Jeffrey H. Sinsky and Peter J.

Winzer

A Microwave Engineers Perspective

t is evident from the explosive evolution of data trafc and multimedia services over the past several years that sending massive amounts of data around the globe is becoming the next communications paradigm. Millions of end users have become obsessed with sending rich content between one another from wherever they are. Starting with simple text-based e-mails and Web pages, the offered content and services have evolved to encompass high-resolution pictures, high-denition video, voice, music, gaming, e-commerce, and peer-to-peer le sharing, to name just a few. The dominance of video traffic in todays networks is illustrated in Figure 1, which shows the change in traffic composition over the last three years [1]. Future corporate and personal communications products will include telepresence, a dataintensive network application that provides an immersive and realistic communications experience by allowing multiple users in different places to feel as if they are interacting in the same room, as their next frontier [2]. Such systems require significant bandwidth in order to send high-resolution, low-latency video to multiple locations and, due to their nature, make content caching difficult to impossible. In our current environment of multinational corporations

and limited energy resources, telecommunications has the opportunity to redefine the way in which the people of the world interact. In order to support the applications listed above, optical transport has evolved as the enabling technology for high-bandwidth communications across the board, from interconnects in data centers and massive f iber-to -t he-home (FTTH) build-outs over transparent metropolitan and regional optical networks all the way to nationwide long-haul optical backbones a n d s u b m a r i n e s y s t e m s spanning the globe [3], [4]. While optical transport has traditionally been tightly coupled to the Synchronous

Jeffrey H. Sinsky ([email protected]) and Peter J. Winzer are with Bell Labs, Alcatel-Lucent, 791 Holmdel-Keyport Road, Holmdel, NJ 07733 USA.

Digital Object Identifier 10.1109/MMM.2008.931669

44

1527-3342/09/$25.002009 IEEE

April 2009

 IMAGESTATE

Optical Network (SONET) and the Synchronous Digital Hierarchy (SDH) standards, Ethernet [5] started to enter optical networking at the first common SONET/ Ethernet bit rate of 10 Gb/s (see Figure 2). The success of 10-Gb/s Ethernet and the clear trend from circuit to data traffic have triggered the desire to scale optical networks using Ethernet, which has been reflected by the IEEEs standardizing higherspeed Ethernet at 40 Gb/s and, primarily, at 100 Gb/s. The ITU-Tdefined Optical Transport Network (OTN), the successor to SONET/SDH, is following the trend by departing from the classical scaling from 40 to 160 Gb/s and taking up OTU4 at 112 Gb/s to natively support 100-Gb/s (100G) Ethernet as the probable new standard for optical transport networks [6]. Terabit Ethernet has thus emerged as a very important area of research to be addressed over the next decade [7], with 100G Ethernet solutions being currently heavily investigated in research and development worldwide.

In the design of 100G optical communications systems, microwave knowledge is more critical than ever. As a result, there is an emerging field referred to as microwave signal integrity engineering, which involves the application of microwave design techniques to broadband data signals. In this article, we explain the current state of the art and the research frontiers of 100G optical communication systems, illustrating various optical modulation approaches. Furthermore, we compare and contrast this technology with traditional microwave communication systems. The increasing role of the microwave engineer in high-speed optical communication design will also be discussed.

Optical and Radio Communications Systems: Whats the Difference?

A radio communications engineer studying the methods used today in most digital optical communications products will very likely be surprised by the simplicity

April 2009

45

systems installed today operate at 10 Gb/s and sometimes Private Lines 6% at 40 Gb/s using mostly ONVoice OFF keying (OOK) with some 24% 50% rudimentary pulse shaping Video Distribution or simple partial response fil2005 Data Services tering (also known as phaseshaped binary transmission, BB Internet 7% or PSBT [9]). Square-law deAccess Residential tection is used rather than 3% BB Internet 10% coherent detection, which in Access SMEs 5x a practical setting only en2% ables binary and quaternary differential phase modulation (DBPSK and DQPSK). These 11% modulation schemes are being 25% deployed at 40 Gb/s today. Equalization, transmitVideo 2008 ter preemphasis, and forward 4% error control (FEC) are imple31% mented in a much simpler fashion (if at all) than what we 27% are used to finding in radio systems. To give some numbers, a commercially widely available 10-Gb/s maximum likelihood sequence estimator (MLSE) Figure 1. Evolution of data traffic composition. BB denotes broadband; SME denotes small operates at 3-b A/D resolution and medium-sized enterprises (from [1]). over a four-state trellis [10]; linear transmitter predistortion at 10 Gb/s has recently been implemented with binary-inof the underlying concepts and will conclude that digiput FIR filters [11], and FEC schemes commonly use a 7% tal optical communications is in its infancy when comoverhead and hard-decision decoding (typically based pared with modern digital radio techniques. Ironically, on well-known Reed-Solomon and BCH codes) at both despite this fact, the data ratetransmission distance 10 and 40 Gb/s [12]. product is significantly higher for optical systems than This brief glimpse at commercially available opit is for fixed wireless, mobile wireless, or wireless LAN, tical transmission technologies is in stark contrast to as illustrated in Figure 3 [8]. The vast majority of legacy radio communications systems, which have been taking advantage of multilevel sig100 Gb/s naling and higher-order constellations for many decades. TN Cellular communication sysO DH/ T/S tems use complex modulation E 10 Gb/s SON formats including p/4 DQPSK (differential quadrature phaseshift keying), OQPSK (offset quadrature phase-shift keying), 1 Gb/s and GMSK (Gaussian mini mu m-sh i f t key i ng) for IS-54/IS-136, IS-95, and GSM [13], [14]. Similarly, satellite 100 Mb/s communications have used 1990 1992 1994 1996 1998 2000 2002 2004 2006 2006 2010 various different modulation Year formats for many years, including QPSK, DQPSK, and OQPSK [15][18]. M-ary quadrature Figure 2. Circuit-based transport standards versus Ethernet.
Channels Data Rate

46

Et

he

rn

et

April 2009

amplitude modulation (QAM) is used in personal wireless 100,000 km communications, such as the high-rate wireless personal Satellites 10,000 km area network (WPAN) stanLamdaXtreme dard, IEEE 802.15.3 [19]. 1,000 km Due to the fact that data Optical Communications rates for radio systems typically dont exceed MHz rates and 100 km Fixed Wirele ss A c ces microprocessor clock speeds Fiber to the Home s Coaxial (FTTH) are typically one to two orders e Cab 10 km le of magnitude faster than this, Mo Twi bil ste digital signal processing has e d 1 km proven to be a very attractive tool for addressing channel Wi 100 m impairments and general sigre LA l nal processing requirements. As two examples, consider 10 m 1 kb/s 1 Mb/s 1 Gb/s 1 Tb/s OFDM, which requires FouAggregate Link Capacity rier transforms, and softwaredefined radio, which effectively carries out all demodulation Figure 3. Reach versus data rate for radio, wireline, and optical communications and signal processing digital- technologies (from [8]). ly. Additionally, FEC has been fundamental distinctions between radio and optical widely used in many radio communications systems. In communications. short, due to the sheer difference in data rates, digital radio communications today are more algorithm-heavy by far than optical communications. Optical Communications Systems Require The natural question arising from this comparison is, Ultrawide-Bandwidth Electronics why have optical communications been lagging behind? Optical communications systems operate at carrier freAs one would expect, the answer has many parts, as exquencies of around 193 THz (see Figure 4), since at midplained in the following subsections, which outline three infrared wavelengths of around 1.5 m
Regeneration-Free Transmission Distance

ir Pa

s les ire

s es N

0.6 Fiber Loss (dB/km) 0.5 0.4 0.3 0.2 0.1 1,200 LF MF VHF

50 THz Optical Bandwidth (e.g., 500 Channels at 100-GHz Spacing) Erbium Doped Fiber Amplifier 4.5 THz C-Band L-Band

1,300

1,400

1,500

1,600 Visible Light 1 PHz 1 m

1,700

Wavelength (nm) h = kT 1 THz VLF HF UHF SHF EHF

1 kHz

1 MHz

1 GHz

1 km

1m 1 mm Frequency/Wavelength of the Electromagnetic Field

Figure 4. Frequency/wavelength of electromagnetic fields.

April 2009 47

It is much easier to design a system with a small bandwidth relative to the carrier frequency than one with a large relative bandwidth.
silica glass happens to have its lowest loss (less than 0.2 dB/km) efficient optical amplification is available through the Erbium-doped fiber amplifier (EDFA) semiconductor properties allow for the construction of high-quality laser sources and photodetectors. Now, as every microwave engineer knows very well, it is much easier to design a system with a small bandwidth relative to the carrier frequency than one with a large relative bandwidth. Since this fact is independent of the carrier frequency, an easily implemented 5% relative bandwidth at 193 THz corresponds to a vast 10 THz of available channel bandwidth! This value is so huge that optical fiber bandwidth has until very recently been considered to be infinite. The downside of this huge bandwidth offering is that there are technological hurdles that must be overcome to utilize it. With a reasonable number (~100) of frequency-multiplexed channelswhich in optics is referred to as wavelength-division multiplexing (WDM) even though channels are spaced on a grid of constant frequency increments rather than constant wavelength incrementsthe absolute bandwidth of each WDM channel becomes ~100 GHz (10 THz/100 channels). Note that this is the bandwidth

at which electro-optic modulation and opto-electronic detection has to take place. Stated in another way, the baseband electronics and opto-electronics needed to modulate a single WDM channel have to operate over a frequency range from ~100 kHz to ~50 GHz. This requires a single electronic or opto-electronic device, which covers almost all radio frequencies of practical interest (see Figure 4)! Needless to say, the resulting device is extremely broadband, which places enormous demands on microwave design. As a result, optical communications have always been pushing the limits of electronic and opto-electronic high-speed technologies to generate, manipulate, and detect signals, and until recently, digital sampling and processing were considered infeasible at 10-gigabaud symbol rates.

Local Oscillators in Optics Are Lasers

As opposed to their widespread use in radio systems, the use of optical coherent receivers has, until recently, been impractical for commercial applications. This is due to the requirement of a local oscillator (which in optics is a laser!) and the associated optical phaselocked loop at the receiver. While commercially viable semiconductor lasers feature very narrow linewidths (analogous to low phase noise oscillators) compared with the 193-THz carrier frequency, they are associated with a significant amount of phase noise when compared with the underlying symbol rates (typical lasers have linewidths between 100 kHz and 10 MHz, which at 10 gigabaud results in ratios of linewidth to symbol rate of ~10 3 10 5). This has made the implementation of analog PLLs a difficult task and in turn has forced many of the current BPSK and QPSK optical systems to rely on differential demodulation (DBPSK and DQPSK) followed by envelope detection using

Optical Transmitter Data In Electronic Multiplexer Bits to Symbols Electrical Driver Amplifier(s) Electro-Optic Modulator Optical Filtering/ Pre-Emphasis To Channel

From Channel

Optical Filtering/ Equalization

Optical to Electrical Conversion

Symbols to Bits

Electronic Demultiplexer

Data Out

Optional LO (for Coherent)

Optical Receiver

Figure 5. Block diagram of a generic optical communications system.

48

April 2009

optical interferometric techniques [8], [20]. However, with the recent availability of A/D converters at 20 gigasamples per second, the implementation of digital PLLs for coherent receivers running at 10 gigabaud has become feasible. Digital PLLs are commonly used in radio systems today and have pointed the way towards coherent detection of multilevel modulation also in optics, where they are generally known by the term intradyne detection [21]. The main driver behind coherent detection, however, is the growing need for optical capacity, which is turning the fiber bandwidth into a finite resource of bandwidth and as in radio necessitates high spectral efficiencies. To overcome capacity scaling bottlenecks, multilevel modulation together with polarization multiplexing has become imperative. The latter is possible because an optical fiber, as a circular dielectric waveguide, inherently supports two orthogonally polarized propagation modes, both of which may carry independent data streams. Coherent detection with digital postprocessing currently seems the most viable technique for cost-effectively exploiting this polarization dimension. It is important to note that intradyne receivers for 40 Gb/s and beyond require a lot of parallel processing, which is carried out in dedicated ASICs [22] or, at lower speeds, using FPGA prototypes [23]. It is clear that the flexibility derived from the digital receiver architecture comes at the cost of a more complex mixedsignal system design, which in turn puts constraints on the algorithmic complexities that can be implemented at high speeds. Nevertheless, as digital logic speeds continue to increase, this approach may eventually be as commonplace and practical as current-day software radio solutions.

Data rates for digital logic circuits for optical communications applications have increased steadily over the past decades at a fairly constant rate of 14% per year.
in comparing radio and optical systems, both from a practical perspective (see [24] and [25]) and from the perspective of calculating Shannon capacity bounds for the fiber channel [26].

The Anatomy of an Optical Communications System

Figure 5 illustrates the basic building blocks of a modern optical communications system. Starting at the transmitter, the data is usually aggregated from many subscribers using an electronic multiplexer. After or in combination with FEC encoding, the data bits are mapped to symbols according to the underlying modulation format. Next, the digital symbols are converted into an analog transmit waveform, which must be amplified adequately to present sufficient drive voltage for the electro-optic modulator. For some forms of modulation, especially for multilevel formats, driver linearity can be an important factor (e.g., 16-QAM [27]), while for others, it is much less important (e.g., NRZ-OOK). In general, one tries to avoid driver nonlinearity issues by working as much as possible with binary high-speed signals in combination with multiple optical modulators [8], [28]. Today, the most popular device technologies used for designing electro-optic modulators are based on indium phosphide (InP) and lithium niobate (LiNiO3). Typical design approaches include electroabsorption modulators and Mach-Zehnder interferometer modulators (MZMs) [8], [25]. Electroabsorption modulators are backward-biased PIN diodes whose band-edge absorption is varied by changing the bias voltage. In contrast, the MZM works by voltage-controlled interference, as described in the next section. Optical filtering and equalization can be accomplished using, for example, planar lightwave technologies [29]. During propagation, the signal is impaired by additive white Gaussian noise from periodically placed optical amplifiers as well as by fiber nonlinearity, chromatic dispersion, polarization-mode dispersion, or bandwidth truncation due to in-line optical add/drop filters, to name the most important effects [8]. On the receive side, an optical filter is typically used, consisting of technology similar to the type employed on the transmit side. For some modulation formats, such as DPSK and DQPSK, an optical interferometer converts differential phase modulation back to intensity modulation amenable to optical-to-electrical conversion using just a PIN diode [20], [30] or sometimesalthough

Optical Fibers Are Nonlinear

Apart from the above two main differences between optical and radio transmission, various forms of fiber nonlinearity represent a very significant impairment that must be accounted for in optical communications. Nonlinear crosstalk arises between different WDM channels, between bits within a single WDM channel, and even between signal and noise, due to the very narrow confinement of the optical field into a fiber core of ~8 m diameter, leading to optical intensities in the MW/cm2 range over hundreds or thousands of kilometers of transmission. This is of course quite a different type of impairment than the multipath interference problems plaguing radio links. Among the intricacies of fiber nonlinearities is the large nonlinear channel memory induced by chromatic dispersion. As a result, many well-proven equalization and source separation techniques cannot be ported directly from radio systems but require substantial changes and are sometimes even rendered useless in optics. This important fact must never be overlooked

April 2009

49

To overcome capacity scaling bottlenecks, multilevel modulation together with polarization multiplexing has become imperative.
typically at lower speedsan APD (avalanche photodiode) [31], [32]. In the case of coherent demodulation, a reference laser is combined with the incoming signal such that the sum is presented to the photodiode. The photodiode acts as a square-law detector that also filters out the sum-frequency component (at 2 3 193 THz) and hence operates much like an RF mixer to down-convert bandpass signals in a radio system. The resulting baseband signal is passed on to an electronic detection device, which in high-speed systems is still mostly implemented as a hard-decision threshold device. Symbols are then mapped back to bits, and data is demultiplexed for further processing using an electronic high-speed demultiplexer. Clearly, there are many similarities to RF communication systems. However, the technologies and techniques used to build modulators, filters, and detectors differ significantly.

Optical Transmission Formats

As explained earlier, many commercially deployed optical communications systems still use binary

intensity modulation (ON-OFF keying, or OOK) because this format is simple to generate, simple to detect, and exhibits reasonable fiber transmission properties at low-to-moderate spectral efficiencies and bit rates [8]. Driven by the increased demand for fiber-optic capacity (and hence for spectral efficiency), improvements in high-speed electronic and optoelectronic technologies are on the brink of enabling much more complex modulation schemes. These are now starting to appear in research and development. Within the myriad optical modulation formats [8], [25], each technique has its distinct advantages and applications in optical networking, when considering delicate trade-offs between transponder cost, bit rate, spectral efficiency, supported reach, transparent routing functionalities using reconfigurable optical add/drop multiplexers (ROADMs), and other specifics related to the type of network the format is designed for. Applications range from short-range interconnects (~10 m or even less in intraoffice connections) to submarine links, using some 10,000 km of cable to bridge, for example, the Pacific Ocean from the United States to Asia. To try to address all of these aspects would be impossible here, and the reader is referred to more detailed texts, for example, to the reviews [8] and [25] and the references cited therein. Here we will give a brief overview of some of the most popular schemes, with some discussion of basic system architecture considerations.

Data Electrode Light In Light Out

Modulation by Interference: The Mach-Zehnder Structure

The Mach-Zehnder modulator (MZM) is so key to almost all advanced optical modulation schemes that it is worth spending a few minutes to understand it. Figure 6 illustrates the basic structure of this simple yet highly effective device. The interferometer has two legs, and the input light is split equally between them. The idea is to change the optical length (d n) of each leg ever so slightly by imposing a voltage that affects the refractive index (n) of the optical waveguide of physical length (d). This results in an optical phase shift between the two legs. When added back together at an output coupler, the two light paths interfere according to their phase difference, which then results in amplitude modulation. Looked at in more detail [8], one actually finds that the phase difference between the two legs modulates the output amplitude, and the commonmode phase shift modulates the output phase. Therefore, one can address the entire complex plane of the optical

Data Electrode

Transmission Line Terminations

NRZ-OOK DPSK DQPSK 33% RZ-OOK Duo (PSBT) 1 Optical Power 0.8 0.6 0.4 0.2 0 Bias Point Swing 0 V 2V 0 3V 4V

Differential Voltage (V)

Figure 6. Mach-Zehnder modulator operation [8].

50

April 2009

field by changing the differential and commonThe dramatic increase in the speed mode voltages applied to the two legs, at least in of semiconductor device performance principle. In practice, when generating multilevel complex modulation formats, one is often restricted over the past decade is in part a by the available drive amplitudes and by the someresult of improved fabrication times inconvenient coupling between in-phase (I) processes. and quadrature (Q) components [34], so that one typically resorts to a different modulator structure known as an optical I/Q modulator, originally deschemes (e.g., DPSK and DQPSK), a delay interfervised as an optical single-sideband modulator [35]. ometer with the appropriate amount of delay (on Here, one combines two MZMs that work on an amthe order of the symbol duration) is used to convert plitude-only basis in a superstructure interferomphase modulation back to amplitude modulation, eter in such a way that their outputs are rotated by where it is then converted to an electrical signal us90. In this way, one builds an optical I/Q modulaing a simple PIN diode detector. tor (see TX2 in Figure 7). High-speed MZMs are able to respond to broadband data modulation voltages or to narrowband Overview of Some Modulation sinusoidal modulation at the clock frequency, thereOptions at 100 Gb/s by providing a way to generate various types of A significant amount of research has gone into the modulation depending on the voltage swings that exploration of various optical modulation formats result from the input data and/or clock. F i g u r e 7 for 100-Gb/s transmission over the last few years. illustrates several of the most common modu l at ion A selection of results is summarized in Figure 9, schemes [8]. The MZM Drive Voltage column ilwhich illustrates measured bit error ratio (BER) lustrates the required voltage extremes for the input performance curves as a function of the optical data signal necessary to achieve the specified modsignal-to-noise ratio (OSNR) for various experimenulation. In the case of return-to-zero (RZ) modulatally implemented modulation schemes at 107 Gb/s tion, a second MZM is used to carve out an optical pulse at the cenMZM Drive Voltage Circuit Configuration Modulation ter of each original NRZ bit. This second device is (0, V) NRZ-OOK TX1, RX1 in cascade with the data (0, 2V) 33% RZ-OOK TX1, Carver, RX1 modulator and due to Duobinary (PSBT) TX1, RX1 (V, 3V) its simplicity is the most DPSK TX1, RX2 (V, 3V) frequently encountered DQPSK TX2, RX3 (0, V) waveform-shaping element in optical commuS1 nications. Figure 8 illustrates the optical power LASER MZM RX1 spectrum and optical S1 intensity eye diagrams clk TX1 for a selection of optical modulation formats. DL1 MZM For a more complete clk understanding of these RX2 S1 Carver modu lat ion sc hemes MZM and many others, the LASER DL1 reader is referred to [8] MZM /2 /4 and [25]. S2 For intensity-moduTX2 DL1 lated for mats, a simple /4 PIN diode serves as a RX3 s q u a r e - l aw d e t e c t o r Receivers Transmitters Pulse Carver converting the optical signal into an electri- Figure 7. Overview of common optical modulation formats and circuit configurations. MZM and cal signal. For differen- DLI denote Mach-Zehnder modulators and delay line interferometers, respectively. TXi and RXi tial phase modulation denote the i th transmitter and receiver configuration, respectively.

April 2009

51

NRZ-OOK Optical Spectrum 10 dB Optical Spectrum

DPSK

Frequency 33% RZ-OOK Optical Spectrum Optical Spectrum

Frequency DQPSK

Frequency Duobinary (PSBT) Optical Spectrum Optical Spectrum

Frequency 33% RZ-DPSK

Frequency

Frequency

Figure 8. Spectra and eye diagrams for several popular optical modulation formats.

2 NRZ-OOK 3

4 DB (PSBT) 5 6 7 8 9 10 18 22 26 30 OSNR [dB] 34 38 NRZ + OEQ RZ-DQPSK (pol-mux, NRZ-DQPSK coherent) RZ-DQPSK

(i.e., 100 Gb/s plus 7% overhead for FEC). The BER is shown as a function of the optical signal-to-noise ratio (OSNR), which, in contrast to the typical radio SNR definition, is given by the signal power divided by the total noise power in a constant 0.1-nm noise reference bandwidth [25]. It is not too surprising that the coherent dualpolarization RZ-DQPSK modulation scheme provides the best BER performance. This, however, comes at the expense of significant algorithmic processing, which requires dedicated ASICs to work in real time. As a consequence, current coherent detection experiments at 100 Gb/s can only be performed using off-line processing, where one emulates ASIC processing by capturing a portion of, say, a million samples of a received optical signal using a high-speed real-time oscilloscope. One then transfers the captured data to a PC and finally processes them to arrive at a BER. To date, transmission at 100 Gb/s using real-time traffic has only been demonstrated using DQPSK with delay demodulation, both in laboratory experiments and in field trials [36]. Nevertheless, given the current improvements in high-speed ASICs, in the near future real-time processing will be possible also at 100 Gb/s and will certainly be adopted for longhaul optical transmission. Regarding traditional binary modulation for 100-Gb/s transmission [NRZ, possibly with an optical equalizer (OEQ), duobinary (DB), or phaseshaped binary transmission (PSBT)], the real problem is not so much the availability of electrical components but rather the need for improved spectral compression provided by the higher-order modulation formats and the difficulty compensating for impairments over long spans of fiber. Therefore, binary modulation may be most important for future short-reach access links using highly integrated and low-cost optical transponders. A further approach to 100-Gb/s transport is the use of multitone techniques such as OFDM, which have recently been explored for optical communications [3739] through off-line processing in research labs. So far, however, there is no clear indication of the true benefits that could be harvested when using OFDM as compared with single-carrier optical transmission [40], [41].

log10(BER)

100 Gb/s Component Technologies: Challenges and Solutions

Data rates for digital logic circuits for optical communications applications have increased steadily over the past decades at a fairly constant rate of 14% per year [42]. Amazingly, there are now many sources of digital logic circuits that operate at data rates higher than 50 Gb/s, and in fact, current research, as along with limited product offerings, are providing circuits

Figure 9. Comparison of 107-Gb/s back-to-back BER performance of various optical modulation formats [33]; DB denotes optical duobinary, or phase-shaped binary transmission (PSBT).

52

April 2009

that operate up to and even somewhat beyond 100 Gb/s [43][47]. As an example of an ultra-highspeed electronic logic circuit, Figure 10 illustrates a silicon germanium (SiGe) electronic demultiplexer [48]. This device takes a single 85-Gb/s data stream and generates two half-rate substreams using a halfrate external clock. At these data rates, maintaining signal integrity on chip can be a difficult task. Furthermore, it becomes hard to maintain good input sensitivity, as signals are attenuated and reflected much more readily than at lower speeds. Research in high-speed opto-electronic design has also enabled 100-Gb/s electro-optic circuits, both for use at the transmitter and at the receiver. Figure 11 illustrates an InP compact electroabsorption modulator (EAM)based DQPSK modulator demonstrated to operate to 80 Gb/s [49]. This optical integrated circuit takes I and Q electrical channels with bit rates of 40 Gb/s and generates optical 80-Gb/s DQPSK in a single integrated circuit. Another example of an EAM architecture designed to operate directly at 100 Gb/s is discussed in [50]. With high-speed modulators, it is challenging to design the circuit such that a low drive voltage can be used. At 100 Gb/s, it is more difficult to design electronic driver amplifiers that can output high voltages (310 V). On the receive side, new high-speed photodiodes have emerged that are capable of bandwidths beyond 100 GHz. Figure 12 illustrates a device with a 3-dB bandwidth of better than 110 GHz and a responsivity of greater than 0.6 at a wavelength of 1,550 nm [51]. In addition to the challenge of obtaining high responsivity with this bandwidth, it is extremely important to design an integrated wideband bias tee structure on the device. Otherwise 100-Gb/s photodiode performance will be significantly degraded, as it is extremely difficult maintain a flat roll-off response over 100 GHz with discrete bias tee components. The dramatic increase in the speed of semiconductor device performance over the past decade is in part a result of improved fabrication processes and in part due to successful solutions to many challenges in designing and packaging these ultra-high-speed components. As data rates approach 100 Gb/s, even small integrated circuits become vulnerable to distributed circuit effects. Care must be taken to avoid mismatch reflections on integrated circuits. Also, signal paths on and off the chips must be designed with great care so as to provide broad bandwidth signal paths to the outside world and other neighboring integrated circuits. The high-speed electronic and electro-optic circuit technology described in this section is providing design engineers with new building blocks that enable innovative solutions for next-generation optical communications systems.

Over the next decade, we expect the field to mature substantially as high-speed components enable more advanced modulation and digital signal processing.

Figure 10. 85-Gb/s electronic demultiplexer [48].

Figure 11. Compact electroabsorption-modulator-based InP DQPSK modulator (80 Gb/s) [49].
Bias Pad

RidgeWaveguide Contact Pads (GSG) 55 pF Blocking Capacitor GND Signal GND 50 Termination

Diode Waveguide Taper Fiber PIN Vbias Rbias R50 GND Cbias SIG

Figure 12. 110-GHz bandwidth InP photodiode [51].

April 2009

53

Today, the most popular device technologies used for designing electro-optic modulators are based on indium phosphide and lithium niobate.
The Challenge of Packaging for 100 Gb/s: Applying Microwave Techniques to Data Communications
The traditional solution for migrating to higher data rates is to shrink the transmission structures so that they once again look like lumped circuit elements, such as capacitors and inductors. On a single integrated circuit (IC), lumped model analysis may be possible, even at 100 Gb/s. However, when connecting multiple ICs, the required mechanical tolerances and physical distances make ultrashort interconnects with wirelike performance unobtainable. Instead, the designer is forced to work with broadband distributed microwave phenomena.

At a frequency of 100 GHz, the wavelength (l) in air is only 3 mm. Of course, in most transmission structures a dielectric is used that will scale this wavelength by a factor of 1/ !ereff , where ereff is the effective relative dielectric constant of material and transmission line used (coax, coplanar waveguide, microstrip, and so on). The result is that the wavelengths of the electromagnetic fields in dielectrics are shorter, making the problem even worse. A common rule of thumb is to maintain an electrical length of less than l/8 to maintain lumped circuit element properties. At 100 GHz, in air, this would be 375 m. Keeping wire bonds significantly shorter than this becomes difficult in practice. Hence, it becomes clear that the problem must be addressed and not avoided.

Hybrid Opto-Electronic Packaging

Most high-speed serial solutions to date have relied on building separate packages for opto-electronic and electronic components, with high-frequency connectors between them. As an example, consider the reported 100-Gb/s demultiplexing demonstrations [52], [53] using optical receivers that rely on separate components for the optiTraditional cal front end (OFE) and the Integration demultiplexer. The assumpTransmission Lines Transmission Lines tion is that adequate perfor1 2 4 Cable mance can be obtained with coaxial interconnects. At 100 Coaxial Gb/s, however, this is not Interconnects usually the case, as will be explained below. Figure 13(a) illustrates how high-speed Wirebonds 4 Connectors components are typically Wirebonds +1 Cable connected together. The key +2 Transmission Lines to success with this approach +2 Wirebonds is maintaining signal integrity from one box to the next, (a) which becomes a formidable challenge at these speeds. Wirebonds As illustrated in Figure 13(a), for the example multibox design, the signal must traverse a total of four coaxiHybrid Optoelectronic al connectors, one cable, two Integration printed transmission lines, and two wirebonds to travel Transmission Lines from one integrated circuit to the next. At low data rates, Only this is not a problem; however, 1 Transmission Line+ No Connectors at data rates of 50 Gb/s and 2 Wirebonds higher, transmission loss and (b) signal integrity become significant issues. For example, at 100 GHz, the loss of a single Figure 13. (a) Illustration of typical high-speed module interconnects. (b) Hybrid high-quality connectorized optoelectronic integration approach [54].

54

April 2009

cable can approach 2 dB, while the loss at dc will be nearly 0 dB. This spectral roll-off results in significant degradation of signal integrity, as the 3-dB rolloff frequency of a cascaded 100 GHz device will now become a 5-dB roll-off frequency. In comparison, a hybrid integration approach is illustrated in Figure 13(b). A single printed transmission line and two wirebonds replace the multiple interconnects shown in Figure 13(a). Such an approach results in greatly improved signal integrity and less signal attenuation. It becomes clear from examples like this that hybrid opto-electronic integration is key to designing high-performance 100-Gb/s subsystems [54] and will also reduce manufacturing costs. Figure 14 illustrates the first integrated optoelectronic receiver demonstrated at 107 Gb/s [55], [56]. Using an ETDM transmitter, the lowest required OSNR for 107-Gb/s carrier-suppressed return-to-zero (CSRZ) OOK, 21 dB for a 10 -3 BER and a 231-1 pattern length, was reported [55]. This receiver integrates a photodiode with more than 100 GHz of bandwidth [51] with a 1:2 electronic demultiplexer, which reduces the output data rate relative to the input bit rate by a factor of two. Therefore, the output data streams are much more manageable, as they are only operating at 53.5 Gb/s. The input interface is an optical fiber, imposing no bandwidth limitations on the 107-Gb/s signal. The required 100-GHz electrical interface between photodiode and demultiplexer is inside the package, where design parameters can be optimized and carefully controlled. The structure illustrated in Figure 15 interfaces a ground-signal-ground (GSG) device with a groundsignal (GS) device while providing a 3-dB bandwidth of 100 GHz. An edge via is used to immediately transition the ground current on the GS (right side) of the circuit to provide improved insertion loss and return loss.

The microwave signal integrity engineer needs to understand and apply distributed circuit design and have knowledge of electromagnetics and radiating systems to properly design ultra-high-speed subsystems.
properly design ultra-high-speed subsystems. Unlike the microwave radio engineer, the microwave signal integrity engineer must solve broadband transmission problems, typically from 20 kHz up to 100 GHz. While some of the circuit design techniques commonly employed in microwave radio engineering will not be relevant, the analysis skills and microwave design tools can be applied to this new class of problems. Furthermore, the designer of high-speed, mixedsignal ASICs must be able to connect many high-speed

Q1

Q1

CLK

The Growing Role of the Microwave Engineer in Optical Communications System Design
As optical communications systems have moved from data rates on the order of Mb/s to Gb/s, the need for microwave engineers has become more and more critical. At the lower data rates, standard electronic circuit design techniques are sufficient, but in the 50100-Gb/s range, traditional lumped circuit design approaches are of limited use. Specifically, low-frequency design approaches result in ineffective packaging and electronic system design and the inability to realize the performance possible from the new collection of ultra-high-speed electronic and opto-electronic integrated circuits. A new field of expertise is emerging in the industry: microwave signal integrity engineering. The microwave signal integrity engineer needs to understand and apply distributed circuit design and have knowledge of electromagnetics and radiating systems to

Q2

Q2

Figure 14. First integrated 107-Gb/s opto-electronic receiver [55].

Via Holes Edge Via

Ground Ground

Signal

Signal

Ground ~0.6 mm

Figure 15. Dissimilar launch interconnect methodology for 100-Gb/s data path.

April 2009

DC Power

55

digital signals from the integrated circuit to the outside world. At data rates of 50 Gb/s and beyond, great care must be taken to provide matched impedance transitions from the integrated circuit to the multilayer circuit board as well as to manage physical dimensions and electrical crosstalk among many hundreds of parallel interface lines. Properly laying out via holes, circuit traces, and so on in order to minimize crosstalk and radiation and maintain a controlled impedance across the dense array of signals is crucial. This technological challenge also applies to high-speed backplane transmission, where low-cost multilayer boards are still being used but line rates are now approaching 25 Gb/s [57]. These challenges open the door for new research exploring innovative ways to transmit signals over dense low-cost transmission structures [58], including high-data-rate equalization and advanced electrical modulation formats. Microwave analysis, including 3-D field simulations, is essential for this work. As data rates continue to increase and optical modulation schemes continue to employ more and more signal processing, the microwave engineer with a signal processing background will become an even more valuable asset for optical communications system companies.

Conclusions
Optical communications is a quickly evolving field, driven by the ever increasing need for communication bandwidth that has beenand still isprofoundly transforming our society, with telepresence offering exciting future possibilities. Over the next decade, we expect the field to mature substantially as high-speed components enable more advanced modulation and digital signal processing. With the standardization of 100-Gb/s Ethernet well in progress, terabit Ethernet is just over the horizon. As data rates continue to increase above 100 Gb/s, the role of the microwave engineer will become more and more critical to the success of next-generation optical communications gear. There is a tremendous opportunity for the radio engineer to migrate his or her skills to work in the field of microwave signal integrity. This is an up-and-coming field with lots of room for innovative research.

References
[1] J.-L. Beylat, How will optical networks continue to transform the way the world communicates?, in Proc. APOC 2007, Plenary Presentation, Wuhan, China, 2007. [2] D. W. Schloerb, A quantitative measure of telepresence, Presence: Teleoperators Virtual Environments, vol. 4, no. 1, pp. 6480, Feb. 1995. [3] R. Mack, Global landscape in broadband: Politics, economics, and applications, in Optical Fiber Telecommunications V, vol. B, Systems and Networks, I. P. Kaminov, T. Li, and A. E. Willner, Eds. New York: Academic, 2008, pp. 437476. [4] R. E. Wagner, Fiber-based broadband access technology and deployment, in Optical Fiber Telecommunications V, vol. B, Systems and

Networks, I. P. Kaminov, T. Li, and A. E. Willner, Eds. New York: Academic, 2008, pp. 401436. [5] C. F. Lam and W. I. Way, Optical ethernet: Protocols, management, and 1-100G technologies, in Optical Fiber Telecommunications V, vol. B, Systems and Networks, I. P. Kaminov, T. Li, and A. E. Willner, Eds. New York: Academic, 2008, pp. 345400. [6] S. J. Trowbridge, High-speed Ethernet transport, IEEE Commun. Mag., vol. 45, no. 12, pp. 120125, Dec. 2007. [7] R. Metcalfe, Towards Terabit Ethernet, in Proc. Optical Fiber Communication/National Fiber Optic Engineers Conf. (OFC/NFOEC) 2009, Plenary Presentation. [Online]. Available: http://www.ofcnfoec. org/about_ofc/archive/2008/2008-plenary-video.aspx [8] P. J. Winzer and R.-J. Essiambre, Advanced optical modulation formats, Proc. IEEE, vol. 94, no. 5, pp. 952985, 2006. [9] D. Penninckx, M. Chbat, L. Pierre, and J.-P. Thiery, The phaseshaped binary transmission (PSBT): A new technique to transmit far beyond the chromatic dispersion limit, IEEE Photon. Technol. Lett., vol. 9, no. 2, pp. 259261, Feb. 1997. [10] A. Frbert, S. Langenbach, N. Stojanovic, C. Dorschky, T. Kupfer, C. Schulien, J.-P. Elbers, H. Wernz, H. Griesser, and C. Glingener, Performance of a 10.7 Gb/s receiver with digital equaliser using maximum likelihood sequence estimation, in Proc. ECOC, Sept. 2004, pp. 1011, Session Th4.1.5. [11] D. McGhan, C. Laperle, A. Savchenko, C. Li, G. Mak, and M. OSullivan, 5120-km RZ-DPSK transmission over G.652 fiber at 10 Gb/s without optical dispersion compensation, Photon. Technol. Lett., vol. 18, no. 2, pp. 400402, Jan. 15, 2006. [12] T. Mizuochi, Next generation FEC for optical communication, in Proc. Optical Fiber Communication/National Fiber Optic Engineers Conf. (OFC/NFOEC) 2008, pp. 133, paper OTuE5. [13] T. Rappaport, Modulation techniques for mobile radio, in Wireless Communications: Principles and Practice. Englewood Cliffs, NJ: Prentice Hall, 1996, ch. 5. [14] S. Yang, CDMA RF Systems Engineering. Boston, London: Artech House, 1998. [15] J. H. Yuen, M. K. Simon, W. Miller, F. Pollara, C. R. Ryan, D. Divsalar, and J. C. Morakis, Modulation and coding for satellite and space communications, Proc. IEEE, vol. 78, no. 7, pp. 12501266, July 1990. [16] M. J. Reinhart, J. R. Jensen, J. M. Cloeren, C. C. DeBoy, K. B. Fielhauer, and R. C. Schulze, The CONTOUR radio communications system, in Proc. IEEE Aerospace Conf. 2001, vol. 1, pp. 1/811/91. [17] P. C. Marth, TIMED integrated electronics module (IEM), Johns Hopkins APL Tech. Dig., vol. 24, no. 2, pp. 194200, 2003. [18] C. C. DeBoy, C. B. Haskins, T. A. Brown, R. C. Schulze, M. A. Bernacik, J. R. Jensen, W. Millard, D. Duven, and S. Hill, The RF telecommunications system for the new horizons mission to Pluto, in Proc. IEEE Aerospace Conf., Mar. 2004, vol. 3, pp. 613. [19] J. Karaoguz, High-rate wireless personal area networks, IEEE Commun. Mag., vol. 39, no. 12, pp. 96102, Dec. 2001. [20] A. Gnauck and P. J. Winzer, Optical phase-shift-keyed transmission, J. Lightwave Technol., vol. 23, no. 1, pp. 115130, Jan. 2005. [21] K. Kikuchi, Coherent optical communication systems, in Optical Fiber Telecommunications V, vol. B, Systems and Networks, I. P. Kaminov, T. Li, and A. E. Willner, Eds. New York: Academic, 2008, pp. 95129. [22] H. Sun, K. Wu, and K. Roberts, Real-time measurements of a 40 Gb/s coherent system, Opt. Express, vol. 16, no. 2, pp. 873879, 2008. [23] A. Leven, N. Kaneda, and Y.-K. Chen, A real-time CMA-based 10 Gb/s polarization demultiplexing coherent receiver implemented in an FPGA, in Proc. Optical Fiber Communication/National Fiber Optic Engineers Conf. (OFC/NFOEC) 2008, paper OTuO2. [24] R.-J. Essiambre, G. Raybon, and B. Mikkelsen, Pseudo-linear transmission of high-speed TDM signals: 40 and 160 Gb/s, in Optical Fiber Telecommunication IV, Vol. B: Systems and Impairments, I. Kaminow and T. Li, Eds. New York: Academic, 2002, pp. 232304. [25] P. J. Winzer and R.-J. Essiambre, Advanced optical modulation formats, in Optical Fiber Telecommunications V, vol. B, Systems and

56

April 2009

Networks, I. P. Kaminov, T. Li, and A. E. Willner, Eds. New York: Academic, 2008. [26] R.-J. Essiambre, G. J. Foschini, G. Kramer, and P. J. Winzer, Capacity limits of fiber-optic networks, Phys. Rev. Lett., vol. 101, no. 16, Oct. 2008. [27] P. J. Winzer and A. H. Gnauck, 112-Gb/s polarization-multiplexed 16-QAM on a 25-GHz WDM grid, in Proc. ECOC 2008, Brussels. [28] T. Kawanishi, T. Sakamoto, and A. Chiba, High-speed vectorial lightwave modulation techniques, in Proc. IEEE/LEOS Summer Topical Meetings, July 2123, 2008, pp. 9192. [29] C. R. Doerr and K. Okamoto, Planar lightwave circuits in fiberoptic communications, in Optical Fiber Telecommunications V, I. Kaminow, T. Li, and A. Willner, Eds. Amsterdam, The Netherlands: Elsevier, 2008, vol. A, ch. 9, pp. 269341. [30] J. H. Sinsky, A. Adamiecki, A. Gnauck, C. A. Burrus, Jr., J. Leuthold, O. Wohlgemuth, S. Chandrasekhar, and A. Umbach, RZ-DPSK transmission using a 42.7-Gb/s integrated balanced optical front end with record sensitivity, J. Lightwave Technol., vol. 22, no. 1, pp. 180185, Jan. 2004. [31] T. Nakata, T. Takeuchi, K. Makita, and T. Torikai, High-speed and high-sensitivity waveguide InAlAs avalanche photodiodes for 10-40 Gb/s receivers, in Proc. Lasers and Electro-Optics Society (LEOS) 2001, vol. 2, pp. 770771. [32] J. Campbell, Advances in Photodectors, in Optical Fiber Telecommunications V, I. Kaminow, T. Li, and A. Willner, Eds. Amsterdam, The Netherlands: Elsevier, vol. A, pp. 221268, 2008. [33] G. Raybon and P. J. Winzer, Optical transport of 100G Ethernet serial formats, in Proc. 20th Annual Meeting Lasers and Electro-Optics Society (LEOS 2007), Oct. 2007, pp. 417418, paper TuEE3. [34] D. McGhan, M. OSullivan, M. Sotoodeh, A. Savchenko, C. Bontu, M. Belanger, and K. Roberts, Electronic dispersion compensation, in Proc. Optical Fiber Communication Conf. (OFC) 2006, Mar. 510, 2006, paper OWK1 [Online]. Available: http://www.opticsinfobase.org/DirectPDFAccess/BD43BD97-BDB9-137E-CF4F7F5FB28AFC8D_107664.pdf?da=1&id=107664&seq=0&CFID=26892923 &CFTOKEN=46055409 [35] M. Izutsu, S. Shikama, and T. Sueta, Integrated optical SSB modulator/frequency shifter, IEEE J. Quantum Electron., vol. 17, no. 11, pp. 22252227, Nov. 1981. [36] P. J. Winzer, G. Raybon, H. Song, A. Adamiecki, S. Corteselli, A. H. Gnauck, D. A. Fishman, C. R. Doerr, S. Chandrasekhar, L. L. Buhl, T. J. Xia, G. Wellbrock, W. Lee, B. Basch, T. Kawanishi, K. Higuma, and Y. Painchaud, 100-Gb/s DQPSK transmission: From laboratory experiments to field trials, IEEE J. Lightwave Technol., vol. 26, no. 20, pp. 33883402, Oct. 2008. [37] Q. Yang, Y. Ma, and W. Shieh, 107Gb/s coherent optical OFDM reception using orthogonal band multiplexing, in Proc. Optical Fiber Communication/National Fiber Optic Engineers Conf. (OFC/ NFOEC) 2008, paper PDP7. [38] S. L. Jansen, I. Morita, and H. Tanaka, 10121.9-Gb/s PDMOFDM transmission with 2-b/s/Hz spectral efficiency over 1,000 km of SSMF, in Proc. Optical Fiber Communication/National Fiber Optic Engineers Conf. (OFC/NFOEC) 2008, paper PDP2. [39] A. Sano, E. Yamada, H. Masuda, E. Yamazaki, T. Kobayashi, E. Yoshida, Y. Miyamoto, S. Matsuoka, R. Kudo, K. Ishihara, Y. Takatori, M. Mizoguchi, K. Okada, K. Hagimoto, H. Yamazaki, S. Kamei, and H. Ishii, 13.4-Tb/s (134 111-Gb/s/ch) no-guard-interval coherent OFDM transmission over 3,600 km of SMF with 19-ps average PMD, in Proc. ECOC 2008, paper PDP Th3.E.1. [40] S. Jansen, Optical OFDM, a hype or is it for real?, in Proc. IEEE ECOC 2008, Invited Paper, Brussels, Belgium, Sept. 2125, 2008, paper Mo.3.E.3. [41] B. Spinnler, F. N. Hauske, and M. Kuschnerov, Adaptive equalizer complexity in coherent optical receivers in Proc. ECOC 2008, paper We.2.E.4. [42] H. Kogelnik, Perspectives on optical communications, in Proc. OFC 2008, Plenary Talk [Online]. Available: http://www.ofcnfoec. org/Materials/08_Plenary_Kogelnik.pdf

[43] H. Knapp, M. Wurzer, W. Perndl, K. Aufinger, J. Bock, and T. F. Meister, 100-Gb/s 27-1 and 54-Gb/s 211-1 PRBS generators in SiGe bipolar technology, IEEE J. Solid-State Circuits, vol. 40, no. 10, pp. 21182125, Oct. 2005. [44] M. Le, G. He, R. Hess, P. Partyka, B. Li, R. Bryie, S. Rustomji, G. Kim, R. Lee, J. Pepper, M. Helix, R. Milano, R. Elder, D. Jansen, F. Stroili, J.-W. Lai, and M. Feng, Self-aligned InP DHBTs for 150GHz digital and mixed signal circuits, in Proc. Int. Conf. Indium Phosphide and Related Materials, May 812, 2005, pp. 325330. [45] Y. Suzuki, Z. Yamazaki, Y. Amamiya, S. Wada, H. Uchida, C. Kurioka, S. Tanaka, and H. Hida, 120-Gb/s multiplexing and 110Gb/s demultiplexing ICs, IEEE J. Solid-State Circuits, vol. 39, no. 12, pp. 23972402, Dec. 2004. [46] M. Mokhtari, C. Fields, R. D. Rajavel, M. Sokolich, J. F. Jensen, and W. E. Stanchina, 100+ GHz static divide-by-2 circuit in InP-DHBT technology, IEEE J. Solid-State Circuits, vol. 38, no. 9, pp. 15401544, Sept. 2003. [47] Datasheet SHF 408 MUX100 Gbps Multiplexer Module. [Online]. Available: http://www.shf.de/fileadmin/download/modules/datasheet_shf_408_mux_v002.pdf [48] O. Wohlgemuth, W. Mller, P. Paschke, T. Link, R. Lederer, B. Kolb, and H. Dotzauer, Digital SiGe-Chips for data transmission up to 85 Gbit/s, in Proc. EGAAS 2005, Oct. 34, 2005, pp. 245248. [49] C. R. Doerr, L. Zhang, A. L. Adamiecki, N. J. Sauer, and J. H. Sinsky, Compact EAM-based InP DQPSK modulator and demonstration at 80 Gb/s, in Proc. OFC 2007, paper PDP33. [50] T. Johansen, C. Kazmierski, C. Jany, C. Jiang, and V. Krozer, Optimization of integrated electro-absorption modulated laser structures for 100 Gbit/s Ethernet using electromagnetic simulation, in Proc. 2007 SBMO/IEEE MTT-S International Microwave and Optoelectronics Conf. (IMOC 2007). [51] A. Beling, H.-G. Bach, G. G. Mekonnen, R. Kunkel, and D. Schmidt, Miniaturized waveguide-integrated p-i-n photodetector with 120-GHz bandwidth and high respon sivity, IEEE Photon. Technol. Lett., vol. 17, no. 10, pp. 21522154, Oct. 2005. [52] R. H. Derksen, G. Lehmann, C.-J. Weiske, C. Schubert, R. Ludwig, S. Ferber, C. Schmidt-Langhorst, M. Moller, and J. Lutz, Integrated 100 Gbit/s ETDM receiver in a transmission experiment over 480 km DMF, in Proc. OFC 2006, Mar. 510, 2006, paper PDP37. [53] C. Schubert, R. H. Derksen, M. Mller, R. Ludwig, C.-J. Weiske, J. Lutz, S. Ferber, and C. Schmidt-Langhorst, 107 Gbit/s transmission using an integrated ETDM receiver, in Proc. ECOC 2006, Sept. 2006, paper Tu1.5.5. [54] J. H. Sinsky, 100-Gb/s hybrid opto-electronic integration, in Proc. ECOC 2008, Invited Paper, Brussels, Sept. 2125, 2008. [55] J. H. Sinsky, A. L. Adamiecki, L. Buhl, G. Raybon, P. Winzer, O. Wohlgemuth, M. Duelk, C. R. Doerr, A. Umbach, H.-G. Bach, and D. Schmidt, 107-Gbit/s opto-electronic receiver with hybrid integrated photodetector and demultiplexer, in Proc. OFC 2007, paper PDP-30. [56] J. H. Sinsky, A. Adamiecki, L. Buhl, G. Raybon, P. Winzer, O. Wohlgemuth, M. Duelk, C.R. Doerr, A. Umbach, H.-G. Bach, and D. Schmidt, A 107-Gbit/s optoelectronic receiver utilizing hybrid integration of a photodetector and electronic demultiplexer, J. Lightwave Technol., vol. 26, no. 1, Jan. 2008. [57] A. Adamiecki, M. Duelk, and J. H. Sinsky, 25 Gbit/s electrical duobinary transmission over FR-4 backplanes, Electron. Lett., vol. 41, no. 14, pp. 826827, July 7, 2005. [58] J. H. Sinsky, M. Duelk, and A. Adamiecki, High-speed electrical backplane transmission using duobinary signaling, IEEE Trans. Microwave Theory Tech., vol. 53, no. 1, pp. 152160, Jan. 2005.

April 2009

57