40 Gb/s CAP32 short reach transmission over 80
km single mode fiber
Yuliang Gao,1,2,* Qunbi Zhuge,2 Wei Wang,2 Xian Xu,2 Jonathan M. Buset,2 Meng Qiu,2
Mohamed Morsy-Osman,2 Mathieu Chagnon,2 Feng Li,3 Liang Wang,3 Chao Lu,3 Alan
Pak Tao Lau,1 and David V. Plant2
1
Photonics Research Center, Department of Electrical Engineering, The Hong Kong Polytechnic University, Hong
Kong, China
2
Department of Electrical and Computer Engineering, McGill University, Montreal, QC, H3A 2A7, Canada
3
Photonics Research Center, Department of Electronic and Information Engineering, The Hong Kong Polytechnic
University, Hong Kong, China
*
[email protected]
Abstract: We present a method to mitigate the chromatic dispersion (CD)induced power fading effect (PFE) in high-speed and short-reach carrierless amplitude and phase (CAP) systems using the degenerate four-wave
mixing (DFWM) effect and a decision feedback equalizer (DFE).
Theoretical and numerical investigations reveal that DFWM components
produced by the interaction between the main carrier and the signal
sideband help to mitigate PFE in direct detection systems. By optimizing
the launch power, a maximum reach of 60 km in single mode fiber (SMF-e
+ ) at 1530nm is experimentally demonstrated for a 40 Gbit/s CAP32
system. In addition, we study the performance of a decision feedback
equalizer (DFE) and a traditional linear equalizer (LE) in a channel with
non-flat in-band frequency response. The superior PFE tolerance of DFE is
experimentally validated, and thereby, the maximum reach is extended to
80 km. To the best of our knowledge, this is the twice the longest
transmission distance reported so far for a single-carrier 40 Gbit/s CAP
system around 1550 nm.
©2015 Optical Society of America
OCIS codes: (060.2330) Fiber optics communications; (060.4080) Modulation.
References and links
1.
2.
3.
4.
5.
6.
7.
8.
9.
J. L. Wei, J. D. Ingham, D. G. Cunningham, R. V. Penty, and I. H. White, “Performance and power dissipation
comparisons between 28 Gb/s NRZ, PAM, CAP and optical OFDM systems for data communication
applications,” J. Lightwave Technol. 30(20), 3273–3280 (2012).
A. Ghiasi and B. Welch, IEEE 802.3bm Fiber Optic Task Force, Sept. 2012.
T. Gui, C. Li, Q. Yang, X. Xiao, L. Meng, C. Li, X. Yi, C. Jin, and Z. Li, “Auto bias control technique for
optical OFDM transmitter with bias dithering,” Opt. Express 21(5), 5833–5841 (2013).
Y. Bao, Z. Li, J. Li, X. Feng, B. O. Guan, and G. Li, “Nonlinearity mitigation for high-speed optical OFDM
transmitters using digital pre-distortion,” Opt. Express 21(6), 7354–7361 (2013).
J. D. Ingham, R. V. Penty, and I. H. White, “40 Gb/s carrierless amplitude and phase modulation for low-cost
optical data communication links,” in Proc. Opt. Fiber Commun. (OFC), Los Angeles, CA, Mar. 2011, Paper
OThZ3.
L. Tao, Y. Wang, Y. Gao, A. P. T. Lau, N. Chi, and C. Lu, “Experimental demonstration of 10 Gb/s multi-level
carrier-less amplitude and phase modulation for short range optical communication systems,” Opt. Express
21(5), 6459–6465 (2013).
R. Rodes, M. Wieckowski, T. T. Pham, J. B. Jensen, J. Turkiewicz, J. Siuzdak, and I. T. Monroy, “Carrierless
amplitude phase modulation of VCSEL with 4 bit/s/Hz spectral efficiency for use in WDM-PON,” Opt. Express
19(27), 26551–26556 (2011).
L. Tao, Y. Ji, J. Liu, A. P. T. Lau, N. Chi, and C. Lu, “Advanced modulation formats for short reach optical
communication systems,” IEEE Netw. 27(6), 6–13 (2013).
L. Tao, Y. Wang, Y. Gao, A. P. T. Lau, N. Chi, and C. Lu, “40 Gb/s CAP32 system with DD-LMS equalizer for
short reach optical transmissions,” IEEE Photon. Technol. Lett. 25(23), 2346–2349 (2013).
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11412
10. J. Zhang, J. Yu, F. Li, N. Chi, Z. Dong, and X. Li, “11 × 5 × 9.3Gb/s WDM-CAP-PON based on optical singleside band multi-level multi-band carrier-less amplitude and phase modulation with direct detection,” Opt.
Express 21(16), 18842–18848 (2013).
11. M. I. Olmedo, T. Zuo, J. B. Jensen, Q. Zhong, X. Xu, and I. T. Monroy, “Towards 400GBASE 4-lane solution
using direct detection of multiCAP signal in 14 GHz bandwidth per lane,” in Proc. Opt. Fiber Commun. (OFC),
Anaheim, CA, Mar. 2013, Paper PDP5C.10.
12. J. Wei, Q. Cheng, D. G. Cunningham, R. V. Penty, and I. H. White, “100-Gb/s hybrid multiband CAP/QAM
signal transmission over a single wavelength,” Opt. Express 33(2), 415–423 (2015).
13. H. Kim, “EML-based optical single sideband transmitter,” IEEE Photon. Technol. Lett. 20(4), 243–245 (2008).
14. A. S. Karar, J. C. Cartledge, J. Harley, and K. Roberts, “Electronic pre-compensation for a 10.7 Gb/s system
employing a directly modulated laser,” J. Lightwave Technol. 29(13), 2069–2076 (2011).
15. J. L. Wei, J. D. Ingham, Q. Cheng, D. G. Cunningham, R. V. Penty, and I. H. White, “Experimental
demonstration of optical data links using a hybrid CAP/QAM modulation scheme,” Opt. Lett. 39(6), 1402–1405
(2014).
16. R. Zlatanovici, A. Manolescu, L. Kabulepa, and M. Glesner, “Decision feedback equalizers for carrierless
amplitude/phase modulation receivers,” in Proc. Semicond. Conf., Sinaia, Romania, 1, 127–130 (1999).
17. ITU, Rec. G.975.1: Forward error correction for high bit-rate DWDM submarine systems (2004).
18. D. J. F. Barros and J. M. Kahn, “Comparison of orthogonal frequency-division multiplexing and on-off keying in
amplified direct-detection single-mode fiber systems,” J. Lightwave Technol. 28(12), 1811–1820 (2010).
19. K. Inoue, “Phase-mismatching characteristic of four-wave mixing in fiber lines with multistage optical
amplifiers,” Opt. Lett. 17(11), 801–803 (1992).
20. G. P. Arawal, Nonlinear Fiber Optics (Elsevier, 2013), 5th ed.
21. J. He, A. Li, X. Chen, and W. Shieh, “Fiber nonlinearity impact on direct-detection block-wise phase switched
systems,” J. Lightwave Technol. 32(10), 1872–1876 (2014).
22. J. M. Buset, Z. A. El-Sahn, and D. V. Plant, “Experimental demonstration of a 10 Gb/s subcarrier multiplexed
WDM PON,” IEEE Photon. Technol. Lett. 25(15), 1435–1438 (2013).
23. K. Kikuchi, “Clock recovering characteristics of adaptive finite-impulse-response filters in digital coherent
optical receivers,” Opt. Express 19(6), 5611–5619 (2011).
1. Introduction
The capacity of access networks must evolve as the proliferation of more bandwidth intensive
services has been provided such as streaming Internet video, video on demand and cloudbased storage and computing. With accessibility to optical phase and polarization, coherent
transceivers supporting 100 Gb/s per channel are widely available for commercial
deployment. However, it is not likely for coherent techniques to be used in short reach
communications as the IQ modulator, optical hybrid and local oscillator not only increase cost
but also footprint and, as a result, impede system integration. Instead, intensity modulationdirect detection (IM-DD) techniques are widely favored for the short reach scenario.
Numerous such schemes have been investigated including pulse amplitude modulation [1,2],
direct detection orthogonal frequency-division multiplexing/discrete multi-tone modulation
[3,4] and carrier-less amplitude/phase modulation (CAP) [5–7].
Among these techniques, CAP modulation is an attractive candidate since it doesn’t
require a complex mixer and a radio frequency (RF) source for down conversion [8]. Neither
does it need the discrete Fourier transform (DFT) therefore the use of high speed digital-toanalog converter (DAC) can be avoided in commercial products. Recently, many CAP
systems with large capacities have been reported in the literature, such as a CAP16 system
achieving bit rates up to 40 Gb/s [5]. However, single-band CAP systems have stringent
requirements for flat channel frequency responses [9]. Multiband CAP technique (MultiCAP)
is proposed to mitigate the impact of channel responses by carefully manipulating the
modulation formats and power loading on each sub-band [10–12]. However, considerable
power penalties can still be observed for the sub-bands near the feeding dips when the
transmission distance is longer than 40km [10]. Thus, the CD induced non-flat in-band
frequency response, also called the power fading effect (PFE), is considered as a critical
limiting factor to system capacity and maximum reach. On the other hand, single-side band
(SSB) techniques have been introduced to avoid the cut-off effect [10]. However, to generate
an optical SSB signal requires an additional sharp optical filter precisely aligned with half of
the signal sideband. Otherwise, an IQ modulator or specially designed transmitter is required
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11413
to perform the Hilbert transform [13]. Another alternative is pre-compensating CD at the
transmitter (Tx) [14] which requires high speed DACs where simple transmitter architecture
of CAP systems such as analogue implementation cannot be adopted [15]. On the other hand,
we have presented various receiver digital equalization techniques to compensate for PFE [6,
8, 9]. However, one drawback of traditional transversal receiver equalizers is that they
introduce extra noise to the recovered signal also known as noise enhancement [16].
Furthermore, the maximum reach for previous 40 Gbit/s CAP systems is still not more than
40 km even with the aid of digital signal equalization techniques.
In this paper, we first theoretically and numerically investigate the PFE in CAP systems.
Then we propose the use of degenerate four-wave mixing (DFWM) to mitigate PFE. In
particular, we find that by using properly adjusted launched powers the DFWM in the fiber
link can effectively reduce the phase shift introduced by CD and thus alleviate fading effects.
Experiment results show that at the optimum launch power in our system produces a 60 km
reach over single mode fiber (SMF) at the 7% FEC threshold of 3.8 × 10−3 [17]. In addition,
we evaluate the performance of the decision feedback equalizer (DFE) in such a frequencyselective optical channel. It is shown that the DFE equalizer provides superior tolerance to the
CD-induced PFE compared with a linear equalizer (LE), and thereby, the maximum reach is
further extended to 80 km which is twice the longest reach for a single carrier 40 Gbit/s CAP
system reported to date at the wavelength of approximately 1530 nm.
2. Operating principle
2.1 Carrier-less amplitude and phase modulation
CAP signal can be generated using two orthogonal shaping filters gI(t) and gQ(t). The impulse
responses of the two filters form a Hilbert pair, expressed as [6]
g I (t ) = f (t ) ⋅ sin(2π f c t )
(1)
gQ (t ) = f (t ) ⋅ cos(2π f c t ),
(2)
where f(t) is the baseband square-root raised-cosine shaping filter, the roll-off factor is set to
be 0.1in this paper, fc is the carrier frequency. The in-phase and quadrature symbols an and bn
are first sent into gI(t) and gQ(t), respectively. The two filtered signals are then added together
to generate the electrical CAP signal as shown in Fig. 1. The transmitted signal can be
expressed as
s (t ) =
∞
 a g
n =−∞
n
I
(t − nT ) − bn gQ (t − nT ) ,
(3)
where an and bn denotes the in-phase and quadrature components, n is the symbol index, and
T is the symbol period. At the receiver side, the signal r(t) after the photodetector (PD) is first
passed through the two matched filters which are the time reverse of gI(t) and gQ(t) to separate
the in-phase and quadrature components
rI (t ) = r (t ) ∗ g I (−t )
(4)
rQ (t ) = r (t ) ∗ gQ (−t ).
(5)
Then, an adaptive filter updated by least mean square algorithm is used to equalize
channel distortion before the final data decision.
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11414
Fig. 1. Schematic structure of CAP modulation and demodulation.
2.2 Chromatic dispersion induced power fading and degenerate four-wave mixing
After transmission, each frequency of the signal has experienced a different phase shift. This
results in a PFE after direct detection. To evaluate the CD-induced PFE, the fiber is modeled
as an all-pass filter with a transfer function given by
H(f)=e
− jπ D
λ2
c
Lf 2
(6)
,
where f is the frequency offset from the optical carrier, D is the chromatic dispersion, λ is the
optical wavelength, L is the length of the fiber and c is the speed of light in vacuum. The
optical signal at the output of the fiber, E(t), is then given by
E (t ) =
A + s (t ) ∗ h(t ),
(7)
where A represents the DC bias and h(t) is the inverse Fourier transform of H(f). At the PD,
the optical signal is captured by the square-law detection
r (t ) =
2
A + s (t ) ∗ h(t ) .
(8)
The signal can be expanded into Taylor series on the square-root term. The detected signal
can be expressed as
*

 



s (t ) s 2 (t )
s (t ) s 2 (t )
− 3/ 2 + ...  ∗ h(t )  ⋅  A +
− 3/ 2 + ...  ∗ h(t )  . (9)
r (t ) =  A +
A
A
8
8
2 A
2 A



 

After expanding Eq. (9), the detected electrical signal can be expressed as
s (t ) ⊗ h(t ) s *(t ) ⊗ h *(t ) s 2 (t ) ⊗ h(t ) ( s *(t ) ) ⊗ h *(t )
+
−
−
+ ..., (10)
2
2
8A
8A
2
r (t ) = A +
where the operation ⊗ stands for convolution and the superscript * is conjugation. In Eq.
(10), the first term is the DC component. The second and third terms are the linear outputs of
interest obtained by convoluting signal s(t) with the channel response h(t). The rest are
regarded as the noise induced by the interaction of the detector nonlinearity and fiber CD.
Note that these noise terms can be considered sufficiently small when bias level A is set to be
much larger than signal s(t) [18]. Then, the linear terms can be further simplified since the
CAP signal is real valued, (i.e., s(t) = s*(t)). Thus, Eq. (10) can be simplified as
r (t ) ≈ A + s(t ) ∗
h(t ) + h* (t )
= A + s (t ) ∗ heq (t ),
2
(11)
where the equivalent transfer function Heq(f) is
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11415
jπ D
λ2
Lf 2
− jπ D
λ2
Lf 2

λ2 2 
Lf  .
(12)
= cos  π D
c


To validate Eq. (12), a simulation has been conducted on a 40 Gbit/s CAP32 system. Fiber
nonlinearity is first turned off to focus on the CD impact. The RF spectral density of r(t) is
shown in Fig. 2 for different transmission distances. The dashed lines display the squared
equivalent transfer function |Heq(f)|2 obtained in Eq. (12). The solid curves represent the
simulated RF spectrum with various transmission distances. Note that all the simulation
results in this section are obtained using data sets of 216 symbols. One can see that the high
frequency components start to experience a cut-off effect as the distance is increased above
40km. When the distance is 60 km and 80 km, power dips can be observed at 8.4 GHz and
7.3 GHz respectively which agree well with the analytical transfer function Heq(f).
H eq ( f ) =
e
c
+e
2
c
Fig. 2. RF spectrum of received 40 Gbit/s CAP32 signals with different transmission distances.
Solid lines are the RF spectrum. Dashed lines represent the analytical transfer function
obtained in Eq. (12). L denotes the fiber length.
To study the DFWM, the nonlinear Schrodinger equation in the absence of polarization
effects is given by [19,20]
i
∂E ( z , t ) α
1 ∂2 E
2
+ i E − β 2 2 + γ E E = 0,
∂z
2
2
∂t
(13)
where E(z,t) denotes the optical field propagating in the fiber, γ is the fiber nonlinearity
coefficient, z is the transmission distance, t is time, α is the coefficient for fiber loss and β 2
is the dispersion of group velocity.
Assuming that three random frequency components are located at f p , f q , f r , , then the
FWM component produced by these frequencies will be located at the frequency of
f g = f p + f q − f r , and the generated FWM product can be described as [21]
EFWM , g = iγ E p Eq Er*
exp ( −α + iΔβ ) z  − 1
i Δβ − α
,
(14)
where Ep, Eq and Er are the optical fields of any three random frequency components in the
system, EFWM,g represents the optical field of the FWM product at frequency f g .
Δβ = β p + β q − β r − β g denotes the phase mismatching term, where β i (i = p, q, r , g ) is the
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11416
propagation constant for the carrier at frequencies f p , f q , f r , f g . In the CAP system, the DC
component is usually much larger than other signal components and thus the DC involved
degenerated FWM product contributes the majority of the entire nonlinear effect.
In this case, the most significant DFWM products are caused by the interaction of the DC
component, which means Ep and Eq both equal to the optical carrier, and one signal frequency
component Er in the signal sideband. The degenerated FWM component that is located on the
frequency f can be expressed as
EFWM ( f ) = iγ E 2 (0) E * (− f )
exp ( −α + iΔβ ) z  − 1
i Δβ − α
,
(15)
where EFWM ( f ) is the FWM product generated by the signal component at frequency –f. For
an intensity modulated signal, the spectrum on one sideband is the Hermitian conjugate of the
spectrum component on the other side (e.g., E ( f ) = E * (− f ) ). Thus, the signal with DFWM
can be written as
Etotal ( f ) = E ( f ) + EFWM ( f )

exp ( −α + iΔβ ) z  − 1 
(16)
= 1 + iγ E 2 (0)
 ⋅ E ( f ).
iΔβ − α


Therefore, each frequency experiences both phase and amplitude changes depending on
the Δβ value of its corresponding EFWM ( f ) . This term introduces a nonlinear phase rotation
to each frequency component that is opposite to the phase shift induced by CD. To illustrate
the interaction between CD and FWM, the optical phase shift Δϕ is obtained by Eq. (17)
Δϕ ( f ) = arg ( Eout ( f ) Ein ( f ) ) ,
(17)
where Ein ( f ) and Eout ( f ) are the optical fields at frequency f before and after 80 km SMF
transmission. Figure 3 shows the simulation results of the optical spectrum and phase shift
Δϕ ( f ) at various launch powers. Different from Fig. 2, both the CD and fiber nonlinearities
are considered. At frequency of 8.5 GHz, 1.55, 1.67, 2.00 and 2.12 rad phase shifts are
observed respectively for 10, 12, 14 and 15dBm launch power cases. The launch power in the
linear case is 0 dBm. For the 10dBm launch power case, the phase shift is almost identical to
the linear transmission where only CD is considered. However, the phase rotation is reduced
by 0.57 rad for the 15 dBm launch power case. Also, the π/2 phase shift point (i.e., the dip
frequency in RF spectrum) is significantly increased by 1.15GHz indicating effective
mitigation of the PFE. As can be noticed, the power density function is quite flat over the
whole optical spectrum. Thus, the phase shift introduced by self-phase modulation and crossphase modulation effects should be identical for different frequency components. Since we
only care about the relative phase-shift differences for different frequency components and
the minimum phase shift will be set to zero for comparison. Thus, the phase shift in Fig. 3 is
only contributed by DFWM.
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11417
Fig. 3. Optical power spectrum at Tx and phase shift after 80 km SMF transmission for a 40
Gbit/s CAP32 signal with various launch powers. Fiber nonlinearity effects are turned off in
the linear transmission case. The LP is launch power.
Figure 4 shows the simulation results of the RF spectrum of the CAP32 signal after direct
detection with various launch powers. It can be seen that the CD-induced power fading effect
is mitigated with increased launch power. The power dip is increased from 7.49 GHz to 8.18
GHz for the launch power of 0 dBm and 9 dBm respectively. When the launch power is
increased to 12 dBm, no power dip is observed in the RF spectrum anymore. Thus, it can be
concluded that the degenerate FWM is effective for the CD-induced power fading mitigation
in IM-DD systems. Here, the DFWM induced phase shift is obtained in a single-carrier
system. If wavelength-multiplexing CAP (WDM-CAP) technique is incorporated, multiple
power peaks should be observed in the optical spectrum as shown in [10]. With such multiple
wavelengths carriers, inter-channel FWMs will not accumulate because the phases of
different carriers are uncorrelated and the CD-induced walk-off between different WDM
channels will destroy the phase-matching condition. Thus, for a WDM-CAP system, only
intra-channel FWM should be considered and similar phenomenon should be expected.
Fig. 4. RF spectrum of received 40 Gbit/s CAP32 signals with different launch powers.
3. Experimental results
Figure 5 shows the experiment setup of the CAP32 system. At the transmitter side, the bit
sequence is mapped onto in-phase and quadrature components an and bn, respectively. Then,
the coded sequence is up-sampled by a factor of four in order to match the operating rate of
the shaping filters that follows. After in-phase/quadrature component separation, the two
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11418
signals are sent into two shaping filters. The filter center frequency f c is given by
(1 + α ) ⋅ B / 2 + Δf . Here, the roll-off factor α is 0.1, B is symbol rate and Δf is the 100
MHz frequency offset. Then, the signal is pre-emphasized to mitigate the DAC analog
frequency response using the method in [22]. Two high speed Micram DACs with 6 bit
resolution together with two field-programmable gate array (FPGA) boards are used to
generate the 8 Gbaud CAP32 RF signal at a sampling rate of 32 GSa/s. It should be noted that
this is only a proof of concept experiment and the high-speed DAC can be replaced by two
electrical filters in the future commercially available products. The output power from laser is
12 dBm. The output amplitude after the electrical amplifier is 400 mV peak-to-peak and is
used to drive an intensity modulator (IM) whose Vπ and bandwidth are 5V and 33GHz
respectively. Since external modulator is incorporated for the data loading, phase or
frequency chirp effect will be mitigated in the laser output to focus on the interaction between
CD and DFWM effect. Contrarily, if a directed modulated laser is adopted for data
modulation, a frequency chirp will be introduced to the output signal. The dominant linear
part of the chirp is equivalent to a CD induced chirp term [20] which will not affect the PFE
compensation from DFWM. The residual small nonlinear chirp will slightly degrade the PFE
compensation due to the unbalanced phase distortion on the sidebands on different sides of
the central frequency. However, the DFWM effect is still helpful in compensating the CD
induced PFE. The modulated signal is then boosted to compensate for the insertion loss of the
PC and the IM. Then, a variable optical attenuator (VOA) is used to adjust the launch power
from 0 dBm to 15dBm. After the fiber link, another VOA is used to keep the received power
to be constant even the launch power has been changed. The signal is then detected by a wide
bandwidth 10 GHz PD with a sensitivity of −25dBm. In the end, 1.6 × 105 CAP32 symbols
are sampled by an oscilloscope at a sampling rate of 80 GSa/s for offline processing.
Fig. 5. Schematic of the 40 Gbit/s CAP32 experiment setup. DAC: digital-to-analog converter;
VOA: variable optical attenuator; SMF: single mode fiber; T-T BPF: tunable bandwidth and
tunable central wavelength bandpass filter; PD: photo detector; Rx: receiver; BER: bit error
rate.
For the offline processing, the captured signal is first resampled to 4 samples per symbol.
Then, two time-reversed of matched filters gI(t) and gQ(t) filters comparing with the
transmitter are used to separate the in-phase and quadrature signals. These signals are
subsequently down-sampled to 2 samples per symbol and processed by the digital equalizer to
compensate for the channel distortions. In such a frequency-selective channel, decision
feedback equalizer (DFE) can be used for the data equalization to alleviate the noise
enhancement effect [16]. A transversal LE is also studied for comparison. To achieve a fair
comparison, both equalizers have the same tap length, step size and are both updated with
least mean square algorithm. Figure 6 shows the block diagram of DFE
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11419
Fig. 6. Schematic diagram of decision feedback equalizer. For linear equalization, only the
forward equalizer C(z) is used while both forward equalizer C(z) and feedback equalizer F(z)
are enabled for decision feedback equalization. y(n): input signal; C(z): transfer function of
forward equalizer; F(z): transfer function of feedback equalizer; d(n): transmitted symbol;
dˆ (n) : data decision; e(n): error signal.
where y(n) is the input signal, d(n) is the transmitted symbol, dˆ (n) is the data decision and
e(n) is the error signal. The feedforward and feedback equalizers are denoted as C(z) and F(z)
whose tap spaces are Ts / 2 and Ts , respectively. The input is first equalized by a 30-tap
forward equalizer C(z). Since the channel response is not flat as shown in Fig. 2, the linear
channel equalizer will over-amplify the noise at the fading dips which is also known as the
noise enhancement effect [16]. To mitigate its impact, another 5-tap feedback equalizer F(z)
can be enabled for estimating the enhanced noise n(n) by looking into the difference between
C(z)’s output and data decisions. The final output of the DFE is obtained by subtracting the
equalized signal of the forward equalizer from the equalized output of the feedback equalizer.
In commercialized realization of our scheme, phase-lock-loop or QAM receivers [12] should
be adopted for clock-jitter compensation. However, the timing jitter is sufficiently small in
our experiment so we rely on the DFE to compensate for the timing jitter [23].
First, bit error rates (BERs) versus different launch powers for 40 km, 60 km and 80 km
transmission links have been studied. Figure 7(a) shows the 40 km transmission result for LE
and DFE. In order to maintain a fixed sensitivity, the received power for all points displayed
are kept at −10dBm using the VOA. The optimum launch power in Fig. 7(a) is 6 dBm for
both LE and DFE as the fading effect is not severe enough while the dominant distortion
comes from the fiber nonlinearity. The DFE increases the BER from 10−4 at 6 dBm to 8 ×
10−4 at 15 dBm. For the LE, the BER increases from 2.4 × 10−4 at 6dBm to 1.7 × 10−3 at
15dBm. Figure 7(b) experimentally demonstrates that the power fading effect has been
mitigated and the high frequency components are elevated, as predicted by the simulations in
Fig. 4.
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11420
Fig. 7. (a) Launch power versus BER (b) RF power spectrum after 40 km SMF transmission.
LE: linear equalizer; DFE: decision feedback equalizer.
In the 60 km transmission, larger launched powers are required to mitigate the PFE and
the optimum launch power is increased to 9 dBm for DFE and 12 dBm for LE as can be seen
in Fig. 8(a). The 3 dB difference is because DFE has a superior tolerance to power fading.
The received power is kept at −10dBm for all the test cases. At high launch powers, we note
that the performance difference between the two techniques becomes smaller. This is because
the high frequencies are over amplified by DFWM, and therefore nonlinear noise becomes
dominant. Thus, a balance has to be struck between PFE and fiber nonlinearities. The
optimum BER obtained using DFE is 2 × 10−4 at 9 dBm launch power while the optimal BER
for LE is 7.8 × 10−4 at 12 dBm. We also observe that the high frequency components are
enhanced with increased launch power in Fig. 8(b).
Fig. 8. (a) Launch power versus BER (b) RF power spectrum after 60 km SMF transmission.
LE: linear equalizer; DFE: decision feedback equalizer.
Figure 9(a) shows the 80 km transmission result for LE and DFE. The optimum power is
increased to 12 dBm and 14 dBm for DFE and LE respectively because a larger FWM effect
is required to mitigate the severe PFE existing at this reach. The optimum BER for DFE is
10−3 at a launch power of 12 dBm while the optimum BER for LE is 4.5 × 10−3 at 14 dBm.
Although the optimum launch power of 12 dBm is much larger compared to long-haul
transmission cases, it is still an acceptable value in short reach communication systems [10].
The received power is set to be −10dBm. From Fig. 9(b), we note that the fading dip is more
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11421
severe compared to the 40 and 60 km cases and thus requires a higher launch power beyond
14 dBm. However, this introduces too much nonlinear noise and even with the optimum
launch power of 14 dBm, LE still cannot achieve a 7% FEC threshold of 3.8 × 10−3.
Fig. 9. (a) Launch power versus BER (b) RF power spectrum after 80 km SMF transmission.
LE: linear equalizer; DFE: decision feedback equalizer.
Figure 10 shows the received power versus BER for (a) 40 km, (b) 60 km and (c) 80 km
fiber links. For the 40 km link, the optimum launch power of 6 dBm is chosen and the
required received power at BER of 3.8 × 10−3 are −15.22 dBm for DFE and −14.75 dBm for
LE with a power penalty of 0.47 dBm. For the 60 km transmission, the launch power is 9
dBm to mitigate PFE and the required received power are −13.41 dBm and −12.82 dBm for
DFE and LE, respectively, resulting in a 0.59 dBm power penalty. For the 80 km
transmission, the optimum launch power of is set to be 12 dBm and the required received
power for DFE at BER threshold of 3.8 × 10−3 is −11.83 dBm. However, LE cannot achieve a
FEC threshold of at the maximum received power of −10 dBm. As a conclusion, using the
DFWM and DFE to mitigate the CD-induced PFE we successfully transmit a 40 Gb/s CAP32
signal over 80 km distance with a received power of −11.83 dBm at the BER threshold of 3.8
× 10−3.
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11422
Fig. 10. BER versus the received signal power for (a) 40 km, (b) 60 km and (c) 80 km SMF
transmissions. LE: linear equalizer; DFE: decision feedback equalizer.
4. Conclusion
We studied the chromatic dispersion CD-induced PFE in short reach IM-DD CAP systems.
DFWM is found to be beneficial to mitigate this effect. The inter-relationship between power
fading and DFWM is theoretically studied and verified through simulation. Also, we have
experimentally demonstrated a 40 Gbit/s CAP32 transmission over 40 km, 60 km and 80 km
SMF fiber links. To recover the distorted signal, both traditional LE and DFE are fully
investigated and their robustness with respect to CD-induced PFE is experimentally verified.
With optimized launch power, traditional LE can achieve a BER threshold of 3.8 × 10−3 for
40km and 60km transmission at received powers of −14.75 dBm and −12.82 dBm,
respectively. For DFE, 40 km, 60 km and 80 km transmission can be achieved with received
powers of −15.22, −12.82 and −11.83 dBm.
Acknowledgments
The authors would like to acknowledge the support of the Hong Kong Government General
Research Fund (GRF) under project number PolyU 152079/14E.
#228929 - $15.00 USD
©2015 OSA
Received 3 Dec 2014; revised 13 Apr 2015; accepted 16 Apr 2015; published 23 Apr 2015
4 May 2015 | Vol. 23, No. 9 | DOI:10.1364/OE.23.011412 | OPTICS EXPRESS 11423