All-Optical and Opto-Electronic Signal Regenerators Masayuki Graduate School of Engineering Osaka University 15th International SAOT Workshop on All-Optical Signal Regeneration September 28-29, 2011 1
Outline Introduction All-optical regeneration of (D)BPSK and (D)QPSK signals Phase-preserving amplitude regeneration using saturation of FWM PSK signal regenerators using amplitude regenerators Regenerators using opto-electronic device nonlinearities Application of signal processing such as MLSE (Maximum Likelihood Sequence Estimation) for mitigation of regenerator-induced signal distortion 2
Introduction Optical signals in long-distance transmission systems are impaired by: 1. Chromatic and polarization dispersion. 2. Transmission-fiber nonlinearities. 3. Noise introduced by optical amplifiers. Dispersion effects are linear and deterministic so that they can be compensated by optical or electrical means. Impairments caused by fiber nonlinearities are also deterministic (in the case of SPM). Its compensation is in principle possible. Impairments caused by random noise will ultimately determine the performance of the system. 3
Introduction Methods of reduction of noise accumulation: Use of low-loss fibers Use of low-noise amplifiers (low-noise EDFA, Raman, PSA,...) Use of signal regenerators TX RX TX RX TX RX Some or all of the complex electrical regenerators are desired to be replaced by all-optical regenerators. Reach reduction caused by noise is severer for multi-level format signals. Future all-optical regenerators should be able to deal with advanced modulation format signals (QPSK, 8PSK, QAM...). 4
Issues: Introduction 1. How we realize regenerators for PSK and/or multi-level signals? Detection and manipulation of optical phase are difficult. Realization of multi-level (staircase) transfer functions is difficult. (More fundamental) issues: 2. Regenerators work (only) when signal pulses to be regenerated are isolated in time with each other. DEMUX/MUX and per channel operation are needed for regeneration of WDM signals. Precise dispersion compensation is needed in the optical domain before the regenerator. How we apply all-optical regenerators in WDM systems? How we relax the transmission-line management required for the system using inline regenerators? 5
Introduction All-optical regeneration of (D)BPSK and (D)QPSK signals (1) Phase-preserving amplitude regeneration using saturation of FWM (2) PSK signal regenerators using amplitude regenerators Regenerators using opto-electronic device nonlinearities Application of signal processing such as MLSE (Maximum Likelihood Sequence Estimation) for mitigation of regenerator-induced signal distortion 6
All-Optical Regeneration of PSK Signals (1) Phase-preserving amplitude regeneration using saturation of FWM Phase-preserving amplitude regenerator only suppresses amplitude noise. However, it is still effective in reducing phase noise of PSK signals in long-distance systems. ASE Im[E] signal Re[E] 1. Linear phase noise Im[E] = + Re[E] Phase noise Im[E] Re[E] Amplitude noise TX 2. Nonlinear phase noise induced by amplitude noise "P and fiber nonlinearity 2 "P 1 TX "# 1 "# 2 "# 3 RX < "# 2 >$ N "P 3 "# 2 $(N % 2)"P 3 "# 2 $(N %1)"P 2 "# 1 $N"P 1 RX < "# 2 >$ N 3 Amplitude regeneration of PSK signals suppresses nonlinear phase noise. 7
Phase-Preserving Amplitude Regeneration Fiber-based amplitude regenerator (limiter) Mamyshev regenerator Soliton filtering pulse energy (pj) 0.8 0.6 0.4 0.2 decaying pulses (a) 0 0 5 10 15 20 25 30 number of regenerators phase deviation (deg) 500 250 0 (b) -250 decaying pulses -500 0 5 10 15 20 25 30 number of regenerators Phase spread is large. Saturation of FWM pulse energy (pj) 0.4 (c) 0.3 0.2 0.1 decaying pulses 0 0 5 10 15 20 25 30 number of regenerators phase deviation (deg) 500 250 0-250 Pump (d) decaying pulses -500 0 5 10 15 20 25 30 number of regenerators Phase spread is small. 8
Phase-Preserving Amplitude Regeneration Phase-preserving amplitude regeneration using saturation of FWM (parametric amplification) Three-wave coupled equations for partially degenerate FWM ω s ω p ω i ω de p de s de i dz = i" [ 2 E p E p + 2(E 2 s + E 2 i )E p + 2E # p E s E i e i$%z ] dz = i" [ E 2 2 s E s + 2(E p + E 2 i )E s + E 2 p E # i e $i%&z ] dz = i" [ E 2 2 i E i + 2(E p + 2 Es )E i + E 2 p E # s e $i%&z ] E s "# = $2#(% p )+#(% s )+#(% i ) = # 2 (% p )("%) 2 When and E i are much smaller than E p, the equations can be linearized with respect to E s and E i. The system behaves as a linear system with which the gain P s (L) is a constant. P s (0) = E s(l) 2 E s (0) = G 2 9
Phase-Preserving Amplitude Regeneration Saturation of parametric amplification Output signal power P s,out G =1+ ("P p g) 2 sinh 2 gl P s,out = P s,in + ( ) g = "(#$/2)(#$/2 + 2%P p ) " 1 " 3 sn 2 # " 1 $ " 3 + " 3 sn 2 # " = 7# 4 (# 3 $ # 1 ) %L /2 Input signal power P s,in (Analytical solution using Jacobi elliptical functions is available: Y.Chen, JOSA B6,1986(1989), G. Cappellini and S. Trillo, JOSA B8, 824(1991)) For high input signal powers and/or long fiber lengths, is not satisfied. E s, E i << E p The system is no longer linear with respect to and. The gain G changes with the input signal power. E s E i 10
Phase-Preserving Amplitude Regeneration Output signal power P s,out Input signal power P s,in Saturation of parametric amplification is detrimental in amplifier applications. Crosstalk between channels mediated by pump depletion in WDM signal amplification Intra-pulse gain saturation in high-speed signal amplification It can be used for ultrafast signal processing. Ultrafast all-optical switching (e.g., H. Sunnerud et al.,ecoc2007,5.3.5.) Ultrafast regeneration of amplitude levels 11
Experiment of Amplitude Reshaping 10GHz 1558nm output signal power (mw) MLLD 1 0.8 0.6 0.4 0.2 LD 1561nm 7.5ps LNM 1GHz phase mod. ATT OBPF1 OBPF2 0 0 0 2 4 6 input signal power to HNLF (mw) 40 35 30 25 20 15 10 5 Q factor PC PC HNLF: length 1.5km, λ 0 1556nm γ ~ 12/W/km HNLF POL Pump power: ~ 40mW at HNLF OSNR of input signal 13dB/0.1nm 15dB/0.1nm 17dB/0.1nm µ PM Q = µ/σ Q is increased by a factor ~2.8. SO σ 12
Experiment of DQPSK Signal Transmission 10 G symbol/s Pulse shapes DDMF(Densely Dispersion-Managed Fiber) 40km SMF 50km +DCF Dispersion: ±3ps/nm/km 2km x 20sections γ~3.5/w/km Before limiter After limiter 13
Experiment of DQPSK Signal Transmission -2-2 log 10 (BER) -4-6 -8 5dB log 10 (BER) -4-6 -8 1.5dB -10-5 0 5 10 launched signal power (dbm) DDMF -10 8 9 10 11 12 launched signal power (dbm) SMF+DCF Blue curves : amplitude limiter removed Red curves : amplitude limiter inserted 14
Experiment of DPSK Signal Transmission 40km x 5span transmission of 10Gb/s short-pulse DPSK signal A limiter is inserted either before recirculating loop (A) or inside the loop (B). -2 solid: pump ON, dashed: pump OFF A log(ber) -4-6 -8-10 A TX OSNR: 21.5dB/0.1nm B log(ber) 0 0.2 0.4 0.6 0.8 signal power (mw) 1 1.2-2 -4-6 -8-10 B TX OSNR: 25.7dB/0.1nm 0 0.2 0.4 0.6 0.8 1 1.2 signal power (mw) 15
Generation of Extra Phase Noise in the Limiter In the (phase-insensitive) parametric regenerator, amplitude noise is suppressed while the phase noise is preserved. σ θ 2, in σθ 2, out = σ θ 2, in + α Extra phase noise should be minimized. [A p + "A p ]e i(# p +"# p (t)) [A s + "A s ]e i# s A s e i(" s +#" s FOPA ) Nonlinear processes in fiber (SPM, XPM, FWM) translate fluctuations in pump and signal amplitudes and frequencies into output signal phase fluctuation. M. Sköld et al., OE16, 5974(2008), M., OL33, 1638 (2008), R. Elschner and K. Petermann, ECOC2009, 3.3.4 (2009), S. Moro et al., OE18, 21449 (2010). 16
Generation of Extra Phase Noise in the Limiter 1. Translation from pump noise to signal phase noise de p de s de i dz = i"e p 2 E p ( ) ( ) dz = i" 2 E p 2 Es + E p 2 E i # e $i%&z dz = i" 2 E p 2 E i + E p 2 E s # e $i%&z Solution for the output signal phase [ ] + $P p L # %& 2 " s (L) = " s (0)+ tan #1 ($P p g) ( 1+ %& (2$P p ))tanh(gl) 2nd term 3rd term phase noise originated from the dependency of the parametric gain on the instantaneous pump amplitude and frequency. phase noise originated from the pump nonlinear phase noise. S. Moro et al., OE18, 21449 (2010). 17
Generation of Extra Phase Noise in the Limiter 2. Phase noise generated by the input signal amplitude noise de s [ ] dz = i" E s 2 E s + 2(E p 2 + E i 2 )E s + E p 2 E i # e $i%&z E m = A m e i" m (m = s,p,i) [ ] d" s dz = # A s 2 + 2A p 2 + 2A i 2 + (A p 2 A i / A s )cos" Output phase noise vs pump power output phase noise!phase [deg] 25 20 15 10 5 0 pump SNR "pump,in = 30dB 40dB 50dB 0 100 200 300 400 500 pump power Pp [mw] " sig,in = 20dB signal-induced phase noise dominates. pump-induced! phase noise dominates. 18
Multi-Channel Reshaping Multi-channel signals can share a cw pump if the signals are in RZ format and are properly time-interleaved. λ 1 time λ 2 time Δτ Pump Adaptive delay control is required for proper time interleaving. 19
Multi-Channel Reshaping 2ch x 10Gb/s RZ-DPSK reshaping and transmission experiment 2ch RZ-DPSK signal source ch.1 (random delay) ch.2 (adaptively controlled delay) 40km transmission fiber N.S.Mohd Shah, M. Sato, and M., submitted. Amplitude limiter 20
2ch signal power transfer functions Multi-Channel Reshaping Pulse shapes before and after the limiter before the limiter after the limiter 21
Multi-Channel Reshaping Spectra after the HNLF signals pump idlers solid curve: Pulses are time-interlearved. dashed curve: Pulses are overlapped. Pulse overlapping generates sidebands around signal, pump, and idlers. overlap Optical power at a certain wavelength can be used for monitoring signal overlapping condition. interleaved 22
Multi-Channel Reshaping BER after transmission with adaptive delay control on and off channel 1 off on channel 2 off on 23
Contents All-optical regeneration of (D)BPSK and (D)QPSK signals (1) Phase-preserving amplitude regeneration using saturation of FWM (2) PSK signal regenerators using amplitude regenerators Regenerators using opto-electronic device nonlinearities Application of MLSE for mitigation of regeneratorinduced signal distortion 24
Phase and Amplitude Regeneration Using Amplitude-Only Regenerator Phase modulation Demodulation using delay interferometer DPSK OOK Amplitude modulation Amplitude regenerator Coherent demodulation BPSK (Amplitude noise is removed.) Local Oscillator OOK Amplitude information is transformed back to phase information. 25
DPSK Regenerator Using a Straight-Line Phase Modulator input 1-bit DI CR / Optical pulse source 2R amplitude regenerator All-optical phase modulator output Clock recovery / Optical pulse source x N λ' s HNLF Phase modulator λ's λ s 1-bit delay interferometer HNLF λs + Δλ Amplitude regenerator M., PTL17, 213 (2007). 26
DQPSK Regenerator Using a Straight-Line Phase Modulator input θ DI =π/4 CR / Optical pulse source 2R regenerator 0 / π 0 / π/2 phase modulator phase modulator output θ DI =-π/4 2R regenerator " #,input Imaginary part [mw 1/2 ] 2 1 0-1 Numerical Results: 80Gsymbol/s(160Gbit/s) 2.5ps RZ-DQPSK (c) -2-2 -1 0 1 2 Real part [mw 1/2 ] Imaginary part [mw 1/2 ] 1.5 0 (d) -1.5-1.5 0 1.5 Real part [mw 1/2 ] Input OSNR=24 db/0.1nm standard deviation of phase fluctuation (deg) 10 8 6 4 2!! " #,output 0 20 22 24 26 28 30 32 OSNR (db/0.1nm) simulation using 1024 symbols 27
DPSK Regenerator Using a Mach-Zehnder Interferometer Modulator DPSK OOK All-optical output CR / Optical modulator pulse source A in exp(" in ) 1-bit DI All-optical A out exp(" out ) modulator OOK Complementary OOK signals drive all-optical modulators in MZI. I. Kang et al.,th4.3.3,ecoc2005 (2005) P. Vorreau et al.,ptl18, 1970 (2006) Ch. Kouloumentas et al.,omt5, OFC2010 (2010) Additional 2R regenerators may be needed either before or after the DI for suppression of both phase and amplitude noise. R. Elschner et al., OL32, 112 (2007) R. Elschner et al.,thp3, 2007LEOS Annual Meeting (2007) 28
DQPSK Regenerator Using MZI Modulators θ DI = π/4 2R All-optical modulator input CR/Pulse source θ DI = -π/4 2R 2R 2R All-optical modulator All-optical modulator All-optical modulator π/2 output M. Spyropoulou et al., OFC2011, OThY2 (2011). M., in Impact of Nonlinearities on Fiber Optic Communications, S. Kumar Ed., Springer (2011). 29
BPSK and QPSK Regenerators Using Coherent Demodulation Demodulation from DPSK to OOK by DI may be replaced by coherent demodulation. input CR / Optical pulse source 1-bit DI 2R phase modulator output input CR / Optical pulse source 2R phase modulator output Local oscillator Required strength of 2R can be halved. Phase data on the signal are preserved. Phase-locked local oscillator is needed. 30
BPSK and QPSK Regenerators Using Coherent Demodulation Coherent BPSK regenerator / wavelength converter input LO 2R CR/PS 2R All-optical modulator All-optical modulator output Coherent QPSK regenerator / wavelength converter 2R All-optical modulator input 2R All-optical modulator CR/PS LO 2R 2R X. Yi et al., JLT28, 587 (2010) All-optical modulator All-optical modulator π/2 output 31
Contents All-optical regeneration of (D)BPSK and (D)QPSK signals (1) Phase-preserving amplitude regeneration using saturation of FWM (2) PSK signal regenerators using amplitude regenerators Regenerators using opto-electronic device nonlinearities Application of MLSE for mitigation of regeneratorinduced signal distortion 32
Regenerators Using Opto-Electronic Device Nonlinearities Regeneration of digital signals in general requires the use of nonlinear thresholding function of some form. All-optical regenerators use all-optical nonlinear media / devices. ultrafast Available nonlinearities are generally weak. High signal power and/or long interaction length are needed. All-optical regenerators are most suitable for noise reduction of ultrahigh speed signals in simple modulation formats. For moderately high-speed signals, nonlinearities in opto-electronic or electronic devices may be conveniently utilized. 33
Regenerators Using Opto-Electronic Device Nonlinearities Nonlinear opto-electronic and electronic devices: electo-optic modulator electro-absorption modulator RF limiting amplifier low-resolution AD converter K. Inoue, IEEE PTL8, 1322(1996) P. Öhlén and E. Berglind, IEEE PTL9, 1011(1997) W. Kuebart et al.,ecoc2003,mo4.3.1(2003) H. -F. Chou and J. E. Bowers, Opt.Exp.13,2742(2005) input CR / Optical pulse source phase modulator output input CR / Optical pulse source MZI EO modulator output 1-bit DI 2R 1-bit DI push-pull MZI modulator output amplitude φ φ input voltage 34
DPSK Regenerator Using MZI EO Modulator 10Gb/s NRZ-DPSK experiment LD φ φ output input G DI G TI Input phase noise is transformed to amplitude noise by DI, but it is suppressed by cosδφ response of the DI and balanced detector (for DPSK format). Amplitude noise is suppressed by MZI. unregenerated signal regenerated signal noise unloaded noise loaded 35
(D)QPSK Regenerator Using MZI EO Modulators DQPSK regenerator θ DI = π/4 optical IQ modulator input signal LD MZI Modulator MZI Modulator π/2 output signal θ DI = -π/4 QPSK coherent regenerator optical IQ modulator input signal LO 90 hybrid LD MZI Modulator MZI Modulator π/2 output signal 36
Coherent QAM Regenerator input signal LO 90 hybrid LD MZI Modulator MZI Modulator π/2 output signal Multi-level nonlinear transfer function will be realized more easily in the electronic domain? Integration will be expected. 37
Contents All-optical regeneration of (D)BPSK and (D)QPSK signals (1) Phase-preserving amplitude regeneration using saturation of FWM (2) PSK signal regenerators using amplitude regenerators Regenerators using opto-electronic device nonlinearities Application of MLSE for mitigation of regeneratorinduced signal distortion 38
Issues in Signal Regeneration All-Optical Regenerator nonlinear signal processing All-optical regenerator works well when symbols are isolated in time. When symbols are overlapped in time at the input of the regenerator, severe inter-symbol interference may appear because of the nonlinearity. Pulses are isolated in time. Regen. Pulses are overlapped by, e.g., dispersion. Regen.? Regeneration is successful. Regeneration will be unsuccessful. 39
Issues in Signal Regeneration In the case of dispersion-induced pulse broadening, this indicates that the transmission line must be precisely dispersion-compensated at the location of the regenerator, which requires strict management of the transmission line. The inter-symbol interference due to the symbol overlap, however, may be mitigated by suitable signal processing at the receiver, e.g. using MLSE (Maximum Likelihood Sequence Estimation). Performance analysis in terms of information rate, or channel capacity with modulation format fixed, considering channel memory will reveal the effectiveness of MLSE. 40
Information Rate Analysis Transmission system model 24spans, 2400km DCF OBPF x Ns x NR TX 2R RX 40Gb/s NRZ-OOK 2R transfer function 100km NZDSF (2ps/nm/km, 2/W/km) Direct detection Low-pass filter 1 sample/symbol Pout 0 Pin P0 [ ] (P in # P 0 ) $ P out = P sin 2 0 "P in (2P 0 ) % & P 0 (P in > P 0 ) (Transparent in optical phase) 41
Information Rate Analysis Information rate (mutual information) I = H(Y) " H(Y X) Input symbol train Output sample train In this analysis: X = (X 1,X 2,...,,X n ) Y = (Y 1,Y 2,...,,Y n ) H(Y), H(Y X)Entropies per symbol The input symbols X i s are chosen from either 0 or 1. (Binary modulation is assumed.) The output samples Y i s are real variables. Channel memory is considered. (Y i is influenced by X i-m,.., X i,.., X i+m.) 42
Results of Information Rate Analysis 40Gb/s, NRZ-OOK, 100km x 24spans (1) Dispersion is 100% compensated at every span. IR vs Signal power Blue curves: 2R regenerators are not used. Red curves: 2R regenerators are inserted every 4 spans. : no memory considered (m=0) : m=3 43
Results of Information Rate Analysis Received eye patterns P sig =0dBm, IR= 1.0 (both with m=0 and 3) P sig =5dBm, IR= 0.7 (m=0), 1.0(m=3) 44
BER Performance Using MLSE BER vs Signal power Blue curves: 2R regenerators are not used. Red curves: 2R regenerators are inserted every 4 spans. : Data decision is made symbol by symbol (m=0). : MLSE is performed with m=3 (128states). 45
Results of Information Rate Analysis 40Gb/s, NRZ-OOK, 100km x 24spans (2) Span dispersion is not fully compensated. Accumulated residual dispersion is compensated at RX. (Transmission fiber nonlinearity is disregarded.) IR vs Residual dispersion per span Blue curves: 2R regenerators are not used. Red curves: 2R regenerators are inserted every 4 spans. IR is decreased because of the regenerator-induced signal distortion. 46
BER Performance Using MLSE BER vs Residual dispersion per span 2R regenerators are inserted every 4 spans. m=0: Data decision is made symbol by symbol. m=1: MLSE (8 states) m=2: MLSE (32 states) m=3: MLSE (128 states) BER degradation caused by regenerator-induced signal distortion can be mitigated by MLSE to some extent. 47
Conclusion All-optical regeneration of (D)BPSK and (D)QPSK signals Phase-preserving amplitude regeneration using saturation of FWM PSK signal regenerators using amplitude regenerators Regenerators using opto-electronic device nonlinearities Application of signal processing such as MLSE for mitigation of regenerator-induced signal distortion 48