Circuit maker

Conard Holton

Tatsuo Izawa is president and CEO of NTT Electronics (NEL). He began working in the research division of NTT in 1970, developing vapor-phase axial deposition for optical fiber manufacture and the silica-based planar lightwave circuit (PLC). He was presented with the John Tyndall Award at OFC 2001 for his pioneering work in these fields. Izawa has a doctorate in electronics engineering from the University of Tokyo.

WDM: NEL has established itself as a leading supplier of components for WDM systems. What core technologies do you believe are most essential for future innovations?

Our core technology for passive components is the PLC. Based on this technology, the technical development headquarters is planning a hybrid-integration project. Specifically, integration of AWGs [arrayed waveguide gratings] with SOAs [semiconductor optical amplifiers], EA [electro-absorption] modulators, and MEMS devices into one board. For future innovation, large-scale integration of many functional components on a single chip will be needed. Our goal is not to reach the target by a single technology but to utilize hybrid technologies by combining the best materials that match the target.

WDM: Planar waveguides can be made from several different materials. Which do you believe is the best for now and which for the future?

Izawa: In our opinion, silica-on-silicon will be the best material for the near term and long term. Indium phosphide circuits based on wafer fusion and/or heteroepitaxy would be one possibility for small-size and low-cost integrated-optic devices. However, very high controllability issues related to refractive indices and lithography remain to be solved.

WDM: Do you feel that channel density or channel speed is more important for the next generation of WDM systems?

Izawa: Although it will depend upon the application and both are important, we are now focusing more on developing higher channel density than speed. This is because the high-density WDM market is much larger than the high-speed one. WDM systems having 80 to 160 channels will dominate the long-haul market for the time being. NEL is able to provide 40-channel AWGs with 50-GHz channel spacing in large volume, either in the C- or L-band. As many as 80 channels have been realized in an AWG chip.

We already have requirements for 25-GHz spacing from various laboratories. Some of them have been shipped, and we have received positive feedback from the customers. Since the modulation speed primarily limits the channel spacing, 25-GHz spacing may be the minimum for 10-GHz modulation. For high-speed WDM systems, NEL has shown 40-channel AWGs with PANDA fibers at NFOEC 2001.

Let me explain our approach to AWGs for next-generation WDM systems. One type is a very low-loss AWG, which enables the system manufacturers to reduce the number of in-line amplifiers in long-haul systems. We are hoping that it will replace thin-film filters in metro systems as well.

The other type is a motherboard that integrates optical components and electronic circuits on one board, which was triggered by the demands of system manufacturers. NEL has recently hired electronics engineers and is having them cooperate with our optical engineers. So far, we have received orders for more than 100 boards. We think this trend of integration will become more popular year by year.

WDM: There are substantial challenges to producing high-speed systems. When do you think we will see commercial 40-Gbit/s systems and where?

Izawa: 40-Gbit/s systems are very attractive in the market. Honestly speaking, there are some obstacles to be overcome even in the component levels such as ultra-high-speed electronic chips, compensators for chromatic and slope dispersion, gain-flattening filters, high-power lasers, and so on. In addition, the carriers are cautious about introducing new systems owing to the recent economic downturn. It is, therefore, too early to say when we will see commercial 40-Gbit/s systems—in particular, long-haul WDM telecommunication systems. 40-Gbit/s systems need at least two more years to compete with the conventional 10-Gbit/s in prices.

NEL has spent a lot of resources on the development of 40-Gbit/s electronic devices, particularly indium phosphide FETs [field-effect transistors]. Some of the test and measurement companies have recently released new products for 40-Gbit/s systems. One of them uses our FETs.

We think that 40 Gbit/s will be first deployed in short-reach applications—for example, interface modules for communications between boards, transponders in routers, and switches because such short-distance systems require no additional components such as dispersion compensators or amplifiers.

WDM: Which WDM market is your primary goal: long-haul or metro? And do you see North America, Asia, or Europe as your largest market?

According to market reports, North America remains the largest market for WDM. Its share, however, will drop to less than 50% in the next few years, while Europe and Asia obtain more than a 25% share each. By communicating closely with our subsidiary company, NEL-America, we dynamically and in a timely way make our market strategy meet the regional demands.

WDM: I understand that in the past Japan has not been a strong market for WDM systems because of the large number of fibers available and the fact that many are dispersion-shifted fiber, which is not ideal for WDM. Is this true and is it changing now?

Izawa: It is true that dispersion-shifted fiber [DSF] has mainly been laid in trunk lines in Japan. Recent economic situations have forced NTT to gradually introduce WDM systems for building up cost-effective telecommunication networks. To avoid the problem of four-wave mixing induced in DSF, NTT' s system uses AWGs with unequal channel spacing around zero-dispersion wavelengths. NEL has shipped more than two hundred AWGs in total for this purpose.

As you know, internet traffic is exponentially increasing this year. For example, the number of DSL subscribers is over one million this November and FTTH [fiber-to-the-home] service is commercially available this August, ahead of other countries. Due to these developments, the metro network may become a bottleneck. In regional cities in Japan, DSF and NZDSF [nonzero dispersion-shifted fiber] are mixed. Metro WDM systems will be popular in the near future.

WDM: The race is on for high-speed electronics in optical networks. What roles do you see for gallium arsenide [GaAs], InP, SiGe [silicon germanium], or silicon at 10 and 40 Gbit/s?

For 40 Gbit/s and above, NEL is preparing to have two lineups based on HEMT [high-electron mobility transistor] and HBT [heterojunction bipolar transistor] chips based on InP. The HEMT will be applied to front-end ICs, taking advantage of InP' s ultra-high-speed, up to 50 Gbit/s. The HBT will be applied to multifunctional LSIs because of its low power consumption feature. Samples of the InP-HEMT IC have just been released.

WDM: Will tunable lasers and wavelength converters find widespread application? In what roles and what sort of performance will be required?

Izawa: We are sure that tunable lasers and wavelength converters will play a key role in future optical networks. Wavelength tuning range, stability, and switching time are very important. The application of these devices can be divided depending on their switching time.

The lasers with a switching time of one second will get a large market as spare devices or backup lasers used for conventional WDM systems if they have a low-price and enough reliability, like DFB lasers. It is expected that these tunable lasers will have been tested practically in a communications system.

In the case of lasers with a switching time on the order of millisecond, tunable lasers can be used to change the optical path for system restoration. When the switching speed increases on the order of a nanosecond, then these devices can be used in packet-by-packet wavelength conversion systems, for example. In the early stage, it is expected that simultaneous wavelength conversion of several channels will be put into practical use. Then, channel-by-channel conversion will be used in the next stage.

WDM: You have just opened a new 20,000-m2 production facility in Ibaraki. Do you believe that the worldwide market for optical components will be strong enough to support it in the near future or is it more of a long-term investment?

Izawa: NEL was founded to sell the state-of-the-art devices developed in NTT laboratories. Electronic devices and active optical devices have been fabricated in the production lines near NTT Atsugi laboratories, while passive optical devices have been manufactured in a factory near the NTT Ibaraki Labs.

This arrangement was made because close communication between the researchers in NTT and the engineers in NEL is critical in order to transfer the techniques to realize high-volume production. Last year, NEL board members decided to consolidate these two production lines to streamline production facilities such as cleanroom maintenance and disposal of industrial wastes. In addition, a new factory capable of high-volume production was needed to meet the orders from customers.

We launched the new facility to fabricate optical components as well as ultra-high-speed electronic devices and started shipping them in May 2001. The demand for lasers continues to be strong so that the production lines are full. The sales of this year for the laser will be double that of last year. On the other hand, the orders for AWGs are not as much as we expected.

For GaAs IC fabrication, we are processing 6-in. wafers, which enables us to be competitive in metro or 10 Gigabit Ethernet systems. We expect strong pressure to reduce prices in this market. Finally, we plan to polish our fabrication techniques on the unused, new 6-in. wafer lines to increase yield, preparing for the market recovery expected for the end of 2002.

WDM: Which areas of your assembly and testing are automated? How much of your manufacturing will be automated in the future?

Izawa: Automation is the biggest issue for optical-component manufacturers. Roughly speaking, optical components are manufactured after being processed in three main steps: film deposition, assembly, and inspection. Among the three, the automation of inspection is the most advanced, followed by deposition. Assembly is the most difficult part to automate, primarily because packaging and/or pin distribution are not standardized. In other words, most of our components are custom designed, including optical specifications.

Despite these tough conditions, we have been making efforts to automate common processes in assembly. Specifically, connecting many fibers with an AWG chip has been automated using internally developed equipment. NEL has created a new group this year that is responsible for developing mass production techniques. This group plans to start with automating production lines for active devices.