ICTON 2013

We.C1.1

Performance Limits of All-Optical OFDM Systems

Julian Hoxha, Gabriella Cincotti
Engineering Department, via della Vasca Navale 84, I-00146 Rome, Italy
e-mail: gabriella.cincotti@uniroma3.it
ABSTRACT
Limits of the bit error rate (BER), information spectral efficiency and transmission reach are analyzed for an alloptical orthogonal frequency division multiplexing (AO-OFDM) with 200Gbaud/s symbol rate and quadrature
phase-shift keying (QPSK) or quadrature amplitude modulation (QAM).
Keywords: All optical OFDM, coherent receivers, ASE noise, phase noise.
1. INTRODUCTION
The throughput of a single-wavelength, single-polarization high-speed optical transmission can be increased by
enhancing the spectral efficiency and using higher-order modulation formats. Orthogonal frequency division
multiplexing (OFDM) can provide the maximum 1 bit/s/Hz spectral efficiency, by distributing data over a set of
orthogonal subchannels, that partially overlap in the frequency domain. Higher-order modulation formats, such
QAM, pulse amplitude modulation (PAM) and phase shift keying (PSK) can further increase the system
throughput by transmitting multiple bits over a single symbol. In this case, coherent optical detection is required,
that presents higher receiver sensitivity, but is severely affected by phase noise.
Novel approaches to implement OFDM directly in the optical domain have been proposed and experimentally
demonstrated [1-3]. An AO-OFDM system can scale Tb/s-rate signals into lower speed optical subcarriers to
reduce the receiver complexity; in addition, these systems do not require expensive and power-consuming highspeed digital signal processing (DSP), and they are more spectral efficient, compared to coherent-OFDM (COOFDM), because no guard interval or training symbols are needed. The AO-OFDM system allows the same
flexible bandwidth allocation as other CO-OFDM approaches, but with lower sampling rate and complexity
equalization. In addition, AO-OFDM performs the optical discrete Fourier transform (ODFT) by using an
arrayed waveguide grating (AWG)-based passive device, so that the subchannels are generated directly in the
optical domain, with a larger power efficiency with respect to CO-OFDM.
Although AO-OFDM schemes have been largely investigated during the past five years, a comprehensive
analysis of the system performances is still lacking in literature, in particular considering the phase noise effects.
In the present paper, we present a detailed investigation, through numerical simulations, of coherently modulated
AO-OFDM systems over a multispan transmission, in terms error vector magnitude (EVM) and bit error rate
(BER). The maximum spectral efficiency has been evaluated, as well as the influence of phase noise, amplified
spontaneous emission (ASE) noise from inline amplifiers and fiber nonlinear effects.
2. SYSTEM SETUP
Figure 1 shows the AO-OFDM system architecture: the source is a mode-locked laser (MLL) that generates
a train of optical Gaussian pulses with 200 GHz rep-rate and 2 ps full width half maximum (FWHM); the optical
spectrum is wide enough to cover the bandwidth of all the N = 8 subcarriers. A 1:8 power splitter sends the
optical pulses into eight independent IQ Mach-Zehnder modulators (MZM), that are used to encode information
with arbitrary modulation formats. An AWG-based device implements the optical inverse discrete Fourier
transform (OIDFT), and it is used to simultaneously filter and multiplex the optical subcarriers. The passive
device has eight input/output ports, eight grating waveguides and 200 GHz free spectral range (FSR); the
orthogonal AO-OFDM subcarriers are 25 GHz frequency spaced [4, 5]. Eight independent pseudo random
binary sequences (PRBS) of 65536 symbols are either quadrature phase skit keying (QPSK) or 16-QAM
modulated and digital-to-analog converted; the RF signals, with 2V peak- to-peak voltage, drive the IQ-MZMs,
that are biased at the minimum transmission point -V. The optical signal generated at one output of the AWGbased device is transmitted over a fiber link composed of 10 spans; each span is a 80 Km standard single mode
fiber (SSMF), with = 0.2 dB/km attenuation coefficient, n2 = 2.7 10-20 nonlinear index, Aeff = 80 m2 fiber core
effective area, and D = 17 ps/nm/km dispersion coefficient at 1550 nm. In each span, an Erbium Doped Fiber
Amplifier (EDFA) with 4 dB noise figure (NF) compensates the in-span losses, and a 13.6 km dispersion
compensating fiber (DCF) is used to mitigate chromatic dispersion. A band pass filter (BPF) at the receiver
eliminates the out-of-band noise. The transmission of the optical signal through the fiber is simulated by using
the Optilux software [6].
At the receiver side, an identical AWG-based device performs the ODFT to demultiplex the AO-OFDM
subchannels at its outputs. In a conventional OFDM scheme, coherent demodulation is performed using eight 90
hybrids, a CW local oscillator (LO) and time gating, to sample the received symbols. Therefore, eight additional
electro-absorption modulators (EAM) are required in the coherent receiver [3], increasing the power

978-1-4799-0683-3/13/$31.00 2013 IEEE

ICTON 2013

We.C1.1

consumption and the system complexity. In the present approach, we propose a simplified coherent receiver with
a pulsed LO, to both coherent detect and sample the received signals; therefore time gating is no longer
requested at the receiver [7]. In a coherent optical time-domain sampling (COTDS) receiver, the LO signal is
generated by a MLL, identical to the one used at the transmitter, that can be synchronized with the received
signal; we assume a zero frequency offset between the transmitter and receiver lasers. The optical signal from
each AWG outputs is mixed with the pulsed LO in a 2 x 4 90 hybrid and detected with two balanced
photodetectors for the I and Q components. The photodiodes are assumed noiseless and the output signals are
sent to DSP circuits.

Figure 1. AO-OFDM system architecture.

3. SIMULATION RESULTS
We have investigated the system performances by evaluating the BER as function of the overall optical power
sent to the first fiber span, using Monte-Carlo simulations. From an inspection of Fig. 2, we observe that BER
increases with the optical power, due to the nonlinear effects, that become dominant with respect to simplified
spontaneous emission (ASE) noise [8].

Figure 2. BER versus launched optical power.

Figure 3 reports the maximum transmission distance versus the launched power for BER=10-3. For optical power
larger than -1 dBm, the nonlinearity effects becomes dominant over the ASE noise limitations.
The optical signal to noise ratio (OSNR), evaluated over a 0.1 nm bandwidth is [9]:

OSNR0.1 = Pch L NF 10log M + 58

(1)

where Pch is the power launched in the first fiber span, L = 18.7 dB the overall fiber span loss, NF = 4 the
amplifier noise figure and M = 8 the number of spans. In our case, we have Pch = -1 dBm, and the OSNR
is 24.3 dB.
The maximum spectral efficiency, can be evaluated according to the Shannon-Hartley formula, when the
OSNR is evaluated over the 200 GHz bandwidth

=
SE log 2 (1 + OSNR1.6 )
Considering only ASE noise, the maximum SE is 3.22 bit/s/Hz.

(2)

ICTON 2013

We.C1.1

Figure 3. Maximum transmission distance versus launched power.

4. PHASE NOISE EFFECTS
The influence of the phase noise has been evaluated in terms of the OSNR, as

OSNR =

(3)

N 1 sin ( n )

p2 1 +
n
n =1
sin

where N is the number of OFDM subchannels and the sum takes into account the crosstalk among the
subchannels [4]. The phase noise variance is p = 2, where is LO linewidth and is the detected symbol
duration.
We consider a 16-QAM transmission, analysing two different cases, when time gating is used to sample the
signals at the outputs of the 90 hybrids, and when a MLL is used as pulsed LO. In the first case, the EAMs are
driven by a pulse sequence of = 2 ps duration and 40 ps time interval and the CW LO has a = 100 kHz
linewidth. On the other hand, the pulsed LO is the signal from a MLL with a = 2510-9 GHz bandwidth.
The EVM as a function of the OSNR and the BER versus the EVM are shown in Fig. 4(a) and 4(b), respectively
[10, 11]. We observe that the phase noise creates a floor in the system performance and that the LO laser
linewidth is the main critical parameter in the coherent receivers.
0

45

w time gating
w/o time gating

40

-1
-2

35

log(BER)

EVM[%]

30
25
20
15

-4
-5
-6
-7

10

-8

-9

w time gating
w/o time gating

-3

10

20
OSNR[dB]

30

-10

40

10

20

EVM[%]

30

40

50

(a)
(b)
Figure 4: (a) EVM versus OSNR;. (b) BER versus EVM.
5. CONCLUSIONS
The performance limits of 200 Gbaud QPSK AO-OFDM systems in long-haul transmission have been
numerically investigated. The optimum launched power has been evaluated, by observing a BER worsening due
to nonlinear effects. The maximum transmission distance, using forward error correction (FEC) is 1100 km and
the maximum SE is 3.22 bit/s/Hz.
In addition, phase noise effects have been evaluated for 16-QAM modulation, considering both time gating and
a pulse LO; in the latter case, the BER presents a floor and the LO laser linewidth becomes the main critical
parameter.

ICTON 2013

We.C1.1

ACKNOWLEDGEMENTS
This work was supported by the European Commission through ICT-ASTRON project (Contract No. 318714)
funded under the 7th Framework Programme and the Italian Ministry of University and Research through
ROAD-NGN project (PRIN2010-2011).
REFERENCES
[1] D. Hillerkuss et al.: Simple all-optical FFT scheme enabling Tbit/s real-time signal processing, Opt. Exp.,
vol. 18, pp. 9324-9340, 2010.
[2] H. Chen et al.: 5200Gbit/s all-optical OFDM transmission using a single optical source and optical
Fourier transform real-time detection, Opt. Exp. vol. 19, pp. 21199-21204, 2011.
[3] S. Shimizu et al.: Demonstration and performance investigation of all-optical OFDM systems based on
arrayed waveguide gratings, Opt. Exp., vol. 20, pp. B525-B534, 2012.
[4] G. Cincotti et al.: Characterization of a full encoder/decoder in the AWG configuration for code-based
photonic routers Part I: modelling and design, J. Lightwave Technol., vol. 24, pp. 103-112, 2006.
[5] N. Wada et al.: Characterization of a full encoder/decoder in the AWG configuration for code-based
photonic routers-Part II: experimental results, J. Lightwave Technol., vol. 24, pp. 113-121, 2006.
[6] http://optilux.sourceforge.net/
[7] X. Chen et al.: Coherent detection using optical time-domain sampling, Photon. Technol. Lett., vol. 21,
pp. 286-288, 2009.
[8] J. Tang et al.: Transmission performance of a 400 Gbits-1 all-optical orthogonal frequency division
multiplexing system, J. of Optics, vol. 15, 055401, 2013.
[9] J. L. Zyskind, J. A. Nagel, H. D. Kidorf: Erbium-doped fiber amplifiers for optical communications,
in Optical Fiber Communications IIIB, I. P. Kaminov and T. L. Koch Ed., San Diego, CA, Academic Press,
pp. 13-68, 1997.
[10] R. Schmogrow et al.: Error vector magnitude as a performance measure for advanced modulation formats,
Photon. Technol. Lett., vol. 24, pp. 61-63, 2012.
[11] H. A. Mahmoud and H. Arslan: Error vector magnitude to SNR conversion for nondata-aided receivers,
Trans. on Wireless Comm., vol. 8, pp. 2694-2704, 2009.