1. Introduction

The wavelength conversion in wavelength division multiplexed optical networks is a key function for proper network traffic management. For this application, all-optical wavelength conversion schemes have been largely studied in recent years exploiting a number of different techniques, but, they never reached enough technological maturity to be used in installed transmission systems. Indeed, a number of practical drawbacks are common to most of the proposed schemes: low conversion efficiency, wavelength dependence, large operating optical power, modulation format dependence, stability issues and others. On the other hand, the introduction in the market of coherent optical transmission systems makes even more desirable the development of reliable and transparent wavelength conversion techniques. In this way, a single all-optical element could be used instead of a pair of complex coherent receiver and transmitter, thus reducing, at the same time, opto-electronic component count and overall power consumption of the network nodes.

Quantum Dot Semiconductor Optical Amplifiers (QD-SOA) have been studied and developed in order to overcome most of the limitations of conventional bulk and quantum-well SOAs. In fact, they show extremely promising features and, we have recently demonstrated efficient high-speed all-optical processing of intensity modulated signals using QD-SOAs [1–3].

Novel coherent modulation formats used nowadays require instead the use of coherent non linear techniques for all-optical processing [4–6]. Among the various techniques, Four Wave Mixing (FWM) in QD-SOA has been exploited in recent years for wavelength conversion of phase modulated signals [7, 8]. In this paper, we experimentally demonstrate that thanks to QD-SOA broadband and extremely large gain, it is possible to obtain coherent wavelength conversion via FWM in the spectral region between 1470 and 1570 nm with positive conversion efficiency using only few dBm input optical pump power. In addition, we show the complete transparency of the scheme to 10 GBaud coherent modulation formats, including Quadrature Phase Shift Keying (QPSK), 8- Phase Shift Keying (PSK), 16- Quadrature Amplitude Modulation (QAM) and Orthogonal Frequency-Division Multiplexing (OFDM) with 16 QAM sub-channels.

2. QD-SOA and FWM characterization

The QD-SOA used in the experiment, whose schematic structure is reported in Fig. 1 , is a 6.15 mm long device with 8° tilted waveguide, antireflection coated and made by 22 stacks of InAs QDs and InGaAsP-tensile barrier layers. The dot layers are sandwiched between 160-nm-thick InGaAsP separate confinement hetero-structure layers. The density of the dots is 9x10¹⁰ cm⁻², height of the dots is 25 nm and total active region thickness about 350 nm. The device has a strain-controlled columnar QD structure [9] and is transverse magnetic (TM) mode gain dominated even if it shows also significant transverse electric (TE) gain. The device was thermally controlled by a Peltier cooler and custom packaged and pigtailed with polarization maintaining input and output fibers. It was operated at 22°C. At 1.85 A bias current it has the fiber-to-fiber small signal gain reported in Fig. 2(a) and measured for an input power of −23 dBm. The gain peak is at 1480 nm and maximum gain is 41 dB for the TM mode and 30 dB for the TE mode.

 

Fig. 1 Schematic image of a columnar QD-SOA structure

Download Full Size | PDF

 

Fig. 2 (a) TM and TE Small signal gain; (b) Optical sampling oscilloscope trace showing device gain recovery.

Download Full Size | PDF

The 3-dB bandwidth is 65 nm, and bandwidth with more than 20 dB gain is more than 120 nm for the TM mode (our experimental set-up was limited to 1445 nm minimum operating wavelength). Complete (100%) gain recovery time at 1550 nm is 30 ps (see Fig. 2(b)) following the peculiar triple exponential trend reported in [10]. Output saturation power is in the range of 19-21 dBm for the different operating wavelengths (1470-1570 nm). While conventional single pump FWM exhibits, even in QD-SOAs, the well known conversion efficiency degradation at increasing the conversion detuning [4], dual pump FWM is a scheme for obtaining broadband wavelength conversion in SOAs without suffering from this large degradation [11, 12]. In common SOAs, however, broadband operation is limited by the device gain profile to typically 30-40 nm [11].

Indeed, within this scheme, the conversion efficiency mainly depends on the fixed wavelength detuning between pump 1 (P1) and signal (S). Efficiency is then almost flat as pump 2 (P2) is moved in the flat saturated part of the gain spectrum of the SOA. Indeed, around P2, two copies of the original signal are generated; S2 and FWM2 (see Figs. 3(b) , 3(c)). S2 has the same phase as S, FWM2 is phase conjugated. The conversion efficiency of the two copies is different and, due to physical reasons related to the interplay between the nonlinear effects in the QD-SOA, it depends on the sign of the P1-S detuning. In Fig. 3(a), we report typical conversion efficiency (measured with CW signals) and relative Optical Signal to Noise Ratio (OSNR on 0.1 nm) trend for the wavelength conversion from the input signal S to the converted S2 with the same phase. Those results are for the case λs = 1510 nm, λp1 = 1511.5 nm and using the following input signal powers: P1 = 3 dBm, P2 = 3 dBm and S = −5 dBm. All the signals have same polarization set on the TM mode of the device in order to maximize the FWM interactions.

 

Fig. 3 (a) Dual pump FWM conversion efficiency and output OSNR; Example spectra (relative power) of (b) 36.5 nm down-conversion, (c) 48.5 nm up-conversion

Download Full Size | PDF

Due to the highly efficient S-P1 beating and to the broadband QD-SOA gain, 100 nm wavelength conversion spanning the entire C- and S-bands is obtained with positive conversion efficiency (we use the conventional conversion efficiency definition: converted signal power/input signal power) and more than 30 dB output OSNR. Noteworthy, 140 nm conversion span is also possible with more than −10 dB conversion efficiency. Two optical spectra showing a 36.5 nm wavelength down-conversion and a 48.5 nm up-conversion are reported as examples of the broadband operation in Figs. 3(b)-3(c) (note that power levels in the spectra are not absolute values). Similar conversion efficiencies can be obtained for different input signal wavelengths lying in the saturated gain spectrum of the QD-SOA, i.e. the region 1470-1570 nm.

3. Wavelength conversion of coherent signals

As FWM is a coherent wavelength conversion technique, it is capable, in principle, to work with any kind of advanced multilevel coherent modulation format. In order to test this transparency to coherent signals, we employed the set-up of Fig. 4 . We used a programmable 10 Gbaud transmitter made by a 100 kHz line-width tunable laser source modulated by a dual-drive dual-parallel Mach-Zehnder modulator directly driven by a 12 GHz arbitrary waveform generator (AWG). The programmable transmitter can generate in turn QPSK, 8-PSK, 16-QAM and OFDM signals. In this way 20, 30 and 40 Gbs signals were generated. In particular the 40 Gbs OFDM signal was made by 128 subchannels, 79 MHz spaced and modulated with 16-QAM signals. For all formats, 9% cyclic prefix is appended for the channel equalization. The coherent signals were coupled with two tunable lasers having 500 kHz line-width that act as pumps. The signals were polarization aligned and injected in the QD-SOA. The device output was filtered by a wavelength and bandwidth tunable filter with 50 dB out-of-band suppression-ratio and sent to the pre-amplified coherent heterodyne receiver where an optical attenuator was used to adjust the received OSNR. Heterodyne coherent receiver was made by a tunable laser local oscillator with 100 kHz line-width, a 90°-hybrid and balanced photodetectors. The receiver outputs were sampled with a 20 GHz, 50 GSample/s real time oscilloscope (DSO).

 

Fig. 4 Coherent conversion experimental set-up. TL = Tunable Laser; AWG = Arbitrary Waveform generator; IQMOD = Phase-Quadrature Modulator; Coh. Rx = Coherent Receiver; DSO = real time oscilloscope; LO = Local Oscillator.

Download Full Size | PDF

The digital signal processing, including carrier frequency offset compensation [13] and frequency domain equalization [14], are performed off-line based on the channel state information (CSI) acquired via the preambles appended on the top of each transmitted data block. The preamble has constant amplitude so that CSI has not any significant impact on the SOA gain modulation. As the coherent transmission system operating range was limited to the C-band, we set the input signal at 1545 nm and tested the wavelength conversion operation in the whole band, from 1530 to 1565 nm. We used the same input powers for all the modulation formats as in the following, P1 = P2 = 4 dBm, and S = −7 dBm. Corresponding flat conversion efficiency at the SOA output was + 4 dB.

The results are summarized in Fig. 5 : We compare there the input and output signal Error Vector Magnitude (EVM) at a fixed OSNR for all the different modulation formats and for wavelength conversion operation spanning all the C-band. We set at the receiver input for all the signals an OSNR = 25 dB (on 0.1 nm) corresponding to FEC error-free operation for all the modulation formats (as a reference, it corresponds to 1.8*10⁻⁵ BER for 16-QAM). Input and output constellations for the larger conversion detuning are also reported in the insets for comparison. OFDM channel by channel EVM and a relative constellation are reported separately in Figs. 5(b)-5(d). All the modulation formats have similar input EVM~11.1 ÷ 11.4%.

 

Fig. 5 (a) Input and output EVM values for wavelength conversion in the C-band of QPSK, 8-PSK, 16-QAM and OFDM-16-QAM signals. OSNR = 25dB (on 0.1 nm resolution bandwidth) for all the signals. QPSK, 8-PSK and 16-QAM constellations for the input and output signals at the larger conversion detuning are reported in the inset. OFDM channel by channel EVM and relative constellation for, (b) the input and, (c) and (d), the converted signals.

Download Full Size | PDF

As can be seen, not any constellation distortion and EVM degradation can be appreciated for the conversion of QPSK and 8-PSK signals in the whole band. Only a maximum 1.3% degradation was measured for the 16-QAM signal, while OFDM was the format performing worst having a maximum 3.4% degradation. As we were using the same input power levels (same pump/signal mean power ratio) for all the modulation formats, this difference in conversion degradation can be mainly attributed to the different peak-to-average power ratios (PAPR) in the different cases. In particular, QPSK and 8-PSK are almost constant envelope signals, 16-QAM has around 4 dB PAPR while OFDM can have a maximum of 12 dB PAPR depending on the particular sequence employed. Therefore, in the case of this last format some excess gain and phase modulations due to the peak powers give undesired signal distortions. An adaptation/optimization (not made here) of the pump power levels, in particular an increase of pump/signal ratio, is expected to reduce the EVM degradation also for OFDM.

4. Conclusions

We have reported what is, to the best of our knowledge, the most efficient broadband wavelength conversion ever demonstrated. The scheme is based on two pumps FWM in an ultra high-gain broadband QD-SOA. 100 nm span wavelength conversion with positive efficiency using low optical pump power, and having more than 30 dB output OSNR has been shown. Conversion transparency for coherent multilevel modulation formats has also been demonstrated in the entire C-band for coherent m-PSK, m-QAM and OFDM modulation formats with only limited EVM degradation. As the FWM process is polarization dependent, dual polarization signals will require the use of a polarization diversity set-up [4, 15].