Migrating to 40-Gbit/sec DWDM networks

Mintera Corp.
Even in the current economic environment, it's not too early to consider the evolution to 40-Gbit/sec transport.

Th 104620
Th 104620
Figure 1. Cost savings of 40 Gbits/sec vs. 10-Gbits/sec at 500 and 1,000 km. Includes all DWDM equipment (transponder, fiber and Raman amplifiers, gain equalizers, optical mux/demux, etc.).

Even in the current economic environment, it's not too early to consider the evolution to 40-Gbit/sec transport.

By Dr. Carsten J. Videcrantz and Dr. Soren Danielsen
Mintera Corp.

With traffic still growing at a pace of more than 100% annually, service providers must eventually increase the capacity of their optical networks. Technology for 40-Gbit/sec transmission is one of the key enablers for a cost-effective capacity upgrade, and current cost estimates of 40-Gbit/sec transponders are very promising. Already, savings in transponder cost favors 40 Gbits/sec over 10 Gbits/sec and even suggest that those service providers who have pushed out the upgrade to 10 Gbits/sec due to the economic downturn may never implement 10 Gbits/sec but migrate directly from 2.5- to 40-Gbits/sec line rates.

However, the complete cost picture is more complex: Service providers will face migration challenges such as interoperability issues with existing equipment like DWDM systems and the already installed fiber plant. These challenges could potentially jeopardize the overall cost savings and must be addressed for 40 Gbits/sec to be attractive. Consequently, 40-Gbit/sec offerings must ensure an evolutionary and cost-effective upgrade path, making it straightforward for service providers to embrace.

This article discusses the technical and commercial aspects of migrating to a 40-Gbit/sec DWDM network. While debating cost and comparing 40-Gbit/sec transmission to 10 Gbits/sec, we will take into account important parameters such as capacity and transmission distance as well as, for example, the issue of Raman- versus non-Raman-assisted optical amplification.

Comparing cost at 40 Gbits/sec and 10 Gbits/sec

Tremendous development progress over the last few years has propelled 40-Gbit/sec solutions from research laboratories to field-deployable alternatives to existing 2.5- and 10-Gbit/sec systems. All of the key building blocks required for 40-Gbit/sec systems are available from several vendors, so now is the time for a closer look at the cost of 40 Gbits/sec compared to 10 Gbits/sec.

The key element in 40-Gbit/sec transmission systems is the 40-Gbit/sec transponder. This device essentially performs the required opto-electronic conversion and multiplexing/demultiplexing of client-side data streams from equipment such as routers or switches into the optical 40-Gbit/sec data stream on the transport/DWDM side. Already today cost estimates suggest that a 40-Gbit/sec transponder will be significantly cheaper than four 10-Gbit/sec transponders. However, the transponder cost represents only part of the total cost picture: The economical measure that matters for service providers is the cost of transmitting bits over a given distance, i.e., the total cost per bit per km. This figure of merit depends on numerous factors such as the transmission distance and number of 40-Gbit/sec DWDM channels (capacity) required.

To validate the superior economics of going to higher bit rates, Figure 1 shows the cost savings of a 40-Gbit/sec system with 100-GHz channel spacing versus a 10-Gbit/sec system with 50-GHz channel spacing. The comparisons are conducted for transmission distances of 500 and 1,000 km, respectively, while assuming 100-km spans between optical amplifiers.

Clearly, 40 Gbits/sec offers huge cost savings for both 500- and 1,000-km distances. For a 500-km system, 40 Gbits/sec is always more cost-efficient, with savings in the 20-35% range. The step-like increase in savings at 800 Gbits/sec is due to the superior spectral efficiency at 40 Gbits/sec, i.e., the 10-Gbit/sec approach requires addition of the L-band to go beyond 800 Gbits/sec. This requires installation of additional optical multiplexers and demultiplexers as well as optical inline amplifiers and dispersion-compensating fiber.

When going to 1,000-km distances, the model takes into account that 40 Gbits/sec requires Raman amplification to achieve the necessary performance, whereas 10-Gbit/sec systems do not. Hence, for low-capacity demands where the higher cost of Raman amplification dominates, 40 Gbits/sec is more expensive.

As capacity demands increase, 40 Gbits/sec becomes more competitive -- and for capacities above 250-300 Gbits/sec, the 40-Gbit/sec solution is the winner. The initial cost is an essential parameter for all new systems, but long-term planning must take the entire lifecycle cost (and therefore operational expenditures) into account. Consequently, for 1,000-km routes, it is important for carriers to consider both first-day bandwidth needs as well as the period of time where the system will be loaded above the capacity that renders 40 Gbits/sec more economical. It is also worth noting that 40-Gbit/sec technology is on a steeper cost-reduction curve compared to 10 Gbits/sec. This plays a role in the evaluation, as does the benefits of managing fewer wavelengths using fewer spare parts while also reducing power and footprint.

Taking both raw materials cost as well as lifetime operational expenses into account, one could rightfully argue that the analysis supports the deployment of 40 Gbits/sec for distances up to 1,000 km. In combination with the fact that up to 70% of fiber routes are less than 600 to 700 km in distance this conclusion should ease the decision process and make the transition from 10 Gbits/sec and even directly from 2.5 Gbits/sec to 40 Gbits/sec a natural next step when increasing the capacity of the optical backbone.

With today's 40-Gbit/sec technology supporting distances up to 1,000 km with compelling economics, most carrier networks are covered from a technology point of view. Obviously, future iterations of 40-Gbit/sec solutions must ensure further improvements and continue to drive down size, power, and cost. Of equal importance, the distances over which 40-Gbit/sec systems are viable must be increased. When considering, e.g., nationwide U.S. networks one will often find that the overall distance sweet spot is around 1,500 km. What this means is that the most economical solution should support 1,500-km transmission without costly opto-electronic regenerators.

Most 10-Gbit/sec DWDM systems today support these distances, with some vendors claiming an impressive reach of 4,000 km. So how do we get to 1,500 km with 40-Gbit/sec solutions and, going forward, how do we get to ultra-long-haul distances of 2,000 km and beyond while ensuring cost leadership per bit per kilometer?

Getting to 1,500 km with 40 Gbits/sec is a well-understood task and, although it is not trivial, it is not a daunting technical challenge. For the most part it will be accomplished via forward error correction (FEC) techniques that -- without increasing the cost -- provide higher coding gain compared to standard Reed-Solomon codes. One of the first practical uses of such an enhanced FEC (EFEC) technique is demonstrated in the next section. It is noted that for 1,500-km distances both 10- and 40-Gbit/sec implementations take advantage of Raman amplification and thus with EFEC 40 Gbits/sec will be more economical than 10 Gbits/sec all the way up to the sweet spot distance of 1,500 km!

For very demanding applications, ultra-long-haul 10-Gbit/sec systems only require low- to medium-gain conventional Raman solutions for distances all the way up to 3,000 to 4,000 km. To accomplish this with 40-Gbit/sec systems, enhanced Raman performance is required, and a few additional improvements are needed for 40 Gbits/sec to work without opto-electronic regenerators over 3,000-km distances.

Validating ultra-long-haul 40-Gbit/sec transmission

Numerous transmission tests using typical fiber types (standard singlemode fiber, LEAF, and TrueWave) have already validated that 40-Gbit/sec transmission over 1,000 and 1,500 km is viable with field-deployable components. These results are important since they demonstrate that 40 Gbits/sec will be able to cover the sweet-spot distance required for carriers' networks. In addition, since these results are obtained over typical fiber types, it suggests a smooth upgrade path to 40 Gbits/sec without requiring an entirely new fiber plant.

It must be emphasized, though, that 40 Gbits/sec is not limited to these distances; 40-Gbit/sec DWDM regenerator-less transmission is possible over 2,000 km and longer using current technologies. Some of the enabling techniques include enhanced coding techniques (EFEC), fibers with moderate to relatively high local chromatic dispersion (to counter act non-linearities), low-noise optical amplification, and negligible impact from PMD.

When it comes to actually demonstrating the ultra-long-haul performance of 40-Gbit/sec systems, most 40-Gbits/sec trials over 2,000 km or more conducted so far have been lab-trials, where non-commercial, one-off equipment has been set up to test and to some degree demonstrate the capabilities of 40-Gbit/sec technology including amplifiers, electronics, and opto-electronic components. In fact, most reported trials have used either optical time demultiplexing receivers (fundamentally not cost-competitive compared to 10 Gbits/sec) or without actually demonstrating error-free performance (i.e., FEC assumed but not implemented).

In Figure 2, new 40-Gbit/sec experiments demonstrate ultra-long-haul transmission with as many as 40 DWDM channels using field-deployable 40-Gbit/sec components in electronic time-division multiplexing (ETDM) type 40-Gbit/sec transmitters and receivers. Error-free transmission of 1.6 Tbits/sec (40x40 Gbits/sec, 100-GHz channel spacing) over of 5,200 km of UltraWave dispersion-managed fiber was obtained. The bit-error rate (BER) without FEC is 1.5x10E-3 for all channels and with EFEC enabled, all channels have a BER below 10E-15. The impressive performance is ascribed to the 40-Gbit/sec transponder design including a new EFEC coding scheme, the use of dispersion-managed fiber that minimizes the impact of non-linearities, advanced Raman amplification giving low-noise optical amplification, as well as use of fibers with low PMD.

PMD is perhaps the most debated and controversial subject when discussing 40 Gbits/sec. Often claimed to prohibit 40-Gbit/sec transmission, PMD causes the two possible polarization states of the fiber to propagate at slightly different speeds. This difference in propagation speed between the slow and fast fiber axes (called "differential group delay," or DGD for short) leads to broadening of the transmitted bits. It is important to note that PMD is not a catastrophic factor that prohibits 40-Gbit/sec transmission. Rather, PMD introduces a power penalty just like any other system imperfection and can be accounted for via adequate system margins.

Still, it is important to discuss PMD and address the concerns raised by many people in the industry. In general an instantaneous DGD of 10 psec (average DGD around 3.5 psec) for a 40-Gbit/sec return-to-zero (RZ) signal will result in 1 dB of power penalty. Singlemode fibers fabricated today are of such high quality that the influence from PMD in the fiber alone will be negligible for distances of more than 6,000 km (Figure 3). In addition, several studies of DGD in fibers installed after 1994/1995 show that only a small fraction of the fibers would not allow transmission over 2,000 km or more.

However, it is not only the PMD in the fiber that contributes to DGD and hence the PMD penalty. Common equipment in long-haul DWDM systems such as optical multiplexers and demultiplexers, optical amplifiers, and in particular the dispersion compensating fibers (DCFs), adds to the DGD as well. Still, taking into account typical PMD characteristics for these devices, 40-Gbit/sec transmission up to more than 2,000 km is possible on most routes today.

Finally, it is emphasized that for those fiber routes in a service provider's network that exhibit high PMD, compact PMD compensating devices are emerging from several vendors with very promising cost and performance specifications. In addition, the tremendous progress in electronic dispersion compensating techniques and improvements in FEC schemes as well as PMD-tolerant modulation formats will ultimately make the impact from PMD negligible both from a performance as well as cost perspective in optical networks.


As pure bandwidth services are becoming more and more of a commodity, more than ever service providers are forced to produce bandwidth to their customers at the lowest possible cost to stay competitive. However, until now huge investments have been required and fundamental changes -- even forklift system upgrades -- have been mandatory to take full advantage of new and more cost-effective technologies.

This will change with the introduction of 40-Gbit/sec technologies that seamlessly integrate with the 10-Gbit/sec optical infrastructure in place today. Combined with recent advances in optic and electronic technologies, 40 Gbits/sec will soon enable service providers to deliver bandwidth at a fraction of the cost per bit for traditional 10-Gbit/sec technology.

Dr. Carsten J. Videcrantz is director of product marketing and Dr. Soren Danielsen is director for product management Mintera Corp. (Lowell, MA). Tel: 978 937 0700; fax: 978 937 9790; www.mintera.com.

Figure 2. Bit-error rate (BER) measurements for each of 40 channels at 43 Gbits/sec after 5,200 km regenerator-less transmission over dispersion managed fiber (DMF). The testing is performed both with and without the new enhanced forward error correction (EFEC) enabled in the receiver. With EFEC enabled all channels measured BER below 10E-15.
Figure 3. Figure 3. Allowed PMD in the fiber versus un-regenerated transmission distance at 40 Gbits/sec. Results are shown for the fiber only as well as the case including PMD of inline equipment such as dispersion compensating fiber (DCF). Most fibers installed after 1995 allow transmission of more than 1,500 km. Fibers fabricated today typically allow transmission up to 3,000 km.Th 104672

More in Transport