The high performance and simple structure of 50-GHz thin-film filters provide an effective alternative to arrayed-waveguide-grating and fiber-Bragg-grating technologies.
By J. J. Pan, Feng Qing Zhou, And Ming Zhou
Given its low cost and high reliability, it is not surprising that thin-film-filter technology has come to dominate the 200- and 100-GHz dense-wavelength-division-multiplexing (DWDM) markets.1 Competing solutions such as arrayed-waveguide-grating (AWG) and fiber-Bragg-grating (FBG) technologies have trailed behind as a result of system complexity and manufacturing difficulties. Surprisingly, in the newly emerging market of 50-GHz systems (which is ushering in a new era of high-bit-rate data communications), the situation is reversed. For all their potential benefits, thin-film filters have yet to catch on as a result of the difficulty in achieving the close manufacturing tolerances necessary to ensure a high level of performance from 50-GHz thin-film devices.
Controlling temperature stability, passband flatness, insertion loss, skirt sharpness, and chromatic dispersion becomes both more crucial and more difficult in the manufacture of 50-GHz thin-film devices than in the manufacture of their 100- and 200-GHz counterparts. As the channel spacing narrows, even the smallest degradation in a thin-film filter can dramatically impact performance. Narrow passband width, for example, distorts the signal pulse shape when it passes through the filter, while chromatic dispersion broadens the pulse width. And thermal wavelength drift not only shrinks the useful bandwidth, but also decreases channel isolation. Unfortunately, most thin-film manufacturers have not yet modified their manufacturing processes to address these issues. As a result, 50-GHz DWDMs currently on the market incorporate AWG technology and/or optical interleaver technology complemented by FBGs rather than thin-film filters.
This does not mean that AWG or FBG technology is better suited to 50-GHz applications than thin film. On the contrary, AWG and FBG technologies both have significant performance limitations that cannot be eliminated without incurring much higher costs and system complexity. This is illustrated in a comparison of four different 1 x 4 50-GHz multiplexers/demultiplexers (see table). The first uses FBGs with 100-GHz filters and a circulator; the second uses FBGs in transmission mode, 100-GHz filters, and a 3-dB optical coupler with an isolator to block the back reflection; the third is a 50-GHz AWG; and the fourth is a 50-GHz thin-film filter. The thin-film filter exhibits the best performance and simplest structure among the four.
By selecting chemically and mechanically stable substrates, improving film-deposition techniques, optimizing optical monitoring systems, ensuring coating uniformity, and minimizing stress buildup in filter coatings, thin-film technology provides a cost-effective and reliable 50-GHz solution, especially in multiplexing and demultiplexing, optical add/drop multiplexer, and dynamic add/drop multiplexer modules (see Fig. 1).
Comparing AWG and FBG technologies
Although DWDMs using AWG technology can meet the requirements of 50-GHz devices, they frequently suffer from performance limitations and manufacturing yield issues. Phase errors arising from manufacturing mistakes, lack of material uniformity, temperature variation, and stress-related performance degradations are all commonplace. Arrayed waveguide gratings are also cost-prohibitive unless used in large-channel-count devices.
Fiber Bragg gratings, too, have their drawbacks. They can only be used in small-channel-count 50-GHz devices and suffer from high insertion loss and chromatic dispersion. Although recent developments in FBG technology can greatly improve the dispersion performance, their complexity with regard to fabrication and packaging and the frequent need for circulators to extract the signals dramatically increase system costs.
Thin-film filters
Typically, today's high-capacity optical-transport systems use a 10-Gbit/s signal rate with 50-GHz-spaced DWDM systems. If thin-film filters are to provide these systems with cost and performance benefits over competing technologies such as FBG and AWG, they must deliver a high level of performance, particularly minimal insertion loss of less than 1.5 dB for a single-filter device, a chromatic dispersion of less than 50 ps/nm, channel isolation of at least 25 dB, and a clear bandwidth of ±10 GHz around the International Telecommunications Union (ITU) grid.
Given these parameters, what makes manufacturing high-performance 50-GHz thin-film filters so difficult as opposed to 100-GHz filters? To understand this, it is necessary to first observe the bandwidth requirements of 50- and 100-GHz DWDM filters. These requirements include margins set aside for center-wavelength thermal drift, measurement error, ITU-grid alignment error, and the aging effect. They also assume a 1-pm/°C thermal wavelength shift. The bandwidth requirements for a 100-GHz filter, then, are 0.4 nm at -0.5 dB and 1.2 nm at -25 dB. For a 50-GHz filter, the requirements are 0.3 nm at -0.5 dB and 0.5 nm at -25 dB.
The difficulty in manufacturing a filter can be measured, expressed as the figure of merit (FOM), which equals the bandwidth at -25 dB divided by the bandwidth at -0.5 dB. The smaller the FOM, the more difficult it is to manufacture the filter. By this calculation, a 100-GHz filter has an FOM of 3.0, whereas a 50-GHz filter has an FOM of 1.7, making 50-GHz filters more difficult to manufacture than 100-GHz filters. It is still possible to manufacture high-performance 50-GHz filters, however, if five manufacturing issues are addressed. These five issues are filter-substrate stability, stress in the filter introduced during the film-deposition process, stress in the filter induced by packaging errors, optical-monitoring-system accuracy, and thin-film-coating uniformity.
To make the manufacture of 50-GHz filters easier, their thermal-wavelength drift performance must be improved. This can be achieved through selecting a filter substrate with a high coefficient of thermal expansion (CTE) and by reducing stress buildup in the filter coating during film deposition. To reduce the thermal wavelength drift of the filter, the substrate should be a chemically and mechanically stable optical glass that has the highest available thermal expansion coefficient (typically a CTE between 9 and 11 ppm/°C).2 By doing this, the thermal wavelength-drift value changes to less than 0.5pm/°C and thereby relaxes the -0.5-dB bandwidth to 0.25 nm and the -25-dB bandwidth to 0.55 nm, which results in an FOM of 2.2. This is still lower than the FOM for 100-GHz filters, but is a significant improvement over the original FOM.
Ensuring the environmental stability of the thin-film layer is also key to performance, and necessitates the use of an energetic deposition process. The most commonly used techniques are ion-assisted deposition and plasma-assisted electron-beam evaporation processes.3 Ion-beam-sputtering technology is also widely used, and some thin-film manufacturers are using proprietary coating machines that use various low-pressure reactive magnetron sputtering deposition processes. No matter which technique is used, however, stress buildup in the filter coatings is a major consideration. Stress not only tears apart the substrate, but also affects filter performance directly by causing spectrum shape deformation and center-wavelength thermal drift. Reducing this stress works toward increasing the FOM of 50-GHz filters.
Packaging can introduce as much stress into the thin-film layer as the thin-film deposition process itself but is frequently overlooked as a source of stress. If it is not eliminated, it can result in greater thermal center-wavelength drift and spectrum shape distortion in the packaged device than in the raw filter.
The optical monitoring system
An effective optical monitoring system is crucial to ensuring thin-film filter quality because it determines whether a coating run is a failure or a success. For multicavity Fabry-Perot filters, previous studies have shown that direct transmission turning-point monitoring is the best approach because of its error-compensation mechanism.4, 5, 6 Total compensation is only achievable when the monitoring light is monochromatic, however. In a typical coating machine, the monitoring light always has a finite bandwidth, making the compensation incomplete. The bandwidth of the optical monitoring light comes from three sources: the bandwidth of the monochromator (or linewidth of the laser source), the effective bandwidth caused by nonuniformity of thin-film thickness inside the monitoring light spot, and the effective bandwidth caused by the incident angle cone of the monitoring light on the substrate. For error compensation, a smaller bandwidth is preferable.
The intensity of the monitoring light, however, decreases proportionally with bandwidth for a white-light source, resulting in an increase in the turning-point error. A tradeoff must be made to optimize the coating monitoring process. For a laser source, problems arising from long coherent length and polarization instability must be solved first to fully exploit its narrow-linewidth advantage.
Thin-film-coating uniformity
The useful area of a thin-film-filter coating is determined by its uniformity. Uniformity here has multiple meanings. The basic uniformity requirement is that the thickness distribution within one coating layer must be very consistent at the substrate area. Because the substrate is relatively small compared to the coating chamber, this is not difficult to achieve as long as the evaporation distribution and plasma distribution are under control. The most critical uniformity issue is the thickness-distribution variation from layer to layer. Although the thickness from the substrate center differs from that of the substrate edge, only the filter's center wavelength will shift and its shape will not change as long as the layer-to-layer distribution does not change. If layer-to-layer distribution does not remain consistent, then the filter shape will be corrupted when moving away from the monitoring point.
By improving film-deposition techniques, optimizing optical monitoring systems, ensuring coating uniformity, and minimizing stress buildup in filter coatings, thin-film technology can provide a more cost-effective and reliable 50-GHz solution than any other technology (see Fig. 2). Thin film dominated the 100- and 200-GHz markets. As the demand for 50-GHz devices grows, thin-film-filter technology can conquer this market as well.
REFERENCES
1. Global Market Forecast: Optical Components for Terrestrial DWDM Applications, RHK (December 2000).
2. H. Takashashi, Appl. Opt. 34, 667 (1995).
3. W. Latimer, WDM Solutions, 51 (September 2001).
4. H. A. Macleod, Optica Acta 19, 1 (1972).
5. J. J. Pan et al., Proc. 41st Annual Tech. Conf., SVC (Boston; 1998).
6. J. J. Pan et al., "Optical coating computer simulation of narrow bandpass filters for dense wavelength division multiplexing," Optical Interference Coatings, OSA Topic Meeting (Tucson; 1998).
J. J. PAN is CEO and president, and FENG QING ZHOU and MING ZHOU are engineers at Lightwaves2020 Inc., 1323 Great Mall Dr., Milpitas, CA 95035; e-mail: [email protected].