Fiber pairs provide submarine solution

Dan Philen, Stig Nissen Knudsen, and David W. Peckham

Dispersion-slope-managed fiber pairs, consisting of nondispersion-shifted fiber with matching dispersion-compensating fiber, are being developed to meet broadband WDM submarine application challenges. The nondispersion-shifted fiber has a very large effective area, greater than 100 µm², and excellent bend-loss properties. The compensating fiber exactly matches the nondispersion-shifted fiber for wideband dispersion compensation. For a fiber of this type, it also exhibits a low loss and a large effective area, greater than 30 µm².

Transmission experiments using the conventional negative-dispersion, nonzero-dispersion fibers (NZDF) across total unregenerated distances of at least 6000 km, to date have been limited to below 1 Tbit/s per fiber capacity. Since the dispersion slope of NZDF and of the nondispersion-shifted fiber (NDSF) used for dispersion compensation in these systems are both positive, the net dispersion slope of these dispersion maps is also positive and typically quite large. The accumulation of dispersion of the WDM channels at the edge of the transmission bands results in signal degradation that limits the usable optical bandwidth to approximately 15 nm and, therefore, the number of 10 Gbit/s WDM channels to around 64.

Transmission of 180 channels, each 10 Gbit/s (1.8 Tbit/s), across 7000 km, and, recently, 211 channels, each 10 Gbit/s (2.1 Tbit/s), across 7200 km have been demonstrated in experiments using new dispersion-slope-managed fiber pairs. These matched fiber pairs have been developed to provide simultaneous dispersion and dispersion- slope compensation in support of WDM transmission over a much wider band than the conventional approaches (see Fig. 1). An optical bandwidth greater than 64 nm in the combined C- and L-bands has been demonstrated.¹

The dispersion-slope-managed fiber pairs typically consist of a specially optimized NDSF with very large effective area (with positive dispersion and dispersion slope), followed by a negative-dispersion, negative dispersion-slope fiber, sometimes referred to as inverse-dispersion fiber (IDF). These fibers are designed to have equal magnitudes of the relative dispersion slope (RDS = slope/dispersion), so when the fibers are combined in a transmission path for zero net dispersion, the net dispersion slope is also zero.

The positive-dispersion fiber has properties that place it in the generic class of NDSF; however, it has been optimized for submarine systems applications at 1550 nm. The primary differences between the new fiber and conventional NDSF are the considerably larger mode field diameter—such as larger effective area (A_eff)—and the longer cutoff wavelength. Not requiring the fiber to support operation in the 1300 nm transmission window allows the maximum cable cutoff wavelength to be 1500 nm rather than the usual 1260 nm, and opens up fiber designs that have A_eff greater than 100 µm² and have very good micro- and macro-bending performance. The primary design trade-off of this type of fiber solution is that the fiber dispersion properties are dominated by the material dispersion and there is little room for adjustment.

When the compensating fiber is included in the cable and thus is part of the actual transmission distance it has, however, not been clear what the optimum dispersion of this fiber should be. In general, the lower the dispersion, and therefore shorter length compared to NDSF, the higher the attenuation and lower effective area. In a recent study four different compensating fibers, called inverse dispersion fibers (IDF), were produced in larger volumes to obtain experimental production-type average values for the optical parameters.³ Based on these, it was found that when combined with the NDSF described in the previous section, minimum span attenuation and accumulated nonlinear phase-shift of the span due to self-phase modulation was obtained for a dispersion of approximately -40 ps/nm/km. This ratio is equivalent to an NDSF-to-IDF length ratio of 2:1 and the IDF is therefore called IDFx2.

As the IDF is to be cabled, another important design trade-off is between effective area and bending losses. In general, larger effective area will give higher bending losses. As the effective area of an IDF type of fiber is significantly smaller than that of the NDSF, it is important to keep its effective area as large as possible to mitigate fiber nonlinear effects.

It has previously been demonstrated that dispersion-compensating fibers with similar, or worse, macro-bending properties can be cabled in loose tube cables.⁴ Preliminary results from loose tube cabling of the NDSF and IDF described here show no increase in attenuation after cabling, even up to 1600 nm.

SPLICING
Dispersion-compensating fibers (DCF) are, in general, well known to have a relatively high splice loss to NDSF. The IDF has a spotsize of just above 6 µm compared to NDSF spotsizes of above 11 µm. Assuming a butt joint splice, this level of spotsize results in a coupling loss above 1.5 dB. With some modification of the fiber dopants, and splicing programs, loss values for arc fusion splices directly between IDF and NDSF are typically on the order of 1.0 dB.

One technique to reduce splice loss is the so-called bridge technique. In this technique another fiber, known as the bridge, is spliced between the IDF and the NDSF. When the bridge is spliced to the NDSF, a longer fusion time is used to expand the spotsize of the bridge to match the spotsize of the NDSF. In this way one can obtain an overall splice loss of about 0.5 dB at 1550 nm between the IDF and the NDSF while maintaining the splice strength of 200 kpsi required for undersea systems.

For an NDSF-IDF link the group delay data can be fitted by a five-term Sellmeier function and results in a dispersion curve described by a third-order polynomial (see Fig. 2). Because of the shape of the curve, to minimize the variation in net dispersion across a broad wavelength range, it can be advantageous to slightly over-compensate the dispersion slope at 1550 nm as shown in Series 2 in Figure 2 in which the RDS is over-compensated by 3%. Also shown are measured curves for spans with 95% and 111% slope compensation at 1550 nm. Such compensation ratios must be expected to lie within the distribution of the compensation ratios that can be assembled from the individual fiber dispersion distributions.

A general feature for practical, highly manufacturable fiber refractive index profiles is that the more negative the dispersion of the compensating fiber, the more curvature to the net dispersion curve of the compensated span.

To our knowledge, one of the broadest compensated spans reported was obtained with an IDF of dispersion -19 ps/nm/km.⁵ A typical spectral attenuation curve for an NDSF-IDFx2 span, including splices, is shown in Figure 3.

PAIR SELECTION
Individual spools of fibers, called constituents, are selected and spliced into longer spans of fiber that may be 50 to 100 km long. These long lengths are called unit fibers, and several units make a set of fibers. The number of fibers in a set is the same as the number of fibers in the finished cable, and constitutes a cross-section of the cable at any point. The length of a set is usually the length of an amplifier spacing but could be a multiple of that for ease of manufacture in the cable factory. In traditional selection methods, one selects based on loss only, if the dispersion is uniform as in standard NDSF, or dispersion only, with a target dispersion for DSF.

Splicing IDF to NSDF in the same unit fiber requires that the two different fiber types be selected from two different inventory distributions and spliced into the final unit. The goal is to balance the dispersion to be zero, or near zero, in the final unit. Since there are two distinct fiber types involved, one can think of each as a subunit of the final unit length. Each subunit has dispersion, which, while opposite in sign, is large in magnitude. Thus, one can no longer trim the length or cut arbitrarily and maintain the desired final dispersion. It should also be clear that the length is now one amplifier spacing instead of a multiple.

A final consideration is how to identify the two different fiber types in the set, and later in the cable. The IDF fibers have one color sequence, and the NDSF fibers have a different color sequence. Once the fibers are spliced, the color of the outside constituent on the final spool will tell the cable manufacturer which spools are which. Some will have IDF (receive fiber) on the outside and some NDSF (transmit fiber) on the outside, since the final cable contains receive and transmit paths.

ACKNOWLEDGMENTS
The authors wish to acknowledge Lucent engineers Morten Østergaard Pedersen, Torben Veng, Louis R. Pritchett, Lars Grüner-Nielsen, and David Bain for their contributions to this article.

REFERENCES

1. T. Tanaka et al., ECOC 2000, post deadline paper 1.8, Munich.
2. K. Mukasa et al., ECOC 1997, Edinburgh.
3. S. N. Knudsen et al., Elec. Lett., 36, 25 (2000).
4. L. Grüner-Nielsen et al., Proc. IWCS (1999).
5. S. N. Knudsen et al., OFC 2000, Baltimore.

Dan L. Philen is a distinguished member of the technical staff in fiber measurements and undersea fiber technology and David W. Peckham is a consulting member of the technical staff at Lucent Technologies, 2000 NE Expressway, Norcross, GA 30071. Stig N. Knudsen is a member of the technical staff at specialty fiber devices, Lucent Technologies, Denmark. Philen can be reached at 770-798-3910 or [email protected].