By ANIL GANDHI, SAMAN BEHTASH, JOHN KEATING, and JUDITH MEESTER
Santel Networks Inc.
Fiber-optic impairments, such as chromatic dispersion (CD) and polarization mode dispersion (PMD), can limit the performance of networks, measured in terms of reach or span length, bandwidth, and outage.
CD arises from the relative transmission delay among spectral components of the optical signal. This results in pulse broadening, causing bit errors such that the optical receiver is unable to discriminate between adjacent light pulses. PMD occurs when two orthogonal polarization components of an optical pulse travel at different rates through the fiber, caused by the asymmetry of the fiber core.
The resulting spread between the two polarization modes is known as differential group delay (DGD). PMD is highly dependent on environmental conditions that can vary with time. In general, DGB varies statistically with a Maxwellian distribution around the mean. Figure 1 shows the span lengths as limited by CD and PMD.
Battle for incumbency
The transmission distance in singlemode fiber (SMF)--for < 1dB penalty--using external modulators for 10-Gbits/sec data rates is limited to about 60 km. Dispersion compensation is necessary to extend reach beyond this limit.
Dispersion compensating fiber (DCF) can be used in tandem with and for each 80- to 100-km span of SMF. DCF, having a high magnitude negative dispersion coefficient, reverses the relative transmission delay between the spectral components, thus negating the effect of CD caused in SMF.
However, an incumbent DCF solution for CD compensation adds significant cost. Additionally, it is technically not ideal for links that use DWDM. DCF is not able to compensate for CD in all the channels in the transmission band with equal effectiveness. Partial slope compensating fibers are available in the market to alleviate this problem, but cost hinders their use.
Also, DCF has the disadvantage of high insertion loss and a high non-linear coefficient, thus reducing the signal-to-noise ratio and causing signal impairment due to non-linearities. Fiber Bragg grating (FBG)-based technologies have been considered as potential candidates to solve the residual dispersion problem. However, they add cost as well as complexity to the network.
Negatively pre-chirping the pulses on the transmitter side can be used to reduce the effect of CD. Chirp arises when the signal modulates the instantaneous frequency of the transmitter. However, introducing chirp in the system increases the bandwidth of the signal, exacerbating the CD-related distortion in longer spans. This limits its use in metro networks. In addition, electro-absorption modulators (EAMs) employed for this purpose allow lower launch powers, which can limit the reach due to attenuative loss in the fiber.
Optical PMD compensation techniques have been proposed and developed, but have not found a receptive market due in part to the bulkiness and cost of such devices ($5,000 per wavelength). The technique primarily relies on separating the orthogonal polarization components and using polarization maintaining fiber (PMF), or similar optical element, to reverse the relative delay between the polarization components caused in transmission through SMF.
Electronic equalization compensates for both CD and PMD and other impairments arising from fiber non-linearities. Additionally, electronic dispersion compensation is rapidly adaptive to changes in magnitude of impairment, whereas solutions like DCF handle only fixed amounts of CD. PMD solutions must necessarily be adaptive given the dynamic nature of the phenomena.
Electronic dispersion compensation
Inter-symbol interference (ISI) can be cost effectively compensated by electronic implementation of filter structures that nearly match (or can adapt to match) the transmission channel characteristics of a fiber link with ISI. A typical compensator of this kind is composed of a feed forward equalizer (FFE) and a decision feedback equalizer (DFE).
The FFE section is a finite impulse response (FIR) filter that takes as input the sampled values of the electrical signal, for example, after optical-to-electrical (OE) conversion. The filtering operation corresponds to scaling the different delayed samples by the filter coefficients and then summing them up once per sample clock.
The DFE section consists of an FIR filter with previous decisions as its input. The output of the filter is fed back to combine with the FFE output. The DFE, therefore, removes that part of the ISI generated by the previously detected symbols from the present estimate.
The combined FFE/DFE structure implements the filter response that optimally matches the ISI channel. The advantage of using a DFE section as compared with a completely linear structure is that the nonlinearity in the DFE loop prevents the undesirable enhancement of the noise in the input signal, allowing better overall performance.
Electronic equalization performed immediately after the OE conversion point is effective for all kinds of systematic impairments. An electronic equalizer can reduce a 4 dB ISI penalty to 1 dB in case of CD and to 1.5 dB in the case of PMD (see Figure 2).
In the case of long haul systems, electronic equalization can be used in combination with DCF to overcome residual dispersion in the presence of non-linearities. In metro networks, electronic equalization can supplant DCF or other solutions entirely, allowing a reach of 93-km SMF with a resultant post compensation penalty of < 2 dB.
While optical compensation techniques have been applied to overcome impairments that limit high-speed transmission, equipment vendors have objected to the cost of such techniques. Electronic techniques for impairment management provide a significantly cheaper alternative and offer other benefits commonly associated with electronic technology, such as ease of integration, scalability, and a manageable footprint.
Electronic equalization offers carriers a viable choice to run high data rates (10 Gbits/sec) on links with legacy fiber. Additionally, electronic equalization enables the use of cheaper components. With electronic compensation solutions priced in the same range as a current CDR-demultiplexer, network operators cannot afford to ignore this new technology.
Anil Gandhi, PhD., is a co-founder and senior director of optical systems at Santel Networks Inc. (www.santelnetworks.com), headquartered in Newark, CA. Saman Behtash is vice president of systems engineering, John Keating is chief operating officer, and Judith Meester is director of marketing.