Integration enables new 100G metro/long haul coherent pluggable transceivers

The groundwork has been laid to develop integrated components that will reduce the cost and power consumption of future 100-Gbps systems.

by Robert Blum

The past year has seen remarkable growth in the deployment of coherent 100G DWDM systems, with about 10,000 total ports shipped into the core optical network. In 2013, these volumes are forecast to more than double, and the rapid growth rate is expected to continue into next year. Expectations are that some 60,000 coherent 100G ports will ship in 2014.

What's enabling the success of coherent 100G from a technical perspective? On the one hand, various DSP engines and other electronic components (such as modulator drivers or transimpedance amplifiers -- TIAs) have now become available with new variants under development that have been tailored for low power dissipation and less demanding applications in the metro space. On the other hand, significant progress has been made on the optical side. Lithium niobate (LiNbO3) modulators are available with low insertion loss and high bandwidth to support polarization-multiplexed quadrature phase-shift keying (PM-QPSK), and tunable lasers with low phase noise and narrow linewidth are starting to reach cost points similar to those in 10G applications. In addition, coherent receivers have become widely available in the industry. In a way, these first generation deployments have also benefited from the successful standardization of these key optical components through OIF implementation agreements.

FIGURE 1. Low phase noise micro-ITLA module (left) and its Lorentzian linewidth (right).

But similar to what has happened in the 10G market, we expect that a different set of optical components will dominate the second and third generation of coherent 100G deployments, where power dissipation, density, and cost are key. Only highly integrated components will be able to meet the challenging requirements of the high-volume metro and regional-network market, specifically when it comes to providing 100G coherent functionality in a pluggable form factor such as a digital CFP or an analog CFP2. We'll discuss how photonic integration based on indium phosphide (InP) will make these next generation coherent networks possible.

Key optical building blocks

There are several challenges in the introduction of compact, monolithic devices for 100G -- with these key optical components being required:

• Full-band tunable laser with high power, narrow linewidth, and low phase noise, both for the signal transmission laser and to act as a local oscillator in the receiver.
• I&Q modulator with low electrical drive power, low insertion loss, low chirp, and high extinction ratio.
• Dual-polarization I&Q coherent receiver, incorporating 90° optical hybrids and dual balanced photodetectors for each polarization state, coupled to TIAs that provide the electrical outputs to feed the analog-to-digital converters (ADCs) for subsequent digital processing of the incoming signals.

Only a successful combination of all these building blocks will enable the next generation of pluggable coherent optics. Let's look at these components in more detail.

Tunable lasers. Narrow-linewidth tunable lasers are now available based on the Digital Supermode Distributed Bragg Reflector (DS-DBR) design, which has been the basis for a range of ITLA, ITTA, and small-form-factor (SFF) transponder products for 10 and 40 Gbps for many years.1

A key recent accomplishment has been to advance this monolithic design to achieve the low phase noise characteristics needed for coherent systems, without the need for an external cavity. First generation coherent systems have generally employed external-cavity lasers, which tend to be larger and more complex than single-chip approaches and have limited potential for photonic integration.

A typical DBR laser consists of an optical gain section, phase control section, front reflector, and DBR reflector section to provide wavelength selectivity of allowed longitudinal cavity modes. Current injection into the DBR and phase control sections produces a change in refractive index, providing wavelength tuning over a wavelength range of up to 10 nm. To achieve full-band tunability, the rear DBR reflector provides a number of peaks (typically seven or nine), corresponding to "super modes" that are selected in turn by injection of a small current into the appropriate sections of the chirped front grating.1 Tuning via current provides precise frequency control and rapid response but has a number of implications for the spectral purity of the laser:

• Modification of the refractive index is simultaneously accompanied by an increase in tuning-related losses, therefore reducing the amplitude of the reflection peaks and consequently reducing the cavity Q-factor. As a result, the laser Lorentzian linewidth shows a trend of increasing linewidth between minimum and maximum DBR tuning current.
• High tuning efficiency and rapid response mean that there's potential for introduction of extrinsic frequency modulation (FM) noise onto the optical field through circuit noise or from EMI. This contribution can be minimized over a wide frequency range by good design with careful decoupling of tuning currents.
• There's also a small contribution to FM noise at low frequencies due to fundamental shot noise associated with the tuning current. This contribution cannot be filtered electrically and has a frequency span of ~100 MHz, dictated by carrier lifetimes.

The DS-DBR design can be adapted for coherent transmission by optimizing the cavity geometry and the multi-quantum-well gain section, offering a lower intrinsic linewidth enhancement factor ("Henry factor") as well as superior operational characteristics at high operating temperature, which also reduces overall power dissipation.2, 3

To characterize and optimize the frequency noise of the laser, a dual-line interferometer can be used to capture transient effects in the time and frequency domain. Measurement of the Lorentzian linewidth across multiple channels is best done through PM-AM conversion in a dispersive fiber, with good agreement to FM noise spectrum measurements.2 Figure 1 shows the Lorentzian linewidth data for all ITU channels of a DS-DBR-based micro-ITLA demonstrating a Lorentzian linewidth between 130 and 400 kHz.

The impact of linewidth and low-frequency FM noise on system performance will depend on details of the modulation format, baud rate, and DSP implementation. Testing of the micro-ITLA described here in PM-QPSK links at 43 and 128 Gbps has indicated little to no measurable penalty due to linewidth or low-frequency FM noise.

Thus DS-DBR lasers have demonstrated their compatibility with coherent-system requirements and offer all the advantages that accrue from a single-chip design. Such lasers can therefore be integrated with an InP-based I&Q modulator, a key enabling step for next generation form factors.

High-speed compact modulators. The last decade has seen pioneering work on the introduction of advanced modulation formats such as DQPSK using semiconductor modulator technology. In common with typical LiNbO3 designs, the InP I&Q modulators employ Mach-Zehnder interferometers as the fundamental phase switching elements, taking advantage of the saturating response of these devices when used in this mode.

Figure 2(a) shows the RZ-DQPSK transmitter chip employed in an SFF 40-Gbps DQPSK transponder, alongside the co-packaged tunable-transmitter component. In this design, the RF modulator bandwidth is just above 20 GHz.4 For 100-Gbps systems, the design can be adapted to provide parallel channels, thereby providing for two independent polarizations. The driver interface, drive voltage, and bandwidth also can be optimized for the 28-Gbaud system requirement. Polarization handling is achieved in micro-optics external to the chip. Figure 2(b) shows a fully integrated device that provides 100-Gbps dual-polarization QPSK functionality developed for next generation coherent pluggable modules.

Integrated coherent receivers. In a digital coherent receiver, the incoming signal is mixed with a local oscillator in a wide-bandwidth photodetector. That downconverts the optical signal to baseband or low intermediate frequency, which is then digitized in an ADC and passed to the signal processing subsystem for carrier and data recovery, digital filtering for propagation impairment mitigation, and other data processing functions. To recover a complete electrical representation of the incoming optical field, we need to recover the in-phase and quadrature components of both orthogonal polarization states. That's accomplished first through polarization splitting (diversity), where each polarization is passed through an optical network that delivers the in-phase and quadrature components to a dual-balanced photodetector pair, a function generally known as a 90° optical hybrid.

FIGURE 3. 100G integrated coherent receiver shown with an adapter board to match the minimum allowable OIF footprint (LO denotes local oscillator).

A monolithic integration approach also is possible for an integrated coherent receiver (see Figure 3). Polarization splitting and combining is again performed in micro-optics, with the optical network and photodetectors integrated in a single InP chip. This approach yields a very compact receiver that meets MSA performance requirements but is considerably smaller in size. Developments are now underway to further shrink this package into a form factor compatible with a CFP2 transceiver using essentially the same approach.

Bringing it all together

What else is needed to put coherent functionality into a pluggable transceiver? For one, the analog CFP2 only allows enough space to accommodate a transmitter and receiver package and limits overall maximum power dissipation to 12 W. That imposes some very clear design constraints onto the optical building blocks, such as:

• The transmit and local oscillator (LO) signal need to be provided by the same narrow linewidth laser.
• The laser has to have sufficient optical output power to accommodate the optical tap for the LO.
• The InP modulator has to be co-packaged with the laser.
• The modulator cannot be on the same chip as the laser.

These requirements are accompanied by several other design limitations, such as those on the electrical and RF side. Defining and understanding the impairments of the analog CFP2 interface and associated signal integrity issues are probably the biggest remaining engineering tasks.

But the good news is that with the recent progress on the optical-component side, the goal of a pluggable coherent transceiver is in sight. Probably the biggest accomplishments at the chip level were (1) the development of a DS-DBR laser with high output power and a phase noise profile designed for coherent transmission and (2) the availability of high-speed InP modulators with sufficient bandwidth for 28-Gbaud or higher transmission that at the same time keep the drive voltage within the bounds of the overall power budget.

Still, without years of prior experience in packaging these components into compact optical subassemblies (sometimes referred to as "gold boxes") or without the high-speed electronics and DSP experience accumulated over the past several years, a pluggable coherent transceiver would be far from becoming reality and the industry would be a lot less excited about the capabilities of InP components.

References

1. A. Ward, et al., "Widely tunable DS-DBR laser with monolithically integrated SOA: design and performance," IEEE J. Select. Topics Quantum Electron,, vol. 11, pp. 149--156 (2005).
2. S. Davies, et al., "Reduced Lorentzian linewidth for monolithic widely tunable C-band lasers utilising InGaAlAs/InP active region," CLEO Europe, Munich (May 2011).
3. N.D. Whitbread, et al., "AlGaInAs-InP C-band tunable DS-DBR laser for semi-cooled operation at 55°C," ECOC, Brussels, Belgium (Sept. 2008), We.3.C4, vol. 3, pp. 177--178.
4. C.F. Clarke, et al., "Highly Integrated DQPSK Modules for 40 Gb/s Transmission," OFC (2009), NWD3.
   

ROBERT BLUM is product marketing director at Oclaro.