Research indicates next steps for 40 Gbits/sec
Optical communication equipment with 40-Gbit/sec interfaces became available for commercial networks in 2004. Deployment of these systems has followed, with many equipment providers looking to provide follow-on generations of interfaces.
Meanwhile, over the past 2 years the optical communications market has accelerated the adoption of 10-Gbit/sec technology. According to industry analysts Ovum-RHK, the 10-Gigabit Ethernet module market grew at a CAGR greater than 150% from 2003 to 2005 (“Datacom Market Update,” April 2006), and the 10-Gbit/sec telecom market grew by more than 40% over the same period (“WAN Market Update,” July 2006). This rapid growth has increased the need for higher-speed interfaces such as 40 Gbits/sec to aggregate the growing number of 10-Gbit/sec links. Such 40-Gbit/sec interfaces are also required to enable bandwidth provisioning for future expansion of capacity in large data centers and central offices.
Wider deployment of 40-Gbit/sec links is driving demand for next-generation 40-Gbit/sec interfaces with reduced footprint and power dissipation across the full range of distances required in communications networks. However, significant technical hurdles must be overcome before this new generation of 40-Gbit/sec devices can meet these requirements. Optical module vendors are experimenting with new modulation formats that promise the ability to overcome these challenges and fulfill carrier needs.
Three primary impediments limit the transmission distance of 40-Gbit/sec signals: chromatic dispersion, optical attenuation, and polarization-mode dispersion. Chromatic dispersion is the tendency for the velocity of the optical signal to be dependent upon the wavelength. Optical attenuation is the loss of optical power through scattering and absorption along the length of the fiber. Polarization-mode dispersion is the tendency for the velocity of the optical signal to depend upon the polarization of the signal.
Chromatic dispersion poses the primary challenge for 40-Gbit/sec optical interfaces. Chromatic dispersion effects are proportional to the square of the data rate. That means that while 40-Gbit/sec transmission increases data rate by a factor of four compared with 10-Gbit/sec transmission, its tolerance to chromatic dispersion decreases by a factor of 16. Thus, simply scaling the traditional TDM technologies used at 10 Gbits/sec for transmission over an 80-km link would result in a transmission range of less than 10 km at 40 Gbits/sec.
These traditional TDM technologies have used a modulation scheme known as “non-return to zero” (NRZ). Recently, however, the OIF proposed a very long reach (VR) 10-Gbit/sec link specification that the ITU then standardized in G.959 as the interoperability code P1V1-2C2. This standard incorporates modulation schemes other than NRZ to achieve up to 120-km transmission over fiber without any external compensation for chromatic dispersion. Conventionally, such non-NRZ schemes had only been used in proprietary long-distance DWDM transmission, but their adoption by the ITU indicates broader industry support for standardization of non-NRZ optical transmission. The ITU’s action also creates an opportunity to standardize modulation formats other than NRZ that would extend 40-Gbit/sec transmission to distances comparable to 10 Gbits/sec.
Due to the severity of the impairments at 40-Gbit/sec transmission rates, the only standardized interfaces available to the market have been for short-reach links. For example, the only commonly implemented optical interface to date has been the ITU VSR2000-3R2 at a wavelength of nominal 1,550 nm, which is specified for transmission link distances of less than 2 km. No standard has yet been adopted that enables longer-reach 40-Gbit/sec data transmission.
Many communication equipment manufacturers have deployed custom approaches to support longer transmission distances, but these are not standard interfaces. As the industry looks toward standardized alternatives, two likely technologies are duobinary modulation and differential quadrature phase-shift key (DQPSK) modulation.
Duobinary modulation was first proposed by Adam Lender in 1963.1 The key advantage of duobinary modulation is that the bandwidth requirements are halved, such that to support a 40-Gbit/sec data rate a bandwidth of only 20 GHz is required. In duobinary signaling the receiver does not sample the bit bk at time instance kT, but rather bk + bk-1. This function is shown in Figure 1, along with a block diagram of a generic implementation of duobinary signaling for fiber optics in Figure 2.
This algorithm produces a three-level signal from the original binary signal. For example, if the binary values for each bit are “0” and “1,” then there are three possible generated values of 0 (0,0), 1 (0,1 or 1,0), and 2 (1,1). Figure 1 shows an example of converting NRZ data to duobinary format. Additionally, this process of sampling two bits to generate the output data lengthens the sampling time, thus narrowing the signal in frequency space or bandwidth. Narrower bandwidth provides better transmission characteristics to tolerate higher levels of chromatic dispersion. (For more detail on duobinary signaling see “Duobinary Modulation for Optical Systems,” Lightwave, June 2006, page 11.)
Duobinary signaling for optical communication has many positive merits. Since the optical power is the square of the electric (E) field the optical receiver is receiving only two levels of data (|E2|,0) rather than the three coded levels of (E, 0, -E). This enables standard NRZ receiver technology to be used by duobinary interfaces for fiber-optic communications.
The drawback is that the overall sensitivity of the receiver is reduced relative to the same average optical power delivered in an NRZ modulation format. Furthermore, duobinary modulation allows for the use of a lower-bandwidth optical modulator, but since the modulator has to make an optical signal using both the positive and negative phase there may be tighter constraints on optical phase to achieve higher dispersion tolerance. A higher modulation voltage is currently required in order to provide the full range of phase control necessary for duobinary signaling-companies are working to provide drivers with higher voltage and modulators with lower voltage requirements. Construction of a precoder IC with sufficient performance to operate at 40 Gbits/sec is also being developed within the industry.
Another likely candidate for longer 40-Gbit/sec transmission is DQPSK. DQPSK has the advantage of better transmission characteristics than duobinary. However, the improved transmission distance is traded off with increased complexity in the construction of the optical interface.
Figure 3 illustrates this complexity. For DQPSK transmission, the 40-Gbit/sec signal is encoded and split into two data streams of 20 Gbits/sec each. This is done because during each bit time two symbols are being sent with different phase characteristics. The precoder segments the data and differentially encodes the data. The dual Mach-Zehnder modulator then adjusts the phase to provide the correct signal pattern. Unlike duobinary, a special decoder is required to achieve proper reception of the signal. The quadrature points are separated with a 1-bit optical interferometer delay line with each optical output connected to a 20-Gbit/sec photodiode and receiver chain. (For more details on the DQPSK modulation format please see Griffin and Carter’s publication.2)
A DQPSK-modulated optical interface provides longer reach capabilities, but will require the development of more new components with the capability to operate at 40-Gbit/sec transmission rates. Additionally, these components require more physical space and electrical power consumption, which will limit the density of 40-Gbit/sec interfaces on communications equipment. The industry is still evaluating both the DQPSK modulation scheme as well as the performance attributes of duobinary-modulated interfaces.
Communications equipment manufacturers are also striving to reduce the power consumption and increase the density and capacity of their systems. Conventionally, the laser and its accompanying driving electronics are the most expensive items in the communication link. However, the initial generation of 300-pin MSA 40-Gbit/sec transponder uses very power-hungry interface chips to enable the SFI-5 interface between the module and host board.
Intermediate and long-reach modules will require high-performance modulators with phase control. This type of performance is only feasible with lithium niobate based Mach-Zehnder (LN-MZ) modulators. For 2-km 40-Gbit/sec links, Opnext and others have migrated away from initial designs that also used LN-MZ modulators to more integrated components such as electroabsorption modulator and integrated distributed feedback laser (EA-DFB). Such an integrated laser and modulator source provides good performance and with improved power dissipation, smaller component sizes, and no need for expensive optical alignment processes. The smaller size and lower power dissipation lead to better thermal characteristics of the overall communications interface. Figure 4 shows the eye diagram of a 40-Gbit/sec EA-DFB module lasing at a nominal wavelength of 1,550 nm while in a high temperature condition of 70°C.
Additionally, laser researchers are working to extend the bandwidth of directly modulated DFB lasers to achieve 40-Gbit/sec transmission performance. This type of light source would enable smaller size and reduced power consumption. Finally, with a directly modulated light source such as a DFB, it is likely that an uncooled laser could eventually be developed to provide even smaller size, better thermal performance, and lower power consumption.
In the near term, designs with reduced size and power consumption are being developed in the 300-pin multisource agreement (MSA) form factor. The 40-Gbit/sec 300-pin MSA uses an SFI-5 electrical interface to communicate between the host board and the optical module. The increased demand for 40-Gbit/sec interfaces is leading more IC manufacturers to develop lower-power-consumption devices that provide the SFI-5 interface. Module manufacturers are targeting the same size and power dissipation for next-generation 40-Gbit/sec as were used in the original 10-Gbit/sec 300-pin MSA. Some of the larger communications systems developed using the original 10-Gbit/sec 300-pin MSA could accommodate four modules-the intention is to duplicate this density with next-generation 40-Gbit/sec modules.
Increased data and communications network traffic is driving demand for more density and higher speeds in network equipment. New technologies such as non-NRZ modulation schemes for greater transmission distance and compact light sources for smaller size and power consumption are enabling these next-generation applications and networks.
Matt Traverso is senior product marketing manager at Opnext Inc. (www.opnext.com).
1. A. Lender, “The Duobinary Technique for High-speed Data Transmission,” IEEE Trans. Commun. Electron., 82:214-18, 1963.
2. R.A. Griffin and A.C. Carter, “Optical Differential Quadrature Phase-shift Key (oDQPSK) for High Capacity Optical Transmission,” OFC, WX6, 2002.