Holzhaider has had a ringside seat to watch and participate in the electronics
design revolution of the past four decades. The project manager for technology
development and technical board member at ams
AG (formerly Austria Mikro Systeme)
has been with that company for 32 years, after collecting a degree in computer
science from Johannes Kepler University in Linz, Austria, and a master's in computer
engineering from Stanford.
As part of our
coverage of Cadence's 25-year anniversary, I talked to this longtime customer (pictured below) by phone in Unterpremstätten, Austria, to get his perspective on the evolution of the industry,
electronics design, EDA tools, and the future of electronics innovation.
Q: From your
perspective, what's been the most surprising technology advancement you've seen
in the past 25 years?
Holzhaider: After almost
50 years, Moore's Law is still intact. It was formulated in 1965 based on an
observation in which Gordon Moore stated the number of components will double
for at least another 8-10 years. Almost 50 years later that's still the case.
It went from doubling every year to every18 months, but it's still intact. It will
last another 8-10 years. This is a major thing.
the context of ams, of course, sensor integration has been another major
advancement. Magnetic sensor integration, MEMS microphones, opto sensors,
ambient light sensing, to name a few. They basically added new features to an
integrated system. That was significant.
technology advancement--which is also significant for ams and others--has been
the whole topic of 3D integration. This is one of the reasons Moore's Law is
intact. At ams, we started that in 2005. We've worked on it for almost a
decade, and we have 3D products in production for more than three years. This
was a very significant advancement for us. 3D integration is barely beyond infancy
and has a very long way to go.
the electronic design automation side? Complete top-down EDA systems and flows
were a major advancement. Twenty-five years ago when Cadence was formed, it was
more or less a collection of single tools. We'd gone beyond layout verification
on a layout plot. Of course SPICE was around in analog. Designers were more
artists than engineers. I guess there was a reason for Cadence to name their
first integrated custom IC design solution "Analog Artist." In general, we've
seen a lot of integration which has made the life of the engineer much easier. That
was closely linked to the concept of handling complexity. Complexity is
increasing every day. There had to be a concept that handled higher abstraction
well and with that, hand in hand, went digital synthesis. That was a key advancement
in 25 years. It enabled that paradigm shift: You could write code and let the
tool translate it into gates and devices, rather than designing at device and
gate level. That had a tremendous impact on productivity.
technology advancement were you looking forward to 25 years ago that either
never came to pass or did so in a way you didn't expect?
that advanced in a different way was a separation of technology spaces. Twenty-five
years ago until 15 years ago, ams was a small company, always a little bit behind
the bleeding edge of technology. But in the 1990s, we were still targeting to
fall behind not more than one and a half to two process generations. This has changed
completely in the last 10-15 years-this separation on the one hand of "More
Moore" and "More than Moore" technologies. We're in the latter category. We're
not in the leading-edge of digital technologies... we look at intelligently integrating
additional components like sensors and delivering comprehensive solutions.
another thing, and if you think back it isn't so much a surprise but it was
unexpected for some: the "Gigahertz Race." In the 1990s, each new technology
generation enabled a doubling of microprocessor speed. Think back 20 years,
each new technology generation produced faster MPUs. Speed of computing in
general went up. This suddenly stopped around 2005 at a process speed of 3-4 GHz.
The reason? Power. Significant processor speed improvements at the expense of
power simply became unfeasible. So we accelerated the concepts of parallelization,
which of course requires software partitioning as well. Cadence APS is a good
example. SPICE was not capable of running many threads in parallel but newer
tools are able to do that.
Q: Talk a little
bit about how ams anticipated or evolved in the context of the integration of
analog and digital into single solutions.
Holzhaider: We found out
very quickly that it wasn't digital as such or analog as such but how to handle
the interfaces between the two. ams was always quite analog or mixed-signal
centric. That's why in our environment methods like hardware acceleration
simulators never really worked out well. Quickly we hit the boundaries to the
analog world and then all of the digital speed-up didn't help. The complete
mixed-signal toolsets and environments are a major achievement recently. Twenty-five
years ago, we hoped that at least something would come up here. Over the years,
ams has worked with a number of tool vendors to help advance the state of the
art in the industry of the world.
Q: Talk about
the company's evolution from ASIC to integrated solutions, and
Holzhaider: ams was founded
more than 30 years ago as a joint venture between an American tech company, AMI
(American Microsystems Inc.), an early ASIC company, and voestalpine AG, an
Austrian steel company that wanted to diversify into electronics. We were
founded as AMI Austria.
started more or less as an ASIC company, but over time, we started to look into
standard products. For about 15 years, we had basically one design department
that served the whole customer space. But 15 years ago, we installed product
lines and business units that allowed us to get into more specialization, more
focus and do more evolutionary programs.
years ago, we pushed into the ASSP arena. We had standard linear product lines, but
we went more to the communication segment, which was more standardized than
automotive or industrial segments. We tried more and more to not only rely on
application-specific ICs but application-specific standard products, which we
defined, developed, and placed in the open market. We always worked with a lead customer
to develop those, so it wasn't a complete blind shot.
more specialized and focused on customers now. That enables us to be a lot more
productive. And we've benefitted from much
more powerful electronic design automation based on more powerful workstations
Q: Where do you
see technology evolving in the next 10-15 years?
Holzhaider: We try to
keep in mind the big picture--where do we want to go as a company? We want to
contribute, to produce electronics systems that serve megatrends. One key trend
is mobility. It could be mobile health for aging people, general mobile
society, autonomous electric cars. These are what we have in mind. For ams,
that means more sensor integration. We integrate quite a few sensors already,
including ambient light, proximity, and magnetic positioning sensors. Sensing
technologies for gases and fluids is an area of growing importance.
the digital world I think overall we'll see more 3D integration. There's more
to be done with active interposers and also 3D integration the way ams does it.
It has many facets. The kind of integration we do is completely different than
what companies do when they stack memories on CPUs with interposers. The common
denominator is that we are all going beyond the X-Y optimization and moving in the
analog world, we'll see more process optimization toward better process
technologies. Feature size is not so much the key, compared with power
consumption, leakage, or noise performance. Of course for leading-edge CMOS
technologies, I think the signs are getting more that indicate CMOS will sooner
or later meet its final boundaries. Then we'll need to talk about technologies
like optical computing or organic materials, and there's a wide field to
Q: What do you need to see from your EDA
vendors to keep your engineering teams more productive?
Holzhaider: With ams, we
have a unique approach to 3D integration and we can sort of handle that from
the development point of view. Generally we are starting to offer this 3D
technology to the outside world. I think a 3D business process for non-IDMs is
one big challenge for the years to come. For example, how do you enable a
fabless company to start designing and making use of 3D technology? Different
parts of the technology will come from different vendors. It will be a
challenge when it comes to the methodology, the collaboration. Whose fault is
it when the system fails? How do you test the whole system? This will require a
advancements in design automation technology would you like to see in the
Holzhaider: Certainly an
integrated analog-mixed signal environment was key. Digital synthesis was, of
course, a major step. What we really gained was, as the analog and digital worlds got
closer together, more comprehensive design environments emerged. When we were
founded 30 years ago, our parent company AMI was a big player. Like other big
companies, they had their own in-house design automation department. So we used
AMI proprietary tools. It was then that the first workstation companies came
out -- Daisy, Mentor, and Valid, and only Mentor is around today. The formation
of Cadence was a major step in our analog-centric mixed-signal world.
decision we took in the ‘90s to go with Cadence as a comprehensive tool
supplier was a key decision for ams. Overall, it has been a good decision. We've
had our challenges and discussions, but altogether we've fared well.
forward, ESL and higher-level design abstraction is a big topic, of course. But
I have to say at ams, due to our "More than Moore" specialization, it is not our
key issue. The processes that we're integrating are sometimes mature and older
technologies, but we're not at that top level where hardware-software
partitioning is our key issue. Of course when you're talking systems that are
maybe barely possible at 28nm and force you do 20 or 14nm, then in the billion-gate
arena it's a key issue.
our environment, we have different priorities. If we can improve signal-to-noise
level by a factor of two, or if we can decrease leakage or improve matching a
little bit, then that may have a much higher impact on our business than
talking about system partitioning.
a MEMS microphone, for example. We shipped over one billion pieces of MEMS
microphone amplifiers last year. An advancement of a few percent in sensitivity
or leakage may have quite a significant impact to our bottom line.
-- 25th Anniversary: Hogan on EDA History and Three
-- Video: Cadence Founder Jim Solomon on Company History,
and What EDA Needs Today
-- 25 Years of Innovation: Then, Now, and the Road Ahead