view paper - OECC/ACOFT 2014

1-Tb/s Stokes Vector Direct Detection over 480-km SSMF
Di Che1,2, Qian Hu2, Xi Chen2, An Li2, and William Shieh2
National ICT Australia – Victoria Research Laboratory (NICTA-VRL), VIC 3010, Australia
Dept. of Electrical and Electronic Engineering, The University of Melbourne, VIC 3010, Australia
Abstract Summary
We demonstrate the first 1-Tb/s (10×100Gb/s) optical
direct detection over 480-km SSMF transmission using
the Stokes vector direct detection (SV-DD) with 7.76bit/s/Hz electrical spectrum efficiency.
With the popularization of personal Internet devices and
the development of medium-rich applications, people are
now thirsty for more high-performance multimedia
experience at their fingertip. As the 100-Gb/s Ethernet
has been standardized in 2010 [ 1 ] and increasingly
become a commercial reality, the next urgent issue is the
migration path toward 1-Tb/s Ethernet transport. Highspeed long-haul transport has achieved the Terabit
transmission primarily thanks to the coherent detection
[2-3]. For the cost sensitive short-reach networks such as
data center interconnects within distance of hundreds of
kilometers, the optical direct detection (DD) technology
[ 4 - 9 ] has long been regarded as one of the suitable
solution, because they can significantly lower the
expense compared with coherent counterpart while
achieving both high data rate and moderate reach.
However, the conventional intensity modulation direct
detection (IM-DD) systems cannot undertake the 100
Gb/s to Tb/s transmission due to the two fundamental
bottlenecks: (i) chromatic dispersion (CD) induced
signal fading due to the lack of phase diversity [4],
which limits the transmission distance less than 20 km;
and (ii) 2nd order nonlinearity due to photo-detection,
which limits the system capacity. The first one can be
overcame by using the single sideband modulation (SSB)
[5] while the second by leaving frequency gaps in
spectrum to separate the signal and 2nd nonlinear
component [5-6]. Both approaches reduce the electrical
spectrum efficiency (SE) by half, resulting in the total
SE only one fourth of single polarization (POL) coherent
detection. By applying the double sideband modulation
(DSB) and standard balance receiver [2] to the DD
system, the signal carrier interleaved DD (SCI-DD) [8]
overcomes the above problems and achieves the first
100G single-wavelength single-polarization transmission,
with the sacrifice of one third spectral efficiency.
To further increase the SE, the Stokes vector DD
(SV-DD) was proposed recently [9]. The Stokes vector
receiver offers the following advantages: (i) achieving 3dimension detection with phase diversity; (ii) eliminating
the 2nd order noise by the balance photo detectors (PD);
(iii) realizing polarization independence by conducting
the polarization tracking using the digital signal
processing (DSP). SV-DD achieves 100% SE with single
polarization modulation while reduces the cost by: (i)
simpler transmitter design compared with polarization
multiplexing (POL-MUX) systems [6]; (ii) no need of
local oscillator (LO) at receiver; (iii) simpler DSP due to
using less number of FFT operations [7], and no need to
track laser frequency offset and phase noise [2]. These
merits reveal the potential for SV-DD to be deployed in
Terabit short-reach networks. The 160-Gb/s SV-DD
transmission of 160-km standard single mode fiber
(SSMF) has successfully been demonstrated [9] using
single wavelength. In this paper, we will demonstrate the
first Terabit wavelength-division multiplexed (WDM)
optical direct detection with transmission over 480-km
SSMF. This work represents the record reach for singlepolarization modulated 100 Gb/s per wavelength allelectronic
amplification and without electronic pre-compensation
or optical compensation.
Principle of Stokes Vector Direct Detection
As the name suggests, the Stokes vector (SV) receiver
detects the 3 (or 4) components of a Stokes vector. It
consists of one 90o optical hybrid and three balance PDs
as shown in Fig. 1. Given the 2-D complex signal in
Jones space:
, the outputs of the receiver
correspond to the three components of the SV:
 X Y
 S1   X  X  Y  Y
S   S2    X  Y *  X *  Y    2 Re( X  Y * )  (1)
 S3   j ( X  Y *  X *  Y )   2 Im( X  Y * ) 
 
where Re() and Im() represent for the real part and
imaginary part of a complex number.
3(Y) Optical
Fig. 1 Structure of an SV receiver. PBS: Polarization beam
splitter; B-PD: Balance photo-detector.
In order to recover the signal, the receiver needs to
acquire the polarization rotation after the fiber
transmission, which is represented by the 3x3 rotation
matrix (RM) in the Stokes space. We use the training
symbol aided estimation to get the RM. One training
period contains 3 orthogonal training symbols with the
Jones space representation of (0, 1), (1, 1) and (i, 1),
corresponding to the SVs of (-1, 0, 0), (0, 1, 0) and (0, 0,
1). Three columns of the RM are acquired respectively.
We multiply the received SV with the inverse of the RM:
1 
 X T 2  YT 2 
 r11 r12 r13   X R  YR 
 2 Re(XT  YT )    r21 r22 r23    2 Re(X R  YR )  (2)
 
* 
* 
 2 Im(XT  YT )   r31 r32 r33   2 Im(X R  YR ) 
where rij is the matrix element of the ith row and jth
column of the rotation matrix RM, the subscripts “T”
and “R” represent for the transmitter and receiver
respectively and the superscript “-1” represents for the
inverse. To guarantee the channel is linear, the Y-POL is
sent with a constant power, namely carrier (C) while
only X-POL is modulated with the signal (S). Therefore,
by combining the 2nd and 3rd component of the SV, we
have the final output of
Experimental Setup for 1-Tb/s Transmission
In this work, we experimentally demonstrate the 1-Tb/s
(10 × 100 Gb/s) direct detection with 480-km SSMF
transmission using SV-DD. Fig. 2 illustrates the
experimental setup. At the transmitter, 10 CW lasers are
multiplexed by the coupler with the channel spacing of
50 GHz. The combined light is first split into two
branches respectively for the signal and carrier. For the
signal branch, the optical signal is fed into a 3-tone
generator [9] to achieve wider optical bandwidth, then
fed into an I/Q modulator driven by an arbitrary
waveform generator (AWG). The RF OFDM signal with
a 16-QAM modulation is loaded into the AWG. The FFT
size of the OFDM signal is 4096 in which 3420
subcarriers are filled. The cyclic prefix (CP) is 128
points. SV rotation training symbols are added before
each OFDM frame with symbol length of 192 points.
The AWG operates at a sampling rate of 10 GSa/s,
leading to the optical bandwidth of 8.33 GHz for 1-band
and total optical bandwidth 25 GHz for 3-band. The raw
data rate is 25×4=100 Gb/s for one channel and 1 Tb/s
for 10 channels. Counting the OFDM overhead, the data
rate is decreased to 969.7 Gb/s before 20 % FEC and
808.1 Gb/s after FEC. This corresponds to pre-FEC
electrical spectral efficiency (SE) of 7.76 bits/s/Hz and
post-FEC SE of 6.47 bits/s/Hz. The lower branch is a
delay line whose fiber length is matched with the upper
line to cancel the phase noise between the signal and the
carrier. Signal carrier power ratio (CSPR) is maintained
to be 0 dB. These two branches are combined with a
polarization beam combiner (PBC) and then launched
into a recirculation loop which consists of two spans of
80-km SSMF whose loss is compensated by the EDFAs.
The inset (i) of Fig. 2 shows the spectrum at the
transmitter captured by the optical spectrum analyzer
(OSA). Each channel has the baud rate of 25 Gbaud/s,
corresponding to 0.2 nm in terms of wavelength; the
channel spacing is 0.4 nm (or 50 GHz). Totally 4 nm
bandwidth is occupied. At reception, the light is fed into
a band pass filter to filter out one channel each time
which occupies a 40 GHz bandwidth and carries 100
Gb/s data. The inset (iii) in Fig. 2 shows the spectrum
after the filter. The spectrum presents a double sideband
modulation (DSB), and the carrier is shown by the power
peak at the center. The optical signal is then spilt using a
polarization beam splitter (PBS). Polarizations X and Y
are equally split into two branches by two 3-dB couplers.
Ports 2 and 3 are fed into a standard coherent receiver.
Ports 1 and 4 can be fed into a balanced PD, while in
experiment we send them to two single-ended PDs and
balance them in DSP. The electrical signal is sampled by
a real-time oscilloscope at a sampling rate of 50 GSa/s
with 15-GHz electrical bandwidth. DSP of the received
signal includes: (1) OFDM window synchronization [2];
(2) SV rotation matrix training and polarization
recovery; (3) cyclic prefix removal and FFT; (4) channel
equalization; (5) constellation reconstruction and BER
calculation. Totally 5.24 million bits are collected for
BER calculation of one channel.
Results and Discussion
We first measure the bit error rate (BER) as the function
Fig. 2 Experimental setup for 1-Tb/s SV-DD. (Inset (i) spectrum before transmission, (ii) spectrum after 480-km transmission, and
(iii) spectrum of channel 6 after wave-shaper.) MUX: multiplexer (10x1 coupler); IM: intensity modulator; I/Q mod. I/Q modulator;
AWG: arbitrary waveform generator; PBC/S: polarization beam combiner/splitter; EDFA: erbium doped fiber amplifier; SW: optical
switch; BPF: band-pass filter; WSS: wavelength selective switch; PD: photo-detector; B-PD: balance Photo-detector.
of the fiber launch power as shown in Fig. 3 to identify
the fiber nonlinearity tolerance for this 1-Tb/s system.
The optimum launch power is 8 dBm. Then, we measure
the BER performance for all the 10 channels at the reach
of 480 km with the launch power of 8 dBm. As shown in
Fig. 4, all the bands can achieve the BER lower than
0.024, the threshold of 20% FEC [10]. The inset of Fig.
4 is a constellation measured for channel 6 at an OSNR
of 34 dB with a BER of 8.8×10-3.
Fig. 3 BER performance as a function of fiber launch power for
1 Tb/s SV-DD signal.
Fig. 4 BER performance for 100-Gb/s SV-DD tributary
channel after 480-km transmission.
about 31 dB for 10×25-Gbaud 16-QAM signals to
achieve the BER below 20% forward error correction
(FEC) threshold. Considering the highest OSNR of 34
dB achieved in the experiment, the system OSNR
margin is about 3 dB between 31 dB (20% FEC
threshold) and this value. Compared with previous 1Tb/s POL-MUX coherent detection, the required OSNR
at 20% FEC threshold is about 25 dB for 1 Tb/s coherent
system in [2]. The experimental result indicates a 6 dB
OSNR sensitivity penalty for SV-DD, which agrees with
the theoretical prediction: since half of the optical power
is shared by the carrier at the SV-DD transmitter, there is
3 dB intrinsic OSNR penalty; both carrier and signal
suffer from noise degradation in SV-DD, leading to
another 3 dB OSNR penalty compared with the coherent
system. Nevertheless, this OSNR sensitivity is still much
better than the conventional single-end PD based DD [7],
because SV-DD does not need a high CSPR to suppress
the 2nd order nonlinearity.
The performance for this initial 1-Tb/s SV-DD
transmission is limited primarily by the multi-tone
generator, which has an EDFA inside sacrificing the
system OSNR and therefore degrading the OSNR
sensitivity. In practice, the 25 Gbaud signal can be
generated with a higher sampling rate digital-to-analog
converter (DAC). Moreover, by using higher baud rate
transmitter with lower modulation format, SV-DD can
even support a transmission distance of more than 1000
kilometers, which reveals the flexible capability for SVDD to be deployed in short reach applications.
We have experimentally demonstrated the first Terabit
direct detection reception over 480-km SSMF using the
SV-DD with 100% spectrum efficiency with reference to
coherent detection. Predictably SV-DD cannot replace
coherent detection for long-haul applications; however,
it provides a cost-effective solution for short-reach
applications, especially those using single polarization
Fig. 5 BER performance for 1 Tb/s transmission over 480-km
SSMF. The OSNR are for 10 wavelengths, and should be
reduced by 10 dB for single wavelength.
Fig. 5 shows the system OSNR sensitivity. For 480
km SSFM transmission, SV-DD requires an OSNR of
1. P. Winzer, IEEE Comm. Magazine, 48(7), 26-30, 2010.
2. Y. Ma, Q. Yang, Y. Tang, S. Chen, and W. Shieh,
OFC’2009, PDPC1.
3. S. Chandrasekhar, X. Liu, B. Zhu and D. W. Peckham,
ECOC’2009, PDP2.6.
4. M. Schuster, S. Randel, C. A. Bunge, S. C. J. Lee, F.
Breyer, B. Spinnler, and K. Petermann, IEEE Photon.
Technol. Lett. 20(9), 670–672, 2008.
5. W. Peng, X. Wu, V. Arbab, B. Shamee, L. Christen, J.
Yang, K. Feng, A. Willner, and S. Chi, OFC’2008, OMU2.
6. M. Nazarathy and A. Agmon, ECOC’2013, Mo.4.C.4.
7. X. Chen, A. Li, D. Che, Q. Hu, Y. Wang, J. He, and W.
Shieh, OFC’2013, PDP5B.7.
8. D. Che, X. Chen, J. He, A. Li and W. Shieh, OFC’2014,
9. D. Che, A. Li, X. Chen, Q. Hu, Y. Wang and W. Shieh,
OFC’2014, Th5C.7.
10. D. Chang, F. Yu, Z. Xiao, Y. Li, N. Stojanovic, C. Xie, X.
Shi, X. Xu, and Q. Xiong, OFC’2011, OTuN2.