Ampalavanapillai Nirmalathas2

1

2

Networked Systems NICTA Victoria Research Laboratory Victoria, 3010, Australia [email protected], [email protected]

Department of Electrical and Electronic Engineering The University of Melbourne Victoria, 3010, Australia [email protected]

Abstract— Forthcoming 100 Gb/s Ethernet (100 GbE) standard has targeted 100 Gb/s transmission over 10 km and 40 km of single mode fiber (SMF) using four channel (4 x 25 Gb/s) wavelength division multiplexed systems, which is neither cost effective nor spectrally efficient compared to a potential single channel system exploiting higher order modulation such as optical orthogonal frequency division multiplexing (O-OFDM). This paper investigates a single channel 100 Gb/s system based on directly detected 64-QAM O-OFDM that has an effective OFDM bandwidth of 24 GHz. It is found that the system can effectively enable error-free (at BER 10-3 without FEC) transmission of data at the desired distances of SMF without any in line amplification. Keywords—Optical OFDM; OFDM; detection; spectral efficiency; 100 GbE.

I.

incoherent;

direct

INTRODUCTION

The explosive growth in bandwidth hungry applications has demanded a bandwidth requirement of 100 Gb/s in the next generation optical access networks [1-2]. The IEEE P802.3ba Ethernet Task Force has been formed with a target to standardize current technologies and offer a common platform for commercial deployment [3]. The taskforce has been considering 4 (four) parallel WDM data stream transmission at 1310 nanometer (nm) band each carrying 25 Gb/s to construct a 100 Gb/s link over a maximum distance of 40 km. This proposition allows them to exploit the benefits of 25 GHz optoelectronics at the expense of additional cost and complexity from 3 (three) sets of additional transmission and receiving opto-electronics and MUX/DEMUX devices [3-4]. As an alternative, if 100 Gb/s transmission, with a single optical carrier and opto-electronics devices with bandwidths of 25 GHz, can be achieved by using a combination of higher order modulation and optical orthogonal frequency division multiplexing (O-OFDM), the additional cost and complexity for multiple parallel transmissions can be greatly reduced. In addition, as O-OFDM systems are immune to fiber impairments, they can operate in both 1300 nm and 1550 nm bands, and exploit the benefits of matured high-speed longhaul technologies. Recently 100 Gb/s and beyond optical transmission systems based on coherent and incoherent OOFDM have been demonstrated using a combination of

NICTA is funded by the Australian Government as represented by the Department of Broadband, Communications and the Digital Economy and the Australian Research Council through the ICT Centre of Excellence program

polarization division multiplexing (POL-MUX), higher order modulations and/or self-coherence [5-10]. Most of these demonstrations have been focused in long–haul systems and adopted system architectures enabling the transmissions at longer distances, where systems’ cost and complexity were never of paramount importance. However, unlike the long-haul systems, the picture is quite different in access domain, where cost-effectiveness and simplicity are the key enablers of the prospective technologies [11-12]. Since higher order modulations such as 128 and 64 quadrature amplitude modulation (QAM) are readily possible with the current stateof-arts [13-15], this paper explores a combination of 64-QAM with low-cost and simpler directly detected O-OFDM (DDOOFDM) system to achieve a spectrally efficient single channel 100 Gb/s transmission system, where a single optical carrier is used to impress and transport 100 Gb/s data over a 40 km unamplified single mode fiber (SMF) link, limiting the OFDMsignal bandwidth within 24 GHz. Also, the proposed system is designed by using the basic IEEE 802.11a OFDM PHY [16] specifications such as the numbers of subcarriers and pilot carriers, training sequences, the choice of modulation formats, the amount of cyclic prefix etc., so that the existing wireless OFDM standard may potentially be considered as a benchmark for future selection of 100 Gb/s O-OFDM specifications in access and metro domain [17].

Figure 1. 100 Gb/s DDO-OFDM simulation setup

II.

SYSTEM DESCRIPTION

size, as shown in (2).

Shown in Fig. 1, functionality of the proposed single channel 100 Gb/s DDO-OFDM system is verified with a simulation model developed in the commercial simulation tool VPItransmissionMakerTM V7.6. The key simulation steps and parameters are described below in more detail for clarity. A. Baseband OFDM signal generation To generate the baseband OFDM signal, the basic OFDM parameters such as FFT size, number of data subcarriers, number of pilot subcarriers, cyclic prefix (CP) length and allocation of subcarriers for data and pilot etc. are chosen according to the IEEE 802.11a OFDM PHY specification [16]. Therefore, a FFT block size of 64 is chosen to map the pseudorandom-bit-sequence (PRBS) data, among which the center subcarrier is filled with zero, 48 subcarriers are associated with 64-QAM modulated data and 4 subcarriers are dedicated for binary-phase-shift-keyed (BPSK) pilot symbols. Two OFDM symbols in the preamble are used for training and a CP of sixteen samples is used to provide a guard interval between OFDM symbols. Assuming the OFDM subcarriers to be numbered from -32 to +32, the subcarrier allocation is shown in Fig. 2. Fig. 2 shows that 48 data subcarriers (d0 to d47) are mapped on the subcarriers numbered -26 to -22, -20 to -8, -6 to -1, 1 to 6, 8 to 20, 22 to 26 and four pilot subcarriers are mapped on the subcarriers numbered -21, -7, 7, and 21. The remaining higher frequency subcarriers are filled with zeros for oversampling that facilitate the low pass filtering of the signal after digital-to-analog-converter (DAC). Since BPSK modulation is used for pilots and preamble (training symbols), whereas higher order QAMs are used in data subcarriers, modulation formats vary across OFDM subcarriers that results in power variation across subcarriers. Therefore normalization is performed to ensure a constant average power across all subcarriers for all mappings [16]. Normalized data symbols ( Cnorm ) are generated by Cnorm = Ci × K mod , where modulation dependent normalization factor K mod is calculated by K mod =

R = 2 × NFFT ×

η= Since

1 Tsym

× (1 + η )

N CP

(1)

(2)

NFFT

1 denote the symbol-rate/subcarrier, the useful Tsym

bandwidth (B) of the signal containing data, pilot and the unused center subcarrier is calculated according to the (3), where NSD and NPILOT denote the number of data and pilot subcarriers respectively. Then the effective bit rate (b) can be calculated using (4), where M is the highest number of constellation points of the used QAM modulation format in each data subcarrier.

B = ( NSD + NPILOT + 1) × b = NSD ×

1 × (1 + η ) Tsym

1 × log 2 M Tsym

(3) (4)

Now, as the number of subcarrier (NSD) and η are fixed in this work according to the IEEE 802.11a OFDM PHY, a symbol rate of 350 MBd/subcarrier is chosen to achieve an overall bit rate of 100 Gb/s where data are encoded using 64-QAM (e.g. 48 × 350e 6 × 6 ). Then the Nyquist sampling rate (R) of the baseband signal is 28 GS/s ( 64 × 350e 6 × 1.25 ) with a useful signal bandwidth (B) of 24 GHz ( 53 × 350e 6 × 1.25 ) allowing an effective spectral efficiency of 4 bit/s/Hz. It is to be noted that the used sampling rate of 28 GS/s is not unrealistic since the commercially available analog-to-digital converters (ADCs) already have a maximum sampling rate of 56 GS/s for future 100 Gb/s systems [18].

, M is the highest number of constellation

M M

∑C

2

n

n =1

points, |Cn| is the amplitude of the nth ideal constellation symbol in the complex plain (i.e. Cn = In + jQn) and Ci is the ith data symbol in complex plain. The values of Kmod for different modulation formats are summarized in Table 1 for reference. A total of 502 OFDM symbols including two OFDM training symbols are generated for the transmission simulation. The Nyquist sampling rate of the baseband OFDM signal is calculated using (1), where NFFT is the FFT size, Tsym is the OFDM symbol time including the guard interval and η is the ratio of the number of samples used as CP (Ncp) and the FFT

Figure 2. Subcarrier allocation for data and pilots

B.

DDO-OFDM transmission and detection In DDO-OFDM setup, radio frequency (RF) up-conversion is needed to ensure the required spectral gap between the optical carrier and the lowest OFDM subcarrier to accommodate the intermodulation products that generate during direct detection [5-6]. Hence, for the desired 100 Gb/s DDO-OFDM signal, the base band O-OFDM signal, having a TABLE 1: MODULATION DEPENDENT NORMALIZATION FACTOR KMOD Modulation

Kmod

BPSK

1

4-QAM

1/√2

8-QAM

1/√6

16-QAM

1/√10

32-QAM

1/√20

64-QAM

1/√42

128-QAM

1/√82

Nyquist sampling rate of 28 GHz, is mixed with a 36 GHz (24 GHz × 1.5) RF signal to generate the required 24 GHz spectral gap between the carrier and the lowest OFDM subcarrier. The composite RF signal is then used to drive a Mach-Zehnder modulator (MZM) biased just above the intensity null to maintain an optimal carrier to OFDM sideband power ratio [56]. A distributed feedback (DFB) laser module (LD) with a linewidth of 1 MHz is used as the optical source. A Butterworth optical band pass filter (OBPF) module is used to suppress the lower sideband of the optical signal. The optical single side band (SSB) O-OFDM signal is then amplified using an optical gain block (5 dB noise figure) to maintain an optical power of +4 dBm at the input of the fiber, which is the maximum launch power suggested by the 100 GbE task force [3]. The generated SSB DDO-OFDM signal is then transported over a block of SMF and directly detected using a PIN photodiode at the receiver. The SMF has an attenuation of 0.2 dB/km, dispersion of 16 ps/nm/km, nonlinear index of −20 2.6 × 10 m2/W, and PMD of 0.1 ps/√km. Another OBPF is used before the PIN photodiode (PD) to avoid any out of band noise. The PIN photodiode has a responsivity of 0.9 A/W with shot noise and thermal noise enabled. The detected OFDM signal is then amplified and down mixed with another 36 GHz local oscillator (LO) to recover the I and Q channels. C. Baseband OFDM detection at the receiver Assuming perfect I/Q-imbalance compensation at the receiver, the received I and Q channels are combined to form the string of complex samples. The string of complex numbers is then divided into groups of 80 (i.e. 64 + 16) complex numbers to form the OFDM blocks. The CP, comprising of 16 complex numbers in the beginning of each block, is removed and each OFDM block is fed into the inputs of the FFT module. FFT operation then helps to recover 64-QAM encoded data symbols associated with the subcarriers in an OFDM block. As O-OFDM uses narrowband subcarriers, the subcarriers are subject to flat fading while passing through the fiber. Therefore the frequency domain channel transfer function hk of the k-th subcarrier is estimated once for every 500 OFDM symbols using the leading two OFDM training symbols as give by (7)

hk

(y =

tr 1 k

+ yktr 2 ) xktr 2

*

(7)

Since, both OFDM training symbols tr1 and tr2 use the tr same BPSK sequence, xk denotes the frequency domain common BPSK modulated symbol on the k-th subcarriers of the two training symbols and * denotes the complex conjugate.

yktr1 and yktr 2 denote the frequency domain received BPSK symbols on the k-th subcarrier corresponding to the common tr training symbol xk . Since the DDO-OFDM system uses local oscillator to down-convert the OFDM signal to baseband, the frequency offset between the local oscillator and the transmitter’s oscillator cause common phase rotation of the subcarriers, which is estimated and compensated by the four dedicated pilot subcarriers, as give by (8)

Φi =

1 S pilot

∑

{arg( y

ik

k ∈S pilot

}

) − arg( x ik )

(8)

i = 1 ,..., L p and k = 1 ,..., NST Here, xik and yik are transmitted and received symbol in frequency domain respectively at i-th OFDM symbol and the kth subcarrier. Spilot denotes the set of pilot subcarrier indices,

ik is a arg(.) corresponds to the phase angle of a symbol and x known frequency domain pilot symbol at the i-th OFDM symbol and k-th subcarrier. Lp is the maximum number of OFDM symbols in a frame (e.g. 500 in this work) and NST is the maximum number of used subcarriers (i.e. data subcarriers, pilot subcarriers and the zero contained center subcarrier) in an OFDM symbol. Finally the estimated transmitted symbol

xjik in frequency domain is found by (9). To recover the data j are then denormalized by bits, the estimated data symbols x ik

dividing the symbols with Kmod and passed to the symbol decision algorithm such as QAM demodulator.

y × exp(− jΦ i ) × hk* xjik = ik 2 hk

(9)

D. Bit-error-ratio (BER) calculation using Error vector magnitude (EVM) Since 64-QAM encoded 500 OFDM symbols are transmitted in the simulation, the total number of transmitted bits is only 144000 (i.e. 500 × 48 × 6 ) that does not allow the BER measurement below 10-5. Therefore BER is estimated from the value of the root-mean-square EVM (EVMRMS) of the transmission system using (10) [19]: ⎛ 1⎞ 2 ⎜1 − ⎟ ⎛ ⎛ 3log L ⎞ ⎛ 2 L⎠ BER = ⎝ ×Q⎜ ⎜ 2 2 ⎟⎜ 2 ⎜ ⎝ L − 1 ⎠ ⎝ EVM RMS log 2 L × log 2 M ⎝

⎞⎞ ⎟⎟ ⎠ ⎟⎠

(10)

Here, L is the maximum number of levels between I and Q dimensions of the M-QAM (i.e. L is 4 for 8-QAM) whereas M is the highest number of constellation points of the modulation format M-QAM (i.e. M is 8 for 8-QAM). The function Q (x) is evaluated by

⎛ x ⎞ 0.5 × erfc ⎜ ⎟ , where erfc (.) is the ⎝ 2⎠

complementary error function. The value of EVMRMS is calculated by (11) [16], [19] Lp

EVM RMS

2⎤ ⎡ NST j ∑ ⎢ ∑ xik − xik ⎥ 1 ⎦ = = i =1 ⎣ k =1 NST × Lp × Pavg SNR

(11)

xik is the normalized ideal constellation symbol in the j, complex plain corresponding to the estimated symbol x where

ik

xjik − xik denotes the magnitude of error vector, Pavg is the average power of the constellation that is 1 (one) since normalization is applied (or otherwise Pavg =

1 ) and 2 K mod

SNR is the electrical signal-to-noise ratio. It is interesting to note that in decible (dB) unit, EVM is the negative of the SNR according to (12). 2 EVM (in dB) =10log10(EVMRMS )= - SNR (in dB)

III.

(12)

SIMULATION RESULTS

To validate the simulation model, the system is at first simulated in a back-to-back setup with 0 Hz laser linewidth, 0 km of fiber and varying the additive white Gaussian noise (AWGN) of the system. The BER performance of the system in such ideal back-to-back setup is plotted against the EVM of the system in Fig. 3 where corresponding theoretical BERs estimate from (10) are also plotted as lines for reference. As shown in Fig. 3, the performance of the simulation model closely matches with the theoretical prediction as expected. Fig. 3 also confirms the EVM thresholds for different modulation orders to achieve a certain BER. For example, for usual forward error correction (FEC) limit of 10-3 BER, the EVM thresholds are roughly -23 dB and -10 dB for 64 QAM and 4-QAM respectively. Now, to measure the optical transmission performance, the laser linewidth is set to 1 MHz and the signal is transmitted over 0 km to 40 km of SMF without any inline amplifier or dispersion compensation. The relevant spectra of the signal at different points of the link are shown in Fig. 4. As shown in Fig. 4 (a), I channel of the signal has a bandwidth of 12 GHz after suppressing the aliasing signals using a low pass filter after the DAC. Then after upconverting the OFDM signal to 36 GHz, the RF spectrum has an OFDM signal bandwidth of 24 GHz and a total signal bandwidth of 48 GHz as shown in Fig. 4 (b). The optical spectra after the modulator and before photodetector are shown

Figure 3. BER vs. EVM in an optical back-to-back setup with 0 Hz laser linewidth and 0 km fiber along with theoretical BERs as lines. EVM threshold for different BER can be found from this figure.

Figure 4. Spectra of 100 Gb/s DDO-OFDM signal

in Fig. 4 (c) and (d), which confirm that lower side band is successfully suppressed using optical band-pass filter. The RF spectrum after photodetector in Fig. 4 (e) clearly shows that the intermodulation products are effectively accommodated within the spectral gap leaving the actual OFDM signal uncontaminated. The last RF spectra in Fig. 4 (f) is the downconverted and low pass filtered I channel, which is subsequently sampled and processed using OFDM receiver digital signal processing module. The performances of this 100 Gb/s DDO-OFDM system at back-to-back (0 km SMF) and over 40 km length of SMF are shown in Fig. 5. The insets of Fig. 5 are the unequalized and equalized constellations after 40 of SMF at a received optical power of -16 dBm. As shown in Fig. 5, the sensitivity of the receiver after 40 km SMF is -18 dBm at the FEC limit. Thus there is a power margin of 14 dB since the maximum received optical power after 40 km is -4 dBm. Hence this 14 dB power margin will ensure successful operation of the proposed 100 Gb/s system over a 1:16 split passive optical network (PON). Fig. 5 also shows that at the

Figure 5. Performance of the 100 Gb/s DDO-OFDM system over 40 km SMF with 1 MHz laser linewidth.

[7]

[8]

[9]

[10]

[11] Figure 6. Effect of laser linewidth on the 100 Gb/s DDO-OFDM signal over 40 km SMF

FEC limit receiver’s sensitivity decreases by 2 dB after 40 km SMF. This performance penalty after 40 km can be attributed to chromatic dispersion since the phase coherency between the optical carrier and the OFDM subcarriers vanishes due to fiber dispersion and it introduces significant phase noise in direct detection of the signal [20]. Therefore, to further improve the performance of the system after 40 km of SMF, we simulate the system with different laser line-widths (LLWs) and the performance is shown in Fig. 6. As expected, the receiver’s sensitivity improves by 1 dB and 2 dB respectively with LLW of 500 kHz and 100 kHz. Thus the power budget increases to 15 dB and 16 dB with laser line-widths of 500 kHz and 100 kHz respectively, which will enable deployment of the proposed 100 Gb/s DDO-OFDM system in traditional FTTH infrastructure with a split ratio of 1:32. IV. CONCLUSION Simulation results confirm that a single channel 100 Gb/s DDO-OFDM system can effectively operate over 40 km SMF with an OFDM bandwidth of 24 GHz, which could be a viable alternative to 4 x 25 Gb/s WDM approaches currently being considered by the IEEE P802.3ba task force for future 100 Gb/s short-haul deployments. REFERENCES [1]

[2]

[3] [4]

[5]

[6]

J. D'Ambrosia, "40 gigabit ethernet and 100 gigabit ethernet: the development of a flexible architecture - [Commentary]," IEEE Communications Magazine, vol.47, no.3, pp.S8-S14, March 2009 M. Cvijetic, "Towards 100 GbE introduction: Challenges and practical aspects," in the proc. of the 10th Anniversary International Conference on Transparent Optical Networks (ICTON 2008), vol.1, no., pp.1-4, 2226 June 2008 IEEE P802.3ba Task Force, http://www.ieee802.org/3/ba M. Duelk, R. Gutierrez-Castrejon, "4x25-Gb/s 40-km PHY at 1310 nm for 100 GbE Using SOA-Based Preamplifier," J. Lightwave Technology, vol.26, no.12, pp.1681-1689, June 15, 2008 B. J. C. Schmidt, A. J. Lowery, and J. Armstrong, "Experimental Demonstrations of Electronic Dispersion Compensation for Long Haul Transmission Using Direct-Detection Optical OFDM," J. Lightwave Technol., vol. 26, pp. 196-203, 2008. Wei-Ren Peng, X. Wu, V. R. Arbab, Kai-Ming Feng, B. Shamee, L. C. Christen, Jeng-Yuan Yang, A. E. Willner, Sien Chi, "Theoretical and Experimental Investigations of Direct-Detected RF-Tone-Assisted

[12]

[13]

[14]

[15]

[16]

[17]

[18] [19]

[20]

Optical OFDM Systems," J. Lightwave Technol., vol.27, no.10, pp.1332-1339, May15, 2009 Y. Ma, Qi Yang, Y. Tang, S. Chen, W. Shieh, "1-Tb/s per channel coherent optical OFDM transmission with subwavelength bandwidth access," in proc. Conference on Optical Fiber Communication (OFC 2009), vol., no., pp.1-3, 22-26 March 2009 S. L. Jansen, I. Morita, T. C. W. Schenk, and H. Tanaka,"121.9-Gb/s PDM-OFDM transmission with 2-b/s/Hz spectral efficiency over 1000 km of SSMF," J. Lightwave Technol., vol.27, pp. 177-188, 2009. B. Schmidt, Z. Zan, L. B. Du, A. J. Lowery, "120 Gbit/s Over 500-km Using Single-Band Polarization-Multiplexed Self-Coherent Optical OFDM," J. Lightwave Technol., vol.28, no.4, pp.328-335, Feb.15, 2010 D. Qian, N. Cvijetic, J. Hu, T. Wang, "108 Gb/s OFDMA-PON with polarization multiplexing and direct-detection," in proc. Optical Fiber Communication Conference (OFC 2009), vol., no., pp.1-3, 22-26 March 2009 M. Bakaul, A. Nirmalathas, C. Lim, D. Novak, and R. Waterhouse, “Spectrally efficient hybrid multiplexing and demultiplexing schemes toward the integration of microwave and millimeterwave radio-overfiber systems in a WDM-PON infrastructure,” OSA Journal of Optical Networking (JON), vol. 8, no. 5, pp. 462-470, May 2009. M. Bakaul, A. Nirmalathas, C. Lim, D. Novak, and R. Waterhouse, “Experimental characterization of single as well as cascaded WDM optical interfaces in millimeter-wave fiber-radio networks,” IEEE Photonics Technology Letters, vol. 18, no. 1, pp. 115-117, 2006. F. Buchali, R. Dischler, “Optimized sensitivity direct detection OOFDM with multi level subcarrier modulation”, In proc. of Optical Fiber Communication Conference (OFC 2008), 2008, Paper OMU3. J. Yu, X. Zhou, Yue-Kai Huang, S. Gupta, Ming-Fang Huang, T. Wang, P. Magill, “112.8-Gb/s PM-RZ-64QAM Optical Signal Generation and Transmission on a 12.5GHz WDM Grid”, In proc. of Optical Fiber Communication Conference (OFC 2010), 2010, Paper OThM1. S. Okamoto, T. Omiya, K. Kasai, M. Yoshida, M. Nakazawa, “140 Gbit/s Coherent Optical Transmission over 150 km with a 10 Gsymbol/s Polarization-Multiplexed 128 QAM Signal”, In proc. of Optical Fiber Communication Conference (OFC 2010), 2010, Paper OThD5. IEEE Std 802.11a-1999(R2003), Part 11: Wireless LAN MAC and PHY Specifications, Clause 17, [online] Available: http://standards.ieee.org/getieee802/download/802.11a-1999.pdf L. Mehedy, M. Bakaul, A. Nirmalathas, "115.2 Gb/s optical OFDM transmission with 4 bit/s/Hz spectral efficiency using IEEE 802.11a OFDM PHY," in proc. the 14th OptoElectronics and Communications Conference, 2009 (OECC 2009), July 2009 I. Dedic, “56Gs/s ADC : Enabling 100GbE”, In proc. of Optical Fiber Communication Conference (OFC 2010), 2010, Paper OThT6. R. A. Shafik, M. S. Rahman, and A. H. M. R. Islam, “On the Extended Relationships Among EVM, BER and SNR as Performance Metrics”, In proc. 4th International Conference on Electrical and Computer Engineering, pp. 408-411, December, 2006 Z. Zan, M. Premaratne, and A. J. Lowery, “Laser RIN and linewidth requirements for direct detection optical OFDM,” CLEO’08, Paper CWN2