A wavelength-preserving photonic integrated ... - OSA Publishing

A wavelength-preserving photonic integrated regenerator for NRZ and RZ signals Nicola Andriolli,* Stefano Faralli, Francesca Bontempi, and Giampiero Contestabile Scuola Superiore Sant’Anna, Via Moruzzi 1, 56124 Pisa, Italy *[email protected]

Abstract: This paper presents a novel Indium Phosphide based photonic integrated circuit (PIC) for all-optical regeneration of both nonreturn-tozero (NRZ) and return-to-zero (RZ) on-off-keying (OOK) signals. The PIC exploits cross gain compression in two semiconductor optical amplifiers to simultaneously obtain a wavelength-preserved and reshaped copy, and a wavelength-converted yet inverted copy of the input signal. Regeneration of 10 Gb/s signals on multiple wavelengths is demonstrated, showing a Qfactor improvement from 1.5 to 4 for NRZ-OOK signals and from 2.3 to 3.6 for RZ-OOK signals, and a BER improvement up to 1.5 decades. ©2013 Optical Society of America OCIS codes: (250.5300) Photonic integrated circuits; (200.6015) Signal regeneration; (250.5980) Semiconductor optical amplifiers.

References and links 1.

S. Azodolmolky, M. Klinkowski, E. Marin, D. Careglio, J. Pareta, and I. Tomkos, “A survey on physical layer impairments aware routing and wavelength assignment algorithms in optical networks,” Comput. Netw. 53(7), 926–944 (2009). 2. O. Leclerc, B. Lavigne, E. Balmefrezol, P. Brindel, L. Pierre, D. Rouvillain, and F. Seguineau, “Optical regeneration at 40 Gb/s and beyond,” J. Lightwave Technol. 21(11), 2779–2790 (2003). 3. M. Matsumoto, “Fiber-based all-optical signal regeneration,” IEEE J. Sel. Top. Quantum Electron. 18, 738–752 (2012). 4. E. Ciaramella, “Wavelength conversion and all-optical regeneration: achievements and open Issues,” J. Lightwave Technol. 30(4), 572–582 (2012). 5. J. Suzuki, T. Tanemura, K. Taira, Y. Ozeki, and K. Kikuchi, “All-optical regenerator using wavelength shift induced by cross-phase modulation in highly nonlinear dispersion-Shifted Fiber,” IEEE Photon. Technol. Lett. 17(2), 423–425 (2005). 6. J. De Merlier, G. Morthier, P. Van Daele, I. Moerman, and R. Baets, “All-optical 2R regeneration based on integrated asymmetric Mach-Zehnder interferometer incorporating MMI-SOA,” Electron. Lett. 38(5), 238–239 (2002). 7. G. Contestabile, R. Proietti, N. Calabretta, and E. Ciaramella, “Cross-gain compression in semiconductor optical amplifiers,” J. Lightwave Technol. 25(3), 915–921 (2007). 8. http://paradigm.jeppix.eu 9. N. Andriolli, F. Bontempi, S. Faralli, E. Ciaramella, and G. Contestabile, “A novel photonic integrated regenerator,” in OFC/NFOEC2013 Tech. Dig., Los Angeles, CA, Mar. 17–21, 2013, paper OW3J.4. 10. G. Contestabile, M. Presi, R. Proietti, and E. Ciaramella, “Optical reshaping of 40-Gb/s NRZ and RZ signals without wavelength conversion,” IEEE Photon. Technol. Lett. 20(13), 1133–1135 (2008). 11. G. Contestabile, M. Presi, R. Proietti, N. Calabretta, and E. Ciaramella, “A simple and low-power optical limiter for multi-GHz pulse trains,” Opt. Express 15(15), 9849–9858 (2007). 12. 86100A/B/C Inniium DCA, Agilent [Online]. Available: http://www.home.agilent.com/agilent/editorial.jspx?cc=US&lc=eng&ckey=98996

1. Introduction Advancements in the field of wavelength-switched optical networks aim to keep the signal in the optical domain as far as possible, to minimize the issues related to electronic processing and switching. However, after traveling long fiber spans and crossing multiple optical amplification and switching nodes, the quality of transmission is degraded, therefore signal regeneration is required. Optoelectronic regeneration, where the optical signal is converted in the electronic domain for the regeneration and then converted back for further transmission, presents challenges in cost, heat dissipation, power consumption, physical footprint, and

#191830 - $15.00 USD Received 5 Jun 2013; revised 5 Aug 2013; accepted 9 Aug 2013; published 27 Aug 2013 (C) 2013 OSA 9 September 2013 | Vol. 21, No. 18 | DOI:10.1364/OE.21.020649 | OPTICS EXPRESS 20649

operation and maintenance costs [1]. On the other hand, all-optical regeneration allows to amplify, reshape and, in some cases, perform retiming of the degraded signals directly in the optical domain, thus avoiding the use of optoelectronic regenerators. BPF2 (λ2)

Input signal (λ1)

SOA1

Coupler

SOA2

Signal Inverter

In CW (λ2)

MMI 1x2

BPF1 (λ1)

ODL VOA

Out

MMI 2x2

Monitor

VOA

Test SOA

Reshaped signal (λ1)

Monitor

Wavelengthconverted signal (λ2)

MMI

SOA2

In

BPF1

Out

BPF2

MMI

SOA1

Photonic Integrated Circuit (PIC)

(a)

(b)

Fig. 1. All-optical regenerator. (a) Scheme. (b) Picture of the fabricated PIC (1 × 6 mm2). 1

Normalized Probe Transmission

Output power (dBm)

10 5 0 -5 -10 -15 -30 -25 -20 -15 -10 -5 0 5 10 Input power (dBm)

15

0.95 0.9 0.85 0.8 0.75 0

20

50 100 150 200 250 300 350 400 Time (ps)

(b)

(a)

Fig. 2. (a) Fiber-to-fiber input/output power for a signal at 1550 nm crossing the test SOA, driven with a current of 150 mA. (b) Recovery time (1/e) of the integrated semiconductor optical amplifier measured by the pulse and probe technique. 0

Normalized output (dB)

Normalized output (dB)

0

-5

-10

-15

-20 1530

BPF1 BPF2 SOA ASE 1535

-5

-10

-15 BPF1 BPF2

1540

1545 1550 Wavelength (nm)

(a)

1555

1560

-20 1530

1535

1540

1545 1550 Wavelength (nm)

1555

1560

(b)

Fig. 3. Normalized spectra. (a) BPF1, BPF2, and ASE from test SOA located in the first test chip. (b) BPF1 and BPF2 located in the second test chip.

A number of schemes for all-optical 2R and 3R regeneration have been proposed in the literature [2–4]. Some of them exploit non-linear effects in fibers [5]: these solutions require large optical powers and are not suitable for integration, thus they cannot overcome power consumption and footprint issues typical of optoelectronic solutions. Other schemes exploit semiconductor optical amplifiers (SOA) as non-linear interferometers: these configurations are intrinsically unstable even if photonic integrated, thus they cannot be easily exploited [6]. In addition, most of these schemes are wavelength converters with a step-like transfer function [4], providing only a wavelength-converted reshaped copy of the input signal. This is often undesirable in wavelength-switched optical networks, when the same carrier wavelength has to be used on successive hops.


In this work we present a novel Indium Phosphide photonic integrated circuit (PIC) for all-optical regeneration that is wavelength-preserving and does not rely on interferometric effects. The PIC exploits two SOAs: in particular, the first SOA creates a wavelengthconverted and inverted copy of the input signal by means of cross gain modulation. Then the second SOA is fed with both the input signal and its inverted copy: by ensuring an almost constant power at the SOA input, the cross gain compression (XGC) effect (related to gain compression and propagation in the strongly saturated SOA) can simultaneously reshape both traveling signals. The scheme was first demonstrated in [7] using discrete components. The integrated device has been designed and fabricated within PARADIGM [8], the European advanced generic integration platform for the manufacturing of Indium Phosphide PICs. The working principle of the fabricated PIC and preliminary results have been previously reported in [9]. In this paper the proper operation of the realized photonic integrated regenerator, is demonstrated with both non-return-to-zero (NRZ) and return-to-zero (RZ) on-off-keying (OOK) signals at 10 Gb/s. The performance of the regenerator has been evaluated in terms of bit error rate (BER) and Q-factor at different emission wavelengths of the modulated and inverted signals. A Q-factor improvement from 1.5 to 4 has been observed on NRZ-OOK signals, and from 2.3 to 3.6 on RZ-OOK signals, for different optical signal to noise ratios (OSNR). Furthermore, BER measurements show a reduction of up to 1.5 decades. 2. Device description and fabrication The realized device implements the scheme sketched in Fig. 1(a). The PIC input port (In) is fed with the degraded input signal at wavelength λ1 coupled with a continuous-wave (CW) signal at wavelength λ2. After entering the PIC, a 1 × 2 multi-mode interference (MMI) splitter sends the signals in two arms. The upper arm acts as a signal inverter by means of cross gain modulation in SOA1, producing an inverted copy at λ2 of the input signal: the latter is then filtered by a band-pass filter (BPF2) centered at λ2. In the lower arm BPF1 centered at λ1 removes the CW at λ2, then an optical delay line (ODL) and a variable optical attenuator (VOA) adjust the power and synchronize this signal with the inverted copy at λ2 traveling along the upper arm. Then a 2 × 2 MMI couples the two complementary waveforms, giving an almost constant overall power at its output (this condition can be verified by inspecting the PIC monitor port). By entering SOA2 with an almost constant power, XGC takes place, thus saturation and limiting amplification can be exploited without any pattern effect [7]. In particular, the SOA saturated gain depends on the gain compression due to the input power and acts as power equalizer of the “ones” amplitude; this effect can be described as high-pass filter of noise for the “ones” amplitude with a cut-off frequency inversely proportional to the SOA gain recovery time [10,11]. Besides, the “zeros” with a specific wavelength allocation can experience compressed gain induced by the copropagating “ones” of the inverted signal: indeed, if the second inverted signal is emitting at a longer wavelength λ2, it produces an amplifier gain tilting toward longer wavelengths and makes a negative differential gain at λ1, that provides an attenuation of noise on the “zeros”. This effect mainly depends on the SOA length and the relative wavelength allocation of the two signals respect to the SOA gain peak [7]. At the PIC output port (Out) a wavelength-preserved reshaped signal on λ1 is obtained, together with a wavelength-converted (yet inverted) copy on λ2. The PIC, shown in Fig. 1(b), was fabricated by Oclaro Technology plc, UK, in a multiproject wafer run of the PARADIGM platform. The platform model enables a generic foundry model, where the process is application-blind and a small set of building blocks are used as basic elements. The portion of the chip with the PIC has a size of 1 × 6 mm2: both short facets are anti-reflection coated and can be used for butt coupling. Tapered fibers with a focused spot size of 2 μm are used to couple light to and from the PIC. High-contrast passive waveguides with a width of 1.5 μm are used, characterized by a minimum bending radius of 150 μm and a loss of around 5 dB/cm. Active elements exploit 2-μm-wide low-contrast waveguides, connected to passive waveguides with proper transition sections. 50-μm-long VOAs and 1-mm-long SOAs have been designed and realized. Figure 2(a) reports the fiber-


to-fiber input/output power characteristic (including all the chip and coupling losses) measured on a test SOA (on the same chip) driven with a current of 150 mA, showing the hard-limiting behavior of the SOAs. SOA recovery time (1/e) has been measured in pulse and probe configuration and is around 80 ps, as shown in Fig. 2(b). BPF1 and BPF2 were realized by means of arrayed waveguide gratings (AWG), designed with a free spectral range of 10 nm and a central frequency of 1550 nm and 1555 nm, respectively. Figure 3(a) presents the normalized SOA ASE spectrum (with the peak around 1534 nm). Figures 3(a) and 3(b) also show the pass-band characteristics of BPF1 and BPF2 for two different test chips utilized in the experimental analysis. 3. Experimental results 3.1 Experimental set-up The experimental setup is sketched in Fig. 4. A 10 Gb/s NRZ-OOK signal is generated using a tunable laser at wavelength λ1 connected to a Mach-Zehnder intensity modulator (MZM) driven by a 231-1 Pseudo Random Bit Sequence. The signal is then combined with ASE noise from an EDFA source, after crossing a BPF centered at λ1 and a VOA. The CW light at λ2 is generated by a tunable laser. To properly set the optical power and the polarization of the two waveforms at λ1 and λ2, an EDFA, a BPF, and a polarization controller (PC) are utilized on each arm before coupling them into the PIC. The total input optical power into the PIC is set to 8 dBm; SOA1, SOA2, and VOA are fed with a current of 110 mA, 180 mA, and 10 mA, respectively. The signal collected from the PIC (around 5 dBm of power) passes a BPF that can be tuned to either λ1 or λ2 and is then detected alternatively by a sampling oscilloscope with a 30 GHz optical head for Q-factor analysis, or by a 10 GHz APD photodetector for BER test. Figure 4 also shows the eye diagrams of the NRZ modulated input signal, of the two corresponding output signals (the reshaped one and converted one) in absence of ASE noise loading, and an output optical spectrum trace before filtering. Input signal (λ1)

Reshaped signal (λ1)

90

Noise loading

PC EDFA

10

BPF

0 -10

λ2

Wavelength-converted signal (λ2)

λ1

-20 -30 -40 -50 -60 1535

VOA EDFA

Normalized power (dB)

10 Gb/s transmitter (λ1)

1540 1545 1550 Wavelength (nm)

1555

BPF

Scope

PIC TL (λ2)

PC EDFA

BPF

BPF BER

Fig. 4. Experimental setup. Insets: eye diagrams of the NRZ signals without noise loading on a time scale of 20 ps/div and normalized output spectrum.

3.2 Regeneration of non-return-to-zero signals The regeneration of a 10 Gb/s NRZ-OOK signal λ1 at 1554.2 nm has been implemented by combining the signal with a CW light λ2 at 1541 nm exploiting the pass-band characteristics of the BPF1 and BPF2 reported in Fig. 3(a). We first assess the Q-factor evolution of the NRZ signal as function of the input OSNR. The Q-factor has been chosen since it allows to show in a compact way the noise redistribution performed by the regenerator, aimed at opening the eye diagram and thus preventing the occurrence of errors on further transmission links. Q-factor is obtained by using the signal-to-noise function of the sampling oscilloscope [12].


The Input-Output Q-factor results are summarized in Fig. 5(a), while the corresponding eye diagrams for three OSNR values are shown in Fig. 5(b). In all cases a net Q-factor improvement can be appreciated, ranging from 1.5 to 4. This eye opening is mainly due to the amplitude limiting and noise compression on the “one” level, which can be clearly appreciated in the eye diagrams of Fig. 5(b). It should also be considered that reshaped signals were suffering slight power instabilities due to vibrations in the fiber coupling set up. The impact of the all-optical regenerator on the BER of NRZ signals has been then assessed. Even though the real application scenario of the device would be in the middle of a transmission line (where the regenerative properties would prevent error accumulation on successive spans), for simplicity we test the all-optical regenerator in front of the error detector, where BER improvements can still be found due to a non-optimized receiver chain. Q = 4.5

IN

OSNR

Q = 6.6

IN

IN

Q = 7.8

40 dB 29 dB

26 dB

15.9 ps/div

OUT

23 dB

20 ps/div

20 ps/div

Q=6

OUT

Q=9.7

OUT

Q=11.8

20 dB

20 ps/div

20 ps/div

OSNR = 20 dB

20 ps/div

OSNR = 26 dB

OSNR = 29 dB

(b)

(a)

Fig. 5. Regeneration of NRZ signals. (a) Input/output Q-factor evolution for different OSNR (measured on 0.1 nm RBW). (b) Input (upper line) and output (lower line) eye diagrams for different OSNR values. B to B Input OSNR=20dB Input OSNR=18dB Input OSNR=15dB

-Log(BER)

2

Regenerated Regenerated Regenerated Regenerated

4

6 8 10 12

-30

-29

-28

-27 -26 -25 -24 Input Power (dBm)

-23

-22

Fig. 6. BER vs. Input Power of the NRZ signal before (solid points) and after (open points) all optical regenerator at different OSNR (measured on 0.1 nm RBW).

Figure 6 shows the measured BER vs. the optical signal input power without (solid points) and with (open points) all-optical regeneration, for different OSNR values: we observe a null power penalty of the regenerator at 10−9 BER for the signal in back-to-back without noise loading. A clear BER improvement with a correspondent lowering of all the BER floors is measured for all the noisy input signals with OSNR values ranging from 15 to 20 dB. 3.3 Regeneration of return-to-zero signals The experimental setup for the regeneration of 10 Gb/s RZ-OOK resembles the one shown in Fig. 4, where the transmitter now includes a first MZM connected to a clock source for producing a continuous stream of RZ pulses, and a second MZM which modulates the RZ pulse amplitudes in accordance with the bit sequence generated by the BPG. We use a


Output Q-Factor

different test chip and according to the normalized spectra of the BPF1 and BPF2, shown in Fig. 3(b), we set the regeneration wavelength λ2 at 1552 nm, and the RZ signal wavelength λ1 at 1538.4 nm, 1547.9 nm, and 1557.7 nm for testing multiple feasible input wavelengths. The Q-factor evolution of the RZ signal at λ1 = 1538.4 nm as a function of the input OSNR is presented in Fig. 7(a), while the corresponding eye diagrams are shown in Fig. 7(b): even in this case a net Q-factor improvement can be appreciated, ranging from 2.3 to 3.6. Eye diagrams confirm the limiting effect and the noise compression on the “one” level of RZ signals, notwithstanding the slight power instabilities due to the coupling. Figure 8(a) shows the BER vs. the optical signal input power without (solid points) and with (open points) alloptical regeneration, for different OSNR values at λ1 = 1538.4 nm. The regenerator introduces no power penalty to the RZ signal at 10−9, and a clear BER improvement and BER floor lowering is attained for all noisy input signals with OSNR in the range from 16.5 to 20 dB. Finally the regenerator wavelength dependence is studied by performing a BER measurement for different input signal wavelengths λ1 at a fixed OSNR = 20 dB. Figure 8(b) shows a consistent BER reduction for all tested input wavelengths, confirming the effectiveness of the proposed regenerator on a wide wavelength range. 14 13 12 11 10 9 8 7 6 5

Q = 3.9

IN

IN

Q = 5.4 IN

Q = 6.9

20 ps/div

20 ps/div

29 dB

OSNR 23 dB 20 dB 20 ps/div

18.5 dB OUT

Q=8.9 OUT

Q=6.2 OUT

Q = 10.5

16 dB

3

4

5

6 7 8 Input Q-Factor

9

10

11

20 ps/div

20 ps/div

OSNR = 16 dB

(a)

OSNR = 18.5 dB

20 ps/div

OSNR = 20 dB

(b)

Fig. 7. Regeneration of RZ signals. (a) Input/output Q-factor evolution for different OSNR (measured on 0.1 nm RBW). (b) Input (upper line) and output (lower line) eye diagrams for different OSNR values. OSNR=20dB OSNR=18.5dB OSNR=17.5dB OSNR=16.5dB

Input Input Input Input BtoB

2

Reg Reg Reg Reg Reg

4

-Log (BER)

-Log (BER)

2

6

6 8

10

10 -29

-28

-27 -26 -25 -24 Input Power (dBm)

(a)

-23

-22

Reg Reg Reg

4

8

12

noiseless B2B Input 1557.7 nm Input 1547.9 nm Input 1538.4 nm

12

-29

-28

-27 -26 -25 Input Power (dBm)

-24

-23

(b)

Fig. 8. BER vs. input power of the RZ signal before (solid points) and after (open points) all optical regeneration. (a) Different OSNR (measured on 0.1 nm RBW) with input signal at 1538.4nm. (b) Different signal wavelengths with OSNR = 20dB (measured on 0.1 nm RBW).


4. Conclusion A novel Indium Phosphide photonic integrated circuit for wavelength-preserving regeneration has been designed and fabricated using the PARADIGM integration platform. Q-factor and BER measurements show the effectiveness of the circuit for the regeneration of both NRZOOK and RZ-OOK signals at 10 Gb/s. Proper operation has been demonstrated at multiple signal input wavelengths. Moreover, due to its working principle and previous results using bulk devices [9], the same circuit is also suitable for operation at higher bit rates provided that SOA recovery time are correspondingly shortened. Acknowledgments Prof. Ernesto Ciaramella is acknowledged for setting up the collaboration within PARADIGM Applications Group and for the helpful discussions. The partners of the PARADIGM platform are acknowledged for their support and for the PIC fabrication. PARADIGM is an ITC funded FP7 project (257210). This work was partially supported by the Regione Toscana through the ARNO T3 project, POR CReO FESR 2007-2013, PAR FAS 2007-2013.
