Why Test for Chromatic Dispersion and Polarization Mode Dispersion

Chromatic Dispersion

Chromatic dispersion (CD) has been a known fact for almost a decade now, yet testing for it on OC-192/STM-64 is not always done de facto. This may lead to major problems, especially today due to the optical distances on the core network reaching unprecedented lengths, and also in the metro environment with the birth of the mesh network topology. But now, with the current penetration of 40 Gbit/s, or OC-768/STM-256, CD is becoming an even bigger issue.

Most system providers find original ways to help deal with CD to some extent, but they cannot completely mitigate this challenge. For example, the native OC-192/STM-64 transmission format is non-return-to-zero transmission (NRZ). If this were to be used at OC-768/STM-256, the CD tolerances would be such that compensation would be required after 6 to 7 km of G.652 singlemode optical fiber—which is not a viable solution. On the other hand, no standards exist to define how 40 Gbit/s should be implemented; hence, every system provider has its own recipe, at very variable costs but also variable CD tolerance levels. The following chart shows some of these differences:

Modulation Format		On-Off NRZ	Duobinary	DQPSK
Symbol rate	Gbaud	43.018	43.018	21.509
CD tolerance for 2 dB OSNR penalty	ps/nm	±65	±100	±125
Max. average PMD for 1 dB OSNR penalty	ps	2.5	2.5	7

So even the most resistant OC-768/STM-256 (40 Gbit/s) can only tolerate a fraction of what OC-192/STM-64 (10 Gbit/s) can accept, which is approximately 1100 ps/nm.

Prior to this, dispersion was compensated via simple dispersion-compensating modules (DCMs), which are deployed automatically and installed by system vendors. Plus, additional pre- or post-testing was not always required, and very seldom performed. In principle, while the DCMs included canceling the dispersion of the optical fiber spans, there are several reasons why this is not exact and why the residual dispersion must be taken into account:

• The change in dispersion with wavelength is often somewhat different for the transmission optical fiber than for the compensators. This means that if the dispersion of the optical fiber is exactly matched for a channel in the middle of the DWDM, there will be a mismatch for channels on the edge, which will lead to a positive residual dispersion at one end and a negative residual dispersion at the other.

• While the optical fiber dispersion is fairly tightly constrained, there is a variation in the actual dispersion per unit length at a single wavelength due to optical fiber variability. For example, for G.652 optical fiber, the dispersion coefficient may vary from 16.9 to 18.2 ps/nm/km at 1550 nm. Likewise, there is a variation from module to module associated with DCMs.

• DCMs are usually only available in a set of values with discrete steps between them. In contrast, installation constraints (e.g., hut positions and available rights of way) mean that the span length in a practical system is often quite different from the nominal value, leading to residual dispersion at all channel wavelengths.

• In a mesh metro network where there is more than one possible route between two particular endpoints, the optical channel may be switched to an alternative route. In this case, while the residual dispersion of the primary path through the network may be acceptable, the residual dispersion of any alternative path that might be required must also be checked as it may be quite different.

Most of the installed optical fiber base, both in long-haul and metro networks, is standard singlemode optical fiber as defined by the ITU-T G.652. CD of such optical fiber can vary such as:

Channel Wavelength (nm)	Min. (ps/nm/km)	Max. (ps/nm/km)
1531.12	Very sensitive	Sensitive
1546.92	Very sensitive	Very sensitive
1562.23	Sensitive	Sensitive

In addition, the aforementioned DCM also has intrinsic uncertainty and variability. Here is an example with a 80 km DCM:

Channel Wavelength (nm)	80 km DCM
	Min. (ps/nm)	Max. (ps/nm)
1531.12	-1278	-1215
1546.92	-1355	-1288
1562.23	-1431	-1361

As previously mentioned, DCMs are usually only available in a set of values with discrete steps between them—not per kilometer. One selects the unit which is closest to the total distance to compensate for.

So for a long-haul route of 495 km of optical fiber, DCMs will be applied for 500 km. To this, one needs to add the dispersion of all network elements, either erbium-doped fiber amplifiers (EDFAs), fixed or reconfigurable optical add-drop multiplexers (ROADM), etc. Typical per-channel CD for such elements is in the ±30 ps/nm range.

Making the calculation for 495 km of optical fiber, plus six DCMs of 80 km and one of 20 km, plus four network elements (EDFAs), we get:

• 1531.12 nm minimum residual dispersion = 15.69 * 495 + 6.25 × –1278 + 4 × –30 = –341 ps/nm
• 1531.12 nm maximum residual dispersion = 17.10 * 495 + 6.25 × –1215 + 4 × 30 = 990.75 ps/nm

These results and those for the other wavelengths are summarized in the table below:

Maximum and Minimum End-to-End Residual Dispersion Values (Worst Case)

Channel Wavelength (nm)	Min. (ps/nm)	Max. (ps/nm)
1531.12	-341	991
1546.92	-342	994
1562.23	-361	974

These values are very close to the acceptable limit for 10 Gbit/s transmission. The calculation can easily be redone with slightly different parameters (a few extra ROADMs in a mesh network, longer links in a long-haul network, etc.) and be over the acceptable limit for 10 Gbit/s.

But as for 40 Gbit/s, depending on the implementation, the typical acceptable CD is often less than 125 ps/nm, making this residual CD much too high for 40 Gbit/s transmission, even though each individual 80 km section may be within tolerance.

The advent of longer optical links and the birth of mesh networks put the poorly compensated residual CD very close to the danger zone for 10 Gbit/s transmission, and beyond the acceptable limit for 40 Gbit/s transmission. On the other hand, accurately testing for this parameter allows one to fine-tune each stage of compensation, as well as perform (if required) end-to-end compensation adjustments to remove excess residual dispersion.

Polarization Mode Dispersion

Polarization mode dispersion (PMD) is a consequence of certain physical properties of optical fiber that result in distortion of optical pulses. These distortions result in dispersion of the optical pulse over time, as well as a reduction in peak power. While the outcome of PMD-induced distortion is similar to CD, it is not accumulated linearly, but stochastically. This means that its value cannot be predicted from one instant to the next, but its distribution (occurrence within a predictable range) can.

Complicating factors that contribute to this unpredictability are as diverse as temperature change on optical fiber, wind load and turbulence on aerial optical fiber, and ice loading changing stresses on aerial optical fiber. One additional and significant constraint associated with characterizing and measuring PMD on optical fiber is the date of manufacture. In fact, optical fiber manufactured before approximately 1994 may not have been designed with PMD mitigation in mind.

Because PMD is somewhat random (stochastic) and because many legacy networks have fiber manufactured before 1994, it is important to understand its contribution to signal degradation and minimize its impact. In short, PMD is not stable or generally predictable based on any known physical characteristic of the optical fiber at the time of manufacture. It must be measured in situ, i.e., after installation.

Many transmission-equipment manufacturers are working on new methods and techniques to minimize the impact of PMD on digital signals transmitted over a variety of optical fiber types. The dominant course of investigation centers on modulation methods. By carefully choosing modulation and demodulation methods, vendors hope to minimize the impact of PMD on the quality of their digital transmissions.

While work continues on modulation formats and methods, these general thresholds must be applied to the decision-making process regarding the fitness of optical fiber links for specific modulation rates.

Bit Rate (Gbit/s)	Max. PMD (ps)	PMD Coefficient (ps/(sqrt(km))
2.5	40	<2.0
10	10 (with no FEC)	<0.5
20	5	<0.25
40	2.5	<0.125

Decision-Making Matrix

Deciding whether or not one should characterize a fiber-optic link or route for dispersion (both CD and PMD) will quickly become a challenging exercise. The dilemma arises when there is significant pressure to cut corners to save time or money. However, one must consider the following: a DWDM route or ring populated with 10 Gbit/s data rates will likely have one or more 40 Gbit/s wavelengths added in the future. At that time, it will be nearly impossible to temporarily remove dozens of active wavelengths from service to characterize the optical fiber carrying them. This one factor alone should motivate all service providers to fully characterize their optical fiber links while they can. Otherwise, they may find themselves with limited possibilities in the very near future and unable to increase bandwidth on their optical fiber links.

Given that several factors must be considered before choosing to test an optical link for dispersion, this matrix will help with the decision-making process.

Conditions	<30 km	>30 km
Increasing bandwidth to 40 Gbit/s	Always	Always
Increasing bandwidth to 10 Gbit/s	Recommended	Always
Network contains pre-1994 fiber	Always	Always
DWDM network designed for 10 Gbit/s per wavelength but may move to 40 Gbit/s on one or more wavelengths in the future	Always	Always

It has been said that characterizing an optical fiber network is optional for short links. In fact, not knowing the other characteristics of the link, such as link loss and optical return loss, can and often will lead to the inability to fully utilize the amazing amount of bandwidth available in optical fiber. So, given the conditions that require testing dispersion and the fact that other physical parameters must be evaluated, it is prudent that service providers test and fully characterize all fiber links as early as possible in their normal lifecycle.