Related press releases
Related research
The world's first complementary metal oxide semiconductor (CMOS), silicon-based, tunable optical waveguide equalizer - an optical filter
:: 05 April, 2007
The fabled Bell Labs, in one of its first major technological tours de
force as a French company, has demonstrated the world's first complementary
metal oxide semiconductor (CMOS), silicon-based, tunable optical waveguide
equalizer - an optical filter, if you will. Translation: In about four years the
world will see 10 Gb/s broadband pumped out by a transceiver that today can
deliver only 2.5 Gb/s, and higher speeds at similar ratios are coming (a 100
Gb/s prototype chip is imminent), with an obvious equal effect on cost.
In other words - broadband may be about to get a heck of a lot cheaper.
What Bell Labs has done until now was only achieved in highly exotic (read
"expensive") technologies - InP, LiNbO3, InGaAs, SiO2-PLCs and MEMS, to name a
few. CMOS, on the other hand, is the workhorse of the semiconductor industry,
the technology used for everything from chips that cost just pennies to advanced
microprocessors.
Broadband Business Forecast met with Alice White, Bell Labs' vice
president for enabling physical technologies, and Sanjay Patel, technical
manager of integrated photonics research, for a look at the newest invention on
the broadband block.
Giving Credit Where It's Due
Before delving into the details, though, it's only fair to point out that
this isn't exclusively a Bell Labs achievement. Rather, it builds on what Bell
Labs itself praises as the "revolutionary" work done under the Electronic and
Photonic Integrated Circuits (EPIC) program, which is funded by the Defense
Advanced Research Projects Agency (DARPA). That program is led by BAE Systems in
partnership with MIT, Applied Wave Research and Bell Labs, with Cornell and
Columbia Universities also on the team.
"The EPIC project achieved the critical first step in building the
foundation for this new breed of devices," says White. "We've applied our core
competency in optics and expertise in chip design and telecommunications
technology towards realizing the full potential of silicon-based optical
networking by not only creating circuits that can carry optical signals, but
providing the control to modify those signals, which is a much more
sophisticated process."
And while the new photonics-on-CMOS circuit unveiled by Bell Labs is just
one of the series of technologies being researched under EPIC, as it turns out,
it could be the one with the biggest payoff.
Standard Technology
That's because of the use of CMOS, indeed standard CMOS processes. Bell
Labs' initial circuits were made in BAE Systems' CMOS foundry - once upon a time
an IBM semiconductor facility - rather than in a prototype lab setting using an
expensive and exotic process. "The component we're talking about was made in a
factory," White says.
Patel picks up the thread, saying, "What excites me about this technology
is we can integrate photonics on the same technology platform" as standard
semiconductors. That, he continues, "opens up a whole vista of applications you
couldn't think of doing with boutique technologies" because of how expensive it
would be. To give a feel for the economics involved, Patel told us 100
modulators can be made using CMOS process, in place of just one that might be
made using the so-called "exotic" technologies.
Indeed, he says, what's now the proved ability to use CMOS to make
optical components opens up the door to "many things I would have blown off as
unfeasible with the old technologies." The optical waveguide equalizer is just
the first of a string of optical components that might be made in CMOS, he says.
Those components, it should be noted, are not great big CMOS chips.
Rather, they are optical cells that can be put on a standard CMOS ASIC along
with electronic cells, in the end creating custom ASICs for a variety of
functions. "You might decide that part of the processing of the signal is best
done photonically and part is best done electronically," White explains, and,
thus, with the existence of optical cells for the first time, the use of ASIC
technology lets designers do that on a single CMOS chip for the first time.
Bye Bye, All-Optical
One obvious question is how Bell Labs' invention fits in with what was
supposed to be the Holy Grail of the use of an optical pipe to transmit
broadband - an all-optical solution or, barring that, an optical-electronic-
optical (OEO) solution with separate components to turn an optical signal into
an electronic one, perhaps for equalization with the equalized signal then sent
to another component for conversion back into an optical signal.
"By the time we're finished, they're not going to be talking about all-
optical or OEO," White says. More precisely, the OEO is going to be on a single,
relatively inexpensive ASIC CMOS chip with the opto and electronic circuitry all
on the same CMOS wafer.
Not that the trick is an easy one, the Bell Labs executives quickly add.
"Photons are a lot more finicky than electrons," says Patel. "The high contrast
of silicon creates a lot of loss (that) you have to reduce to a practical
number."
Without delving too deeply into the technology, we note that Bell Labs
explains it did trick using "a new control configuration that uses a single
voltage to adjust the signal equalization and an innovative architecture to
realize complex responses in a low order filter." A symmetric Mach-Zehnder
interferometer is used with two tunable-ring resonators.
The Path To Market
So an obvious question is: What do you build with this neat technology,
and when can I get it?
Bell Labs outlines some of the implications of its work in its formal
written briefing on the technology: "Synergistically combining such optical
filters with on-chip electronic circuits can provide a commercially viable path
to providing reconfigurable, low-cost, low-power consumption devices that could
fit into small form factor pluggable modules. These new devices are also ideal
for dual-use applications in systems that route data over both electronic and
optical networks - depending on the most appropriate delivery and/or transport
format."
"We're probably looking at three to five years" until such ASICs are
available commercially, White adds. (BBF notes that's as little as only 36
months from now -- fast for a technology to move from the lab into existence as
a commercial product.)
One reason, White says, is that Bell Labs was able to get its first
prototype parts made on what is a commercial CMOS production line, proving it
didn't need some exotic process or equipment to do the trick. By doing that, "we
certainly leapfrogged the tech-transfer piece from the research lab to the
factory," says White. And while initially, like any new technology, the cost
will of course be a bit high, "what we'd hope to do is get on the same cost
curve as CMOS electronics is on," he adds.
The Target Customers
As to who might be a potential customer for the Bell Labs invention - the
list will probably start with Alcatel-Lucent itself, these days the French
entity that owns Bell Labs as a result of Alcatel's acquisition of Lucent in a
"merger of equals." Alcatel-Lucent, it should be noted, doesn't manufacture
semiconductors itself. It buys them and, thus, somewhat obviously "we're not
saying that Alcatel-Lucent is going to introduce a silicon based networking
device," says Patel.
Nor will the technology remain the exclusive preserve of BAE, although it
certainly is expected to make the widgets. Rather, under the DARPA funding, the
inventors of the technology are expected to make it available to others. "Part
of the charter from DARPA was that you should be able to license to other
foundries as well," says Patel. "For us as a user of these foundries, the more
the merrier." >>Alice White and Sanjay Patel, Bell Labs, 908/582-8500<<
BBF's Take On The Situation
The way White and Patel tell it, the new technology could make it to
market in near record time, which is why Broadband Business Forecast was all
ears. We don't normally write about esoteric research. However, when something
may be only 36 months away from market, it isn't esoteric any more.
Indeed, we've been writing massive amounts about lots of technologies
that much longer than three years to become a market reality. For instance, look
how long it's taken 4G to spread around the globe. Then there's 802.16 (the
basis of what's being called WiMAX these days), which was first approved as a
standard six years ago and still hasn't seen commercial reality (with all due
respect to those who think they're selling the stuff).
Interestingly, we've also done a good bit of writing about the Holy Grail
- all-optical technology - that only a few years ago lots of folks thought would
be common by now. Maybe it will be some day, but clearly it isn't yet. Indeed,
thinking back perhaps 20 years, the first research and patents started flowing
on all-optical logic gates - in theory, the stuff that could be used to build
microprocessors not only far denser and faster than today's fastest dual-core
Pentiums, but also a heck of a lot cooler (we mean temperature) compared to
electrical devices. Indeed, if memory serves, Bell Labs did the pioneering work
and holds a clutch of patents on that technology.
It would, of course, be nigh unto impossible to detail all the inventions
that have come out of Bell Labs since it was founded in 1925 by what was once
upon a time American Telephone and Telegraph - the AT&T of old or Ma Bell, if
you will. This latest invention, thus, continues that string and, in our
experience, AT&T researchers have been careful about discussing the market
potential of their inventions, unless they were pretty certain about the issue.
This time, we have to admit we were a bit blindsided. While it was no
secret that DARPA had the EPIC project underway nor were the goals of the
project a secret, it wasn't evident until now that the commercial possibilities
being opened up were so immediate.
Bell Labs, we might note, kept quiet about things until presenting its
work during a session at the Optical Fiber Communication Conference & Exposition
and the National Fiber Optic Engineers Conference (OFC/NFOEC) last week, a move
that gave it something major to tout.
That, right now, of course is a bit of a political issue. There are lots
of folks who are worried about the fact that Bell Labs - named, of course, after
Alexander Graham Bell - no longer is American. It became a French-owned
operation when Alcatel bought Lucent in the famous "merger" of equals. The CMOS-
based optics described here hardly proves anything about what the future might
bring for Bell Labs because the project long predates the decision of Lucent to
get into bed with Alcatel
Latest comments
Researchers from Alcatel-Lucent's Bell Labs in New Jersey described how they had developed a versatile guided-wave equalizing optical filter fabricated entirely in a CMOS manufacturing line -- the sort of technology that produces computer integrated circuits today.
This research will enable telecom providers to move from using specialized and large optical networking devices to a new generation of low-cost, mass-produced silicon chips that combine electronics and Photonics in a single chip, opening the door to new optical networking architectures that could usher in new sorts of broadband deployments and applications. Possible new applications include low-cost, mass deployment of fiber-to-the-home; truly meshed optical networks that cleanly switch optical signals between different transmission formats; and the deployment of optical networking into places unapproachable by today's optical networking devices, such as over short-runs or in confined spaces.
The demonstration of the new tunable filter is a critical step in this field of research.
This work could also enable devices that could be remotely reconfigured, simplifying large-scale optical network management or reconfiguration, it said. By simultaneously equalizing the signal intensity profile, compensating for dispersion and switching transmission formats, such agile devices can also help transform the network from a point-to-point architecture to a mesh topology.
While this work represents only a proof-of-concept demonstration (generally, available technologies are not expected for 3 to 5 years), combining such optical filters with on-chip electronic circuits can provide a commercially viable path to providing reconfigurable, low-cost, low-power-consumption devices that could fit into small-form-factor pluggable modules. These new devices are also ideal for dual-use applications in systems that route data over both electronic and optical networks -- depending on the most appropriate delivery and/or transport format.
Posted by: 09 April, 2007 15:55
|
|
|
