Using complex optical modulation technology to improve optical fiber data transmission rate

In response to the advent of the big data era, data centers are actively upgrading their infrastructure to 100Gbit/s or higher-speed fiber networks, and abandoning the traditional on-off-keying (OOK) data encoding mechanism in favor of complex modulation technology, thereby reducing the bandwidth occupied by optical wave signals and reducing latency, thereby improving the data transmission efficiency of optical fiber links.

Ordinary users still don’t know how “cloud”, “big data”, “data mining”, etc. will affect their lives, and many experts are trying to define these new terms. It is worth noting that a silent revolution has begun to spread in data centers around the world.

Many new data centers are being built around the world, and the latest generation of central processing units (CPUs) are beginning to be used in the new generation of high-performance computing (HPC) servers. As CPU performance and RAM levels continue to increase, and latency is significantly reduced, it is no longer a problem to image large amounts of data across many servers, and it can be completed in less than a second.

In the past, only large companies with their own data centers were able to conduct large-scale data mining to analyze extremely large amounts of structured and unstructured data, such as promoting personal marketing based on the preferences of different customers. However, there are now many infrastructures designed specifically for small and medium-sized enterprises that can store and analyze big data in the cloud, allowing these companies to obtain the best supply chain and marketing campaign benefits. Through cloud services, small businesses that cannot afford the cost of building servers can now perform fast data analysis at any location at the lowest cost. In the near future, even a small cafe will be able to analyze large amounts of structured and unstructured weather data in order to understand customers' consumption in different climates and festivals, and then bake just the right amount of cakes, muffins and cookies on a rainy Sunday.

Optical fiber data transmission rate surges with DQPSK/PDM technology

As mentioned above, data centers are ready to embrace the data revolution, but the more important question is whether the external infrastructure can keep up with this trend. The explosive growth of data volume has brought great challenges to the backbone network. If we do not want the backbone network to become a transmission bottleneck in the future, we must also improve the data transmission efficiency of the optical fiber network. In the near future, the optical fiber infrastructure must be able to support data transmission rates of 100Gbit/s or higher, and traditional data encoding mechanisms will not be able to cope with this change.

Like electronic signal transmission, optical data transmission technology also initially adopted the simplest and lowest-cost digital encoding mechanism, namely return-to-zero (RZ) or non-return-to-zero (NRZ) on-off-keying (OOK). At this time, the signal is an ideal 1 (power on) and 0 (power off) rectangular sequence, but if the transmission rate is as high as 40Gb/s, this concept will face limitations.

Another limiting factor at rates above 40Gbit/s is that the signal occupies a bandwidth greater than the 50GHz ITU channel bandwidth due to the high clock rate. As shown in Figure 1, as the bandwidth channel becomes larger, it begins to overlap with adjacent channels, and wavelength filters change the shape of the signal, causing crosstalk interference and degradation of the modulation information. As a result, developers have to abandon OOK and use complex modulation techniques such as differential quadrature phase shift keying (DQPSK). Complex modulation technology can reduce the bandwidth required, and the actual reduction in occupied bandwidth is related to the different symbol clock rates, that is, the baud rate.

(Baud Rate) and can support higher data transmission rates in 50GHz ITU channels.

Figure 1 When using OOK technology, when the transmission rate reaches 100Gb/s or higher, channel interference or modulation data degradation will begin to occur. Complex modulation technology can solve this problem.

Since coherent detection technology provides complete optical field information, these new concepts also allow users to perform chromatic dispersion (CD) and polarization mode dispersion (PMD) compensation when processing the signal.

The principle of dispersion is that different light waves propagate at different rates according to their frequency and polarization characteristics. Dispersion occurs due to different refraction angles. If it is not compensated, the signal quality will be reduced. The longer the transmission distance, the more serious the dispersion problem.

By using complex optical modulation techniques, developers do not need to use PMD compensators or dispersion compensating fiber (DCF) to compensate, so there is no delay caused by these modules.

Complex modulation schemes, which use parameters of light waves such as amplitude and frequency or phase to encode the light waves to make more efficient use of bandwidth, have been used by wireless engineers for years and more recently by the optical communications industry.

In addition to complex modulation, there are other methods that can also improve the data transmission efficiency of optical fiber links, such as polarization multiplexing (PDM) technology (Figure 2), which can perform orthogonal polarization processing on the second lightwave signal and the first lightwave signal so that different data can be transmitted through the same optical fiber. In this way, users can have a second channel without adding a second optical fiber and double the transmission speed.

Figure 2 Polarization multiplexing technology

Engineers still continue to use other types of multitasking technologies such as wavelength division multiplexing (WDM). What these technologies have in common is that they bundle multiple independent data streams together and transmit them through the same optical cable. In addition, users can also use pulse shaping filters to further reduce the bandwidth occupied by the signal.

Figure 3 provides an idea of ​​how these different techniques can be combined to improve spectral efficiency. At the bottom is simple OOK. If quadrature phase shift keying (QPSK) is used instead, the transmission rate of the OOK symbol rate can be doubled because QPSK can encode two-bit symbols. By using polarization multiplexing (PDM) technology, the transmission rate can be doubled again. QPSK plus PDM allows users to obtain 2×2=4 times the data transmission rate at the same clock rate. Finally, after further reducing the occupied bandwidth using a pulse shaping filter, users can transmit data at a rate of 100Gb/s through a 50GHz channel.

Figure 3 By combining different modulation techniques, spectral efficiency can be rapidly multiplied.

Enhanced bandwidth/SNR performance Spectrum transmission acceleration upgrade

The above method seems to be foolproof, as long as no other problems arise. However, things are certainly not that simple.

As early as the 1940s, Claude Shannon, an American mathematician and electronic engineer, and the father of information theory, discovered that the maximum error-free data transmission rate of a transmission channel depends on the noise and bandwidth. He called this rate "channel capacity", which is well known as the "Shannon limit".

Shannon–Hartley Theorem:

Channel capacity:

Shannon–Hartley theorem

Where B is the bandwidth (Hz), S is the average received signal power (W), and N is the average noise power (W). By increasing the bandwidth or optimizing the signal-to-noise ratio (SNR=S/N), users can increase channel capacity; in fact, the Shannon–Hartley theorem only provides a theoretical maximum value, but does not indicate which signal transmission method can allow users to get closest to this limit.

In practice, SNR is the most fundamental limiting factor. Therefore, the industry must continue to improve this issue from now on to the future to achieve the Shannon limit. When the data transmission rate exceeds 100Gb/s, better SNR performance is required for long-distance transmission under a given bandwidth.

Ellis, Zhao, and Cotter used these example parameters to simulate the information spectral density C/B of the relevant transmission and detection types (Figure 4). When nonlinear transmission is performed, the information spectral density does not increase infinitely with the transmitted power spectral density. Since the optical fiber itself has the saturation effect and nonlinear effect of the power amplifier, its information spectral density has a maximum upper limit. However, if pure linear transmission is performed, this problem will not be encountered.

Figure 4. This figure uses an example of the spectral density limit of the expected information frequency per polarization proposed by A. Ellis, J. Zhao and D. Cotter, “Approaching the nonlinear Shannon limit”, JLT 28(4), 423-433.

In Figure 4, users can clearly see that in terms of information spectral density, OOK direct detection (extracting only amplitude information) is completely unable to compete with coherent detection of complex modulated signals. Undoubtedly, different types of complex modulation methods have a critical impact on how close optical transmission solution developers can get to the Shannon spectral efficiency limit.