Ted Wolcott, Shane Eleniak, and Eric Schmidt
By significantly reducing the cost per bit through advanced modulation and coding, subcarrier multiplexing technology promises higher capacities for single-wavelength and multiwavelength systems.
Escalating demand requires orders-of-magnitude increases in network capacity while simultaneously requiring reduced cost per bit-transmitted to support the new competitive business models rapidly being implemented. Traditional modulation systems over DWDM have proven their past effectiveness in long-haul transport, yet the exponential growth in data services is pushing the limits of metro and long-haul networks and demanding dramatic reductions in cost.
Optical channels are vast compared to other communication channels, and the rapid advances to date in DWDM technology have allowed current fiberoptic networks to use bandwidth-inefficient, but simple and inexpensive, on-off key (OOK) modulation to meet the relatively linear growth requirements of the last five years. This approach has allowed service providers to add more capacity by simply adding additional wavelengths. The situation is changing because of both the challenge and expense of building devices that operate at rates faster than 10 Gbit/s, and because of the complications of deploying and managing sophisticated C- and L-band DWDM systems.
It is commonly believed that technical breakthroughs increase the capacity of systems (bits/fiber), while other breakthroughs dramatically increase transmission distances (and thereby greatly reduce the associated costs of electrical regeneration). Unfortunately, it's not possible to have it both ways. Shannon's Law proves that a system can be either bandwidth-efficient (capacity) or power-efficient (distance), but not both. It is possible, however, for service providers to take advantage of breakthroughs that offer the best of both worlds while significantly outperforming traditional OOK systems.In applications where power is at a premium, like deep-space probes and ultralong-haul transmission, forward-error correction (FEC) coding can be added to increase transmission distances at a cost of decreasing the channel capacity available for uncoded data. Likewise, applications exist in which increasing the capacity of channels is paramount, usually because the channel spectrum is scarce (wireless applications) or still relatively expensive (metro optical applications). A new generation of technology based on subcarrier multiplexing (SCM) offers the potential for higher capacities of single-wavelength and multiwavelength systems, significantly reducing the cost per bit through advanced modulation and coding.
CAPACITY VS. DISTANCEClaude Shannon (1916-2001) did groundbreaking work on communications theory. The Shannon-Hartley bound (see Fig. 1) illustrates the theoretical limit in the tradeoff of bandwidth efficiency (capacity) and power efficiency (directly related to distance) compared to on-off keying. The curve designates the point at which the bit rate, R, equals the capacity of the channel, C. The region to the left, where R>C, is unachievable. The region to the right of the curve, where RFIGURE 2. Different points on the plot indicate where ideal uncoded systems fall on the bandwidth-efficiency plane with uncoded modulation (BER = 1 x 10-12). As the number of points in the PSK constellation increases, the bandwidth efficiency improves, but the required Eb/N0 increases quickly, making them less practical for long-distance transport.
There is an asymptotic bound. As the bandwidth efficiency decreases (through the use of more and more FEC, for example) the required Eb/N0 approaches -1.59 dB. Below that value, no system can be designed with arbitrarily small bit-error rates (BERs). The bound tells us just how well a system could be designed if we knew how to do it. Shannon never told us how to design the systems, however, just that a system can be designed to achieve small BERs as long as the bit rate does not exceed the capacity limit. The goal is to operate near the "knee" of the curve.FIGURE 3. On the bandwidth-efficiency plane, comparing coded and uncoded modulation, the required Eb/N0 is improved by 6.5 dB, while the bandwidth efficiency is only slightly decreased (to 0.94 bits/sec/Hz). Both uncoded and coded BPSK, QPSK, 16QAM and 64QAM are included for comparison.
There are advantages to more sophisticated modulation, in comparison to the basic OOK (see Fig. 2). The different points on the plot indicate where ideal uncoded systems fall on this plane. For fair comparison, the following assumptions were made for all of the modulation formats: ideal Nyquist pulse shaping was used, resulting in the minimum occupied bandwidth, W, equaling the symbol rate. All systems were achieving a BER of 1 x 10-12, and performance is limited by additive white Gaussian noise only.
For practical systems, the bandwidth efficiency must be adjusted for non-ideal pulse shaping. For square pulses, the occupied bandwidth is doubled (bandwidth expansion factor of two), reducing the bandwidth efficiency (bits/sec/Hz) by a factor of two. For systems employing raised-cosine pulse shaping, the bandwidth expansion factor is typically between 1.2 and 1.5. Imperfections in the system also will increase the required Eb/N0 for a given BER.Binary phase-shift keying (BPSK) requires 3-dB less Eb/N0 than OOK. Quadrature phase-shift keying (QPSK) has the same Eb/N0 requirement—QPSK is essentially two orthogonal BPSK signals—while doubling the bandwidth efficiency. As the number of points in the phase-shift-keying (PSK) constellation increases, the bandwidth efficiency improves, but the required Eb/N0 increases quickly, making them less practical for long-distance transport.
Sixteen-state quadrature-amplitude modulation (16QAM) achieves the same bandwidth efficiency as 16PSK, but requires much less Eb/N0 (only slightly more than the 8PSK scheme). Likewise, increasing the order of the QAM constellation improves the bandwidth efficiency at the cost of an increased Eb/N0 requirement.
This bandwidth-efficiency plane allows the designer to select the optimum modulation format for a given channel (available bandwidth and Eb/N0).
BENEFITS OF CODING
By adding FEC coding, the required power can be reduced dramatically with very little decline in bandwidth efficiency.
Power (distance) advantages and slight efficiency (capacity) disadvantages of employing Reed Solomon RS(255,239) FEC on traditional OOK signals can be quantified (see Fig. 3). The required Eb/N0 is improved by 6.5 dB, while the bandwidth efficiency is only slightly decreased (to 0.94 bits/sec/Hz).
The use of an RS(255,239) with other modulation formats has an identical effect—an improvement in Eb/N0 with only a slight penalty in bandwidth efficiency. Both uncoded and coded BPSK, QPSK, 16QAM, and 64QAM are included in the figure for comparison.Greater improvement in required Eb/N0 can be achieved through the use of concatenated FEC. By pairing two or more coding schemes, chosen carefully to work well together, the required Eb/N0 can be decreased dramatically. One common FEC combination is the use of Reed Solomon block-coding with trellis coded modulation (TCM)—an FEC scheme that makes use of the modulation process to further strengthen the codes' error-correction capability. The bandwidth/power efficiency of 16QAM with Rate 3/4 TCM, paired with RS(255,239), is included. In this example, new 16QAM systems with coding are superior to traditional OOK systems in both bandwidth efficiency and power efficiency.
WHAT'S RIGHT FOR YOU?We have shown how coded QAM adds capacity and reduces power compared to traditional OOK systems (see Fig. 4). The next step is to be sure that every span can achieve the desired BER for next-generation data traffic. The benefits of coding to improve span distance can be clearly seen in the plots of BER vs. required power for various types of modulation and coding (see Fig. 5). Modern digital systems require a BER of 1 x 10-12 or better.In practical applications, implementation losses degrade system performance from their theoretical level, forcing carriers to evaluate tradeoffs in actual performance, not theoretical predictions. Operators either have routes that need to be upgraded or are designing future routes that need to be cost-optimized. Where routes exist, the fiberoptic cable properties are well-known and the spans/amps/ regenerators are in place, SCM-based systems offer an increase in capacity eight time above the OC-48 systems that are being replaced, and there is potential for further increases in bandwidth efficiency.
At least two companies are developing QAM-based optical products. They take in optical OOK signals as tributaries, modulate them with QAM, apply coding, multiplex all the tributaries in the frequency domain, and use this waveform to modulate a laser (see Fig. 6). This SCM approach provides up to 20 Gbit/s per wavelength of capacity.
Thus, SCM technology provides great flexibility and scalability in long-haul and regional optical networks. By deploying a frequency-based architecture within a DWDM network, carriers can change traffic type, class of service, and optical paths without major network adjustments, enabling rapid coast-to-coast provisioning, accurate fault isolation and new revenue opportunities. SCM technology offers the best optical performance and lowest cost per bit of any transmission architecture. It not only meets the growing demands for bandwidth, but also supports the associated reduction in cost.
At Centerpoint Broadband Technologies, 1741 Technology Drive, Suite 400, San Jose, CA 95110, Ted Wolcott is director of advanced optical-product development; Shane Eleniak is executive director of optical-product line management and strategy; and Eric Schmidt is director of product marketing. Eric Schmidt can be reached at (408) 452-2913 or [email protected].