We present experimentally and analytically the phase noise characterization of an externally injected gain switched comb source. The results reveal the residual high frequency FM noise in the comb lines, which stays unnoticed in the optical linewidth value but leads to an increased phase-error variance. The potential impact of the residual phase noise is investigated in a 10.7 GBaud optical DQPSK system where a 2 dB power penalty is recorded at BER of 10−9. In a 10.7 GBaud digital coherent QPSK system no penalty is observed but with 5 GBaud 16-QAM format a 3 dBpenalty exists at the FEC limit of 4.4e-3.
© 2014 Optical Society of America
The excessive growth of Internet traffic is pushing wavelength division multiplexed (WDM) optical communication systems towards denser channel spacing and higher capacity per wavelength channel [1–3]. Techniques like Nyquist WDM  and optical orthogonal frequency division multiplexing (OFDM)  have stirred a lot of attention as they potentially offer channel spacing close or equal to the symbol rate. Due to the reduced or eliminated inter-channel guard bands, these techniques are best served by optical comb sources which guarantee constant frequency offset between the carriers. To achieve high capacity on the individual channels, advanced modulation formats can be utilized to increase the overall spectral efficiency . One strict requirement imposed by these formats, however, is the phase noise property of the comb lines in an optical comb source.
Amongst many different comb generation schemes, an interesting approach is the external injection of gain switched (directly modulated) laser diodes. The potential of this approach rests in its suitability for monolithic integration , as well as the flexible tunability in both channel spacing  and central wavelength . More importantly, it has been shown that external injection effectively transfers the narrow emission linewidth of the injection master to the slave comb lines [5,6], making such a comb source particularly attractive for coherent applications requiring low linewidth multi-carrier transmitters. Although there are other comb generation techniques such as the modulator based comb source that also provides flexibility as well as low linewidth [7,8], the relatively complicated configuration, the inherent instability and the large insertion loss might prove disadvantageous for applications in real systems. Therefore, the simple, compact and low noise injected gain switched comb sources could be promising components for future coherent optical communication systems.
Previous studies have shown that the spectral linewidth might not be sufficient to fully describe the phase noise property of an optical light source [9,10]. A more detailed analysis of the FM-noise spectrum, phase-error variance, and the field spectrum can provide a better insight of the phase noise process as well as its potential influence [9–11]. In this work, we present the experimental phase noise characterization of an injected gain switched comb source. The results are then analytically characterized by a simple mathematic model. We also investigate the potential impact of residual comb line phase noise on coherent communications with a self-coherent optical differential quadrature phase shift keyed (DQPSK) system, as well as a digital coherent receiver based system which employs QPSK and 16 quadrature amplitude modulation (QAM) formats.
2. Phase noise characterization of injected gain switched comb lines
The experimental setup for the phase noise characterization of the externally injected gain switched comb source is proposed as in Fig. 1. The injected gain switched comb generation is achieved with a master-slave configuration. An integrated tunable laser assembly (ITLA) master laser injects light via an optical circulator to a distributed feedback (DFB) slave laser, with a polarization controller (PC) in the light path to align the polarization state of the injected light to the optical waveguide of the slave laser. A variable optical attenuator (VOA) controls the injection power to the slave cavity, while the master bias is kept at a constant level throughout the experiment to ensure the consistent phase noise property of the injectingsource. Gain switching of the slave DFB laser is achieved by applying an amplified 15 GHz sinusoidal RF signal (24 dBm) in combination with a DC bias current of 40 mA, while the laser is controlled at room temperature. It is worth noting that the 15 GHz free spectral range (FSR) is only chosen for demonstration purpose and the FSR of the injected gain switched frequency comb source is variable from 5~40 GHz .
The optical comb generation is followed by an Erbium doped fiber amplifier (EDFA) before being passed to a narrow bandwidth (0.03~0.04 nm at 3 dB) tunable optical filter (TOF) to select individual comb lines (one line at a time) for the phase noise characterization. The phase noise characterization utilizes a modified delayed self-heterodyne (DSH) method based on the harmonic detection principle . The detected signal is captured with a real-time oscilloscope and sent for offline processing to recover the complex amplitude of the differential phase error. This allows us to evaluate the FM-noise spectrum, which is the power spectral density (PSD) of instantaneous frequency fluctuations, and the phase-error variance, which is the mean-square value of differential phase variation in a given time interval .
The injected gain switched comb generation depends highly on the injection conditions, namely the optical injection power and the wavelength detuning between the master and slave laser. By carefully choosing these conditions, optical combs with different noise properties are generated due to the various interactions between the injecting signal and the photons generated in the slave cavity. Here we choose two such conditions, with the injection power and wavelength detuning (Δλ = λmaster - λslave) set to (a) 0 dBm, −0.5 nm and (b) + 3 dBm, + 0.17 nm. The comb in Fig. 2(a), corresponding to injection condition (a), demonstrates 10 clearly resolved comb tones in a 3 dB window and we therefore call it the “flat comb”. The second comb in Fig. 2(b), corresponding to injection condition (b), shows relatively poor flatness compared to the previous case, but has a smaller noise profile under the comb lines and we name it the “low noise” comb. Such naming is verified by the phase noise measurement results (in terms of FM-noise spectrum) shown in Fig. 3(a).
In the actual phase noise measurement, we picked three different lines (with similar noise extinction) from each comb (flat and low noise) and found that comb lines from the same comb source also perform the same. Therefore only one line from each comb is demonstrated in Fig. 3(a) for clarity, and they are also compared with the master laser FM-noise spectrum. It is clear in Fig. 3(a) that the FM-noise spectrum of the master is dominated by white FM noise and shows a flat PSD over the frequency range being measured (limited to ~1 GHz due to digital filtering in the offline processing ). This is as expected since the master ITLA is a DFB based laser array . The injected slave lines for flat and low noise combs, on the other hand, have identical FM noise PSD to the master at lower frequencies, but both start to deviate from the master at around 100 MHz, and there is a marked increase for the flat comb line at higher frequencies towards ~1 GHz. Such FM noise deviation/increase, however, does not affect the 3 dB optical linewidth, since we perform a conventional DSH measurement on the injected comb lines and the resulting linewidths all follow the master laser linewidth (~300 kHz), as indicated in Figs. 2(a) and 2(b) by the blue triangles on top of the comb lines. Moreover, the averaged beat signals on the electrical spectrum analyzer (ESA) also yield identical Lorentzian lineshape for the flat, low noise combs and the master as evidenced in Fig. 3(b). In other words, the high frequency FM noise in the comb lines remains largely unnoticed in a conventional DSH measurement.
These experimental observations in our phase noise measurement agree well with the previous study where the slave laser was in a free running condition . The results presented here confirm that insufficient injection (due to low injection power or improper wavelength detuning) leads to an unsuccessful “locking” of the slave FM noise properties to that of the master, at higher frequencies, where the slave FM-noise spectrum is determined by both the master and the slave itself . However, a unique feature for the gain switched slave comb source when compared to its free running case would be the additional phase distortions induced by unwanted frequency modulation during the gain switching (direct modulation) process . Such additional distortions are expected to result in more stringent requirements on the injection conditions and/or more severe deviation in the slave FM-noise spectrum from that of the master laser.
The high frequency FM noise demonstrated in Fig. 3(a) shows quite distinct properties from the FM-noise spectrum of a free running semiconductor laser, which usually exhibits low frequency fluctuations (e.g. 1/f and 1/f 2 types FM noises ). However, the same methodology that utilizes power-law noise analysis can be implemented here to estimate the unusual comb line phase noise. We propose the use of f and f 2 terms to analytically characterize the high frequency FM noise of the injected comb lines. The analytical model of comb line FM-noise spectrum can be therefore expressed as:9]; kb and kp are constants. By carefully choosing kb and kp values (kb = 1.7e-5, kp = 4.6e-12) we find that the FM-noise spectrum of the comb lines can be closely resembled by the proposed analytical form in Eq. (1), as shown in Fig. 3(a) by the green dash line (only flat comb is discussed here for clarity). Furthermore, since the phase-error variance is directly related to the FM-noise spectrum  we also evaluated the analytical form of the comb line phase-error variance via analytical integrations, using similar procedure as in . The result is given as:Fig. 3(a), is the sine integral function and is the cosine integral function. We thus plot the calculated phase-error variance of the flat comb from Eq. (2) and show in Fig. 3(c) by the green dash line.
In Fig. 3(c), we also demonstrate the experimental phase-error variance of the flat comb, low noise comb, and that of the master. In the figure, a first observation is the linear increase of the master phase-error variance with respect to the delay time, which is commonly seen for DFB lasers with the presence of only white FM noise . As for the injected comb lines (both flat and low noise), a very similar linear slope as for the master is recorded at long delay times, thus explaining the identical FM-noise spectrum at low frequencies and the same emission linewidth for the injected comb lines and master laser (the spectral linewidth is determined by the low frequency content of the FM-noise spectrum ). However, for short term delay times of less than ~4 ns an additional damped oscillatory behavior is noticed for both combs sources, with the flat comb suffering a severe distortion from the oscillation. Such oscillatory distortion corresponds to the high frequency increase of the FM-noise spectrum and the consequence is that the phase-error variance of the comb lines being upshifted with respect to that of the master. As such, the comb lines (flat comb in particular) will experience larger phase errors at all delay times compared to the values predicted by the master, potentially leading to unexpected errors in coherent communication systems. In the following section, we experimentally investigate the performance of these comb lines in coherent optical communication systems operating at 10.7 GBaud and 5 GBaud with different modulation formats to evaluate the potential influence from the comb line phase noise.
3. Comb line phase noise impact on coherent communications
The injected gain switched comb lines are tested back-to-back (B2B) in two types of coherent systems, as shown in Fig. 4: a digital coherent receiver based system and a direct detection based self-coherent system. In the digital coherent system two modulation formats with different symbol rates are investigated: QPSK at 10.7 GBaud and 16-QAM at 5 GBaud. The direct detection system employs 10.7 GBaud DQPSK.
At the transmitter side, the individual comb lines are selected by the same TOF as in Fig. 1 to avoid the potential influence from inter-channel crosstalk. A dual-parallel optical I/Q modulator is used for the (D)QPSK modulation (with or without the differential pre-coding), as well as the 16-QAM modulation of the comb lines. The B2B received optical power is controlled by a VOA before going through an optical pre-amplifier. For the digital coherent detection of 10.7 GBaud QPSK an external cavity laser (ECL) with 100 kHz linewidth serves as the local oscillator (LO), and a digital sampling oscilloscope operating at 50 GSa/s captures the intra-dyne I/Q beat signal at the detector output. Regarding the 5 GBaud 16-QAM we use a 25 kHz ECL as the LO and the sample rate is reduced to 20 GSa/s. The QPSK or 16-QAM signal is then sent to offline digital signal processing (DSP) . In terms of the 10.7 GBaud DQPSK reception, an optical 1-bit delay interferometer (DLI) is used for the de-modulation. It is followed by a single-ended photodetector (PD) and then an error detector (ED).
The system performance of the comb lines (3 lines from flat comb and 3 lines from low noise comb) are illustrated in Fig. 5, in terms of BER versus received optical power. Figure 5(a) shows that in the digital coherent receiver based system, the residual phase noise in the flat and low noise comb lines have no visible influence on the system BER performance of the 10.7 GBaud QPSK system, when compared to the master laser. In terms of the 5 GBaud 16-QAM system an overall floor-like behavior is noticed, since the system requirement on phase noise/linewidth is much more stringent in this case (due to the increased modulation order and the reduced symbol rate). Besides, both low noise and flat comb sources demonstrate degraded performance compared to the master laser. At a forward error correction (FEC) limit of 4.4e-3 (assuming 7% overhead ) the penalties are 1 dB and 3 dB for the low noise and flat comb respectively. As such, the low noise comb shows similar performance with the master while the flat comb demonstrates a higher penalty. Therefore in a high order modulated digital coherent system operating at low baud rates, certain trade-off should be made in the comb line flatness to reduce the system impact from the excess phase noise increase in injected comb lines.
Next, the 10.7 GBaud DQPSK system results are shown in Fig. 5(b) with BER calculated from the average of I and Q channels. In this case the low noise comb lines yield identical performance with the master but for the flat comb, there is a notable performance degradation which leads to a 2 dB penalty at a BER of 10−9. In order to find out the potential contribution from intensity noise to this 2 dB penalty, we also perform a 10.7 GBaud on off keying (OOK) modulation experiment, using the same sources (master, low noise and flat comb) and the same single-ended direct detection receiver (PD + ED). The results for OOK B2B system are shown in the inset of Fig. 5(b) and demonstrate the same BER performance between the three sources. Therefore we conclude that the power penalty in the direct detected DQPSK system for the flat comb is only due to the residual comb line phase noise.
In this work we study the phase noise characteristics of an injected gain switched comb source and demonstrate the impact of comb line phase noise on coherent communication systems. We experimentally show that residual phase noise can exist in the comb lines due to certain injection conditions, which stay unnoticed in optical linewidth values but can be observed through a detailed phase noise characterization process. We propose using a power-law FM noise model to analytically describe the residual phase noise and we also derived the analytical expression of the phase-error variance. To test the influence of the comb line phase noise in coherent communications we employ two types of coherent systems working with different modulation formats operating at 5 GBaud and 10.7 GBaud. The results suggest the potential of the injected gain switched comb source while also highlighting the possible limitations caused by the residual comb line phase noise, which needs to be accounted for by optimized optical injection at the transmitter or further reducing the combined linewidth of master + LO (thus down-shifts the overall phase-error variance).
We acknowledge the funding from China Scholarship Council, Science Foundation Ireland through grants 09/IN.1/I2653 and 10/CE/I1853, Higher Education Authority PRTLI 4 and 5, INSPIRE Programs and CTVR Grant.
References and links
1. I. Morita and W. R. Peng, “High capacity WDM transmission using terabit super-channels,” ECOC2012, Th.2.C.3.
2. G. Bosco, V. Curri, A. Carena, P. Poggiolini, and F. Forghieri, “On the performance of Nyquist-WDM terabit superchannels based on PM-BPSK, PM-QPSK, PM-8QAM or PM-16QAM subcarriers,” J. Lightwave Technol. 29(1), 53–61 (2011). [CrossRef]
3. Y. K. Huang, E. Ip, Z. Wang, M. F. Huang, Y. Shao, and T. Wang, “Transmission of spectral efficient super-channels using all-optical OFDM and digital coherent receiver technologies,” J. Lightwave Technol. 29(24), 3838–3844 (2011). [CrossRef]
4. R. Zhou, P. M. Anandarajah, M. D. G. Pascual, J. O’Carroll, R. Phelan, B. Kelly, and L. P. Barry, “Monolithically integrated 2-section lasers for injection locked gain switched comb generation,” OFC2014, Th3A.3.
5. P. M. Anandarajah, R. Zhou, R. Maher, M. D. G. Pascual, F. Smyth, V. Vujicic, and L. P. Barry, “Flexible optical comb source for super channel systems,” OFC2013, OTh3I.8.
6. R. Zhou, S. Latkowski, J. O’Carroll, R. Phelan, L. P. Barry, and P. Anandarajah, “40 nm wavelength tunable gain-switched optical comb source,” Opt. Express 19(26), B415–B420 (2011). [CrossRef] [PubMed]
7. Y. Xing, Q. Wang, L. Huo, and C. Lou, “Frequency chirp linearization for ultraflat optical frequency comb generation based on group velocity dispersion,” Opt. Lett. 38(13), 2188–2190 (2013). [CrossRef] [PubMed]
8. A. K. Mishra, R. Schmogrow, I. Tomkos, D. Hillerkuss, C. Koos, W. Freude, and J. Leuthold, “Flexible RF-based comb generator,” IEEE Photon. Technol. Lett. 25(7), 701–704 (2013). [CrossRef]
9. K. Kikuchi, “Characterization of semiconductor-laser phase noise and estimation of bit-error rate performance with low-speed offline digital coherent receivers,” Opt. Express 20(5), 5291–5302 (2012). [CrossRef] [PubMed]
10. T. N. Huynh, L. Nguyen, and L. P. Barry, “Phase noise characterization of SGDBR lasers using phase modulation detection method with delayed self-heterodyne measurements,” J. Lightwave Technol. 31(8), 1300–1308 (2013). [CrossRef]
11. R. Zhou, V. Vujicic, T. N. Huynh, P. M. Anandarajah, and L. P. Barry, “Effective phase noise suppression in externally injected gain switched comb source for coherent optical communications,” ECOC2013, P.2.5.
12. P. Spano, S. Piazzolla, and M. Tamburrini, “Frequency and intensity noise in injection-locked semiconductor lasers: theory and experiments,” IEEE J. Quantum Electron. QE22(3), 427–435 (1986). [CrossRef]
13. K. Petermann, Laser Diode Modulation and Noise (Kluwer Academic, 1988), Chap. 5.
14. T. N. Huynh, F. Smyth, L. Nguyen, and L. P. Barry, “Effects of phase noise of monolithic tunable laser on coherent communication systems,” Opt. Express 20(26), B244–B249 (2012). [CrossRef] [PubMed]
15. M. Scholten, T. Coe, and J. Dillard, “Continuously-interleaved BCH (CI-BCH) FEC delivers best in class NECG for 40G and 100G metro applications,” OFC2010, NTuB3.