A2.1 The Three Classes of Optical Channel

The first optical channel, Type 1, is the simplest. Here the optical channel uses a single carrier or wavelength with no multiplexing. This was the approach used in the early days of optical transmission where the data rate increased continually using an optical modulator to either pass or block light, known as on-off keying, and direct detection was used at the receiver (see Section 5.5.1). The approach works fine for 2.5, 10 and 40-Gb data rates, but going to higher speeds is problematic as the modulator struggles to keep up (Fig. A2.1).

A Type 2 optical channel tackles the modulator’s limitation by adding multiplexing to the single carrier. For a single carrier, the three ways to perform optical multiplexing are polarization, space, and time. We focus on the first two only; time-based multiplexing has not been commercially exploited.

Light has two polarizations (see Fig. A2.2) that allow data to be sent along separate independent paths [2]. Using polarization, transmission capacity can be doubled or, alternatively, the same data can be sent at half the signaling rate, thereby relaxing the demands made of the modulator and the transmitter’s and receiver’s electrical circuitry but at the expense of more complex optical componentry. This is why Type 2’s use of multiplexing improves on Type 1 by effectively halving the modulator’s bandwidth.

Type 2 is already deployed commercially for long-distance transmission, as discussed in Chapter 5, Metro and Long-Haul Network Growth Demands Exponential Progress, and explained in the next section. Spatial multiplexing is another Type 2 scheme and is viewed as a long-term solution to overcome the nonlinear Shannon limit of fiber.

The Type 3 optical channel also uses modulation and multiplexing but exploits multiple carriers. There is only so much information a single carrier can hold—we talk about 400 Gb and even 600 Gb per wavelength in Section A2.5—so to expand capacity further, multiple carriers are used in parallel to form what is known as a superchannel.

The three schemes are shown in Fig. A2.1, while Table A2.1 summarizes their characteristics.

These optical channel categories are used to help explain how optical engineers are expanding optical transmission capacity.

A2.2 Single-Carrier 100-Gb Transmission With Coherent Detection

Toward the end of the last decade, there was much industry debate as to whether 10-Gb wavelengths should be succeeded by the next natural speed progression—40 Gb—or whether it would be prudent, albeit more challenging, to go straight to 100 Gb/s.

Until the debate was resolved in favor of 100 Gb [3], more complex modulation schemes than simple on-off keying were adopted to achieve 40-Gb data rates. The modulation schemes exploited the amplitude and phase of the light waves and included 40-Gb duo-binary, 40-Gb quadrature phase-shift keying (QPSK) and 40-Gb differential QPSK (DQPSK).

Despite going to 100 Gb/s transmission, the development of these more complex 40-Gb modulation schemes was not wasted work. It helped the industry alight on a Type 2 optical channel based on polarization-multiplexing, quadrature phase-shift keying (PM-QPSK), coupled with coherent detection for 100-Gb. Here two independent polarizations of light halve the signaling rate, while using QPSK, the phase of each of the two polarized streams is modulated, with each of the two components carrying data. Simply put, PM-QPSK encodes 4 bits per signal or symbol; the electronics is at 25 GBd/s as is the symbol rate but the data transported is at 100 Gb/s.

Choosing PM-QPSK allowed components already developed for 40-Gb optics to easily be used for the more demanding 100 Gb. This however results in more complex optical componentry. There are more paths—two polarizations, each with real and imaginary components—that results in a more complex and expensive transmitter (see Fig. A2.1 Type 2) and receiver design.

For long-distance transmission, forward-error correction codes are included alongside the data payload such that when 100-Gb transmission is mentioned, what is being referred to is the data payload only. The forward-error correction adds extra overhead bits so that in reality, the bit rate is between 128 and 140 Gb. This means that, in practice, the symbol or baud rate is between 32-35 GBd.

A2.3 Improving Spectral Efficiency

Another issue optical designers fret about is the spectral efficiency of an optical channel.

The 40-Gb schemes of a decade ago occupied 100- and 50-GHz-wide dense wavelength-division channels; a spectral efficiency of 0.4 and 0.8 bits/s/Hz, respectively. For 100 Gb in a 50-GHz channel, the spectral efficiency improves to 2 bits/s/Hz. That is what PM-QPSK achieves: a higher spectral efficiency, which means more bits squeezed onto the fiber. But this must not be at the expense of overall transmission distance.

This is where coherent detection comes in. Coherent detection allows for the recovery at the receiver of all the information associated with the transmitted lightwave such as its phase and amplitude. This information allows digital signal processing techniques to be used to counter transmission impairments over the fiber. Coherent detection also uses a copy of the signal at the receiver to recover the data. Because a local copy used is “clean” rather than using a signal that has also traversed the fiber, coherent detection delivers an optical gain improvement and therefore improved transmission distances.

The latest coherent systems deliver 25× the capacity-reach product versus traditional 10-Gb wavelengths that use Type 1 optical channels and direct detection, a notable advance that has taken a decade of engineering only to achieve [4].

A2.4 Higher-Order Modulation

Designers have not stopped at 100-Gb PM-QPSK. Using more advanced modulation schemes, 6 bits per symbol is achieved using an 8-point constellation quadrature amplitude modulation scheme (8-QAM) and 8 bits per symbol using 16-QAM. These equate to 150- and 200-Gb data rates sent on a carrier. But sending more bits per second means the receiver has less scope for their correct recovery in the presence of noise, reducing the overall transmission distance that can be achieved. But the benefit is in improved spectral efficiency: doing the math, 150 and 200 Gb over a 50-GHz channel boosts spectral efficiency to 3 and 4 bits/s/Hz.

Optical designers also use pulse-shaping techniques at the transmitter. Shaping the sent signals helps squeeze adjacent optical carriers into narrower channels, better using the spectrum. For example, a 100- or 200-Gb signal can fit in a 37.5-GHz channel rather than a 50-GHz one. Indeed, channel sizes are being segmented into more granular increments, such as 12.5 GHz and even 3.125 GHz, known as flexible grid. Designers have also explored a gridless approach where channels can be arbitrarily sized, not just integer multiples of smaller increments such as 12.5 or 3.125 GHz.

Telecom operator BT and equipment maker Huawei trialed in 2015 a Type 3 optical channel: a 15-carrier superchannel where each carrier held 200 Gb of data in a subchannel slightly wider than 33 GHz. The resulting spectral efficiency achieved was just under 6 bits/s/Hz [5]. Such schemes require changes to the rest of the dense wavelength-division multiplexing line system. For example, the reconfigurable optical add-drop multiplexers used to switch optical carriers between fibers at network nodes also need to handle such flexible channel spacing.

Current work is focused on making transponders—advanced transceivers for long-haul-adaptable by supporting several modulation schemes to maximize the transmitted data for a given transmission distance. Already coherent systems support polarization-multiplexing, binary phase-shift keying (PM-BPSK), PM-3QAM, PM-QPSK, PM-8QAM, and PM-16QAM. And new proprietary modulation schemes are also being introduced [6].

PM-QPSK is the de facto 100-Gb standard that supports transmission distances of several thousand kilometers without optical signal regeneration. The simpler PM-BPSK delivers 50 Gb per carrier while enabling even greater transmission distances. Such a scheme is typically used for the most demanding long-distance fiber routes such as pan-Pacific submarine links. Using PM-8QAM achieves 150 Gb per carrier, 1.5× more than QPSK, and 2000-km-plus distances, twice the 1000 km of PM-16QAM.

A2.5 The Levers Used to Boost Transmission Capacity

Given that the bulk of data center interconnect links are in a metro network, there is an advantage in maximizing the amount of data that can be carried by a carrier as long as it meets the reach requirements. Accordingly, QAM schemes using more than a 16-point constellation (16-QAM) are being explored such as 32-QAM and 64-QAM. But as mentioned, reach drops dramatically the higher order the modulation scheme used.

This approach can be seen as maximizing the bits per symbol—traveling along the y-axis, as shown in Fig. A2.3. The goal is to increase the number of bits carried based on a given symbol rate, for example 32 GBd/s which is used today at 100 Gb with PM-QPSK. As mentioned above, PM-QPSK uses 4 bits per symbol. But the symbol rate is not limited to 32 GBd/s. Increasing the baud or symbol rate—moving along the x-axis—is an additional lever to increase the overall data sent by the carrier.

Optical transport systems have already been announced that use a higher symbol rate than 32 GBd. Cisco’s NCS 1002 data center interconnect platform, discussed in Chapter 5, Metro and Long-Haul Network Growth Demands Exponential Progress, supports 250 Gb on a single wavelength using PM-16QAM and 40 GBd. Ciena's WaveLogic Ai coherent DSP-ASIC supports two baud rates: 35GBd/s and 56GBd/s [7]. And work has started among component makers to double the symbol rate to 64 GBd/s. That would mean each of the two polarization’s real and imaginary components of PM-QPSK would carry on the order of 64 Gb, resulting in a total data rate of 200 Gb on a carrier. Using PM-16QAM, the payload would achieve 400 Gb per wavelength, and 64-QAM would result in a 600-Gb wavelength [8].

Once systems are able to support 64 GBd, 400-Gb transmission using 16-QAM will revert from a Type 3 optical channel to a Type 2 one. That is because instead of using two carriers, each carrying 200 Gb (and two transmitters and two receivers), only one channel will be needed. However, it should be noted that doubling the symbol rate doesn’t improve spectral efficiency: going to higher speeds increases the carrier width in the frequency domain.

But before the baud rate can be doubled to 64 GBd, vast improvements in component performance are needed such as doubling the modulator’s bandwidth and increasing the speed of the photodetectors and the associated modulator driver and receiver amplifier electronic components. The analog-to-digital converters that sample each of the four streams of the PM-QPSK received signal and turn them into digital bit streams for processing also need to operate at twice their current rate.

To recover a signal, it needs to be sampled at twice its highest-frequency component. At 64 GBd this equates to an analog-to-digital converter working at up to 128 giga-samples per second. In other words the converter must take a sample—representing the signal digitally—once every eight-billionth of a second. Equally the coherent digital signal processor must implement its various algorithms on a doubled-rate data stream.

Optical designers thus have at their disposal two knobs—modulation (bits per symbol) and symbol rate—to maximize data rate for the required distance. And this is what the systems vendors are up to: working to develop line-side optical transmission designs that increase the symbol rate beyond 32 GBd toward the goal of 64 GBd, while adding further modulation scheme options between the extremes of BPSK and 64-QAM.

But there is a third knob, as we have seen, to increase capacity. Once a single carrier is optimized, multiple copies can be used in parallel to create a Type 3 optical superchannel. This can be seen as adding an extra dimension to the bit and baud rates. This z-axis is shown in Fig. A2.4.

The Type 2 single-carrier multiplexed approaches highlighted in Fig. A2.5 are Examples A, B, C, and D.

Examples A, B, and C use increasingly complex modulation schemes to increase the data contained in the channel, but this is at the expense of distance. In Example D, PM-64QAM is used. Here the equivalent of 6 bits per symbol are carried on each polarization. If 25-GBd signaling is used (assuming no overhead bits), the result would be a 300-Gb bit-rate using one carrier.

To achieve 400 Gb, a slightly faster symbol rate—33.3 GBd—is needed. This is more demanding than 25 GBd but nowhere near as demanding as the doubling needed to achieve 400 Gb on a single carrier with 50 GBd and PM-16QAM. But again, adopting 64-QAM compromises transmission reach further.

Example E adopts a Type 3 optical channel by bonding two carriers together—each carrying 200 Gb—to form a simple two-carrier 400-Gb superchannel.