Open Access
Add to BibSonomyAbstract
Optical injection locking characteristics of a distributed feedback laser are experimentally investigated for multiple-wavelength injection. Using a three-wavelength source generated by intensity modulation as the injected signal, it was found that the presence of adjacent lines could cause disturbance to the locking if a minimum guard band between the respective locking limits of two adjacent lines was not observed. With a 21-line comb with 20 GHz line spacing as the injected signal, the injection locking range was observed to become asymmetrical in relation to the laser free-running frequency under high-power injection conditions and was found to be dependent on whether the laser was locked to lines located at centre and or edges of the comb. Finally, the use of the injection locked laser as a seed laser for generating a local oscillator (LO) comb for superchannel coherent detection was investigated and the phase error between the input and LO combs analysed.
© 2015 Optical Society of America
1. Introduction
Multi-wavelength systems based on intensity modulation and direct detection (IM/DD) are very close to reaching saturation as electronic limitations on higher data rate modulation are reached and the number of wavelength division multiplexing (WDM) channels approaches the limit set by the required channel spacing and optical amplifier bandwidth. More spectrally efficient modulation formats are now being used to overcome some of these problems, allowing more capacity with the same available infrastructure. To accomplish this, coherent optical techniques are being employed [1–4]. Recent implementations of coherent optical receivers make use of digital signal processing (DSP) to recover information, and of polarization state multiplexing to further increase spectral efficiency. These digital coherent receivers use a local oscillator (LO) laser to recover the field of the transmitted optical signal, enabling transmission impairments, as well as polarization tracking, and phase and frequency error compensation, to be dealt with via DSP. As a result, transmission capacity has been considerably increased [5], although bit rates over 400 Gb/s using more refined and spectrally-efficient modulation formats, such as quadrature amplitude modulation (QAM), impose stringent demands on electronics. To ease these requirements and still attain high capacity transmission, superchannel schemes can be used, which employ a number of closely spaced sub-carriers, each operating at a lower rate. This approach requires multiple transmitters and LO lasers with close, well-defined frequency separation, and with narrow linewidths to compensate for algorithm assumptions regarding phase noise statistics.
A way to generate phase-synchronized and closely-spaced multi-wavelength signals with precise frequency spacing is by using optical frequency comb generators (OFCGs) [6–8]. These also allow a single laser to be used to generate carriers for a number of sub-channels, reducing component count and cost. The same approach can be replicated at the receiver to generate multiple LOs (Fig. 1). Phase tracking between the transmitted and LO combs can be accomplished in the optical domain by using one line in the transmitted comb as a pilot carrier (green line in Fig. 1). As a result, all the LOs in the receiver comb can be synchronized to their respective sub-channel carriers independently of the data rate, in principle allowing homodyne coherent detection of the transmitted signals without the overhead of phase tracking in DSP. Although using a self-coherent receiver scheme, in which the pilot is optically filtered to provide the seed for the LO comb, has been demonstrated to make DSP frequency offset compensation unnecessary [9], the use of optical phase-locking techniques, such as optical phase-lock loop (OPLL), optical injection locking (OIL), or optical injection phase-lock loop (OIPLL) [10,11], should improve phase tracking performance, ease DSP requirements further and allow the use of lasers with wider linewidths.
Fig. 1 Schematic diagram of an optical superchannel receiver scheme using a pilot carrier (green) to generate a comb of phase-locked local oscillators (LO OFCG) for homodyne coherent sub-channel demultiplexing, detection (IQ), and later DSP.
In this paper, an experimental investigation of the OIL characteristics of a DFB laser under single facet multi-wavelength injection is presented. Although optical filtering could be used to select the pilot signal before injecting it into the laser, this adds complexity, especially if the filter needs to be very narrow to select just the pilot carrier, in which case a tracking filter may be required. It is therefore of interest to understand the conditions under which OIL can be used to select the pilot carrier from a comb without pre-filtering. Since OIL is possible only within a certain bandwidth, it can be achieved for different wavelengths by tuning the free-running frequency of the injected (slave) laser accordingly. Locking to a given master signal occurs when the frequency difference between the slave laser and the master signal (detuning) is inside this bandwidth. After carrying out a single-wavelength OIL experiment to obtain injection locking parameters for the distributed feedback (DFB) laser under test, different approaches to multi-wavelength injection were then used to characterize the process. From OIL measurements using three intensity modulation generated lines, it was observed that a minimum frequency spacing (guard band) between the respective locking limits of two adjacent lines is necessary to avoid disturbances (locking crosstalk) that can prevent the laser from locking to the required line. Subsequently, OIL experiments were carried out using a frequency comb with 21 lines separated by 20 GHz, where the locking line (pilot) peak power was kept at the same level whilst those of the other lines were increased by up to 12 dB. With the pilot positioned at the centre and at the extreme right and left edges of the 21-line superchannel comb, locking asymmetry was observed as the injected power was increased, but also showing a dependency on the relative position of the pilot in the comb. The 20-GHz comb line spacing was enough to prevent locking crosstalk. Finally, the use of the locked laser as a seed source for generation of a LO comb was investigated. In this way, the LO comb lines are phase locked to those of the original comb. In order to prove the concept, the 21-line comb was set with the pilot at its long-wavelength extreme and the phase error between one of the LO comb lines and the equivalent line in the input comb was measured. Phase noise spectral density below −110 dBc/Hz was measured at 10 MHz offset or greater, with the measurement limited by LO-ASE noise.
2. Basic OIL experiment
Figure 2(a) shows the experimental arrangement for the OIL receiver used. The LO (slave) laser was a pigtailed bulk-active layer DFB laser supplied without an output isolator. Thus, injection locking was accomplished by single facet laser injection. At 20°C, the laser had a threshold current of around 13 mA, average static tuning sensitivity of 2.5 GHz/mA between 50 and 70 mA, side mode suppression ratio over 50 dB and linewidth just below 2 MHz at 60 mA. Part of the input signal was coupled into the LO laser after passing through a 3-dB optical coupler (C1), a variable optical attenuator (VOA), a polarization controller (PC2), and a circulator (CIR). The VOA controlled the amount of injected power into the LO laser and PC2 permitted input signal polarization alignment to that of the LO laser cavity. The LO laser output was combined with the other portion of the input signal after leaving the CIR using 3-dB optical coupler C2. The circulator was arranged to allow single facet injection and to provide isolation for the input-LO and LO-output paths. A second polarization controller (PC1) provided polarization matching at the receiver photodetector. The additional C2 port was used to monitor the output signal via an optical spectrum analyser (OSA).
Fig. 2 (a) Experimental arrangement for the OIL receiver and (b) measured locking limits as a function of the measured injection ratio.
For the basic OIL experiment, an external cavity laser (ECL) with picometre tuning resolution (Anritsu MG9638A) was used as the master signal. The measured FWHM linewidth of the master laser (ML) was below 500 kHz and its maximum nominal output power was 8 dBm. The ML was directly coupled into the input of the OIL receiver. The insertion loss of each component was measured and corrected for to determine the injection ratio. The VOA in Fig. 2 provided the injected power variation for the locking range characterization. By sweeping the ML wavelength around that of the LO laser, in this case, set at 1554.71 nm and output power just over 3 dBm (I = 58 mA and T = 20°C), the locking range was measured by simultaneously monitoring the optical spectrum and the electrical laser beating spectrum, via an OSA (Agilent 86140B) and a lightwave signal analyser (LSA, HP 71400C), respectively.
In order to obtain the limits of the locking band, the injected power was turned on and off via the VOA and the heterodyne signal frequency without injection (detuning) measured at the LSA. The nominal wavelength of the tuneable laser provided a second estimate of the locking range. From the OSA, it could be observed that the LO laser frequency was pulled to that of the master signal when the detuning lay within the locking range limits, especially for higher injection ratios. Also, inside the OSA resolution (0.06 nm), it was verified that the LO laser practically assumed the spectral shape of the master signal during locking, in particular for stable-lock detuning. At same time, the noisy electrical beating signal between both lasers shifted to baseband in the LSA, with the residual phase noise spectral shape being dependent on the injection level and detuning [10–12].
Figure 2(b) shows the measured locking range as a function of the measured injection ratio, defined as the ratio between the estimated ML power at the input of the LO laser and the LO laser output power, using the corrected readings from LSA and ML. It can be seen that, for the power level available and the experimental conditions, the maximum locking range possible was as wide as 17 GHz. For high injection conditions, unstable locking was observed during positive detuning (ML frequency > LO laser frequency) by means of severe changes in the LO laser optical spectrum and a noticeable frequency-varying peak in the electrical spectrum. This unstable regime is caused by strong changes in the LO laser gain dynamics and material birefringence caused by the excessive injection and must be avoided in laser synchronisation. Figure 2(b) also shows the theoretical injection locking limits, calculated based on the linearization of the injected laser rate equations [10–12]:
where fML is the ML frequency, fLO is the free-running LO laser frequency, c is the speed of light, ng is the group refractive index, α is the linewidth enhancement factor, L is the LO laser cavity length, Ie is the measured injection ratio, and η is a factor that accounts for coupling losses during the injection process. By assuming typical values for DFB semiconductor laser parameters (ng = 4.3, α = 5 and L = 300 μm) and reported injection locking experiments [10–12], η was estimated as 3.2%. Thus, Ie is 15 dB higher than the actual injection ratio (Ir = ηIe). For 0-dB VOA attenuation, Ie = −5.4 dB, Ir is −20.4 dB and ΔfOIL = 8.5 GHz. Although (1) was determined from linearized rate equations, it presents a good agreement with the experimental data up to the injection levels that were presented in Fig. 2(b).3. OIL experiment with adjacent lines
One of the characteristics of a superchannel is that closely-spaced subcarriers are used to transmit information. It is therefore of interest to investigate how susceptible a laser that is injection locked to a pilot carrier is to the influence of closely adjacent spectral lines. On the assumption that the lines immediately adjacent to the pilot carrier will have the biggest influence on the injection locking, locking to a three-line comb generated by intensity modulation was investigated. This allowed the subcarrier spacing to be easily adjusted by changing the frequency of the RF source driving the modulator. The three-line comb also serves as a useful model of a comb that has been pre-filtered using a relatively broad optical filter. The basic OIL experiment was modified by placing a polarization controller and an intensity modulator (JDSU OC-192) between the ECL and the input of the OIL receiver of Fig. 2(a). To compensate for the modulator insertion loss, an EDFA was positioned after the modulator, followed by a 3-nm bandwidth tuneable optical filter (Newport TBF-1550-3.0) to filter out the majority of the amplified spontaneous emission (ASE) noise present at the amplifier output. The bias voltage applied to the intensity modulator was adjusted to change the power of the modulation sidebands relative to that of the carrier (the sideband-to-carrier ratio, SCR). The polarization of the light at the modulator input was adjusted to maximise the modulation depth and fine-tune the SCR. Under these conditions, the three lines at the output of the modulator are co-polarised. A range of SCR values were investigated, but in the following we consider the cases where the sidebands have either the same peak power as the carrier (SCR = 0 dB) or 8-dB higher power (SCR = 8 dB). Figure 3 depicts the LO laser locking range distributions around carrier and sidebands as a function of the SCR and the VOA attenuation, as well as the amplitude distributions for carrier and sidebands used during the experiments described below. To enhance possible locking instabilities, the largest possible injection ratio was used. The modulation frequency was set at 10 GHz while the ML was tuned in such a way that the LO laser could potentially be locked to the carrier or to the sidebands. The LO laser was then biased at 57.6 mA (T = 20°C), corresponding to a wavelength of 1554.69 nm and −1.88 dBm at the OIL receiver output.
Fig. 3 LO laser locking range distributions to carrier and sidebands, having the sideband-to-carrier ratio (SCR) and the VOA attenuation as parameters: VOA = 0 dB, SCR = (a) 0 and (b) 8 dB; VOA = 7 dB, SCR = (c) 0 and (d) 8 dB.
By adjusting the peak power of the carrier and the sidebands to −10 dBm at the OIL receiver OSA output (SCR = 0 dB) and by analysing the results from both OSA and LSA simultaneously, it was possible to observe strong disturbances from the sidebands during the locking of the LO laser to the carrier, in agreement with results reported in [13]. Without modulation, locking to the carrier was measured to be between 1554.669 nm and 1554.721 nm, corresponding to a fully-stable locking range of 6.5 GHz (Ie = −13.6 dB). Practically the same limits were observed with modulation at 10 GHz. However, stable LO laser locking was only possible between 1554.690 nm and 1554.710 nm (2.5-GHz band). To illustrate that, Fig. 3 presents LO laser locking range distribution to carrier and sidebands, having SCR and VOA attenuation as parameters. The long arrows represent the relative zero-detuning positions in relation to carrier (centre) and sidebands, as well as their respective amplitude distributions. The boxes show the extension of the individual locking ranges (positive detuning to the left and negative detuning to the right of the arrows) and the divisions on the horizontal axis are scaled to 2 GHz spacing. In particular, Fig. 3(a) shows how the locking range to the carrier changes in relation to that without modulation, stressing the sideband-induced unstable (magenta) and asymmetric stable (green) regions for VOA and SCR = 0 dB.
Figure 4(a) presents an example of the LSA electrical spectrum in the unstable locking region, at + 1.8-GHz frequency detuning between the carrier and the LO laser, corresponding to the magenta small arrow in Fig. 3(a). The black trace shows the beat signal observed without (w/o) both modulation and injection. With (w/) injection but without modulation (i.e. carrier only injection), the beating disappears and stable locking is observed (red trace), with the phase noise considerably lower around baseband and the noise floor at higher frequencies matching that without modulation and injection. When both injection and modulation were applied, a large increase in the phase noise was observed throughout the frequency range (blue trace), indicating locking crosstalk. On the other hand, when considering locking to the sidebands, there was some difficulty defining the locking range limits closer to the carrier, but no apparent influence on the locking bandwidth was detected.
Fig. 4 Electrical spectra for LO laser locking to the carrier when the carrier peak power is set (a) at same level as the sidebands ( + 1.8 GHz detuning) and (b) 8 dB lower than that of the modulation sidebands ( + 300 MHz detuning).
The measured locking ranges were 6.8 GHz and 7.3 GHz for the sidebands to longer and shorter wavelengths, respectively. Figure 3(a) also illustrates both sideband locking regions. Next, the peak powers of the carrier and sidebands were adjusted to −16 dBm and −8 dBm at the OIL receiver OSA output, respectively, and the spectra simultaneously observed at both OSA and LSA. Figure 3(b) shows the equivalent diagram for this case (SCR = 8 dB). Figure 4(b) presents the electrical spectra for LO laser locking to the carrier at + 300-MHz detuning, which is also indicated (magenta arrow) in Fig. 3(b). Without modulation and injection, the spectrum (black trace) shows the beat signal and the noise floor of the LSA. For the situation with injection but without modulation, the beating disappears and stable locking is observed (red trace). Locking to the carrier was accomplished between 1554.680 and 1554.710 nm, which corresponds to a locking range of 3.7 GHz (Ie = −19.1 dB), keeping a similar spectral behaviour. When injection and modulation were applied, locking to the carrier was no longer possible for the previous wavelength range, with strong disturbances caused by the sidebands in the optical and electrical (phase noise) spectra being observed, in agreement with results reported in [13]. As can be seen in Fig. 4(b), for the same detuning of 300 MHz, there is a strong increase in the phase noise throughout the frequency span (blue trace), with the appearance of a resonance peak around 5 GHz. Similar spectral responses were observed for other detuning values. In contrast, when locking the LO laser to the sidebands was considered, the limits to the locking range were clearly definable because of the lower carrier peak power. No spectral disturbances were observed during locking and the measured locking ranges for the sidebands to longer and shorter wavelength of the carrier were 8.7 GHz and 10 GHz, respectively. Figure 3(b) also shows the sideband locking regions. It is important to point out that no locking range overlap was intended during both of the previous experiments. The main idea was to investigate if the sidebands were still able to influence the LO laser locking to the carrier when guard bands (frequency gaps between the locking limits of adjacent lines) were applied.
In Figs. 3(a) and 3(b), partially- (SCR = 0 dB) and completely- (SCR = 8 dB) unstable locking are observed, respectively, for guard bands of 3 to 4 GHz. This shows that the excess injected power from neighbouring lines can alter the LO laser gain dynamics, leading to non-linear effects that compromise locking stability to the reference line. Thus, to explore a possible reduction in the locking crosstalk, the VOA attenuation was increased from 0 dB and the effects on the LO locking to carrier and sidebands analysed. After reducing the injection ratio and, therefore, the total locking range for each line, it was possible to observe that, for some level of VOA attenuation, locking was possible with no further crosstalk present. A situation with no apparent interference was possible when the attenuation was set at 5 dB. Before that, if the limits of the estimated locking ranges of two adjacent lines were still too close, some degree of disturbance in the electrical spectrum prevented the locking limits from being clearly measured. Disturbance-free locking to any of the lines was possible for attenuation values over 7 dB. Figures 3(c) and 3(d) illustrate the distributions of the fully-stable LO laser locking ranges around carrier and sidebands for VOA = 7 dB, with SCR = 0 and 8 dB, respectively. The 7-dB drop in total injected power led to measured guard bands of over 6 GHz, with stable locking throughout the locking ranges. These results show clearly that a compromise between sideband separation and injected power must be considered during the design of a multi-wavelength injection OIL receiver to prevent locking crosstalk.
4. OIL experiment with optical comb
An optical frequency comb generator (OFCG) based on a stabilized optical fibre recirculation loop [14] provided the frequency comb signal for the next LO laser locking investigations. A narrow linewidth laser (<15 kHz, Redfern Integrated Optics Inc.) was used as the seed laser for the OFCG. To avoid locking crosstalk and to provide spacing typical of a superchannel system, the comb line spacing was set at approximately 20 GHz. By considering a possible system operating with 1.2 Tb/s superchannels and using 15 Gbaud QPSK dual-polarization transmission, the number of comb lines was assumed to be 20 for the OIL experiments. One extra line was necessary for locking purposes (pilot), since carrier suppression is a common characteristic for phase-modulated signals. Thus, with the 20-GHz line spacing, the superchannel bandwidth reached a little over 400 GHz and a system spectral efficiency of just under 3 b/s/Hz would be potentially possible with such a transmission scheme. After the amplification of the OFCG output signal by an EDFA, a programmable optical filter (POF, Finisar WaveShaper 4000S) was used to select the required comb lines, equalize their peak powers and filter out ASE before coupling to the OIL receiver input of Fig. 2(a). At the output, the peak power of each line, including the pilot, was set at approximately −25 dBm.
In order to keep the locking conditions similar to those of the previous investigations, the LO laser was biased at 55.9 mA, which sets its wavelength very close to one particular line of the comb (1554.647 nm). This line was assumed to be the pilot line and the POF was programmed so that the pilot line could be located at the centre, to the right, or to the left of the 20-line comb. Figure 5(a) presents the overlapped optical comb spectra for the three cases at the OIL receiver output. Frequency tuning over the locking bandwidth was accomplished through LO laser current, the LO laser power variation being low enough to prevent a significant effect on the locking bandwidth. Also, in order to investigate possible locking crosstalk and power distribution characteristics for the transmission, the pilot peak power was kept constant at −25 dBm, while the power of the other lines was increased in steps of 3 dB up to 12 dB via POF pilot attenuation and EDFA gain adjustments (line-to-pilot ratio, LPR).
Fig. 5 (a) Overlapped optical frequency comb spectra with the pilot to the left, to the right and at the center of the superchannel and (b) measured locking limits for the different superchannel pilot positions as a function of the increase in the peak powers of the comb lines (LPR).
Under these circumstances, the measured injection ratio for the pilot was about −26 dB (Ir = −41 dB), leading to a locking range around 1.7 GHz, which would be free from locking disturbances from adjacent lines due to the resulting wide guard band (over 16 GHz). Figure 5(b) shows the measured locking bandwidth for distinct pilot positions within the superchannel as a function of the LPR. As can be seen, increasing the line peak power does not significantly change the fully-stable locking bandwidth as the pilot peak power is kept the same for all measurements. However, the excessive amount of power coupled to the LO laser cavity tends to break the apparent initial locking bandwidth symmetry (especially in comparison to an initial case, when the POF selected only the pilot for locking analysis) and shift it towards the negative detuning region. This asymmetry in locking is expected [13], since the excess light coupled into the LO laser can vary its free-running (no injection) frequency, which is used as the reference for locking detuning measurements. Furthermore, for the levels of power used, the resulting asymmetry showed a distinct response to the position of the pilot in the superchannel, suggesting that not only will the power level and comb line spacing influence the receiver locking performance, but also the line within the comb that is chosen as the pilot.
Since single facet injection is used for the LO laser, light reflection is expected not only from its facet but also from the DFB grating. By taking the propagation of the reflected signal into account, the OIL receiver design resembles that of a Mach-Zehnder interferometer for the input signal. Path mismatch interference between the arms can then affect phase noise measurements when synchronization is achieved, although locking range measurements are still possible. Indeed, it was found that the light reflected by the LO laser structure, when combined with the other portion of the input signal (bypass path) at the OIL receiver output, can generate path mismatch interference and become an issue if either LO laser tuning and locking to different comb lines is required or same-wavelength lines are superimposed during detection. The interference effect was clearly noticeable during the experiments as slowly-varying amplitude fluctuations in the comb line spectral distribution when the LO laser was locked to the pilot.
Thus, path mismatch is an important issue in the proposed OIL receiver and a great effort was made to ensure its reduction. The path lengths of the OIL receiver arms were measured and the difference reduced to a point where the spectral fluctuations were no longer noticed at the receiver output. When the OIL receiver was adapted to generate a local comb, as described in the next section, the phase noise was measured and the path difference further compensated to reduce the observed levels, leading to the lowest possible noise floor.
5. Remote comb generation and phase noise measurements
The next step was to investigate the possibility of using the locked LO laser (“locking technique” in Fig. 1) as a seed laser to generate a LO comb, whose lines would then be phase locked in relation to those of the original comb. In order to test the concept, the LO comb was formed by only two lines corresponding to both intensity-modulation sidebands generated by carrier suppressed modulation of the locked LO laser (“LO OFCG” in Fig. 1). The OIL receiver of Fig. 2(a) was modified to include a polarization controller (PC3) and a Mach-Zehnder intensity modulator (MZIM) between CIR and C2, as in Fig. 6(a). The MZIM (JDSU) was driven by a sinusoidal modulation signal about its null at the same frequency as that of the transmitter OFCG comb spacing. Separate RF synthesisers (Rohde & Schwarz SMP 04), synchronised via a 10 MHz reference signal, were used for the transmitter OFCG and the LO MZIM. In a real system, the comb spacing would have to be derived from the received optical signal, either by processing the received superchannel to extract frequency components related to the OFCG spacing, or by encoding a reference signal into the superchannel, perhaps using low amplitude modulation on the pilot itself.
Fig. 6 (a) Experimental arrangement for the LO comb receiver and (b) single-sideband phase noise measurement, having path matching as a reference.
For this experiment, Fig. 6(a) also shows that the input comb (in reality, with 21-lines) was set with the pilot to its right, in such a way that one of the LO comb lines would serve as LO for the comb line immediately to the left of the pilot. The output of the LO MZIM was combined with the other path from the input of the receiver, with fibre added to the bypass path to match the path lengths to less than 1 m. The power in the LO line of interest was approximately 20 dB higher than that in each transmitted comb line. The combined signal after C2 was amplified and directed onto a high-speed photodiode (u2t XPDV2320R, 60 GHz). A 1-nm optical filter was placed after the EDFA to reduce ASE-ASE noise. The output of the photodiode was band limited to the range 15 kHz to 7.4 GHz with a combination of a DC block and a low-pass Bessel filter. This ensures DC and heterodynes between the LO and other lines are rejected. The received electrical signal was analysed using a spectrum analyser (ESA, Rohde &Schwarz 1166.1660.26).
Phase locking between the LO and the line adjacent to the pilot was then assessed. The frequency of the RF drive to the LO MZM was set 500 MHz lower than the transmitter OFCG drive frequency, resulting in a detected tone at 500 MHz. That was done to allow direct phase noise measurements via the ESA built-in software, which are shown in Fig. 6(b). LO-ASE noise imposes a noise floor, but this result shows that the phase noise is generally −110 dBc/Hz or less at higher offset frequencies (> 10 MHz), indicating very good phase locking between the comb line and the LO. The peaks in the 1 to 10 MHz frequency range are related to incompletely suppressed modes of the transmitter OFCG fibre loop, which have a spacing of about 9 MHz. Between 10 kHz and 300 kHz, the phase noise appears to be largely determined by the phase noise of the RF synthesisers used to drive the OFCG and the MZIM used to generate the LO. Below 10 kHz offset, the phase noise increases due to flicker noise, 1/f noise, and electrical noise on the LO laser drive, and variations in the optical path lengths. The impact of these noise sources could be reduced, in this instance, with the addition of an electronic feedback loop path from the photodiode to the LO laser bias current, which would upgrade the arrangement to an OIPLL based receiver [10,11], and by actively controlling the optical path lengths or by using optical integration techniques.
6. Discussion
The use of traditional optical phase-locking techniques offers the possibility of easing DSP requirements on frequency offset compensation and phase tracking in coherent receivers, as demonstrated in several recent works on non-DSP based coherent receivers. For instance, an OIPLL coherent receiver has already been proposed and tested for ASK and BPSK signals [11]. Potentially, such a receiver could be used for 64-QAM signals as long as the full optical injection-locking range is kept at 1 GHz or over and lasers with a linewidth of 130 kHz are employed. Also, a homodyne coherent receiver for phase modulated signals with carrier extraction based on a feed-forward modulation stripping scheme has used optical injection locking to improve performance in comparison to a self-homodyne approach [15]. Other works have already examined the benefits of using optical injection locking in digital coherent receivers for orthogonal frequency division multiplexing (OFDM) transmission [16,17], using carrier recovery via OIL to lock a laser to a central pilot tone in a ~10-GHz bandwidth OFDM signal. It was observed that the DSP load was significantly reduced in both cases.
In this context, this paper has presented an experimental analysis of a potential superchannel coherent receiver based on a laser injection locked to a superchannel pilot line. The approach is similar to that in [17], which dismisses the use of an optical filter to separate the pilot line, but considers a far wider bandwidth multi-wavelength signal. Furthermore, the proposal is to apply the phase-locked LO laser as a seed to generate a LO comb (a two-line comb in this work) and perform coherent detection of superchannel subcarriers. It is important to point out that, with the injection locked seed laser, the generated LO comb lines would be locked in relation to the subcarriers, so that they can be individually demodulated via a less complex DSP. While the introduction of a pilot carrier (which could, in principle, be intensity modulated) and its guard band reduce the spectral efficiency of the superchannel to some extent, this disadvantage should be compensated by the reduction of the number of LO lasers required for digital coherent detection of each subcarrier to only one, saving on power consumption in the DSP and laser temperature control.
Since the amount of injected power into the slave laser is much larger for a 21-line comb than the ones reported in [16,17], in this paper the effect of adjacent lines on the locking range was first investigated. For the power levels used in the experiment, it was found that it is necessary to observe a guard band for interference-free injection locking, especially with the injection-related asymmetric locking range observed. However, if the pilot carrier power is lower than those of other subcarriers and still enough for the required OIL phase noise compensation, the OFCG line spacing for superchannel formation can be chosen to avoid interference. Next, the coherent detection of one of the superchannel lines was performed. In this case, the locked laser was used to generate a two-line comb (sidebands of carrier-supressed amplitude modulation). Phase noise measurements show that, although LO-ASE noise imposes a noise floor, preventing accurate measurement of the residual phase noise associated with the OIL process, the phase noise is as low as the phase noise from the synthesizers used in the OFCG (transmitter) and to modulate the locked laser signal (receiver) at offsets between 10 and 300 kHz, and is below −100 dBc/Hz above 1 MHz offset, indicating good quality locking.
Finally, although laser locking was observed in these experiments to be stable in a controlled environment, it is expected that polarization effects would play an important role in a long-haul fibre link, and optical polarization tracking of the pilot carrier at the receiver would be required prior to OIL. The OSNR after a long fibre link would also be lower than in our back-to-back experiments, but OIL has already been demonstrated to be robust at low OSNR (<8 dB/0.1 nm) [11]. Therefore, further investigation of the phase-locked LO generation scheme is required in a more realistic system environment in relation to both polarization dependence and OSNR.
7. Conclusions
Optical injection locking to a pilot line in an optical comb has been investigated, with the application of generating a phase-locked LO comb for coherent superchannel demodulation in mind. Using a three-line comb generated by intensity modulation, injection crosstalk from adjacent lines was investigated, leading to the conclusion that a relationship between line spacing and the power of adjacent lines has to be established to avoid disturbances preventing the stable locking to the pilot line, with clear guard bands required between the injection locking ranges of adjacent lines. In the case of 10 GHz-spaced lines, with the power in the adjacent lines up to 8 dB higher than the pilot power, a guard band of at least 6 GHz was found to be required.
In experiments with a frequency comb with 21 lines spaced by 20 GHz, the locking line (pilot) peak power was kept at the same level whilst those of the other lines were increased by up to 12 dB. For different pilot positions, stable but asymmetric locking towards the negative frequency detuning region (assuming the DFB free-running frequency as reference) was observed as the injected power was increased, with a dependency on the relative position of the pilot in the comb.
The locked LO laser was employed as a seed laser to generate a two-line LO comb by carrier suppressed intensity modulation. By controlling path mismatch, the LO generated comb lines can be phase locked in relation to those of the original comb. The 21-line comb was set with the pilot at its longest wavelength extreme and the phase noise between one of the LO comb lines and the input comb line immediately adjacent to the pilot was measured. Phase noise spectral density below −110 dBc/Hz was measured at 10 MHz offset or more, with the measurement limited by LO-ASE noise.
Acknowledgments
The authors would like to thank CAPES (grant BEX 3364/11-9), FAPESP, Padtec S.A., and CNPq (grant 308553/2010-1), Brazil, and EPSRC (COSINE project, EP/I012702/1), UK.
References and links
1. L. G. Kazovsky and D. A. Atlas, “A 1320-nm experimental optical phase-locked loop: performance investigation and PSK homodyne experiments at 140 Mb/s and 2 Gb/s,” J. Lightwave Technol. 8(9), 1414–1425 (1990). [CrossRef]
2. M. Fukushima and J. Miura, “Recent progress of erbium-doped fiber amplifiers and their components,” Proc. SPIE 6775, 677502 (2007). [CrossRef]
3. A. R. Chraplyvy, “High-capacity lightwave transmission experiments,” Bell Labs Tech. J. 4(1), 230–245 (1999). [CrossRef]
4. R.-J. Essiambre, G. Foschini, P. Winzer, and G. Kramer, “Capacity limits of fiber-optic communication systems,” in Optical Fiber Communication Conference, OFC 2009 Technical Digest CD-rom (Optical Society of America, 2009), paper OThL1.
5. C. R. S. Fludger, T. Duthel, D. van den Borne, C. Schulien, E.-D. Schmidt, T. Wuth, J. Geyer, E. De Man, G.-D. Khoe, and H. de Waardt, “Coherent equalization and POLMUX-RZ-DQPSK for robust 100-GE transmission,” J, Ligthwave Technol. 26(1), 64–72 (2008). [CrossRef]
6. K. Yonenaga, A. Sano, E. Yamazaki, F. Inuzuka, Y. Miyamoto, A. Takada, and T. Yamada, “100 Gbit/s all-optical OFDM transmission using 4 x 25 Gbit/s optical duobinary signals with phase-controlled optical sub-carriers,” in Optical Fiber Communication Conference, OFC 2008 Technical Digest CD-rom (Optical Society of America, 2008), paper JThA48.
7. D. Hillerkuss, T. Schellinger, R. Schmogrow, M. Winter, T. Vallaitis, R. Bonk, A. Marculescu, J. Li, M. Dreschmann, J. Meyer, S. Ben Ezra, N. Narkiss, B. Nebendahl, F. Parmigiani, P. Petropoulos, B. Resan, K. Weingarten, T. Ellermeyer, J. Lutz, M. Möller, M. Hübner, J. Becker, C. Koos, W. Freude, and J. Leuthold, “Single source optical OFDM transmitter and optical FFT receiver demonstrated at line rates of 5.4 and 10.8 Tbit/s,” in Optical Fiber Communication Conference, OFC 2010 Technical Digest CD-rom (Optical Society of America, 2010), paper PDPC1. [CrossRef]
8. D. Hillerkuss, R. Schmogrow, T. Schellinger, M. Jordan, M. Winter, G. Huber, T. Vallaitis, R. Bonk, P. Kleinow, F. Frey, M. Roeger, S. Koenig, A. Ludwig, A. Marculescu, J. Li, M. Hoh, M. Dreschmann, J. Meyer, S. Ben Ezra, N. Narkiss, B. Nebendahl, F. Parmigiani, P. Petropoulos, B. Resan, A. Oehler, K. Weingarten, T. Ellermeyer, J. Lutz, M. Möller, M. Hübner, J. Becker, C. Koos, W. Freude, and J. Leuthold, “26 Tbit s−1 line-rate super-channel transmission utilizing all-optical fast Fourier transform processing,” Nat. Photonics 5(6), 364–371 (2011). [CrossRef]
9. because of a name mismatch. The CrossRef authors areZ. He, Q. Yang, X. Zhang, T. Gui, R. Hu, Z. Li, X. Xiao, M. Luo, X. Yi, D. Xue, C. Yang, C. Li, and S. Yu, “4 × 2 Tbit/s superchannel self-coherent transmission based on carrier tracking and expanding,” Electron. Lett. 50(3), 195–197 (2014). [CrossRef]
10. A. C. Bordonalli, C. Walton, and A. J. Seeds, “High-performance phase locking of wide linewidth semiconductor lasers by combined use of optical injection locking and optical phase-lock loop,” J. Lightwave Technol. 17(2), 328–342 (1999). [CrossRef]
11. M. J. Fice, A. Chiuchiarelli, E. Ciaramella, and A. J. Seeds, “Homodyne coherent optical receiver using an optical injection phase-lock loop,” J. Lightwave Technol. 29(8), 1152–1164 (2011). [CrossRef]
12. O. Lidoyne, P. Gallion, C. Chabran, and G. Debarge, “Locking range, phase noise and power spectrum of an injection-locked semiconductor laser,” Proc. Inst. Elect. Eng. J. 137, 147–154 (1990).
13. B. Cai, L. A. Johansson, C. F. C. Silva, S. Bennett, and A. J. Seeds, “Crosstalk, noise, and stability analysis of DWDM channels generated by injection locking techniques,” J. Lightwave Technol. 21(12), 3029–3036 (2003). [CrossRef]
14. S. Bennett, B. Cai, E. Burr, O. Gough, and A. J. Seeds, “1.8-THz bandwidth, zero-frequency error, tunable optical comb generator for DWDM applications,” IEEE Photon. Technol. Lett. 11(5), 551–553 (1999). [CrossRef]
15. S. K. Ibrahim, S. Sygletos, R. Weerasuriya, and A. D. Ellis, “Novel real-time homodyne coherent receiver using a feed-forward based carrier extraction scheme for phase modulated signals,” Opt. Express 19(9), 8320–8326 (2011). [CrossRef] [PubMed]
16. S. Adhikari, S. Sygletos, A. D. Ellis, B. Inan, S. L. Jansen, and W. Rosenkranz, “Enhanced self-coherent OFDM by the use of injection locked laser,” in Optical Fiber Communication Conference, OFC 2012 Technical Digest CD-rom (Optical Society of America, 2012), paper JW2A. [CrossRef]
17. Z. Liu, D. S. Wu, D. J. Richardson, and R. Slavík, “Homodyne OFDM using simple optical carrier recovery,” in Optical Fiber Communication Conference, OFC 2014 Technical Digest CD-rom (Optical Society of America, 2014), paper W4K3. [CrossRef]
