Important Issues in CWDM

In last month's issue, in an article entitled Understanding CWDM, we discussed the basics of the coarse wavelength-division multiplex (CWDM) network and its main components. As a follow-up topic, this month, we will be covering the major issues that come into play when testing these networks.  

As mentioned in our previous article, the CWDM network mostly suits short-haul metropolitan environments and provides a cost-effective solution for applications requiring fairly high bit rates, at minimal cost.

As a result, the CWDM network topology is simpler than that of the more powerful and complex DWDM network, as it doesn't contain any optical-electrical-optical (OEO) converters, nor does it contain any erbium-doped fiber amplifiers (EDFAs). The figures below illustrate the three common CWDM topologies currently in use: the point-to-point topology; the linear add/drop topology; and the two-fiber-ring topology.  

Once these basics are understood, several questions arise and a few key points should be kept in mind in order to optimize not only transmission, but loss budget as well.

CWDM Loss Budget

Since the lengths of CWDM links are usually relatively short (max. 80 km), CWDM networks do usually allow a certain amount of extra loss, while still maintaining sufficient power at the receiving end. The following is a summary calculation of the power loss budget for a common CWDM topolgy.

Keeping the loss budget in mind as a general backdrop, there are three important points to consider when operating the CWDM network.

Issue 1─Water Peak

As we know, the entire purpose of CWDM is to get a decent amount of bandwidth at low cost, which is why the tolerance levels of the main components were made less stringent compared to those of DWDM. On the other hand, this means that the fiber has to be used to its maximum capacity.

When fibers are extracted, hydroxy-ion (OH-) molecules from the humidity in the air will get trapped in the silica (SiO2) glass structure, and every molecule has spectral absorption lines;
i.e., energy at a certain frequency that corresponds to the excitation frequency of the molecule. This is the exact same principle used in EDFAs; erbium molecules can absorb energy from pumps at 980 nm and 1480 nm.

It just so happens that the OH- absorption line is centered at 1383 nm, thus creating high absorption at and around this wavelength, which is exactly in the center of the E band. The more humidity there is in the process, the higher the absorption line and the broader the impact.

On typical fiber, the peak can start at about 1360 nm and extend all the way to 1440 nm,  creating high loss (up to 2 dB/nm) and thus eliminating four usable CWDM wavelengths in the process. For those who work with DWDM, four channels may not seem very significant, but on CWDM networks, which typically have only 16 available channels, this is a major concern, as it represents 25% of the usable bandwidth.

Dealing with Water Peak

In the metro/access environment, it is not always clear what is in the ground; first, most of these networks were acquired from providers that are now long gone and, second, they usually contain a lot of legacy fiber. Even though CWDM is less expensive than DWDM, it is still a significant investment; therefore, when deploying CWDM, it is crucial to know if there is water peak on this fiber and, if so, how significant it is.

The best way to check if there is any water peak is to use an OTDR. In fact, hundreds of thousands of dollars can be saved by performing this test, and the network's available bandwidth can be increased by 25%, just by being certain that the fiber being used is actually usable throughout all the transmission bands. The extra cost of a quadruple-wavelength OTDR (standard 1310, 1550 and 1625 nm, plus 1383 nm) is well paid for after the very first deployment.

Issue 2─Drift vs. Wavelength and Power

Since the lasers contained in CWDM networks are uncooled, they can drift by as much as 7.5 nm. The figure below illustrates the typical drift at each wavelength.

Although filters are used to compensate for this high drift, the spectral attenuation curve through the transmission bands is not flat, so a wavelength drift is always accompanied by a power drift. Here are a few examples:

On an 80 km link, this means that power can drift by as much as 8 dB! In addition, even though CWDM uses DFB lasers, since they are uncooled, second-mode effects can be quite powerful. Although, during a wavelength drift,  the central wavelength may stay within the transmission window, if its central peak is at the edge of the tolerance, powerful second-order modes can fall off the range, thus inducing a significant power penalty:

Dealing with Drift

Since power can change quite fast and significantly, a good power meter is a must─but not just any power meter.

Standard power meters are either based on Germanium or InGaAs detectors, calibrated at a few wavelengths (typically 1310, 1490, 1550 and 1625 nm). As shown in the figures below, these detectors do not have a flat response as a function of wavelength:

This means that if CWDM wavelengths at 1352 nm or at 1400 nm are measured with a regular power meter, the reading obtained will not be accurate; this is especially true if Germanium detectors are used. These power meters, while excellent for DWDM since attenuation across the C band is fairly flat, just aren’t useful in CWDM applications.

A CWDM-ready power meter, on the other hand, is optimized for this type of network and will do a much better job. These power meters are calibrated at very tight wavelength spacing, and some even allow for interpolation between calibration wavelengths, thus allowing the user to calibrate it in the field at virtually any desired wavelength.

Another indispensable unit is the CWDM optical spectrum analyzer (OSA). Since channels are widely spaced, the CWDM OSA does not require the narrow FWHM found in a DWDM OSA, nor does it require the same power and sensitivity range because CWDM power levels are between +5 and -40 dBm. In addition, the OSNR reading in CWDM does not need to be as accurate either.

On the other hand, a CWDM OSA does need to cover the entire 1250 to 1650 nm CWDM wavelength range.

It also needs to be able to monitor wavelength and power drift over time, so as to determine the impact of the attenuation slope within a single channel:

Furthermore, if some DWDM OSAs can monitor peak power, what is important in a CWDM OSA is to be able to monitor in-channel power because side modes may slip out of a channel.

Issue 3─Laser Modulation and Chromatic Dispersion

In DWDM systems, the threshold for chromatic dispersion (CD) at 10 Gb/s is about 1100 ps/nm. On G.652 fibers, with CD typically less than 19 ps/nm∙km in the C band, the distance between  regeneration or compensation sites is roughly 60 to 80 km.

Transmission at 2.5 Gb/s is four times slower than at 10 Gb/s, so the CD tolerance for 2.5 Gb/s is about 16 times higher (i.e., around 18 000 ps/nm). This is due to the fact that spacing between bits is four times larger, and modulation creates four times less chirp (signal broadening). Conclusion? DWDM links operating at 2.5 Gb/s have a higher tolerance and therefore do not necessarily require CD testing, whereas links operating at 10 Gb/s benefit greatly from CD testing, especially between 60 and 80 km, as not doing so could lead to unacceptable bit error rate.

For CWDM systems, on the other hand, ITU-T Recommendation G.695 established that the maximum CD tolerance for 2.5 Gb/s transmission is 1600 ps/nm—a much lower value than for DWDM systems. Since CWDM uses the L band, the maximum CD level at the highest CWDM channel (1617 nm) is approximately 21 ps/nm.km (on G.652 fiber), which leads to a limit of 76 km between sites.

In order to save costs, which is the main purpose of the CWDM network, distributed (DFB) lasers in CWDM deployments are directly modulated, whereas DWDM networks use lasers with external modulation. The direct consequence of this is the occurrence of higher chirp in CWDM networks, meaning that the pulse starts broader. CWDM transmission can therefore accept a lot less CD and reaches its threshold limits a lot quicker than DWDM transmission.

Directly modulated lasers have higher chirp because of small changes in the laser material’s refractive index due to the varying electrical current being applied to it. The alternative to direct modulation is external modulation. In this method, the electrical current applied to the laser is kept constant, thus eliminating any issue with direct modulation such as excessive chirp. This extra performance, however, does require a bigger investment, which is why they are not used in cost-sensitive CWDM deplyments.

Dealing with Modulation and Chromatic Dispersion

For the reasons explained above, DWDM guidelines do not apply to CWDM. CD testing is extremely important in CWDM systems, even at a low bit rate such as 2.5 Gb/s.

With regards to testing in the O band, it can definitely be done, but transmissions in this band are not affected much by CD. For attenuation purposes, Recommendation G.695 states that CWDM systems can go up to 80 km for wavelengths of 1471 nm and higher; wavelengths in the O-E bands are limited to 40 km. With smaller distances and lower internal CD, testing in the O to E range is not necessary. Only S, C, and L bands are at risk, since CD and distances are higher.

Testing is an essential part of network operation and maintenance, and considering the concessions required by CWDM technology in order to keep costs low, regular assessments and monitoring the issues described above can help operators keep their transmissions running smoothly.