Japan's Oki Gets in the Groove
Oki's invention is a silicon microlens that sits between a laser diode and singlemode fiber to boost the coupling efficiency (see Oki Invents Mini Microlens). The microlens is unique because it is small enough to be positioned in an optical V-groove, enabling low cost, passive assembly processes to be used, says Takeshi Takamori, project leader of Oki’s Advanced Devices Laboratory.
Up until now, according to Takamori, coupling technologies for optical modules have fallen into two categories:
At the top end of the scale, there's technology used for high-end, gigabit-class optical units for long haul and longer metro systems. This usually relies on complicated laser sub-mounts that typically need two aspherical quartz lenses and an isolator, filter, or splitter to achieve the required 70 to 80 percent coupling efficiency between fiber and laser diode. Assembly is a highly labor-intensive process that requires customized parts and is expensive even for manufacturers that have moved production over to China, Takamori contends.
Modules for 2.5-Gbit/s systems typically ship for about ¥40,000 (about US$325), with about 60 percent of the cost coming from materials and 40 percent from fabrication, according to Oki. About 70 percent of the fabrication costs come from the tricky alignment process between lens and fiber, says Takamori.
At the bottom end of the scale are cheap-and-cheerful modules for, say, access systems. For these, many manufacturers rely on a so-called silicon V-groove SMT (surface mount) process. This uses automated machines to place optical devices such as laser diodes, microlenses, and optical fibers on silicon platforms. The fiber slots into a V-shaped groove etched into a silicon substrate adjacent to the laser diode chip. This cheap process enables Oki and competitors to turn out modules for less than ¥5,000 ($40). But the coupling efficiency is typically only about 10 percent, says Takamori.
Oki's new process attains some middle ground. By adding two 125-micron-diameter silicon lenses in the V-groove between the fiber and the laser diode chip, Oki is able to get about 50 percent coupling efficiency, or -3.0 dB output power. Takamori claims that the production cost is about 50 percent less than the high-end, optical lens assembly method.
One reason the method is so cost effective is that the lenses themselves are very cheap to produce. Diffraction (Fresnel) lenses can be made easily with a standard silicon fab process, such as micromachining and etching. A six-inch wafer yields about 10,000 of the tiny lenses with little waste. Oki already has pick-and-place technology that enables it to accurately line up and drop lenses in their slots.
“Yes, we have filed patents. Yes, we are the first to develop this technology,” Takamori assures Light Reading.
“These units can be easily produced; fabrication can be done on legacy lines; and we can produce high-end modules at reasonable prices. It’s a bit like Toyota making a new model car on one of their existing lines."
However, it does appear that at least one other vendor -- Infineon Technologies AG (NYSE/Frankfurt: IFX) -- has come up with something similar. It sells a range of uncooled, Fabry-Perot and Distributed Feedback (DFB) Lasers for applications up to 1.25 Gbit/s, which it claims are produced using a fully-automated micromechanical submount process. "The rather high coupling efficiencies of optical power into the singlemode fiber is realized by using a high-performance etched silicon lens," claims the company's Website.
Takamori is a bit shy about revealing how Oki’s optical module business is doing, compared with rivals such as Alcatel Optronics (Nasdaq: ALAO; Paris: CGO.PA), JDS Uniphase Corp. (Nasdaq: JDSU; Toronto: JDU), and NEC Corp. (Nasdaq: NIPNY). But he does concede that the market is so horrible that Oki is losing money on its low-end modules -- even with what he says is a competitive manufacturing process.
Oki plans to have samples of its microlens available by the end of this month. The company will license the technology to interested competitors and will go to production of its own modules with the new system by next summer, in time for an anticipated recovery in the metro/access markets in 2004, Takamori says.
— Paul Kallender, special to Light Reading