Good morning everybody, it's my great pleasure to be here today with you.
From the earliest days Honda Civics were known for their fuel efficiency.
Quoting a Honda Civic advertisement, when we built our first Honda Civic back in 1972
we designed it as an answer to the world's transportation problems even that long ago,
fuel economy was one of our prime considerations, now this is from an advertisement announcing
the new Honda Civic that can deliver 42 miles per gallon.
That's pretty good for a non-hybrid car these days, but this is an ad for the 1980 new Honda
Civic.
Now fuel efficiency among all automotives rose rapidly through the 1970s until the early
1980s, at which pointed plateaued or declined for nearly 25 years now.
Granted, the focus on fuel efficiency is subject to the whims of consumer interests, but the
fact that it remained stubbornly stuck is less a function of a lack of engineering focus,
and more a function of the challenges and limitations of internal combustion engine
technology.
In fact, the primary reason that this chart reflects recent improvement is that automotive
manufacturers began seeking technology solutions outside of internal combustion engine design.
New types of drive trains, like hybrid gas electric and all-electric and novel energy
supplies such as lithium ion batteries and hydrogen fuel cells have jump-started fuel
efficiency gains (if you'll pardon the pun).
Listening to the pundits lamenting the end of semiconductor innovation in the twilight
of Moore's law, it's hard not to feel like turn-of-the-century automotive engineers,
whose future was potentially limited by the laws of physics.
So is the innovation party over for semiconductors?
Is ours a future of fine tuning and incremental-ism?
Well as both an optimist and a student of the history of technical advancement my answer
is emphatically, no.
Before I delve into the rationale from my perspective, I'd like to set the stage by
describing a few pervasive constants that are time independent and will be the dynamics
propelling us well into the future.
First, innovation is as essential as energy and matter.
As humans we are programmed for exploration and adventure and we possess an unquenchable
thirst for knowledge and information.
Second is the constant dance between the economy and technology.
Novel technologies shape and reshape the economy which begets further novel technologies, creating
ever new challenges and so on and so forth.
Third, combinatorial innovation, the power of which is demonstrated by the evolution
of automotive drive trains to enable greater fuel efficiency.
As Brian Arthur articulates in his book The Nature of Technology, combinatorial evolution
is foremost and routine.
In many cases.
unexpected applications emerge at the intersection with other technologies and physical domains.
For example, the inventors of the digital camera certainly never had Instagram in mind,
which was created by combining digital camera technology with software and the Internet.
Fourth, the physical and digital worlds continue to mesh, and digitalization is becoming more
pervasive.
To continue to make the kind of social and economic progress to which we have become
accustomed, we require increasing levels of data to better understand and operate in the
world around us.
This is an era of progressively bigger data, as the data being generated and stored continues
to double about every two years from one zettabyte of data in 2010 to a predicted 44 zettabytes
around 2020.
Finally, the result of these dynamics is that complexity will continue to increase at exponential
rates and innovation cycles will continue to speed up.
So now let's consider our context.
The hardware paradigm underpinning the information and communications technology industry has
experienced three major shifts over the past 60 years; the mainframe era, the personal
computing era and the Internet of Things era.
During this period, the user-to-device ratio inverted the mainframe era's many-to-one ratio
and has become one-to-many.
Today users are outnumbered by the devices that they access and even more significantly
outnumbered by the instrumented nodes on which they rely.
This exponential growth in devices and nodes has created a virtuous cycle of pervasive
digitalization and automation that is driving an even greater demand for devices and nodes.
The four-trillion-dollar ICT industry that is changing how people learn, work and live
is both literally and figuratively built upon the four-hundred-billion-dollar semiconductor
industry that enables it.
Technologies such as artificial intelligence and virtual and augmented reality as well
as companies such as Facebook and Google rely upon and gain an order of magnitude benefit
from our industry.
The global economy and indeed society needs us to continue to create and bring new capabilities
to market.
That being said, the direction of pace of our future advancement is uncertain.
As we all know, Dennard scaling ground to a halt a little more than a decade ago when
current leakage grew to such a level that it was no longer possible to further reduce
supply voltages and chips.
As Moore's law reaches its limits at the deep submicron level due to technical pressures
from lithography, power, quantum tunneling and so on, and economic pressures that have
driven the cost of state-of-the-art wafer fabrication facilities such as TSMC's planned
three nanometer fab to more than twenty billion dollars, it's clear that our industry can
no longer rely solely on the classical drivers of advancement in the future.
New paths forward have to be charted and different forms of investments made if our industry
is going to match or even come close to the progress and pace of the past half century.
Beyond these technological issues, business realities are also complicating the path forward.
While innovation has always been the lifeblood of the semi sector, in the aggregate growth
is slowing and ROI (return on investment) has become an issue.
This environment creates a subtle but steady pressure to the emphasize innovation as the
path to future value generation, and shift business models, the ones focused more on
financial engineering, or technical evolution playing a more modest supporting role.
While fiscal discipline of course is critical, the pressure to de-emphasize innovation has
to be resisted vigorously if our industry is to continue to carry its mantle of technological
leadership.
That continued leadership is critical not just for its own sake but for the sake of
the entire information industry, and the even larger aggregation of industries and organizations
that increasingly rely on it for their growth and progress also.
For example, doctors are increasingly leveraging bits and bytes to make and keep people healthier.
Computers are transforming transportation and making it safer greener and of course
more enjoyable.
People can communicate, collaborate and create with virtually anyone anywhere thanks today
to the spread of digital communications.
The reality is that the future will require more, not less semiconductor ingenuity.
We find ourselves at the center of the proverbial perfect storm.
Innovation is being buffeted by technological constraints, business concerns and insatiable
market demand, and yet our Moore's law boat is running out of steam.
The late physicist and philosopher, Thomas Kuhn, revolutionized thinking about the advancement
of science in the 1960s, but he published his book The Structure of Scientific Revolutions
which overturned the conventional wisdom that science advanced in a linear fashion.
One advance built upon another in a more or less steady, but slow march of progress.
His contrarian theory was that sciences advance was more lumpy.
That science lurched from prolonged periods of relative stability, punctuated by dramatic
periods of revolution.
If Kuhn were alive today he might describe our current situation as the early stages
of a crisis that catalyzes a new paradigm our way forward for the industry.
Our industry's existing model will increasingly struggle to respond to market demand within
its current parameters.
This is the pre-paradigm shift phase, where alternative models begin to compete to become
the new model.
The reality is that the advent of the Internet of Things and artificial intelligence is decreasing
the importance of traditional approaches to and limits of innovation.
The Internet of Things and the proliferation of chip scale sensors and computing elements
that intelligently connect the physical and digital worlds are making the Shannon Hartley
limits and information transmission more important than Moore's law.
Artificial intelligence in its push toward neural networks and massively parallel processing
are leaving the sequential processing of Turing and von Neumann behind.
In short, the methods by which we measure and pursue advancement are evolving.
On the digital side of the industry teams began managing thermal and clocking constraints
in their processing engines and moving to multi-core processing architectures years
ago to overcome limitations that couldn't be overcome through scaling alone.
On the analog side issues are being tackled through a strategy referred to as �more-than-Moore�.
Kuhn would say these non-Moore's-law-based models mark the beginning of a paradigm shift
in how semiconductor innovation happens.
Today, the overall industry is increasingly approaching this problem from a perspective
that supplements the traditional technology supply-driven approach, with an application
demand-driven approach.
The technology-supply-driven perspective focuses on driving improvements along the primary
dimensions of performance, size, cost and power efficiency, whereas the demand-driven
perspective starts with the problem to be solved and works backward from there to more
efficiently and effectively align inventions to applications.
In this combined technology supply and application demand era, there are three principal areas
in which ADI has been investing to drive future innovation.
I'd like to share some of our progress with you here today.
The first area is the technology layer.
The primary dimensions of innovation in the Moore's Law era have been reducing gate lengths
and other critical dimensions and increasing wafer sizes to achieve scale and convergence
around the limited number of nodes.
What the ITRS refers to as more Moore.
The sure and rapid pace of hardware advancement in this era enables the entire technology
stack to predictably plan for the future of capabilities and needs quite easily, but on
the downside allowed inefficiencies to creep in.
For example, in the development of software, programmers didn't really need to worry too
much about producing code that made the most efficient use of hardware resources because
they knew the hardware would cover those inefficiencies in a very short period of time.
That won't work in the future.
In contrast to the convergence of more Moore, more-than-Moore is characterized by divergence
of technology scope; it is combining and layering technologies with silicon to achieve something
completely new.
It ranges from new materials and adding more elements from the periodic table, to employing
process and package technologies to create novel passive and active devices and more
complex and complete system-on-chip hardware architectures.
For example, the image you see on the screen here is ADI's chip scale pH sensor technology,
which promises to bring lab instrument quality performance to the field in a 0-pin tiny form
factor, and at a cost that opens up entirely new applications in industrial, medical, pharmaceuticals,
and so on and so forth.
This is only possible in a world where we're leveraging cleanrooms, semiconductor manufacturing
equipment and other types of equipment and processes, to develop extremely small extremely
precise non-silicon-based products like sensors.
In fact, there is no semiconductor content at all in this pH sensor element.
Instead we're leveraging known manufacturing techniques such as electroplating and wafer
bonding, with features you would never see on a CMOS wafer or anywhere near a silicon
cleanroom, such as gold microfluidic channels and even liquids.
This slide gives you a sense of how things are changing.
As you can see, the number of elements upon which we are relying has more than doubled
over the past decade.
That isn't to say we avoid silicon in more-than-Moore, on the contrary we are combining these sensor
elements with ASICs.
Developing the sensor and electronics together opens a variety of possibilities and advantages
the combined solution is ultimately of higher performance and we can closely couple the
sensor and electronics, tailoring one for the specific advantages and disadvantages
of the other.
We can also use the ASIC to monitor the sensor and enable a smarter solution; one that can
compensate for drift in the sensor for example.
Finally, sensors can provide additional indirect and subtle sources of information, what may
appear to be noise, that we can extract from the sensor via the active circuitry and combined
with the primary measurement to make a more precise or valuable measurement.
For example, by incorporating additional electrodes on our sensor we can capture temperature and
conductivity measurements as well as pH, providing additional information on the species or reactions
occurring in the sample.
We're now well beyond the realm of simply miniaturizing a pH sensor.
Aside from materials innovation, advanced packaging technologies are becoming increasingly
valuable in product development as demonstrated by the micro module technology that our linear
technology franchise develops and has refined over the past 10 years.
During this time period, the power density of these products increased by more than a
factor of 10 primarily due to creativity and 3D packaging techniques and magnetics.
Through clever integration of passive components developing novel thermal management structures,
and incorporating advanced interconnect techniques such as copper pillars, the size, efficiency
and thermal performance were all optimized.
Today one small module can deliver the 100A load that previously took 12 modules in 2010,
and do so without frying the board.
This matters not just to our customers but to our customer's customers as well.
For example, in data centers, the operational cost is the electricity needed to cool the
equipment.
If improving the thermal management of power supplies can reduce cooling needs by just
1 degree Celsius, millions of dollars in electricity costs can be saved annually for a data center.
Today we're adding power system health monitoring features that constantly measure and analyze
the status of the power system measuring performance, correcting drift and watching for warning
signs of impending faults.
By communicating with the processor and reporting that data the power system performance can
be adaptively optimized.
In essence more-than-Moore is enabling smarter power.
Once we begin adding capabilities like cloud-based analytics, however, I would argue that we've
entered a new realm that is more-than-more-than-Moore.
Once we begin adding big data analytics, algorithms, software and security, essentially building
out the complete integrated technology stack beyond hardware, we're now in beyond more
territory.
When these three axes of technology innovation are combined, we free ourselves to deliver
more comprehensive and a much wider array of solutions to the market.
Thinking expansively about technology advancement is key to our industry's future.
Secondly, we need to pursue innovation at the physical system layer, shifting our focus
from the internal to the external, from ourselves to the world around us.
This level draws on insights, inspirations and advances, from the world in which semiconductor
technologies are being applied to drive us forward.
I realize that taking an application-centric approach isn't new to this industry, after
all, ASICs have been around for nearly 40 years.
What is new is the level of insight we now have into physical chemical and biological
phenomena, insight that inspires and shapes our product developments.
When we begin to go beyond merely knowing an applications technical problem, to a deeper
understanding of why the problem truly matters, as well as the business context in which our
solution can be applied, a new realm of possibilities opens for us.
That's why ADI invests not just an exceptional analog digital and software engineers, but
also chemists cryptographers, biomedical systems engineers and even physicians to ensure we
fully understand the applications we are seeking to address with our technologies and capabilities.
Now take for example our work in energy; one of the largest industries and markets in the
world.
Recently IOT has become a very hot topic, but utility companies advanced metering infrastructure
or AMI, were IOT before IOT was even cool.
AMIs are the vastly distributed but integrated systems of smart meters communications networks
and data management systems that utilities used to remotely monitor energy generation
and consumption.
All those millions of meters must be periodically monitored to ensure they don't drift an accuracy
and begin over billing customers.
ADI set out to solve the meter accuracy problem with a complete edge-to-cloud non-invasive
monitoring solution, that reports on meter accuracy and sensor health while the meter
remains in the field.
We do this by injecting a tiny but very stable background signal into the sensor.
We then pull that signal back out from the unknown energy signal, which may be more than
130 dB larger, using advanced signal processing capabilities.
This process combined with system level understanding and a cloud-based analytics service allows
us to discern the most minute signal changes and diagnose the sensor and meter irregularities
they can indicate.
The most interesting aspect of this example is what we discovered when we worked with
our customers and began to better understand the industry and its market dynamics.
The energy sector loses almost 100 billion dollars of value every year as a result of
electricity theft - that's 1/4 of the value of our entire industry every year.
We realized that our product could also be deployed to detect and locate tampering and
give utilities the information they needed to stop it.
In this case simply understanding the larger context of our customers and their markets
opened up an entire new application and revenue stream for our technology.
The final side of our triangle of innovation addresses the growing complexity of applications.
To adequately address this complexity, suppliers need to start looking beyond their four walls
and begin truly innovating as an ecosystem.
That's to say, the challenges we�re working to solve today are sometimes beyond the capabilities
of any one supplier, no matter their scale or their scope.
In our world of increasing complexity, collaboration across ecosystems will become increasingly
critical if we are to tame it and create an effective match between technology solutions
and market needs.
When collaborating as an ecosystem, suppliers must be clear about the areas where they are
unique and should strive for leadership, but also the areas where they should take a federated
approach in which they leverage formal and informal partnerships to advance.
As ADI's CTO Peter rail is fond of saying "you have to know when to lead, when to follow,
when to co-invent and when to partner."
Ultimately this needs to be a win-win relationship for all parties in order for the ecosystem
to be successful.
When the three sides of this triangle are coordinated, we can generate inspiring innovation
and importantly create real market impact.
For example, this is our general-purpose software defined radio.
ADI first tackled the SDR challenge by introducing a single-chip, low-power, MIMO transceiver
comprised of two transmitters and two receivers.
This transceiver removed much of the hardware complexity from our customers radio designs
by incorporating that complexity within our chip, and in addition software-oriented customers
could fully leverage the sophistication of the chip in their native software domain.
As a result, this product has been deployed and hundreds of applications ranging from
camera drone video streaming and control, to cellular base stations.
Recognizing the multi-application potential of the transceiver early on, we sought to
build an ecosystem of complementary products and services to better support and ease our
customers design and efforts.
We built an ecosystem of simulation tools with Mathworks and a variety of third-party
hardware and software packages, to shrink the development gap between systems architects,
RF designers and systems software developers at the beginning of their software defined
radio design journey.
Thus, we're enabling more elegant and efficient designs and a faster time to market.
This chart shows the diversity and number of organizations we are working with on the
backend to ease our customers front-end experience.
As we worked with our SDR customers over the years, our teams gained a deeper understanding
of the environment in which we are operating - particularly the new form factors for cellular
base stations and the challenges created by the exponential growth of cellular traffic.
This led to the evolution of a relatively simple general-purpose SDR to the much more
complex solutions for cellular infrastructure applications that we offer today.
Using roughly the same architecture, many of the critical specifications of our product
were improved by up to 30 dB through a combination of circuit level innovation and digitally
assisted algorithms.
Shown on the right of this triangle, the top photo is an RF transceiver for 3 and 4G radios.
The product below adds macro 2G support, which increases the dynamic performance by another
20 dB.
This product, which will be described in paper 9.3 at this conference, provides the footprint
reduction that communications customers crave, reducing a many-chip solution into a single
chip, and eliminating many impossible to integrate components such as IF filters.
Finally, the photo on the left of the triangle demonstrates how we are leveraging more-than-Moore
and beyond more to create our next generation SDR solution.
Returning to our up the stack diagram, we've gone from a solution that covers these dimensions,
to one that covers these dimensions.
Moving to 28 nanometer RF CMOS, we are doubling the number of RF channels and adding algorithms
such as power amp linearization, receive path linearization, crest factor reduction and
noise cancellation to address the challenges of delivering 5G solutions, that improve spectral
efficiency by dramatically increasing the number of antennae.
This is the power of expanding the aperture of your innovation to include technology,
physical system insight and ecosystems.
Now the fusion and expansion are built into the DNA of innovation.
Technology spreads out, combines with other technologies and techniques, continues growing
and sometimes spawns entirely new technologies, businesses and markets.
As we stand at the end of the eras of Dennard scaling and Moore's law, our understanding
of how we can continue to grow and evolve in the future has to become much more expansive.
I've been in this business for about 35 years now since the tail end of the first wave,
and the start of the second wave of our end of the ICT industry.
I've lived through the supposed golden age of semiconductors and have seen some amazing
advances, but as I stand here now and think back in those past 35 years I really wish
I were starting all over again.
I say that because there is such tremendous need for creativity as the physical and digital
worlds increasingly intertwine and create new opportunities to solve problems that have
bedeviled humanity for decades, for centuries and even longer in areas of healthcare, environmental
degradation, security and the need for communication and learning that breaks down barriers and
opens up possibilities everywhere across the globe.
Semiconductors play a critical role in engineering our social world.
Secondly, we have at our disposal a trove of technology, tools of capabilities and building
blocks and insights that are ripe to be combined into novel and more powerful combinations.
This goes further than just semiconductors.
When you consider the combinatorial possibilities with sensing computing and communications
technologies.
All that we need as an industry is imagination and courage.
We need to put aside the cynicism and pessimism, and reclaim the optimism and enthusiasm with
which our industry has been marked since its beginning.
Ours is an industry with a rich history and heritage of ambitious revolutionary characters
the industry needs to reignite its zeal for pushing through what we're perceived to be
impossible limits and its flirtation with the bazaar.
It's time to throw off the shackles of Moore's law, which in some ways has been the antithesis
of creativity.
It was a road map that we slavishly followed into multimillion-dollar masks and billion-dollar
fab.
To be sure the solutions to the technical problems encountered on that journey were
really, truly inspired, and this conference is the annual celebration of that ingenuity.
Compared to the open field of possibilities we face now, however, the Moore's Law era
was a relatively constrained environment about more bits, higher speed, lower power and so
on and so forth.
The direction was clear, and the future was calibrated - not a lot of thinking required.
Now that the Moore's Law train is rolling to a stop we can finally get off look around
and really do some interesting things together.
Thank you.
For more infomation >> A Man Kicks A Stray Dog And What Happens Next Is Unbelievable... - Duration: 2:54. 
For more infomation >> See What Happens When 15-Year-Old 'Spoiled Brat' Is Asked To Feed The Homeless In Los Angeles - Duration: 4:43. 
Không có nhận xét nào:
Đăng nhận xét