Component technology enables high-capacity dwdm systems
Recent advances in components for dwdm and edfas promise to make optical networking more economical and more efficient.
Ronnie Chua and Bo Cai e-tek Dynamics Inc.
The growth of the Internet, global deregulation of the telecommunications industry, and lowering of trade barriers have created an ever-increasing demand for greater bandwidth in optical networks. A cost-effective way to deliver bandwidth is to send multiple wavelengths through a single fiber and at very high data rates. The high data rates are currently realized through time-division multiplexing (tdm) systems operating at up to 10 Gbit/sec (OC-192). Mulitple wavelengths are sent through a single fiber using dense wavelength-division multiplexing (dwdm) technology that operates in the 1550-nm wavelength window.
Enabling dwdm component technology
dwdm could increase data-carrying capacity without having telecommunications companies lay new fiber cables. Unlike tdm, which uses shorter light pulses to achieve higher bandwidths on the same wavelength, dwdm technology combines multiple optical signals onto a single fiber by transporting each signal on a different wavelength or channel. tdm does not fully utilize the intrinsic capacity of fiber compared to dwdm, making the economics of moving beyond OC-192 unattractive compared to the terabit bandwidth potential of dwdm technology.
Today`s dwdm is driven by several component technologies, the most prevalent being interference filters, fiber Bragg gratings, and arrayed waveguide gratings (see Table 1). Interference filter technology uses layers of dielectric thin films coated on a glass substrate to combine or separate specific wavelengths in a dwdm system. By controlling the layers of deposit on the substrate, different types of interference effects could be created to produce a diversity of narrowband, wideband, and gain-flattening filters. This technology is flexible because it allows the filters to be cascaded in a variety of ways to produce dwdm modules for add/drop, high-, and low-channel applications.
Unlike other dwdm component technologies, interference filters are totally passive and do not require the complexity and added cost of temperature control. Furthermore, they can provide 40-dB or higher isolation on adjacent or nonadjacent channels with very low insertion and polarization-dependent losses. It is perhaps the most cost-effective option for up to 16 channels and the only dwdm component technology suitable for wideband applications. But for dwdms with narrowband or dense channel spacing requirements of 100 GHz (0.8 nm) or smaller, controlling the appropriate layers of thin-film deposits becomes increasingly difficult. This method requires a lengthy development phase to produce all the individual wavelength channels in the International Telecommunications Union (itu) grid.
Fiber Bragg gratings are produced from thousands of refractive index modulation sections that are imprinted onto the core of photosensitive fibers using ultraviolet lasers. This process creates interference patterns that reflect specific channels or wavelengths. The technology is capable of producing dwdm filter shapes with very steep slopes and high channel isolation. The drawback is that it requires costly optical circulators or an interferometric Mach-Zehnder setup to pass selective wavelengths. This complication is compounded as channel count increases, often in direct proportion to cost. Another disadvantage is the need for temperature control because of the thermal sensitivity of fiber Bragg gratings.
Arrayed waveguide gratings, however, are fabricated by depositing thin layers of silica glass onto silicon wafers. The fine silica glass layers are processed at high temperatures and patterned into waveguide circuits by using photolithography before the wafers are diced into individual circuit chips. The circuits are designed to direct each wavelength onto an aligned silica glass block containing multiple fiber outputs. Because optical waveguides are patterned onto silicon in much the same way that electronics are integrated on a computer chip, the cost of arrayed waveguide gratings is not proportional to channel count. Thus, this component technology is the most cost-effective for producing dwdm with very high channel counts in a compact package.
The main disadvantage is that performance-arrayed waveguide gratings generally have an inferior filter passband, a higher polarization-dependent loss, and poorer nonadjacent channel isolation compared to other existing component technologies. As a result, moving beyond 16 channels sometimes requires the aid of interference filters. Like fiber Bragg gratings, the thermal sensitivity of arrayed waveguide gratings also necessitates temperature control of its planar substrate.
In addition to the three main technologies, other novel techniques for fabricating dwdm components also are commercially available. A few companies have introduced dwdms using fused biconic taper (fbt) technology combined with fiber Bragg gratings. A hybrid combination of interference filters with fiber Bragg gratings to achieve narrowly spaced 50-GHz dwdms has also been demonstrated.
But the types of component technology used depend on the application of the dwdm system. For instance, long-haul terrestrial or undersea systems designed to carry a more or less predictable range of data traffic require dwdm components that are highly reliable and uncompromising in performance. But in metropolitan applications where service providers must offer a whole range of data formats and bit rates, the main requirements of dwdm components are flexibility, channel scalability, and low cost.
Enabling optical amplifier technology
Theoretically, dwdm could effectively transport multiple-wavelength signals over a band as much as 100 nm wide around the 1550-nm window (see Fig. 1). But today`s wdm system could only take advantage of the conventional band (C band) between 1535 and 1565 nm because of the limitations of the erbium-doped fiber amplifier (edfa).
The edfa is a critical component for extending the reach of wdm networks because it amplifies signals in the same 1550-nm window. It works by relying on either a 980- or 1480-nm pump laser to excite a rare-earth element known as erbium that is doped into a piece of fiber. When a transmission signal in the 1550-nm window passes through the same fiber, the excited ions will amplify the signal as it exits the amplifier. The advantage of the edfa is that it performs this amplification in the optical realm, unlike regenerators that require optical-to-electrical and electrical-to-optical conversions before the signal is amplified.
But the edfa only has sufficient gain in the C band covering a 30-nm bandwidth, limiting the number of channels that a dwdm system could effectively carry. A dwdm system with 100-GHz channel spacings, for instance, could only accommodate about 40 channels on the standard itu wavelength grid. Narrowing the channel spacings to 50 GHz could potentially double the channel count in the same bandwidth, but this poses technical difficulties if each channel has to be upgraded from carrying OC-48 (2.5-Gbit/sec) data rates to OC-192 data rates. Furthermore, a 50-GHz dwdm requires improvements to enabling components such as filters and laser transmitters.
Recent studies have demonstrated a potential solution to the problem by extending the amplification bandwidth of the optical amplifier to incorporate the longer wavelengths (L band) around 1570 to 1610 nm. At ofc `98, Lucent Technologies presented an ultra-broadband amplifier solution that uses silica-based erbium-doped fibers in a dual-stage amplification architecture. It works by amplifying all wavelengths as a group before they are separated into the C and L bands. The L band signals undergo two further amplifications before they are recombined with the C-band signals. The dual-stage amplifier is pumped by higher-power 980- and 1480-nm lasers to achieve a 25-dBm output and 6-dB noise figure.
By broadening the edfa`s amplification bandwidth, the dwdm system capacity could be increased to achieve terabit-per-second transmission rates, as demonstrated by Pirelli`s high-capacity 128-channel transceiver system at the recent supercomm `98. To achieve these rates, optical components used in the edfa must be developed for broadband performance.
Enabling edfa component technology
An edfa system consists of inline fiber isolators and pump wavelength-division multiplexers (see Fig. 2). The isolator prevents backreflection of light that might degrade a transmission signal that passes through the edfa. The wdm is necessary to combine the 980-nm pump energy with the 1550-nm signal in the amplifier. These critical components are mostly available for conventional edfa systems operating in the C band. But recent research by e-tek Dynamics has produced isolators and pump wdms that cover both the C and L bands for edfas in high-capacity dwdm systems.
The ultra-wideband isolator consists of a pair of collimators, wedge polarizers, Faraday rotators, and magnets. As the first collimated beam enters the birefringent wedge, it is separated into two polarization states at an angle normal to the wedge`s incident plane. The Faraday rotator then turns the optical axis of both polarization states by 45 before they are recombined through a second birefringent wedge that has an axis oriented at 45 to the first wedge. This arrangement enables light to pass through with minimal losses before it is refocused into the output fiber collimator. Light propagating in the opposite direction passes through the second wedge and is similarly separated into two polarization states, but at a refracted angle not normal to the wedge`s incident plane. The Faraday rotator then causes both states to rotate by another 45, thus displacing the beams` angular propagation, which prevents light from being refocused into the input fiber.
In this architecture, a polarization-insensitive isolator assembled with specially chosen materials could operate with at least 45 dB of isolation across a wide 1530- to 1620-nm wavelength range. Other critical parameters of an isolator in an edfa include insertion loss, polarization-dependent loss (pdl), and polarization-mode dispersion (pmd). Insertion loss directly affects the gain and gain saturation of the amplifier; pdl defines the change in insertion loss as a result of a change in the polarization of light; and pmd measures the transmission delay of the orthogonal propagation of light in the two polarization states (see Table 2).
The ultra-wideband wdm is produced by using an fbt process to combine light at 980 and 1570 nm. Until now, wideband pump wdms were available only in a costly filter-based solution that required a pair of collimators and a thin-film interference filter to separate or combine the wavelengths. The fbt process is an attractive low-cost alternative that only requires two fibers to be fused to create a taper that allows light at the two wavelengths to transfer between both fibers. This transfer is possible because the fiber core has a higher index of refraction than air. The length and diameter of the fused region directly affect the optimal wavelength and performance of the pump wdm.
Because of the sinusoidal nature of the fbt process, the insertion loss generally is not uniform across the signal passband. Thus, commercially available wdms made using the fbt process typically will have a high 0.5-dB insertion loss at the wavelength band edges of 1520 and 1620 nm. In order to overcome this problem, an advanced fbt process that controls the wavelengths at the length and duration of the tapered region has been developed. The result is a pump wdm with low insertion loss and high signal channel isolation across a 1520- to 1620-nm bandwidth (see Table 3).
As dwdm systems move toward higher channels to accommodate the growing demand for greater bandwidth, the wavelength window of opportunity also must open up possibilities for transmitting a wider range of channel signals. While this is limited by conventional optical components, recent research on ultra-wideband edfa and amplifier components has shown great promise in enabling the reality of a wider wavelength window in high-capacity dwdm systems. Such component technology eventually will enable dwdms to fully utilize the intrinsic capacity of the optical fiber. u
Ronnie Chua is assistant marketing manager and Bo Cai is a senior engineer in the Technical Service Group of e-tek (San Jose, CA).