Mike P. Li
Jan Wilstrup
Today`s infrastructure growth in fiber communication systems is largely driven by data transfer increases over the Internet. It is estimated that the peak data rate at a node of a public network in North America already reaches ~1 Tbit/s and the Internet traffic is increasing at a rate of more than 200% per year. Obviously, copper-based transmission media will not satisfy this need at a reasonable cost. And even if optical fiber offers much better bandwidth (for example, ~5 Tbit/s over the 1540-1565-nm window), it will not keep up with the demanding pace unless more fiber is laid.
In this scenario, optical DWDM has become one of the most promising technologies for delivering higher data rates over the existing and future fiber communication infrastructure. Actually, for a point-to-point-based communication system, there are only two ways to increase the data rate: either increase the data rate for the single-channel point-to-point link (for example, OC-48 to OC-192) or use DWDM to add more point-to-point link-based channels over the same fiber. In practice, the optimal approach is probably to use both methods.
The technology trend is to develop DWDM systems containing hundreds of channels with each one running at ~10 Gbit/s or higher. To accommodate such a high number of channels, channel spacing of <50 GHz is needed. As the data rate over a single channel approaches 10 Gbit/s (OC-192) or higher, however, and channel spacing for a DWDM shrinks down to a few nanometers or less (a few hundred GHz or lower), performance issues become pressing. For instance, although ~5 Tbit/s data rate is a theoretical expectation for an optical fiber carrier at the window of 1540-1565 nm, the actual rate is limited to less than 100 Gbit/s in practice due to the fiber dispersion and switching speed of transmitter and receiver.
Advances have been made, such as development of the erbium-doped fiber amplifier (EDFA), which has increased fiber transmission distance to more than 300 km without regeneration and significantly reduced the power attenuation limitation. In the case of dispersion, which is common to both single-channel point-to-point link and multichannel DWDM, recent developments of dispersion compensators have eased the dispersion problem. But they have not eliminated it.
Overall, dispersion limits higher data rate, and crosstalk constrains smaller channel spacing. Other limiting factors in DWDM optical fiber communication systems include jitter and bit-error rate (BER). Good methods for measuring and characterizing these and other limiting factors are critical to the design and manufacture of reliable DWDM systems. Traditionally, each parameter required a specific instrument to measure it. This article discusses new testing methods currently under development that can measure these critical parameters at the same time, reducing the cost and enhancing the efficiency for DWDM testing. Before getting into that, however, we focus on the three major limiting factors-dispersion, crosstalk, and jitter-and on their inter-relationship.
Limiting mechanisms
In general terms, dispersion is caused by wavelength-dependent changes in signal velocity within a fiber. There are numerous specific cause factors, however. For a single-mode fiber, dispersion mechanisms can be either chromatic or polarization induced. Chromatic dispersion is material-related. For a silica fiber, the optical refractive index is a function of wavelength; therefore different wavelengths of light travel at different speeds. Polarization dispersion is caused by the variations of fiber core geometry along its length. The geometrical variation also leads to socalled birefringence, in which orthogonally polarized components have different refractive indexes/propagation speeds. The two orthogonally polarized modes are usually coupled in a random manner generally characterized by the root-mean-square (rms) value of a Gaussian distribution.
Crosstalk is the interference between simultaneously propagating signals. Crosstalk causes power transfer/amplitude fluctuation between channels. There are basically two types of crosstalk, one is linear and another is nonlinear. Linear crosstalk is composed of out-of-band crosstalk, which is typically associated with the optical filters and demultiplexer; in-band crosstalk is associated with the wavelength router.
Nonlinear crosstalk is caused by stimulated Raman scattering (SRS), stimulated Brillouin scattering (SBS), and four-wave mixing. Nonlinear SRS and SBS are examples of fiber acting as an amplifier for longer wavelengths when the differences between longer wavelengths and short wavelengths fall within a certain range. The probability for SRS to occur is much higher than that for SBS because the gain bandwidth for SRS is ~5 THz, while the gain bandwidth for SBS is ~0.05 GHz. When a DWDM system has more than three channels, a fourth-wave mixing (FWM) will occur at a frequency value equal to a superposition of the original three. For a DWDM system with many channels, there are many FWM possibilities. DWDM can cause both in-band and out-of-band crosstalk.
A communication system is used to carry information, which requires coding and decoding of bit streams for a digital optical communication system. In this context, jitter is used to quantify any displacement in time or amplitude relative to an ideal, and BER is used to quantify the overall performance for a communication system or component. Jitter is a statistical process and has a probability density function (PDF) associated with it. Jitter has two major components, one deterministic (DJ) and the other, random (RJ); each component has its own PDF. Jitter should be viewed as a signal that can be described in the time-domain, the frequency-domain, or the wavelength-domain. Jitter limits the bit rate and degrades the performance of any communication system. As we have discussed above, many different physical mechanisms can cause dispersion and crosstalk (see Fig. 1). Dispersion and crosstalk can cause jitter, which in turn causes BER to increase. Jitter can cause a logical 1 to be detected when a logical 0 is expected or vice versa. BER relates to a jitter PDF through an integration process. Minimal jitter and BER are important requirements for all communication systems and components.
DWDM measurement methods
In general, a fiberoptic system/component testing setup has three parts: a stimulator, such as a tunable laser source (TLS); a device under test (DUT), such as a fiber, filter, amplifier or demultiplexer; and a response-measurement device, such as an optical spectrum analyzer (OSA), a real-time optical analyzer (RTOA), an optical sampling oscilloscope (OSO), or a bit-error-rate tester (BERT; see Fig. 2).
The OSA is, by far, the most commonly used wavelength-domain measurement instrument. A typical OSA combines a high-speed optical detector and a wavelength filter. The wavelength filter scans the wavelength range of interest and measures the amplitude or power at a given wavelength. The optical power as a function of wavelength is obtained and displayed. Key performance parameters for an OSA are spectral resolution, sensitivity, and dynamic range. The major limitations of an OSA are its lack of phase information and time-domain information. Because of these limitations, an OSA is generally not suitable for measuring fiber dispersion. The OSA is used to give an estimate of adjacent channel crosstalk.
Time-domain measuring instruments include the OSO, the RTOA, and the BERT. The OSO has been around for many years. It is composed of a sampling oscilloscope and an optical-to-electrical (O-to-E) converter in the front end. It is trigger-based and it samples the input signal waveform. In contrast, a real-time optical analyzer (Wavecrest Corp.; Edina, MN; patent pending) measures the edge transition corresponded time at a programmed amplitude level in real time. It is composed of a time signal analyzer plus an O-to-E converter. Unlike the OSO, the RTOA measures rising edge and falling edge jitter, as well as channel-to-channel jitter with much better throughput.
A BERT is used to measure the BER by moving the bit clock edge around the data edges. It requires both a bit clock and a data signal to conduct the measurement. Similar to an OSO or a RTOA, an O-to-E converter is needed in order to measure the BER for an optical system or component. With the channel-to-channel jitter measurement capability, a RTOA can be used to measure dispersion-induced jitter and dispersion function under certain conditions (see Fig. 3).
With the developed jitter separation algorithms, a PDF for the deterministic component and RMS value for the random component can be obtained. As the TLS signal sweeps through the wavelength range of interest, the jitter PDF, DJ peak-to-peak, and RJ rms values are measured as a function of wavelength. Because jitter is caused by dispersion and the jitter PDF has certain unique features, jitter as function of wavelength can be used to infer the dispersion as a function of wavelength for the fiber.
The channel-to-channel jitter measurement capability of the RTOA is ideal for the DWDM characterization and testing (see Fig. 4). It provides jitter crosstalk matrices that give instant overall diagnostics of a DWDM system, and it enables rapid identification of bad channels. A jitter crosstalk matrix can be converted to an amplitude (power) crosstalk matrix if the data bit period and amplitude are known. Similarly, if the fiber characteristics are known, the spectrum of the laser source can be determined for the rising and falling edges, respectively.
Mike P. Li is currently a director of R&D/ senior scientist with Wavecrest Corp., 1735 Technology Dr., Ste. 400, San Jose, CA 95110; tel.: 408-436-9000; e-mail: [email protected]. Jan Wilstrup is a corporate consultant at Wavecrest Corp., 7275 Bush Lake Rd., Edina, MN 55439; tel: 612-831-0030; www.wavecrestcorp.com.FIGURE 3. Dispersion and jitter measurements can be performed simultaneously with a real-time optical analyzer.