Bulletins are occasional
leaflets published by the OMI PROMISE project. 0MI,The Open Microprocessor
systems Initiative, is a programme set up by the European Information
Technology industry and the European Commission in the framework
of Esprit the European strategic Programme for Research and Development
in Information Technology. |