optical communication

LU_LAYOUT_Layout 1 11/22/13 12:19 PM Page 6

Advanced Modulation Formats for Short Reach Optical Communication Systems Li Tao, Yu Ji, Jie Liu, Alan Pak Tao Lau, Nan Chi, and Chao Lu Abstract

The use of advanced modulation formats helps to reduce the number of components required for short reach optical communication modules at 100 Gb/s and beyond. Unlike long haul optical communication systems where externally modulated high order modulation formats with coherent detection receivers are used, direct modulated lasers with direct detection are likely to be employed for short reach system implementation. Here we discuss the performance of pulse amplitude modulation, carrierless amplitude/phase modulation, and direct detection orthogonal frequency-division multiplexing for short reach optical communication systems. Their possible use for future high-capacity short reach optical communication systems is explored.

T

he proliferation of bandwidth intensive services and cloud computing has driven the speed of data communication links to higher and higher data rate. The 100 Gb/s bit rate has been standardized, while discussion has been actively carried out for 400 Gb/s and 1 Tb/s data links. Although multiple wavelength channels can be used (e.g., 4 × 25.8 Gb/s for 100 Gb/s connection) for implementing transceiver modules for such communication systems, the demand for system integration to reduce system cost requires the number of wavelengths to be kept to the minimum. One possible approach to realize this is through the use of more advanced modulation formats to increase spectral efficiency by increasing the data rate for a given transmission system bandwidth. High order modulation formats such as quadrature phase shift keying (QPSK), quadrature amplitude modulation (QAM), and orthogonal frequency-division multiplexing (OFDM) with coherent detection and a digital signaling processing (DSP) algorithm have been studied extensively in recent years for long haul optical communication systems. External modulation using IQ modulators is typically employed at the transmitter, while at the receiver, optical hybrid and local oscillators together with DSP algorithms implemented using application-specific integrated circuits (ASICs) are employed for signal detection. Dispersion and polarization mode dispersion as well as other linear impairments can easily be compensated; spectral efficiency as high as 10 b/s/Hz has been achieved, and coherent transceivers supporting 100 Gb/s per channel using polarization-multiplexed QPSK are widely available for commercial deployment. However, for the short to medium term, coherent detection technique is unlikely to be used for short reach systems. The use of IQ modulator and optical hybrid will not only increase system cost, but also increase transceiver footprint and, as a result, increase the difficulty of system integration. The complicated coherent detection algorithm will also significantly increase the power consumption of the transceiver modules, which can be an important concern in short reach optical communication systems. For short reach communication systems, low-cost light sources such as vertical

6

0890-8044/13/$25.00 © 2013 IEEE

cavity surface emitting lasers (VCSELs), directly modulated lasers (DMLs), and externally modulated lasers (EMLs) are likely to be used, and signal will be directly detected at the optical receiver. As a result, no optical phase information will be available. Because of these issues, we need to study solutions that can increase system spectral efficiency while using DML, EML, or VCSEL lasers and direct detection. A number of schemes have been investigated. These include pulse amplitude modulation (PAM) [1, 2], carrierless amplitude/phase modulation (CAP) [3–5], and direct detected orthogonal frequency-division multiplexing (OFDM)/discrete multi-tone (DMT) modulation [6, 7]. Performance as well as power consumption have been studied and compared [1]. In this article, we review the characteristics of these modulation schemes for short reach optical communication systems while highlighting some of the work we have done in the area.

Pulse Amplitude Modulation Systems The simplest scheme for increasing data rate without increasing transmission symbol rate is through the use of PAM. Unlike simple non-return to zero (NRZ) signal, where two signal levels are used, pulses with multiple signal levels are employed for signal transmission. If the number of possible signal levels in a symbol is M and signal transmission speed (symbol rate) is D, data rate R can be increased by a factor of log2M, and R = Dlog2M. Since the bandwidth requirements of transmitters and receivers are decided by the symbol rate, the data rate can be increased with low-cost optical components. However, with the increased number of signal levels, receiver sensitivity will be reduced. At the same time, with the number of increased pulse levels, the signal will be more like an analog signal, and impairments that typically will only affect an analog optical transmission system such as relative intensity noise (RIN) and waveform distortion due to limited component bandwidth will have a more significant effect in such a transmission system. In addition, when the number of levels is doubled, the receiver sensitivity will be reduced by at least 3 dB.

IEEE Network • November/December 2013


SMF TX

O-DEMUX

Level encoder

Optical source

Driver

O-MUX

FEC encoder

Transmitter

SMF RX

Data decision

FEC decoder

Receiver

Equalizer

TIA

PIN

APD (a) 5

1

Power penalty (dB)

-Log(BER)

4

2

3

4 5 -32

PAM4 BTB PAM8 BTB PAM16 BTB PAM4 after 3 km PAM8 after 3 km PAM16 after 3 km -28

-24

-20

PAM2 PAM4 PAM8 PAM16

3

2

1

0 -16

-12

-8

-4

0

-160

-155

-150

145

-140

-135

Received optical power (dBm)

Laser RIN (dB/Hz)

(b)

(c)

-130

-125

-120

Figure 1. 25 Gbaud PAM system: a) simulation model; b) BER curves; c) power penalty due to RIN for PAM4, PAM8, and PAM16 signals.

Trade-off between system performance and number of levels required will be an important concern in a PAM-based system if the bandwidth of system components is fixed. To evaluate the effect of various system parameters on overall system performance, we have simulated the performance of a 25 Gbaud PAM system using the VPI™ Transmission Maker. The schematic diagram of the PAM system is shown in Fig. 1a. The level encoder will convert a binary data stream into 4-level, 8level, and 16-level signals to generate the PAM4, PAM8, and PAM16 signals. The symbol rate is 25 Gbaud; the data rate will then be 50 Gb/s, 75 Gb/s, and 100 Gb/s. We studied system performance with the typical system component parameters for back-to-back and 3 km fiber link transmission. For component parameters, we assume an AC extinction ratio of 10 dB, dark current of 5 nA, RIN of –140 dB/Hz, and receiver TIA equivalent noise input current of 25 —— pA/ Hz and bandwidth of 20 GHz. The performance is shown in Fig. 1b. The results show that under typical system parameters, the bit error rate (BER) floor has significantly limited the usefulness of PAM8 and PAM16 for high-speed signal transmission systems. In addition, the performance of PAM8 and PAM16


systems will be significantly impacted by relative intensity noise (RIN) and dispersion, as shown in Figs. 1b and 1c. Due to the analog nature of the high order PAM signal, other factors such as transmitter linearity will also have significant impact on system performance. It is commonly agreed that modulation formats of higher order than PAM8 will unlikely be used, and an adaptive equalizer at the receiver for impairment equalization is necessary [1, 2].

Carrierless Amplitude and Phase Modulation An attractive alternative scheme that may provide good system performance using low-cost optical components such as DML and VCSEL is CAP modulation [3–5]. It allows relatively high data rate to be achieved using optical components of limited bandwidth. Compared with alternative schemes such as QAM and OFDM, no electrical or optical complexto-real-value conversion is necessary, which involves a complex mixer and radio frequency (RF) source or optical IQ modulator. Neither does it require the discrete Fourier transform (DFT) utilized in OFDM signal generation and demodulation.

7


a(t)

Mapping

A

I/Q separation

↑M B

In-phase shaping filter

Optical source

D/A

b(t)

C

↑M

EDFA

SSMF

Quadrature shaping filter

-20 -40

-20

B

0 -20

2

4 6 8 Frequency (GHz)

10

0 20 Power (dBm)

D

-60 -80

-100

2


2


10

-40 -60

20

E

0 -20 -40

0

C

-80 1552.45

10

Power (dBm)

-40 Power (dBm)

M/T

-40 0

Direct detection

A/D

Power (dBm)

0

D

r(t) Quadrature matched filter

10

B Power (dBm)

Power (dBm)

20

E

Downsampling

Demapping

Equalization

In-phase matched filter

1552.50 1552.55 Wavelength (nm)

F

1552.60 ε=4 ε=8

0 -20 -40

0

2


10

0

2


10

Figure 2. Schematic diagram of system structure based on CAP modulation. The insets are the spectra after A: coding, B: D/A, C: optical up-conversion, D: direct detection, E: matched filters; F: the frequency response of the shaping and matched filters with different e (e is the ratio between the number of taps of shaping or matched filters and the up-sampling factor).

Principle Figure 2 shows the schematic diagram of system structure based on CAP modulation. Although the shaping filters at the transmitter and the match filters at the receiver can be implemented using analog techniques, here we focus on the digital implementation since it is more likely to be used due to its flexibility. At the transmitter, the original bit sequence is first fed to an encoder, which maps blocks of bits into complex symbols where T is the symbol period. The coded sequence is then separated into the in-phase and quadrature components. The two tributaries are first up-sampled by a factor M, that is, M – 1 zeros are inserted between two consecutive input symbols and sent into the two digital shaping filters, respectively. The outputs of the filters are subtracted. The generated CAP signal is then converted into an analog signal using a digitalto-analog (D/A) converter. It is subsequently used to modulate a light source. It should be noted that the D/A converter and digital shaping filters are working at a rate of M/T, but the bandwidth of the CAP signal is decided by the system’s symbol rate 1/T. At the receiver, direct detection is used, and the received signal after analog-to-digital (A/D) conversion is fed into two matched filters to separate the in-phase and

8

quadrature components. The A/D converter and matched filters also operate at a rate of M/T. After down-sampling, an equalization technique is employed, and a decoder is utilized to obtain the original bit sequence. The generated CAP signal can be expressed as s (t) = a(t)  f(t) − b(t)  f2(t)

(1)

where a(t) and b(t) are the in-phase and quadrature components of the transmitted bit sequence after the coding and upsampling process, respectively. The functions f 1 (t) = g(t) ◊ cos(2p fct) and f2(t) = g(t) ◊ sin(2p fct) are the corresponding impulse response of shaping filters and form a so-called Hilbert pair [8], as shown in Fig. 3a. g(t) is the square-root raised-cosine function, which is usually employed as the baseband impulse response. Note that the shaping filter should work at a higher rate than the system symbol rate. Thus, the up-sampling process is used here to match the rate of the shaping filter and obtain the output analog signal without the aliasing products. Assuming that the channel response is ideal, the output of the two matched filters at the receiver is expressed as



1.0

In-phase Quadrature

(a)

1.0

0.5 Amplitude

Amplitude

0.5

0.0

-0.5

-1.0

-1.0 -100

-50

0 t

50

-100

100

h12(t)

(c)

1.0

-50

0 t

50

100

h22(t)

(d)

0.5 Amplitude

0.5 Amplitude

0.0

-0.5

1.0

h11(t)

(b)

0.0

0.0

-0.5

-0.5

-1.0

-1.0 -100

-50

0 t

50

100

-100

-50

0 t

50

100

Figure 3. The impulse responses of: a) in-phase and quadrature shaping filters; b) h11(t); c) h12(t); d) h22(t). ri(t) = s(t)  m1(t) = a(t)  h11(t) – b(t)  h12(t) rq(t) = s(t)  m2(t) = –a(t)  h12(t) + b(t)  h22(t) (2) Here, m 1 (t) = f 1 (–t) and m 2 (t) = f 2 (–t) are the impulse response of the corresponding matched filtesrs, and  denotes convolution: h11(t) = f1(t)  f1(t), h12(t) = f1(t)  f2(t), h22(t) = f 2 (t)  f 2 (t). Figure 3b–3d indicates the joint impulse response h 11 (t), h 12 (t), and h 22 (t), respectively. For the inphase component in Eq. 2, the first term on the left is the desired signal component. However, it is corrupted by the second term, which comes from the quadrature component. It can be observed from the joint impulse response h 11(t) and h 12 (t) that the maximum h 11 (t) coincides with the zero of h 12 (t). Therefore, the desired in-phase component can be extracted without intersymbol interference (ISI) and the distortion that comes from the quadrature components at an appropriate sample time. The same conclusion can be obtained for the quadrature component in Eq. 2. As mentioned above, synchronization is very important in CAP demodulation. However, the appropriate sampling time is hard to decide, and sampling time offsets will lead to subsequent signals seriously affected by intersymbol interference (ISI) and the crosstalk between the in-phase and quadrature components. Considering that the distortions induced by either ISI or crosstalk is linear, an adaptive equalizer is needed to recover the output signal of matched filters. Usually, a fractional space equalizer is adopted, which can equalize the serious channel distortion and time delay. The classic


cascaded modulus algorithm (CMA) [7] is often utilized for the equalizers because it is a blind algorithm and easy to realize. However, it is much less effective for a high order signal modulation system, such as CAP16, which does not have constant amplitude and the error function in the adaptation process will not approach zero even for an ideal multi-level signal without distortion. Therefore, the cascaded multi-modulus algorithm (CMMA) is usually utilized for the multi-level CAP signal [9]. Its error function can be expressed as fc = r (t ) − A1 − A2 − A3

(3)

Here, A1, A2, and A3 are the radius of the chosen reference circles. r(t) = p(t) + j ◊ q(t) is the output of the matched filters. For the multi-level CAP system, there is only one complex input for the equalizer as supposed to polarization-multiplexed (PM) transmission in long-haul systems. Since there is no phase noise in the CAP system because of the use of direct detection, the desired output of the equalizer should have similar square constellation. A method called the multi-modulus algorithm (MMA) is presented with better performance to take advantage of the symbol statistics of square constellations [10]. Thus, we propose an equalization scheme for a multi-level CAP system called the modified CMMA scheme. The error function in Eq. 3 is modified as

9


Received optical power at 10^-3

-6 CAP16, CMMA CAP32, CMMA CAP64, CMMA CAP16, modified CMMA CAP32, modified CMMA CAP64, modified CMMA

-9 -12

CAP64

-15 3 dB

-18 CAP32 -21

1.2 dB CAP16

0.6 dB

-24 0.4

0.6

0.8

1.0

1.2 1.4 BL ratio

1.6

1.8

2.0

Figure 4. Comparison between CMMA and modified CMMA for different bandwidth limitations at the transmitter and coding schemes.

fc′1 = p(t ) − Am1 − Am 2 fc′2 = q(t ) − Am1 − Am 2

(4)

Here A m1 = (L 1 + L 2 )/2 and A m2 = (L 2 – L 1 )/2, L 1 , L 2 represent the two moduli of the real and imaginary values of the desired normalized CAP16 symbols, respectively. The difference between Eqs. 4 and 3 is that both the errors of the real and imaginary components are calculated. It is similar to the MMA scheme but more suitable for multi-level signals. Thus, the coefficients of the transfer function for the IQ components will be updated individually. On the other hand, the separated cost functions do not take into account the phase of each symbol, so the subsequent phase rotation in the previous work [4, 11] is unnecessary. Another benefit from modified CMMA is the reduced number of required reference moduli. If the reference modulus is increasing, the difference of the outer moduli is smaller, which makes it harder to find a correct reference modulus during the convergence process. For CAP16 signal, the number of moduli for CMMA and modified CMMA is 3 and 2, respectively. For CAP32 signal, it is 5 and 3. For CAP64 signal, it is 9 and 4. Thus, the number of required reference moduli is reduced significantly with the increase of the coding level. Note that an initial CMA-based equalization for pre-convergence is utilized for these two schemes [9]. It can be found from Fig. 4 that the modified CMMA scheme outperforms CMMA for different bandwidth ratio and coding schemes. The BL ratio represents the bandwidth limitation of the transmitter, which is the ratio between the effective bandwidth of the transmitter and the system’s symbol rate. It can also be concluded that modified CMMA is more suitable for a high order CAP system because the improvement is increased to 3 dB for a CAP64 system, and 0.6 dB and 1.2 dB for CAP16 and CAP32 systems with little bandwidth limitation.

Experiment Figure 5a shows the experiment setup of a CAP16 system. First, the original data sequence is mapped into four levels and up-sampled by a factor of 4. Then the separated signal is sent into two shaping filters with 32 taps. An arbitrary waveform generator (AWG, Tektronix 7122C) is used to produce the RF signal. The square root raised-cosine function is used as the baseband impulse response with a roll-off coefficient of 0.1. A ligthwave at 1551.6 nm from an external cavity laser is

10

used as the signal source. The output RF signal is used to drive the intensity modulator (IM). The fiber launch power is set at 2 dBm. After the fiber link and PD, the CAP16 signal is sampled by an oscilloscope and processed offline. First, the sampled signal is sent into two matched filters with 32 taps, so the in-phase and quadrature signals are separated. After down-sampling, the equalizer is used and then followed by signal demapping. Finally, bit error counting for CAP16 signal is performed over 2 × 105 bits. Figure 5b shows the spectrum of the received signal of a 10 Gb/s CAP16 system with CMMA-based equalization. Note that phase rotation need to be done after the equalization. The measured BER curves of the CAP16 system with BTB and over different transmission length are shown in Fig. 5c. 2 dB and 3.5 dB receiver power penalties are observed at BER of 10–3 with 20 km and 40 km fiber links, respectively. We also demonstrate a 20 Gb/s CAP16 system. Figure 5d shows the electrical spectrum after PD. Compared with Fig. 5b, the received 5 Gbaud signal is distorted because of the bandwidth limitation of the AWG, electrical amplifier, and oscillator scope (Tektronix TDS6604, 6 GHz analog bandwidth). In fact, the maximum power attenuation can be over 8 dB at the high-frequency components. Compared to the CMMA scheme, 3.2 dB improvement at BER of 10 –3 after BTB transmission is found when the modified CMMA scheme is used, as shown in Fig. 5e. Therefore, the proposed modified CMMA scheme can lower the requirement for in-band frequency response flatness of CAP systems. 0.9 dB and 1.7 dB receiver power penalties are observed at BER of 10–3 with 12 km and 24 km fiber links, respectively, but 5.1 dB penalty for 40 km transmission. The channel response fading is believed to be the reason for this large penalty, which cannot be neglected without dispersion compensation. The demonstrated data rate is limited by the analog bandwidth of the AWG and scope used. A much higher data rate can be expected when higher-speed AWG is used. Using a high-speed AWG, multiband CAP, and EML, 102 Gb/s signal transmission over 15 km single-mode fiber (SMF) have recently been realized successfully [11].

Direct Detected Orthogonal FrequencyDivision Multiplexing Direct detected OFDM (DD-OFDM), also known as DMT, is another attractive scheme for low-cost and short reach communication [12]. The output after inverse fast Fourier transform (IFFT) in a DMT scheme is real-valued, which makes the IQ modulation onto an RF or optical carrier unnecessary, so it reduces the system cost. As a kind of multi-carrier modulation technique, DMT shows high spectrum efficiency, flexible multi-level coding, and obvious tolerance to ISI. Figure 6a shows the schematic diagram of the DMT system. The original bit sequence is first fed to an encoder, which maps blocks of bits into complex symbols. A high-order coding scheme can be used here. Then the serial complex data is divided into parallel streams, and each stream is modulated onto N subcarriers. In order to get a real-valued time-domain output, a new 2N points sequence is constructed, and the second N points is the conjugate symmetric sequence of the first half. After 2N point IFFT, the output signal is of real value. The cyclic prefix is then inserted for each DMT symbol. The generated DMT signal is passed through a D/A converter and subsequently used for optical modulation. At the receiver, direct detection is employed. After A/D conversion, the cyclic



Fiber link

AWG

I/Q matched filter

Offline processing

4-levels demapping

I/Q shaping filter

I/Q separation

Upsampling

EDFA

EA

Downsampling

SSMF

CAP16 generator

4-levels demapping

Oscillator scope

IM

Equalization

ECL

(a)

B

A 0.01

20 BER

Electrical power (dB)

40

C

C B

1E-3

A

0

-20

1E-4

0

1

2

3

4

5

-26

BTB After 20 km After 40 km -22

-24

-20

-18

Frequency (GHz)


(b)

(c)

-16

20

10

CMMA, BTB Modified CMMA, BTB Modified CMMA, 12 km Modified CMMA, 24 km Modified CMMA, 40 km

0.01 8 dB

BER

Power

0

1E-3

-10

0.9 dB 1.7 dB 3.2 dB 5.1 dB

-20 Before AWG After AWG After PD, BTB -30

0

2

1E-4 4

6

8

10

-20

-18

-16

-14

Frequency (GHz)


(d)

(e)

-12

Figure 5. a) Experimental setup of CAP system; b) electrical spectrum of received 10 Gb/s CAP16 signal; c) BER performance of 10 Gb/s CAP16 system with different transmission lengths; d) the spectrum of 20 Gb/s CAP16 signal at different nodes; e) BER performance of 20 Gb/s CAP16 system with different equalization schemes. prefix is removed and followed by 2N points FFT. Only the first half of the 2N points output after FFT is necessary for signal decoding. There are several issues in DMT modulation system. The most important one is the high peak-to-average power ratio


(PAPR) for multi-carrier modulation. This shortcoming would bring serious nonlinear distortion in the electrical and optical domains, especially at the driving power amplifier and in the fiber link. It would also reduce the dynamic range of a DMT system. Usually, a complementary cumulative dis-

11


CP insertion

2N - IFFT

SN S1 S0 S0 S1

Outlook

Optical source

D/A

A number of modulation formats are available for direct modulated short reach optical communication systems. Among them, PAM is the simplest to implement. Reasonable performance can be obtained when PAM4 and PAM8 are used. It is a good candidate for a very short reach optical communication system with distance below 10 km. For longer transmission distances and data rates beyond 100 Gb/s, alternative modulation formats need to be considered. Among them are multicarrier modulations based on discrete multitone modulation and/or optical orthogonal frequency-division multiplexing. 50 Gb/s transmission using DMT over 20 km single mode fiber and DML [14], and 40 Gb/s transmission over 80 km using OFDM and EML have been realized[7]. CAP-based modulation offers much promise as a technology that may offer very high data transmission rate while allowing simple system implementation since it allows simple modulation and direct detection to be utilized. System performance depends significantly on the algorithm employed at the transmitter and the receiver. Further work in this area should help to significantly improve system performance further while simplifying system implementation.

Receiver CP removal

Parallel to serial

Demapping

Conjugate symmetric sequence

SN

Transmitter

2N - FFT

Serial to parallel

Mapping

S0 S1

Direct detection

A/D

SN (a)

100

CCDF

10-1

10-2

0.3 dB

10-3

10-4

1 dB

DMT DFT-S OFDM CAP 5

6

7

8 9 PAPRO (dB)

10

11

12

(b)

Figure 6. a) Schematic diagram of DMT modulation system; b) PAPR performance comparison for different schemes. The number of subcarriers for the OFDM scheme is 256, and there are four modulation levels. tribution function (CCDF) curve is used to evaluate the PAPR performance. CCDF presents the probability distribution that the PAPR of the current signal symbol is higher than a certain threshold, PAPR0. Figure 6b indicates the PAPR performance comparison for different schemes. It can be found that the CAP system has 1 dB better PAPR performance than the DMT system at CCDF of 10–3. We also add a curve to indicate the PAPR performance of a discrete Fourier transform spread (DFT-S) OFDM system, which is well known as an effective PAPR reduction technique in long haul coherent OFDM transmission. Compared to the CAP system, there is 0.3 dB improvement but at the cost of an extra FFT at the transmitter, so the computational complexity is increased. Another key issue that affects the multi-carrier system performance is intercarrier interference (ICI) [13]. Frequency offset between the transmitter and receiver is the key factor that induces ICI. Although there is no carrier frequency offset in a DMT system because of the direct detection, the sampling frequency offset caused by the instability of the crystal oscillator of the sampling module has an obvious influence on a DMT system. This sampling frequency offset will make the mismatch between the frequency grid of subcarriers and FFT, leading to serious distortion. On the other hand, ICI also comes from the FFT process at the receiver. The distortion from one of the subcarriers will spread over the whole frequency band during the FFT process, and this distortion is very hard to equalize. Commercial 10 G-class DML-based 50 Gb/s and 100 Gb/s DMT systems have recently been proposed [14, 15]. The longer distance transmission is limited by the chromatic dispersion in the fiber link. Therefore, new DSP techniques that reduce the PAPR and ICI, and improve chromatic dispersion tolerance for DMT systems are required.

12

Acknowledgment This work was supported by project G-UA44 and G-YL24 of the Hong Kong Polytechnic University and Huawei Technologies Co. Ltd, the NHTRDP (973) of China (Grant No. 2010CB328300), NNSF of China (No. 61107064, No. 61177071), NHTRDP (863 Program) of China (2011AA010302, 2012AA011302), and the National Key Technology R&D Program (2012BAH18B00).

References [1] J. L. Wei et al. , “Performance and Power Dissipation Comparisons Between 28 Gb/s NRZ, PAM, CAP and Optical OFDM Systems for Data Communication Applications,” J. Lightwave Tech., vol. 30, no. 20, Oct. 2012, pp. 3273–80. [2] A. Ghiasi and B Welch, IEEE 802.3bm Fiber Optic Task Force, Sept. 2012. [3] J. D. Ingham et al., “40 Gb/s Carrierless Amplitude and Phase Modulation for Low-Cost Optical Datacommunication Links,” Proc. OFC/NFOEC, Los Angeles, USA, paper OThZ3, Mar. 2012 [4] L. Tao et al., “Experimental Demonstration of 10 Gb/s Multi-level Carrierless Amplitude and Phase Modulation for Short Range Optical Communication Systems,” Optics Express , vol. 21, no. 5, Mar. 2013, pp. 6459-65. [5] R. Rodes et al., “Carrierless Amplitude Phase Modulation of VCSEL with 4 bit/s/Hz Spectral Efficiency for Use in WDM-PON,” Opt. Express, vol. 19, no. 27, Dec. 2012, pp. 26551–56. [6] T. Gui et al., “Auto Bias Control Technique for Optical OFDM Transmitter with Bias Dithering,” Optics Express, vol. 21, no. 5, 2013, pp. 5833–41. [7] Y. Bao et al. , “Nonlinearity Mitigation for High-Speed Optical OFDM Transmitters using Digital Pre-Distortion,” Optics Express, vol. 21, no. 6, 2013, pp. 7354–61. [8] G.-H. Im et al., “51.84 Mb/s 16-CAP ATM LAN Standard,” IEEE JSAC, vol. 13, no. 4, May 1995, pp. 620–32. [9] X. Zhou and J. Yu, “Multi-level, Multi-Dimensional Coding for High-speed and High-Spectral-Efficiency Optical Transmission,” J. Lightwave Tech., vol. 27, no. 16, Aug. 2009, pp. 3641–53. [10] J. Yang, J. J. Werner, and G. A. Dumont, “The Multimodulus Blind Equalization and Its Generalized Algorithms,” IEEE JSAC, vol. 20, no. 5, Jun. 2002, pp. 997–1015. [11] M I Olmedo et al., “Towards 400GBASE 4-lane Solution Using Direct Detection of MultiCAP Signal in 14 GHz Bandwidth per Lane,” Proc. OFC/NFOEC, Anaheim, USA, paper PDP5C.10, Mar. 2013. [12] Z. H. Li et al., “432-Gb/s Direct-Detection Optical OFDM Superchannel Transmission over 3040-km SSMF,” IEEE Photon. Tech. Lett., vol. 25, no. 15, 2013, pp. 1524–26. [13] L. Tao et al., “Reduction of Intercarrier Interference Based on Window Shaping in OFDM RoF Systems,” IEEE Photon. Tech. Lett., vol. 25, no. 9, May 2013, pp. 851–54. [14] T. Tanaka et al. , “50 Gbps Class Transmission in Single Mode Fiber using Discrete Multi-tone Modulation with 10G Directly Modulated Laser,” Proc. OFC/NFOEC, Los Angeles, CA, paper OTh4G.3, Mar. 2012.



[15] W. Yan et al., “100 Gb/s Optical IM-DD Transmission with 10G-Class Devices Enabled by 65 GSamples/s CMOS DAC Core,” Proc. OFC/NFOEC, Anaheim, CA, paper OM3H.1, Mar. 2013.

communications. He is a reviewer and Technical Program Director for various IEEE/OSA journals and international conferences in photonics and communications.

Biographies

NAN CHI received her B.S. and Ph.D. degrees from BUPT in 1996 and 2001, respectively. From July 2001 to December 2004, she worked as an assistant professor at the Research Center COM, Technical University of Denmark, Lyngby. From January 2005 to April 2006, she was a research assistant at the University of Bristol, United Kingdom. She joined Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, from June 2006 to May 2008, where she worked as a full professor. Since May 2008, she joined Fudan University. She is the author or co-author of more than 100 papers. Her research interests are in the area of convergence of wireless and optical networks, all-optical processing, and advanced modulation formats

LI TAO (taoli522930) received a B.S. degree from Huazhong University of Science and Engineering, China, in 2006. Currently, he is a Ph.D. student with the Department of Communication Science and Engineering, Fudan University, Shanghai, China, and an exchange student in the Department of Electronic and Information Engineering, the Hong Kong Polytechnic University. His research interests include advanced modulation format, OFDM, and shortreach optical communications. YU JI ([email protected]) received her Ph.D. degree from Beijing University of Posts and Telecommunications (BUPT), China, in 2013. Her research interests include optical signal processing, and high-speed optical transmission systems. Currently, she is working at China Mobile Group Beijing Company Ltd. JIE LIU’s biography was unavailable when this issue went to press. ALAN PAK TAO LAU ([email protected]) received his B.A.Sc. degree in engineering sciences and M.A.Sc. degree in electrical and computer engineering from the University of Toronto, Ontario, Canada, in 2003 and 2004, respectively. He received his Ph.D. degree in electrical engineering from Stanford University in 2008 and joined the Hong Kong Polytechnic University, where he is currently an associate professor. His current research interests include various aspects of coherent fiber optic communication systems and OPM. He is currently the principal investigator and/or co-investigator of various government and industry funded research projects in optical


CHAO LU ([email protected]) received his B.Eng. degree in electronic engineering from Tsinghua University, China, in 1985, and his M.Sc. and Ph.D. degrees from the University of Manchester, United Kingdom, in 1987 and 1990, respectively. From 1991 to 2006, he was with the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore, as a lecturer, senior lecturer, and associate professor. From June 2002 to December 2005, he was seconded to the Institute for Infocomm Research, Agency for Science, Technology, and Research (A*STAR), Singapore, as program director and department manager. Since April 2006 he has been a professor in the Department of Electronic and Information Engineering, Hong Kong Polytechnic University. His research interests include optical communication systems and networks, fiber devices for optical communication, and sensor systems.

13

optical communication

Recommend Documents