Receiver sensitivity improvement in spectrally-efficient guard-band twin-SSB-OFDM using an optical IQ modulator

To further improve receiver sensitivity of spectrally-efficient guard-band direct-detection optical orthogonal frequency-division multiplexing (OFDM) with twin single-side-band (SSB) mod[r]

(1)

twin-SSB-OFDM using an optical IQ modulator

Ming Chena,*, Miao Penga, Hui Zhoua, Zhiwei Zhenga,b, Xionggui Tanga, Lap Maivanc aCollege of Physics and Information Science, Hunan Normal University, Changsha 410081, China

bSZU-NUS Collaborative Innovation Center for Optoelectronic Science and Technology, Key Laboratory of Optoelectronic Devices and Systems of Ministry of Education

and Guangdong Province, College of Optoelectronic Engineering, Shenzhen University, Shenzhen 518060, China

cElectronic and Electrical Engineering Department, Haiphong Private University, Haiphong 35000, Vietnam

a r t i c l e i n f o

Keywords:

Single-side-band (SSB) modulation Optical orthogonal frequency-division

multiplexing (OFDM)

Carrier-to-signal power ratio (CSPR) Direct-detection

a b s t r a c t

To further improve receiver sensitivity of spectrally-efficient guard-band direct-detection optical orthogonal frequency-division multiplexing (OFDM) with twin single-side-band (SSB) modulation technique, an optical IQ modulator (IQM) is employed to optimize optical carrier-to-signal power ratio (CSPR) The CSPRs for the guard-band twin-SSB-OFDM signal generated by using dual-drive Mach–Zehnder modulator (DD-MZM) and optical IQM are theoretically analyzed and supported by simulations The optimal CSPR for the two types of guard-band twin-SSB-OFDM are identified The simulations exhibit that the error vector magnitude (EVM) performance of the IQM-enabled guard-band OFDM is improved by more than 4-dB compared to that of the twin-SSB-OFDM enabled by DD-MZM after 80-km single-mode fiber (SMF) transmission In addition, more than 3-dB and 10 dB receiver sensitivity improvements in terms of received optical power (ROP) and optical signal-to-noise ratio (OSNR) are also achieved, respectively

1 Introduction

Direct-detection optical orthogonal frequency-division multiplexing (DDO-OFDM) has been widely considered as one of the most promising candidates for next-generation optical fiber access networks [1,2] and data centers [3,4], due to its simple configuration, high spectral effi-ciency (SE) with higher-order modulation formats, and robustness to optical fiber dispersions It has been extensively studied by numerical simulations and experiments [5–8] Two major issues, i.e., chromatic dispersion induced power fading and poor receiver sensitivity due to high optical carrier-to-signal power ratio (CSPR), should be well addressed Single-side-band modulation is an efficient way to overcome the power fading and thus extend the transmission distance [9] To improve receiver sensitivity of DDO-OFDM, many methods have been proposed to reduce CSPR of the SSB-OFDM signals in the literature However, the signal-to-signal beating interference (SSBI) after square-law photo-detection will deteriorate receiver performance Generally, a guard band with a bandwidth equals to the signal bandwidth between the optical carrier and OFDM signal spectrum is required [9], or the data only on the odd subcarriers are interleaved with blank subcarriers [10] to avoid the SSBI Therefore, the spectral efficiency (SE) is reduced by

* Corresponding author.

E-mail address:ming.chen@hunnu.edu.cn(M Chen)

half A spectrally-efficient compatible-SSB modulation without using guard band was proposed in [11] However, this method is sensitive to chromatic dispersion (CD) and has a design trade-off between better receiver sensitivity and CD tolerance In [12], a RF-tone-assisted virtual SSB-OFDM with iterative estimation and cancellation technique has been proposed and experimentally demonstrated, but it increases digital signal processing (DSP) complexity of the receiver Recently, a beat interference cancellation receiver (BICR) [13] and a SSBI cancellation receiver with balanced detection (ICRBD) [14] for the reduced guard-band SSB-OFDM systems were investigated by means of experiments and numerical simulations, respectively However, a complex balanced receiver is required Also, the non-ideal parameters of the devices in the receivers may degrade the receiver performance, and its impacts should be compensated in the optical or digital domain Moreover, a blockwise signal-phase-switching (SPS) technique with additional DSP algorithms in the receiver is employed for SSBI cancellation [15], which requires a modified optical IQ modulator and high optical signal-to-noise ratio (OSNR) In [16], a spectrally-efficient SSBI cancellation method us-ing guard-band twin-SSB technique is experimentally demonstrated in a dual-drive Mach–Zehnder modulator (DD-MZM) based DDO-OFDM

http://dx.doi.org/10.1016/j.optcom.2017.08.033

(2)

Fig 1. Schematic diagram of twin-SSB-OFDM and the proposed dual-SSB-OFDM transceivers transmission system This technique can reduce PD bandwidths and ADC

sampling rate However, the CSPR of the generated twin-SSB-OFDM signal is relatively high and thus results in poor receiver sensitivity

In this paper, the DD-MZM is replaced by an optical IQ modulator (IQM) to optimize the CSPR by adjusting DC bias voltages and the amplitude of the drive signal Here, the IQM-enabled guard-band twin-SSB-OFDM is named as dual-twin-SSB-OFDM Firstly, the principle of the dual-SSB-OFDM transmitter is described Secondly, the CSPRs for the twin-SSB-OFDM and dual-SSB-OFDM signals are theoretically analyzed Thirdly, the optimal CSPRs for the two types of guard-band SSB-OFDM are identified Lastly, the error vector magnitude (EVM) performance and receiver sensitivity in terms of received optical power (ROP) and optical signal-to-noise ratio (OSNR) are studied for both twin-SSB-OFDM and dual-SSB-OFDM

2 Operation principles

The operation principles of the twin-SSB-OFDM and the proposed dual-SSB-OFDM are schematically illustrated inFig In the case of twin-SSB-OFDM, two base-band digital OFDM transmitters with the same DSP procedures are employed to generate two digital guard-band right-side-band (RSB) signals𝐼𝑋(𝑛) +𝑗𝑄𝑋(𝑛)and𝐼𝑌(𝑛) +𝑗𝑄𝑌(𝑛) The corresponding diagrams of the input vectors forN-point IDFT are shown in the insets (i) and (ii) of Fig 1(a), whereDis the number of data subcarriers (SC) The two digital RSB signals are organized into the form

of[(𝐼𝑋(𝑛) +𝐼𝑌(𝑛)) +𝑗(𝑄𝑋(𝑛) −𝑄𝑌(𝑛))], and converted to the electrical

signals by two parallel DACs Then, a DD-MZM is used to generate the guard-band twin-SSB-OFDM signal and its spectrum is plotted in inset (iv) ofFig 1(a) It should be pointed that𝑄𝑋(𝑡)and𝑄𝑌(𝑡)are the Hilbert transforms of𝐼𝑋(𝑡)and𝐼𝑌(𝑡), respectively, so the signal𝐼𝑌(𝑡) −𝑗𝑄𝑌(𝑡) should be a left-side-band (LSB) signal However, a small drive signal should be used to realize IQ modulation by using a DD-MZM biased at its quadrature point, which leads to high CSPR and results in poor receiver sensitivity The detailed descriptions for the twin-SSB-OFDM transceiver proposed by L Zhang can be found in [16]

In this work, to simplify the design of the twin-SSB-OFDM trans-mitter and improve receiver sensitivity, IQM-enabled dual-SSB-OFDM is proposed and presented in Fig 1(b) Firstly, the two digital SSB-OFDM transmitters are simplified into a single digital dual-SSB-SSB-OFDM transmitter The diagram of the input vector forN-point IDFT is shown in the inset (v) of Fig 1(b) Secondly, by adjusting DC bias voltages applied to IQM, the CSPR of the dual-SSB-OFDM signal is expected to

be largely reduced The spectrum of the dual-SSB-OFDM is the same as that of the twin-SSB-OFDM, which is schematically plotted in inset (vi) ofFig 1(b)

The same receiver structure can be used to receive and demodulate the two types of the SSB-OFDM signals The receiver consists of one optical coupler (OC), two optical band-pass filters (OBPFs), two PDs and two ADCs Two side-band signals (i.e., LSB and RSB) can be independently demodulated The received twin (dual)-SSB-OFDM signal is filtered by two OBPFs to obtain optical RSB and LSB signals in the upper and lower branches, respectively The corresponding spectra are shown in insets (vii) and (viii) ofFig 1(c) Then, the optical RSB and LSB signals are directly detected by two PDs, and their spectra are presented in insets (ix) and (x) ofFig 1(c), respectively The recovered signal can avoid being affected by the SSBI which is only located in the guard-band Compared to the coherent optical OFDM (CO-OFDM) system where the additional optical source and optical hybrid at the receiver are required, so it still has a simpler system configuration In addition, CO-OFDM is very sensitive to laser phase noise (LPN) and carrier frequency offset (CFO) To address these issues, the additional digital signal processing (DSP) algorithms should be implemented As a result, CO-OFDM system has higher system cost and higher power consumption, due to the complex system configuration and induced by the additional DSP algorithms for LPN and CFO compensations

3 Theoretical analysis of CSPR

The structures of DD-MZM and IQM are schematically depicted in Fig 1(a) andFig 1(b), respectively, where DD-MZM consists of two parallel phase modulators (PM1 and PM2), while IQM is composed of two single-drive MZMs and one phase modulator (MZM1, MZM2 and PM)

3.1 CSPR for twin-SSB-OFDM

In the case of DD-MZM, the output optical field,𝐸𝐷𝑀 𝑍𝑀(𝑡), after the second Y-branch can be related to the input optical field𝐸𝑖𝑛(𝑡)by [17]

𝐸𝐷𝑀 𝑍𝑀(𝑡) = [

𝑒𝑗

𝜋

𝑉𝜋(𝐼(𝑡)−𝑉𝑏𝑖𝑎𝑠1)+𝑒𝑗 𝜋

𝑉𝜋(𝑄(𝑡)−𝑉𝑏𝑖𝑎𝑠2)]𝐸

𝑖𝑛(𝑡) (1)

(3)

×𝐸𝑖𝑛(𝑡)𝑒 𝑗𝜋

𝑉𝜋𝑉𝑏𝑖𝑎𝑠2

When𝜎of the drive signal is small enough, the second-order inter-modulation item as given in the third term in Eq.(2)can be neglected Then, the complex-valued OFDM signals(t) is linearly converted into the optical field From Eq.(2), the optical carrier power and signal power of the twin-SSB-OFDM signal can be expressed as𝑃𝑐∝|1 −𝑗|2= 2and 𝑃𝑡𝑤𝑖𝑛 ∝ (𝜋∕𝑉𝜋𝜎

)2

, respectively Therefore, the CSPR of the twin-SSB-OFDM signal is defined as

𝐶𝑆𝑃 𝑅𝐷𝑀 𝑍𝑀= 𝑃𝑐 𝑃𝑡𝑤𝑖𝑛= ( 𝜋 𝑉𝜋 𝜎) (3)

According to Eq.(3), the CSPR of the twin-SSB-OFDM signal can be only changed via adjusting the𝜎of the electrical drive signal, i.e., the average power The 𝐶𝑆𝑃 𝑅𝐷𝑀 𝑍𝑀 can be reduced by increasing the average power of the drive signal Unfortunately, the second-order and higher-order intermodulation terms will deteriorate the receiver performance after the square-law photo-detection

Assuming the output power and line-width of the optical source are 10 dBm and Hz, two power spectra of the twin-SSB-OFDM signals with CSPRs of 40 dB and 10 dB are presented inFig 2(a) andFig 2(b), respectively It should be pointed that the power spectra of the 1st, 2nd, and 3rd items given in Eq.(2)are separately plotted inFig On one hand, the second-order and high-order intermodulation terms can be neglected when CSPR is 40 dB, but it results in poor receiver sensitivity; on the other hand, the power of the second-order intermodulation term (3rd item) is significantly increased when CSPR may be reduced to 10 dB, which will degrade the receiver performance Therefore, there is an optimum CSPR for the twin-SSB-OFDM to achieve a good receiver sensitivity, which will be determined by numerical simulations in later section

3.2 CSPR for dual-SSB-OFDM

In the case of IQM, the output optical field,𝐸𝐼 𝑄𝑀(𝑡)can be related to the input optical field𝐸𝑖𝑛(𝑡)by [17]

𝐸𝐼 𝑄𝑀(𝑡) =

2 [

cos (

𝜋 2𝑉𝜋 ⋅

(

𝐼(𝑡) −𝑉𝑏𝑖𝑎𝑠1))

+𝑗cos

( 𝜋 2𝑉𝜋

⋅(𝑄(𝑡) −𝑉𝑏𝑖𝑎𝑠2))]𝐸𝑖𝑛(𝑡)

= [ cos ( 𝜋 2𝑉𝜋 𝐼(𝑡) ) ⋅cos (𝜋𝑉 𝑏𝑖𝑎𝑠1 2𝑉𝜋 ) + sin ( 𝜋 2𝑉𝜋 𝐼(𝑡) ) ⋅sin (𝜋𝑉 𝑏𝑖𝑎𝑠1 2𝑉𝜋 ) +𝑗 ( cos ( 𝜋 2𝑉𝜋 𝑄(𝑡) ) ⋅cos (𝜋𝑉 𝑏𝑖𝑎𝑠2 2𝑉𝜋 ) + sin ( 𝜋 2𝑉𝜋 𝑄(𝑡) ) ⋅sin (𝜋𝑉 𝑏𝑖𝑎𝑠2 2𝑉𝜋 ))] 𝐸𝑖𝑛(𝑡) (4)

where the in-phase componentI (t) and quadrature componentQ (t) of the drive signals(t) are applied to two parallel MZMs A relative phase shift of𝜋/2 is adjusted in the quadrature arm by an additional

=1

2 [

(1 +𝑗)⋅cos (𝜋𝑉 𝑏𝑖𝑎𝑠 2𝑉𝜋 ) + 𝜋 2𝑉𝜋 ⋅sin (𝜋𝑉 𝑏𝑖𝑎𝑠 2𝑉𝜋 ) ⋅(𝐼(𝑡) +𝑗𝑄(𝑡)) −1 ( 𝜋 2𝑉𝜋 )2 ⋅cos (𝜋𝑉 𝑏𝑖𝑎𝑠 2𝑉𝜋 )

⋅(𝐼2(𝑡) +𝑗𝑄2(𝑡)) ]

𝐸𝑖𝑛(𝑡)

Similarly, the third term in Eq.(5)can be neglected when drive signal is small, the ratio of𝑉biasto𝑉𝜋is defined asM and the CSPR of the dual-SSB-OFDM signal can be expressed as

𝐶𝑆𝑃 𝑅𝐼 𝑄𝑀=

2cos2(𝜋𝑉𝑏𝑖𝑎𝑠 2𝑉𝜋

) (

𝜋 2𝑉𝜋 ⋅sin

(𝜋𝑉

𝑏𝑖𝑎𝑠 2𝑉𝜋

)

⋅𝜎)2 =

2𝑐𝑜𝑠2(𝜋 2𝑀

) (

𝜋 2𝑉𝜋 ⋅sin

( 𝜋 2𝑀

)

⋅𝜎)2 (6)

In addition to the average power of the drive signal related to𝜎, the CSPR of the dual-SSB-OFDM signal also can be changed by adjusting the DC bias voltage applied to IQM From Eq.(5), both the second-order intermodulation term (3rd item) and the optical carrier (1st item) can be largely reduced whenM is close to However, the insertion loss becomes very large and thus a high-gain and low-noise-figure optical amplifier is required Especially, the optical carrier and the second-order intermodulation term are completely suppressed whenMis equal to In this case, the useful signal will not be recovered after square-law photo-detection According to Eq.(5), the power spectra of the 1st, 2nd and 3rd items of the dual-SSB-OFDM signal with the same CSPR of dB andMof 0.7 and 0.9 are shown inFig 3(a) andFig 3(b), respectively As we can see clearly, the second-order intermodulation term is reduced, and larger insertion loss is also observed for the case ofM equals to 0.9 The optimal CSPRs for differentMvalues will be also discussed in Section5

4 Simulation setup

(4)

Fig 2. Power spectra of the twin-SSB-OFDM signals (a) CSPR= 40dB and (b) CSPR= 10dB

Fig 3. Power spectra of the dual-SSB-OFDM signals with a CSPR of dB (a)𝑀= 0.7and (b)𝑀= 0.9

Fig 4. Simulation setup of the proposed dual-SSB-OFDM direct-detection systems by an optical amplifier (OA) with the noise figure of 4-dB, which is used

to compensate the insertion loss of the DD-MZM (IQM) The power of the signal coupled into the optical fiber is fixed at 8-dBm using an optical attenuator (ATT) The guard-band dual-SSB-OFDM signal with a CSPR of 3-dB after IQM and ATT are presented in the insets (iii) and (iv) of Fig 4, respectively

Two transmission system configurations, i.e., optical back-to-back (OB2B) and 80-km SMF are studied In OB2B case, the attenuated signal at the optical transmitter is directly connected to the second ATT In the case of 80-km SMF transmission, the attenuated signal is launched into 80-km SMF, and the fiber loss is compensated by the second OA with 16-dB gain At the receiver end, the ROP is measured after the second ATT which is employed to change the received optical power (ROP) In addition, a noise loading module consists of ATT and followed by an OA with 20-dB gain is used to adjust the OSNR in order to

(5)

ELPF Cutoff frequency 40 GHz

Order

Ripple factor 0.05 dB LD

Operation wavelength 1550 nm

Line-width 100 kHz

Output power 10 dBm DD-MZM Half-wave voltage (𝑉𝜋) V

Bias voltage difference (𝑉bias1-𝑉bias2) V

IQM Half-wave voltage (𝑉𝜋) V

OBPF

Filter type Gaussian

Center wavelength 1550.16/1549.84 nm

BW 60 GHz

Order

OA Noise figure dB

SMF

Length 80 km

Attenuation 0.2 dB/km Dispersion 16.75 ps/nm/km Dispersion slope 0.075 ps/nm2/km

Differential group delay 0.2 ps/km PD

Responsivity A/W Thermal noise 1e–22 W/Hz Dark current 10 nA

EBPF

Filter type Chebyshev Center frequency 30 GHz

BW 20 GHz

Order

Ripple factor 0.05 dB

the sampled data captured by the two ADCs are post-processed inde-pendently with the same receiver DSP algorithms including TS-based symbol timing synchronization, CP removal, discrete Fourier transform (DFT), TS-aided channel estimation and one-tap channel equalization, and EVM calculation The detailed DSP descriptions can be found in our previous work [18] The parameters used in our simulations are listed in Table

5 Simulated results and discussions

The theoretical CSPR according to Eqs.(3)and(6), and the simulated CSPR as a function of the standard deviation of the drive signal (𝜎) are presented in Fig 5(a) It indicates that the simulated CSPRs for both twin-SSB and dual-SSB signals are in good agreement with the corresponding theoretical ones The CSPR decreases as the standard deviation of the drive signal increases Moreover, the difference between the simulated CSPR and theoretical CSPR is also given in Fig 5(b) The small difference is mainly induced by the modulator nonlinearity The high-order intermodulation terms become larger when𝜎increases, and might fall on the LSB and RSB signals This leads to an increased difference between the simulated and theoretical CSPR Meanwhile, the IQM in the dual-SSB-OFDM with theMof 0.9 operates at a large linear range, and then the high-order intermodulation terms are much smaller than the outputs of the DD-MZM used in the twin-SSB-OFDM Therefore, the difference for the dual-SSB-OFDM is smaller than the twin-SSB-OFDM

Fig shows that the insertion loss of the DD-MZM and IQM employed in twin-SSB-OFDM and dual-SSB-OFDM, respectively, as a

Fig 5. Theoretical and simulated CSPRs versus standard deviations

Fig 6. Insertion loss of optical modulators versus CSPR

Fig 7. EVM performance versus CSPR in the OB2B configuration

function of difference CSPRs It should be mentioned that the insertion loss is defined as the power difference between input and output powers of optical modulators For twin-SSB-OFDM, the insertion loss of the DD-MZM is about dB for all of the CSPRs In the case of dual-SSB-OFDM, the insertion loss of the IQM becomes larger asMvalue increases More than 25 dB insertion loss is observed whenMis 0.95 For a certain𝑀 value, the insertion loss can be decreased by reducing CSPR

(6)

Fig 8. Optimal CSPR and the corresponding EVM performance versus ROP after 80 km SMF transmission

12, 9, and−2 dB, respectively The corresponding EVM values are −24.4,−26,−27.6,−29.2,−30.5 and−29.7 dB, respectively WhenM is 0.9, the EVM performance can be improved by 5.5 dB compared to the twin-SSB-OFDM For the CSPR less than its optimal CSPR, the EVM performance is mainly limited by the second-order intermodulation term and other high-order intermodulation terms; while the degraded EVM performance is mainly attributed into the OA and PD noises for the CSPR larger than its optimal CSPR In the case of𝑀 = 0.95, the large insertion loss makes the dual-SSB-OFDM signal more sensitive to OA and PD noises, which results in a little worse EVM performance than that of𝑀= 0.9case

After 80-km SMF transmission, the optimal CSPR and the corre-sponding EVM performance as a function of ROP are presented in Fig 8(a) and Fig 8(b), respectively On one hand, the receiver per-formance is limited by the modulator nonlinearity at the high ROPs, therefore, a small drive signal (i.e., with a low𝜎) is applied to overcome this issue; on the other hand, OA and PD noises may be the main factors for the degraded receiver performance at the low ROPs, and thus a large drive signal is used for increasing signal power The optimal CSPR increases as ROP increases, which can be seen from Fig 8(a) At an EVM of−16.5 dB, and the corresponding BER for 16-QAM modulation is 1e−3, the receiver sensitivity of the proposed dual-SSB-OFDM in terms of ROP is improved by more than dB compared to the twin-SSB-OFDM In addition, more than dB improvement in EVM at the same ROPs is also observed inFig 8(b)

The required OSNR when EVM is−16.5 dB for twin-SSB-OFDM at the optimal CSPR is 33.15 dB, while the required OSNR for dual-SSB-OFDM as a function of M values is presented inFig It exhibits that the receiver sensitivity can be improved by more than 10 dB when𝑀 is 0.9 or 0.95 compared to that of twin-SSB-OFDM

Fig 9. The required OSNR (0.1 nm resolution) of dual-SSB-OFDM for EVM= −16.5dB versus different𝑀values

6 Conclusion

A spectrally-efficient guard-band dual-SSB-OFDM by using an optical IQ modulator is investigated by numerical simulations The theoretical CSPRs for both dual-SSB-OFDM and twin-SSB-OFDM are analyzed, and they are in good agreement with the simulated CSPRs The optimal CSPRs are also identified for different ROPs It exhibits that the receiver sensitivity in terms of ROP can be improved by more than dB after 80-km SMF transmission at a BER of 1e−3, and more than dB (10 dB) improvements in EVM (OSNR) compared to the twin-SSB-OFDM It indicates longer transmission distance or higher data rate can be supported by the proposed dual-SSB-OFDM

Acknowledgments

This work is supported in part by Hunan Provincial Natural Sci-ence Foundation of China (Grant Nos 2017JJ3212, 2016JJ6097 and 14JJ6007), in part by Scientific Research Fund of Hunan Provincial Education Department (Grant Nos 17C0957 and 14B119), and in part by the Project Supported for excellent talents in Hunan Normal University (Grant No ET1502)

References

[1] Q Dayou, N Cvijetic, H Junqiang, W Ting, in Proc OFC 2009, Paper OMV1 [2] R Giddings, IEEE/OSA J Lightwave Technol 32 (4) (2014) 553–570

[3] Y Benlachtar, R Bouziane, R.I Killey, C.R Berger, P Milder, R Koutsoyannis, J.C Hoe, M Püschel, M Glick, in Proc ICTON 2010, Paper We.A4.3

[4] C Kachris, E Giacoumidis, I Tomkos, in Proc OFC 2011, Paper JWA87 [5] A.J Lowery, D Liang, J Armstrong, in Proc OFC 2006, Paper PDP39 [6] A Ali, J Leibrich, W Rosenkranz, in Proc OFC 2009, Paper OMT7

[7] Z Cao, J Yu, W Wang, L Chen, Z Dong, IEEE Photon Technol Lett 22 (11) (2010) 736–738

[8] F Li, X Li, L Chen, Y Xia, C Ge, Y Chen, IEEE Photon Technol Lett 26 (9) (2014) 941–944

[9] B.J.C Schmidt, A.J Lowery, J Armstrong, IEEE/OSA J Lightwave Technol 26 (1) (2008) 196–203

[10] W.-R Peng, X Wu, V.R Arbab, K.-M Feng, B Shamee, L.C Christen, J.-Y Yang, A.E Willner, S Chi, IEEE/OSA J Lightwave Technol 27 (10) (2009) 1332–1339 [11] M Schuster, C.A Bunge, B Spinnler, K Petermann, in Proc OFC 2008, Paper OMU7 [12] W.-R Peng, X Wu, K.-M Feng, V.R Arbab, B Shamee, J.-Y Yang, L.C Christen,

A.E Willner, S Chi, Opt Express 17 (11) (2009) 9099–9111

[13] S Alireza Nezamalhosseini, L.R Chen, Q Zhuge, M Malekiha, F Marvasti, D.V Plant, Opt Express 21 (13) (2013) 15237–15246

[14] J Ma, Opt Express 22 (24) (2014) 29636–29654

[15] A Li, D Che, X Chen, Q Hu, Y Wang, W Shieh, Opt Lett 38 (14) (2013) 2614– 2616

[16] L Zhang, T Zuo, Q Zhang, J Zhou, E Zhou, G.N Liu, in Proc ECOC 2016, Paper 1178

[17] M Seimetz, Springer, Berlin, 2009, pp 18–21

Định dạng
Số trang	6
Dung lượng	1,17 MB