During the boom in optical communications over 1999-2000 there was much investment in new technologies, especially in the area of active optoelectronic semiconductor-based components.
These include not just the traditional — and essential — use of III-Vs materials gallium arsenide (GaAs) and indium phosphide (InP) in the lasers inside the transmitters and the detectors inside the receivers, but also in modulators and more integrated devices such as photonic integrated circuits — instead of less costly lithium niobate (LiNbO3) and silicon-based materials.
Consequently, many technology start-ups sprang up. However, as the downturn started to bite and venture capital finance dried up, many III-V start-ups such as Nanovation Technologies went bankrupt as users played safe with less expensive, conventional technology.
Traditionally, such start-ups are acquired by large existing diversified component manufacturers. But, since the continued deepening of the downturn, even those companies — such as Agere Systems — have been seeking to divest their optical component divisions.
This has opened up opportunities for acquisitions by medium-sized companies. One example is integrated optical component and subsystem supplier Bookham Technologies which, unlike other larger companies, had the cash needed to make acquisitions.
Bookham was founded in 1988 to make its Active Silicon Optical Circuit (ASOC) silicon-on-silicon waveguide technology. This was incubated at the Rutherford Appleton Labs, and transferred to its site in Milton, UK. As well as transceivers, Bookham makes smart passive devices such as:
- arrayed waveguide gratings (AWG);
- mux/demux;
- variable optical attenuators (VOAs);
- electrically tunable VOAs (EVOAs);
- multiplexer-VOAs.
In October, Bookham had to downsize by closing its assembly facility in Swindon, UK and its fab in Maryland, USA. But, since it had no capability in long-haul or III-Vs materials (i.e. no lasers or modulators), at the end of 2001 it acquired Marconi Optical Components (MOC) in Caswell, UK from the retrenching Marconi.
MOC originated as Plessey's Allen Clarke Research Centre in 1968. It made GaAs field effect transistors (FETs) in the 1960s and added optical components in the 1970s. Its fabrication plant, built in 1984 for bipolar transistors on 4in silicon wafers, was converted in 1993 to make GaAs monolithic microwave integrated circuits (MMICs). In 1988 the company became GEC/Plessey then Marconi.
Although Caswell had made good revenues from selling MMICs commercially for several years, in 1999/20000 Marconi created Marconi Optical Components to develop commercial manufacturing of components for optical communications. The Caswell site now employs about 300 staff.
As well as still making MMICs, Caswell makes active components. These are complementary to Bookham's silicon-based components and include:
- InP-based fixed-wavelength 1360–1620nm lasers for high-end, long-haul transmitters;
- tunable lasers;
- GaAs-based 10 and 40Gbit/s electro-optic modulators.
The latter are suitable for long-haul in the C or L bands, unlike electro-absorption modulators, which suffer from chirp and limited tunability and are best suited to short and intermediate reach.
These products are now in the commercial introduction phase, prior to production being ramped up in volume. Thanks to much investment by Marconi before its financial troubles, Caswell has the capability for automated manufacturing on 6in GaAs wafers (its five MOCVD reactors include two which can accommodate 5x6in wafers). This makes increased integration and large chip sizes more economically viable. "Optical integration in GaAs is extremely important going forward," says Bookham's Andy Carter, VP Active Components and III-Vs.
Compared to traditional lithium niobate modulators, which are 30–40mm long, GaAs modulators are only 15–20mm long and can be co-packaged or integrated with the transmitter, says Carter.
For example, a 6in wafer can accommodate up to 250 10Gbit/s modulator–VOAs (compared to 50 on a 3in wafer), including the struc-tures for automatic RF-on-wafer validation (see Fig 1 and cover picture).Carter says that Caswell also has two types of widely tunable laser, both monolithically integrated.
- A three-section 10Gbit/s DBR which can tune (by carrier rejection) over a wavelength range of 10nm (i.e. 16–20 50GHz-spaced channels, or a quarter band) with 20mW output.
- A digital supermode DBR laser (DS-DBR, announced at March's OFC 2002 event) which can tune over the full C or L band. This is based on a three-section laser, but with a rear phase grating with seven reflectors and 6nm-spaced comb peaks. A front grating selects the peaks, so it effectively has seven DBRs plus a switch.
Alpha samples of this DS-DBR are now with customers, with qualification and full manufacturing targeted by Q3/2003.
Carter adds that, for widely tunable lasers to be viable, they have to be of comparable cost to single-wavelength lasers. There is therefore a need for them to be monolithically integrated. Such technology developments are "disruptive" for the long-haul sector, he says.
Integration in GaAs has also yielded a dual-section 40Gbit/s RZ transmitter (exhibited at September's ECOC 2002 event). Currently there is no market for such a component, acknowledges Carter, but Bookham wants the parts ready in order to show that it is a "strategic supplier". While such a function can be done electronically at 10Gbit/s, it must be done optically at 40Gbit/s, Carter emphasises.
Also, at OFC 2002 Bookham exhibited a GaAs optical Differential Quadrature Phase Shift Keying (O-DQPSK) transmitter. Rather than simple "on-off" signal coding, DQPSK coding is phase modulated. Four points in phase space and two bits per symbol allow a symbol rate of just 5Gbit/s for a data rate of 10Gbit/s. Hence, reach is increased, and two OC-192 signals can be transmitted on a single channel with a single module.
The use of DQPSK coding has been restricted by the fibre between discrete components in the transmitter being prone to uncontrolled phase changes, but Bookham's O-DPQSK has several elements fully integrated in GaAs. This has increased transmission distance from 100km to a record 250km without dispersion compensation.
At the end of September Bookham announced the acquisition of Nortel Networks Optical Components (NNOC), which was completed in early November. This, reckons Carter, means that Bookham's combined revenues make it the third largest optical component manufacturer (behind JDS Uniphase and Agere, which is selling its optical components business to TriQuint Semiconductor).
NNOC has sites in Ottawa in Canada, Zurich in Switzerland (ex-IBM Research), Paignton and Harlow in the UK, and Poughkeepsie, NY, USA. NNOC's components include:
- direct modulation lasers for 2.5Gbit/s;
- InP-based modulators;
- receivers;
- amplifiers (incorporating 980nm pump lasers from Zurich);
- transponder modules.
There is some complimentarity between the products (i.e. NNOC's directly modulated 2.5Gbit/s lasers for intermediate-reach and Caswell's InP 10Gbit/s lasers with GaAs external modulators for long-haul.
However, while Nortel is closing its electronic module and tunable laser units, there is still some overlap between the products and activities of Caswell and NNOC. For example, there are assembly activities at both Caswell and Paignton. In integrating Caswell and NNOC, Bookham will therefore focus on savings. But most importantly, NNOC has the capability to make transponder modules.
One consequence of the divestments of Marconi Optical Components and Nortel Networks Optical Components to Bookham, says Carter, is the independence gained from a parent which is also a system manufacturer and therefore demands captive supply. Potential customers are no longer reluctant to do business with a supplier which is also a competitor.