A new report by the author, titled The Market for Optical Components: A Seven-Year Forecast, will be published this month by Laurin Publishing.
C. David Chaffee, Contributor
Robust growth is anticipated and, in fact, already occurring in elements of the optical
components market. When compared with past quantities, one area of extraordinary
growth is lasers, the sources or drivers of any fiber optic network, and their accompanying
receivers. In fact, the fiber optics telecommunications industry has never seen
the mass production of these components that is currently under way. These tiny,
grain-of-salt-size devices that pulse billions of times per second deliver a binary
code through optical fiber that enables voice calls, video transmissions and data
messages.
Laurin Publishing (LP) believes that a central reason for this
acceleration is the enormous global market for fiber-to-the-x (FTTx) applications.
However, there are other drivers, including data centers and other short-reach applications.
Laser manufacturers do not always know where the devices they make end up, especially
in the volumes that are being dispensed.
Once, it would have been considered substantial volume if an active
components vendor could make 10,000 transmitters in a year, but the industry has
reached the point where some manufacturers now make more than 1 million lasers per
month. The report details the level of production and the corresponding downward
price spiral that continues to affect this industry.
These lasers remain relatively sophisticated devices despite the
massive volumes being produced. They transmit at a narrow or fixed wavelength, usually
in the C-band for distances of up to 20 km, with very high reliability and operating
at a versatile temperature range with an extremely low margin of error. Manufacturers
say each one of the sources is tested prior to shipment.
The $10 laser is coming
As the report will show, prices have dropped precipitously. In
fact, several manufacturers told us that they now can manufacture the lasers for
about $10. This fabled price goal for the laser driver goes back to the first days
of fiber optics, suggested by founder Charles Kao and other early pioneers as being
necessary if the technology were ever to reach its full commercial potential. The
fact that we are approaching this level of cost for such a complex and multifaceted
device represents a milestone in the optical transport industry.
We must mention one caveat. Retail clients for these transceivers,
including system vendors and, finally, carriers, are not paying these prices, but
rather closer to $25 or $30 per unit. That includes the cost of the transceiver,
plus packaging and shipping.
For the purposes of this article, we will discuss three manufacturers
who make at least more than 1 million of these lasers in some months: BinOptics
(US), CyOptics (US) and Mitsubishi (Japan).
Before we get into the details of these operations, it is important
to make three other points:
First, these vendors make this large quantity of lasers and detectors
directly for the telecom market. Other optical components vendors, such as Finisar
and Oclaro, both based in the US, can make even more lasers – Oclaro has been
known to make 1 million in a week – but these are vertical-cavity surface-emitting
lasers (VCSELs). VCSELs are used mainly in data communications and in consumer applications
and do not have the level of sophistication of the distributed feedback lasers made
in quantity by the aforementioned vendors.
Second, other vendors have large operations that work directly
with these suppliers. Taiwan-based Delta and US-based Ligent, for example, compile
hundreds of thousands of transceivers (a laser and detector packaged in the same
unit) every month, largely from the lasers and detectors that Mitsubishi and other
vendors make. Delta provides transceivers for Gigabit passive optical network (GPON),
Ethernet passive optical network (EPON) and Active Ethernet (AE) and at wavelengths
that include 850, the coarse wavelength division multiplexing slots, and 1310, 1490,
1510 and 1577 nm.
Third, despite the current quantities and price levels, this is
an ongoing growth market, as is further defined in the report. One new vendor soon
to be part of the equation is Onechip Photonics, based in Canada, which formally
announced products at OFC/NFOEC 2011 in March. With a highly experienced team of
optical component experts, the company will be making large quantities of transceivers
using photonic integrated circuit (PIC) technology.
BinOptics and Mitsubishi have significant manufacturing operations
in Asia. CyOptics, on the other hand, operates the last US-based large-scale turnkey
commercial InP fabrication facility in Pennsylvania.
Trend areas: PICs, fabless
Two terms you will hear frequently in coming years when it comes
to these laser and components manufacturers are PICs and fabless. These are both
trends, however. Currently, the three manufacturers cited all have their own fabs
(fabrication facilities) and, as cited below, there are still important reasons
for that control.
PICs allow vendors to make lasers on wafers, so that literally
hundreds can be produced on one board. This is only one of numerous examples where
advances from the silicon industry have immeasurably helped the fiber optics industry.
To make PICs, vendors need fabs to make the wafers. Why, then,
do specific vendors advertise themselves as being fabless? The reason is that costs
can be held down if the vendor outsources the wafer manufacturing to a fabrication
facility that specializes in that work. The vendor, then, does not incur all of
the costs associated with running its own facility.
The relationship between the vendor and the fab owner is a critical
one if vendors are going this route, LP has found. Fabs make chips for many types
of applications, including consumer products such as cell phones, CDs, automobiles
and computers.
We also should point out that some vendors, including BinOptics,
CyOptics and Mitsubishi, do have their own fabs, giving them greater control over
the product and perhaps reducing problems that may occur in facilities that have
multiple customers.
In fact, one reason the optical components industry experienced
a slowdown as it was ramping up in the second half of 2009 and in 2010 was that
these fab facilities already were booked with other business. There are only a finite
number of chip vendors, and the widespread deployments in consumer products have
caused a global drain in some instances. The report observes that some optical transport
vendors have designed their own chips and contracted with fabs to make them to their
specifications and timetables.
What are some of the specifications that customers require for
these popular lasers? LP found that they require long wavelengths, generally in
the 1310- to 1550-nm range. The less expensive devices also must be able to push
signals out to 20 km without the need for a repeater. Two popular materials for
making the lasers are indium phosphide and gallium arsenide.
Another consideration: Both CyOptics and Mitsubishi do the epitaxial
growth internally, while other vendors do not. They again believe that control over
this growth is important for pricing, security and PIC integration.
FTTx lasers: Three types
There are generally three flavors for these lower-priced, higher-quantity
transceivers that conform to the architectures of the fiber-to-the-home (FTTH) networks
they serve: GPON, EPON and AE. These largely are replacing BPON (broadband passive
optical networking), which was the first popular PON architecture in some markets.
Initially, EPON transceivers were the least expensive of the three
to make because they conformed to the looser Ethernet standard. In fact, when Verizon
decided to adopt the GPON standard for its FiOS buildout, it was adding cost to
its network – at least at that time. However, GPON is considered to have higher
reliability.
AE architectures are very exciting to some people because they
use a laser to every residence (whereas PON transceivers are shared with multiple
users). This, of course, will add to the number of lasers that are now being produced,
which manufacturers view as a positive development. AE was seen as being more expensive
than GPON or EPON because more transceivers were being used; it also was seen as
being more problematic because the introduction of additional active devices was
viewed as causing the networks to be less stable, resulting in higher maintenance.
However, the report finds that, as the laser cost comes down,
AE is seen as being more economical and the cost barrier as disappearing, just as
it has between GPON and EPON. It also finds that, as laser reliability improves,
the need to maintain AE networks lessens.
LP believes there are two lessons for the optical components vendor
here. First, the successful vendor will be able to provide lasers to all three markets:
GPON, EPON and AE. LP believes that not only will all three continue to thrive as
the global FTTH market continues to explode, but that a carrier often will require
two for the same job. This probably will be either GPON or EPON together with AE.
To expand on this point, the US has largely used BPON and then GPON because that
is what Verizon used as part of its FiOS buildout, which has accounted for the large
majority of US FTTH connections. However, cable TV companies are starting to bring
fiber to the home in America and are committed to an EPON architecture.
The second lesson is that the successful optical component vendor
will know how to continue to ramp up. AE already is starting to catch on in Europe,
and global growth will continue unabated. If vendors could make hundreds of thousands
of lasers every month five years ago, and millions of lasers monthly now, it is
not a significant stretch to conclude that they will need to make 5 million or even
10 million per month five years from now.
Not surprisingly, the more sophisticated that transceivers become,
the more they cost; e.g., transceivers transmitting signals greater than 20 km cost
more, as do transceivers operating at higher data rates such as 40 or 100 Gb/s.
As we discuss in another section of the report, tunable lasers
also cost significantly more than fixed-wavelength types.
What is the main differentiator between transceivers that cost
more and those that cost less? LP believes it is how well transceiver manufacturers
can stay ahead of the China-based manufacturers. Once the China-based transceiver
manufacturers can make the component in volume, the price reaches its low point,
at least for now. LP therefore encourages the transceiver manufacturer to be innovative,
to continue to build in the right advances that will provide a uniqueness that customers
will find necessary.
Just as important, LP believes that it is critical for optical
component vendors to develop relationships throughout the world as the market evolves.
FTTH transceivers are kind of like telephones – or at least
the way telephones used to be. Someday, every residence will have one. As we point
out in the report, our industry literally is looking at the potential for billions
of these devices.
Meet the author
C. David Chaffee is the principal author of The Market for Optical
Components: A Seven-Year Forecast.