1. Introduction

Because of their outstanding inherent merits of low noise, high beam quality, narrow linewidth and excellent compatibility with fiber systems, single-longitudinal-mode (SLM) fiber lasers, especially for the SLM erbium-doped fiber lasers (EDFLs) in C-band (1530-1565 nm), have been adopted or anticipated as preferred light sources for many important applications. For instance, single-wavelength EDFLs with narrow linewidth output have been used for ultra-long distance coherent optical communication, fiber optic sensing, optical metrology, high resolution spectroscopy and LIDAR applications [1–6], and also are promising for potential applications relating to optical atomic clocks, measurements of fundamental constants and physics [7,8]; The use of dual-wavelength SLM EDFLs have been demonstrated in high resolution fiber sensor [9] and in generation of high spectral purity microwave signals [10,11]; EDFLs with multi-wavelength channels, each being SLM state, are strongly expected to be used in wavelength-division-multiplexing optical communication in the future [12–14]. Therefore, it is attractive if a single SLM EDFL has the capability of being flexibly switched among single-, dual- and multi-wavelength lasing operations.

A number of wavelength-switchable techniques in EDFLs have been reported in recent years, which in general involve a multi-channel optical filter to select the primary lasing wavelengths and a switching mechanism to control the operation mode in the laser cavity. The multi-channel optical filter may be one of the following: a few-mode fiber filter with core-offset structure [15], a superimposed fiber Bragg grating (FBG) [16–18], a silicon-micro-ring-resonator [19], a thin-core fiber comb filter [20], a tapered non-adiabatic microfiber based Sagnac mirror [21], a polarization-maintaining chirped moiré FBG filter [22], a short twin-core photonic crystal fiber based Mach-Zehnder interferometer [23], a FBG embedded fiber modal interferometer [24], an erbium-doped photonic crystal fiber based equivalent filter [25] or a wideband chirped Moiré FBG filter [26], while the switching mechanism can be of polarization controlling [15,17–19,22,23], nonlinear polarization rotation (NPR) combining with strain switching [16,21], intensity-dependent inhomogeneous loss [25], cavity-loss adjusting [20,24] or tunable-filter scanning [26]. In addition, for stable dual-/multi-wavelength lasing in an EDFL, a stabilizing mechanism, such as a NPR based or a nonlinear amplifier loop mirror based amplitude equalizer [17,27] or an enhanced polarization-hole-burning (PHB) effect [28,29], is necessary to suppress the wavelength competition induced by the homogenous broadening effect of erbium-doped fiber (EDF). The enhanced PHB effect can be established by using a high-birefringence FBG (HBFBG) as a cavity reflector [28,29], which inherently reflects two wavelengths with orthogonal states of polarization (SOPs); the reflected orthogonal polarized lights propagating into the EDF can utilize different subsets of gain erbium ions to significantly mitigate the wavelength competition. As described above, a HBFBG not only can bring strong PHB effect but can also act as a dual-channel reflecting filter for making switchable EDFL with excellent polarization characteristics [28], but having only two-wavelength channels to be switched for use. We have verified that the superimposed FBGs can be used as compact, low cost and stable multi-channel filters [17,30,31]. Unfortunately, using a superimposed HBFBG (SI-HBFBG) in a switchable EDFL has not yet been reported.

Numerous configurations of EDFLs can produce SLM output, such as distributed feedback (DFB) [32], short cavity distributed Bragg reflector (DBR) [33], conventional linear and ring cavities [34–36] and compound-cavity [6,17–19,28,31,37,38]. However, most of them have one or more significant shortcomings. For example, both the DFB based and DBR based EDFLs have limited output power and low flexibility; the conventional linear and ring cavities based EDFLs require a complicated and expensive ultra-narrow filter for SLM selection; and in EDFLs using DBR and linear cavities, the spatial gain hole burning induced by the standing-wave effect cannot be avoided. Multi-ring compound-cavity (MR-CC) is a simple fiber laser configuration for SLM operation, which is generally composed of a main-ring cavity (MRC) for providing laser gain and one or more passive subring cavities as filters for selecting SLM. The mode-selecting capabilities of ring-cavity based filters made with a single 2×2 optical coupler (the single-coupler ring, or SCR), with two 2×2 couplers (the dual-coupler ring, or DCR), with three nested couplers and with cascaded or nested SCR/DCR have been studied extensively by our group [17,18,28,31] and others [6,19,37,38]. Although a DCR can provide ultra-narrow passband filtering for both SLM selection and laser linewidth compression by using optical couplers (OCs) with appropriate coupling ratios, its free-spectrum-range (FSR) is too small to be effective, due to the smallest fiber ring length achievable in DCR fabrication. According to vernier effect [6,17–19,28,31,37,38], the small FSR problem may be resolved by cascading two DCRs with a small length difference to form a DCR-based compound-cavity (DCR-CC) filter, which has not been used in an EDFL, to the best of our knowledge. Besides, the vernier effect can only qualitatively explain how such MR-CC filters are utilized for SLM selection, but no a comprehensive method of combining theory and experiment has been found in literatures to quantitatively study how to make a MR-CC filter aiming for SLM selection in a fiber laser.

In this paper, firstly we present in detail how to design, fabricate, and characterize the DCR-CC filter with both theoretical analysis and experimental results, and we believe that is the first systematic approach for making a compound-cavity based filter used for selecting SLM in a fiber laser. Then, using the DCR-CC as an enabling component for SLM selection and a SI-HBFBG as a compact four-channel reflective filter to define wavelength channels, both for the first time, we propose and demonstrate a four-wavelength-switchable erbium-doped fiber laser (4WS-EDFL) with high output performance. By introducing the enhanced PHB effect to mitigate wavelength competition and controlling the SOPs to bring polarization-mismatch-induced losses, the 4WS-EDFL can be switched among fifteen states including four single-wavelength operations, six dual-wavelength operations, four three-wavelength operations and one four-wavelength operation. As part of the demonstration, we show that, in all four single-wavelength operations, the 4WS-EDFL exhibits superb performances in aspects of spectrum stability, optical signal to noise ratio (OSNR), longitudinal-mode, linewidth, power fluctuation, relative intensity noise (RIN) and polarization characteristics.

2. Experimental setup, principle and theory

2.1 Experimental configuration of the 4WS-EDFL and lasing principle

Figure 1 shows the configuration of the proposed 4WS-EDFL which belongs to a typical MR-CC structure formed by a MRC with a subring compound-cavity filter. In the MRC, a 2.9 m long EDF (Fibercore M12-980-125), coiled round the plates of a TL-PC, is pumped by a commercial 980 nm laser diode (LD, Connet VLSS-980) through a 980/1550 nm wavelength division multiplexer (WDM), a SI-HBFBG is innovatively used as a four-channel reflecting filter for primary selection of lasing wavelengths, by cooperating with a three-port optical circulator which also can ensure a unidirectional oscillation, a drop-in polarization controller (DI-PC), combining with the TL-PC, can adjust the light gain and loss inside the laser cavity, and a novel DCR-CC filter, made by two DCRs, DCR-1 and DCR-2, is designed to select one single mode from dense longitudinal-modes of the MRC. The laser is extracted from the 10% port of a 90:10 OC, OC-5, to be measured. The length of MRC is ∼18.20 m corresponding to a longitudinal-mode spacing of ∼11.23 MHz.

 

Fig. 1. Schematic of proposed four-wavelength-switchable erbium-doped fiber laser (4WS-EDFL). LD: laser diode; WDM: wavelength division multiplexer; EDF: erbium-doped fiber; TL-PC: three-loop polarization controller; OC: optical coupler; DI-PC: drop-in polarization controller; SI-HBFBG: superimposed high-birefringence fiber Bragg grating; DCR: dual-coupler ring; DCR-CC: dual-coupler ring based compound-cavity. The TL-PC is made with a length of 2.9 m EDF pigtailed by single-mode fibers (SMF-28) at both sides, and the SI-HBFBG, combining with the gain EDF coiled inside the TL-PC, introduces the enhanced polarization hole burning effect to mitigate the strong wavelength competition. The DCR-CC filter selects one single mode from dense lasing longitudinal-modes. Note that a FC/APC connecter behind the SI-HBFBG is used to avoid the unnecessary reflections.

Download Full Size | PDF

The SI-HBFBG was inscribed in a section of hydrogen-loaded commercial polarization maintaining fiber (PMF, YOFC PM#1550_125-18/250) using the phase mask method [17,18,30] with a KrF excimer laser emitting ultraviolet light at 248 nm. At the exact same fiber position of the PMF, two repeated grating-inscribings were accomplished in turn using a uniform phase mask with a period of 1075 nm while a suitable tensile-force was imposed to the PMF before the second grating was inscribed, which was equal to using two uniform phase masks with different periods to inscribe two gratings in turn. Since the PMF used was with a birefringence $\Delta n$ of ∼2.2×10⁻⁴, according to the equation $\Delta \lambda \textrm{ = }\Delta n \cdot \Lambda $ ($\Lambda $ is the period of phase mask), two reflecting channels with a wavelength spacing of ∼0.237 nm would be obtained after each grating-inscribing, therefore a SI-HBFBG should be with four reflecting channels theoretically. The pure reflecting optical spectrum of the inscribed SI-HBFBG is shown in Fig. 2(a) in normalized linear scale, and to measure that a measurement system, with a customized erbium-doped fiber amplifier (EDFA), an optical circulator and an optical spectrum analyzer (OSA, Yokogawa AQ6370D), was designed as shown in the inset of Fig. 2(a). A measurement process with steps (i) to (iv) as described in the caption of Fig. 2(a) was carried out, and the pure reflecting spectrum of the SI-HBFBG was calculated using the measured spectra through eliminating the effects of the EDFA and the circulator. The peak-reflecting wavelengths of four channels are approximately 1555.830 nm, 1556.073 nm, 1556.332 nm and 1556.573 nm, and the corresponding reflectivities are 39.1%, 36.6%, 39.2% and 36.7%, respectively, measured using a resolution of 0.02 nm and a data sampling interval of 0.001 nm for OSA. Note that, the first two reflecting channels and the last two reflecting channels are respectively with spacings of 0.243 nm and 0.241 nm, consistent with the theoretical estimated value of 0.237 nm; the spacing of 0.259 nm between the second and third channels is induced by the tensile-force applied during the second grating-inscribing. Additionally, it is obvious that the lights reflected by the first and second channels should be with orthogonal SOPs and the same situation is for the third and fourth channels, while the lights reflected by the first and third channels should be with almost same SOPs and the same situation is for the second and fourth channels.

 

Fig. 2. (a) Measured pure optical spectrum of SI-HBFBG in normalized linear scale and corresponding measurement system in inset; (i) measurement of output spectrum of EDFA; (ii) measurement of output spectrum of EDFA involving loss spectrum of port 1 to port 2 of circulator; (iii) measurement of output spectrum of EDFA involving loss spectrum of port 2 to port 3 of circulator; (iv) measurement of reflecting spectrum of SI-HBFBG involving loss spectra of port 1 to port 2 and port 2 to port 3 of circulator both and effect of output spectrum of EDFA. (b) Schematic of the proposed DCR-CC filter; $L1$∼ $L5$ denote lengths of fibers; 1∼15 denote port numbers of optical couplers (OCs).

Download Full Size | PDF

2.2 Principle of wavelength-switchable operation

Generally, due to the strong wavelength competition in the EDF at room temperature, the stable dual- and multi-wavelength operations are extremely difficult to achieve, comparing with the single-wavelength operation, unless an efficient suppressing mechanism is employed inside the laser cavity. In our 4WS-EDFL, the TL-PC made with EDF and the SI-HBFBG can introduce enhanced PHB effect to enable the switching among single-, dual- and multi-wavelength operations. The PHB arises from the randomly distributed orientations of erbium ions in the glass matrix of an EDF and the selective deexcitation of those ions aroused by a polarized light [39]. The lights with different polarization directions can utilize different subsets of the gain ions, indicating that the gains of the lights with different SOPs are contributed by different groups of erbium ions, and consequently the wavelength competition can be mitigated significantly.

The loops inside TL-PC introduce stress in the EDF and therefore a birefringence which depends on the radius of the EDF and the radius of the plates [40]. According to the theory described in [41], by adjusting the polarization direction of incident lights with an almost same SOP into the TL-PC appropriately, the polarization directions of lights at different wavelengths can be different inside the EDF. Secondly, as aforementioned, a HBFBG reflects two wavelengths naturally separated at two orthogonal linear polarization modes, and then they are amplified by the different excited erbium ions of the EDF according to the PHB effect; also thanks to the birefringence of EDF induced by the coils, the PHB can be enhanced to significantly stabilize the dual-wavelength operation further. Although the four wavelengths reflected by the SI-HBFBG are simultaneously with orthogonal and parallel polarization directions each other, as described in Section 2.1, the dual- or multi-wavelength lasing can be achieved according to the above principle, because we can jointly rotate the two polarization axes of the EDF by tuning the TL-PC and control the polarization direction of incident lights into the EDF by tuning the DI-PC to effectively adjust the loss and gain of different lasing wavelengths. In addition, since the reflectivities for the four reflecting channels of the SI-HBFBG are highly polarization dependent, the total cavity-loss for each lasing channel can be changed by adjusting the SOP of the oscillating light inside the cavity. Therefore, by adjusting the TL-PC, we are able to access different polarization-mismatch-induced losses for the four different wavelengths at the SI-HBFBG to allow only one wavelength with the minimum cavity-loss to sustain lasing to achieve single-wavelength operation.

2.3 Theory of DCR-CC filter

Figure 2(b) shows the running schematic of the proposed DCR-CC filter formed by cascading two DCRs, DCR-1 and DCR-2, using four SMF-based OCs, OC-1, OC-2, OC-3 and OC-4. We set $L1$–$L5$ as the lengths of fibers as marked, 1–15 as the port numbers of OCs, ${E_1}$–${E_{15}}$ as the electric field amplitudes of ports of OCs, ${\alpha _i}{\kern 1pt} {\kern 1pt} ({i = 1,2,3,4} )$ and ${\gamma _i}{\kern 1pt} {\kern 1pt} ({i = 1,2,3,4} )$ respectively as the coupling ratio and insertion loss of the $i\textrm{ - th}$ OC, $\beta $ as the fiber loss coefficient, $\delta $ as the fusion splicing loss, $k = {{2\pi } \mathord{\left/ {\vphantom {{2\pi } \lambda }} \right.} \lambda }$ as the wave number and ${n_{eff}}$ as the effective refractive index of the SMF. For DCR-1, the field transmission relationships of OC-1 and OC-2 are as

2.4 Theory of SLM operation and determination of parameters of DCR-CC filter

From Fig. 2(a), we calculated that the full-widths at half-maximum (FWHMs) of the four reflecting peaks of the SI-HBFBG are 12.89 GHz, 13.26 GHz, 13.62 GHz and 13.25 GHz respectively. Firstly, since the SI-HBFBG and the DCR-CC filter are used corporately to select the SLM in the laser cavity, so the free spectrum range (FSR) of the DCR-CC should be 0.5∼1 times of the FWHM of each reflecting peak of the SI-HBFBG [37]. According to the vernier effect [42], when the length difference $\Delta l$ between DCR-1 and DCR-2 is much less than their cavity-lengths, the FSR of DCR-CC can be given as

Secondly, in order to select SLM from dense longitudinal-modes of MRC, the FWHM of the passband of DCR-CC should be 0.5∼1 times of the spacing of adjacent longitudinal-modes [31,37]. Since a MR-CC fiber laser is generally with a MRC length l of 10∼20 m according to our previous work [17,26,28,31], the corresponding longitudinal-mode spacing $\Delta f$ is about 10∼20 MHz according to the equation $\Delta f\textrm{ = }{\raise0.7ex\hbox{$c$} \!\mathord{\left/ {\vphantom {c {{n_{eff}}l}}} \right.}\!\lower0.7ex\hbox{${{n_{eff}}l}$}}$, meaning that a FWHM of <10 MHz for DCR-CC is enough for most situations. On the basis of determined cavity lengths of DCR-1 and DCR-2, according to the theory in Section 2.2, the influences of coupling ratio $\alpha $ to the filtering characteristics of DCR-CC were simulated in detail, as shown in Fig. 3. In calculation, the parameters, $\beta $=0.2 dB/km, $\delta $=0.01 dB, ${\gamma _i}{\kern 1pt} {\kern 1pt} ({i = 1,2,3,4} )$=0.09 dB, ${n_{eff}}$=1.468, $L1$=0.30 m, $L2$=0.30 m, $L3$=0.31 m, $L4$=0.31 m and $L5$=0.50 m were used. Figure 3(a) shows the simulated transmission spectra of DCR-1, and we can see that with the increasing of $\alpha $ from 0.2 to 0.5 the FWHM and transmittance both increase. For the DCR-CC, we also simulated the transmission spectra when the coupling ratio $\alpha $ was 0.2, 0.4 and 0.5 respectively as shown in Figs. 3(b), 3(c) and 3(d). As can be seen, the FWHM of the main resonant peak increases from 19.82 MHz to 50.79 MHz and the suppression ratio (SR) increases from 0.73 to 0.96, while the FSR keeps the constant of 10.22 GHz determined by the $\Delta l$ of 20 mm. Here, the SR is defined as the ratio of the height of the first side-peak to that of the main resonant peak, as marked as $SR = {b \mathord{\left/ {\vphantom {b a}} \right.} a}$ in the figures. Generally, the SR of a filter should be less than 0.50 for good filtering effect. To study the variations of FWHM, transmittance and SR of the DCR-CC filter with the changing of coupling ratio $\alpha $ of four OCs comprehensively, we plotted Fig. 3(e). After making a tradeoff, we decided to use the coupling ratio $\alpha $ of 0.1 for the DCR-CC, and the simulated spectrum with a FWHM of 8.67 MHz and a SR of 0.37 is shown in Fig. 3(f).

 

Fig. 3. (a) Demonstration of FWHM variation of DCR-1 with the changing of coupling ratio $\alpha $ of two OCs used. (b) Transmission spectrum of DCR-CC filter when the coupling ratio $\alpha $ of four OCs used is 0.2; (c) Transmission spectrum of DCR-CC filter when the coupling ratio $\alpha $ of four OCs used is 0.4; (d) Transmission spectrum of DCR-CC filter when the coupling ratio $\alpha $ of four OCs used is 0.5. (e) Variations of FWHM, transmittance and suppression ratio (SR) of DCR-CC filter with changing of coupling ratio $\alpha $ of four OCs used. (f) Transmission spectrum of DCR-CC filter when the coupling ratio $\alpha $ of four OCs used is 0.1. Above, the cavity lengths of DCR-1 and DCR-2 are constant as $C1 = 60$ cm and $C2 = 62$ cm respectively.

Download Full Size | PDF

Based on the determined parameters of DCR-CC, to show the combined filtering effect of SI-HBFBG and DCR-CC, we plotted the simulated spectrum of DCR-CC and the measured reflecting spectrum of SI-HBFBG together in Fig. 4(a), and multiplied them by each other as shown in Fig. 4(b). As can be seen, only four passbands with a FWHM of ∼8.40 MHz, as shown in the inset, are left for final SLM selection of the proposed 4WS-EDFL.

 

Fig. 4. (a) Simulated spectrum of DCR-CC and measured reflecting spectrum of SI-HBFBG; (b) Spectrum obtained by multiplying two spectra in (a); inset showing enlarged spectrum of the third passband.

Download Full Size | PDF

3. Experimental results and discussion

3.1 Fabrication and characterization of DCR-CC filter

Based on the parameters of DCR-CC determined theoretically, using four commercial OCs, we firstly fabricated the DCR-1 and DCR-2 respectively and measured their transmission spectra as shown in Fig. 5(a), and for comparison the simulated transmission spectra of DCR-1 and DCR-2, calculated using the measured $\alpha $ and ${\gamma _i}{\kern 1pt} {\kern 1pt} ({i = 1,2,3,4} )$ of OCs and $\delta $ for every splicing point, are shown in Fig. 5(b). Note that, in this paper, we measured the transmission spectrum of a filter by inputting a sweeping-laser from a wavelength-swept laser source (Yenista T100S-HP) into the filter and then detecting its output using a 400 MHz photodetector (PD, Thorlabs PDB470C) monitored by a data acquisition card (DAQ, Measurement Computing Cor. USB-1602HS); The laser source is with a sweeping-speed of 1 nm/s and a sweeping-range from 1554∼1558 nm, and the DAQ is with a data sampling rate of 500 kHz; In the data processing, the wavelength and the data sampling interval are calibrated each other, and the laser power and the voltage are calibrated each other as well. By carefully adjusting the cavity lengths of DCR-1 and DCR-2 to $C1 = 59.94$ cm and $C2 = 61.90$ cm in simulations, as given in Fig. 5(b), the FSR and FWHM of DCR-1 and DCR-2 same with those of them given in the measured spectra in Fig. 5(a) were obtained, indicating that the handmade errors of DCR-1 and DCR-2 in cavity lengths were 0.10% and 0.16% respectively. Then the DCR-CC filter was obtained by cascading the DCR-1 and DCR-2 with the $L5$ length of 0.50 m.

 

Fig. 5. (a) Measured transmission spectra of DCR-1 and DCR-2. (b) Simulated transmission spectra of DCR-1 and DCR-2 using measured parameters of $\alpha $, ${\gamma _i}{\kern 1pt} {\kern 1pt} ({i = 1,2,3,4} )$, and $\delta $. (c) Measured transmission spectrum of DCR-CC and reflecting spectrum of SI-HBFBG. (d) Simulated transmission spectrum of DCR-CC using same parameters with experiment and measured reflecting spectrum of SI-HBFBG. (e) Spectrum obtained by multiplying two spectra in (c); inset showing enlarged spectrum of the third passband. (f) Spectrum obtained by multiplying two spectra in (d).

Download Full Size | PDF

The transmission spectrum of DCR-CC measured is shown in Fig. 5(c), with the reflecting spectrum of SI-HBFBG for comparison, and the combined filtering effect are given by multiplying the two spectra each other as shown in Fig. 5(e). Additionally, for comparison, the simulated transmission spectrum of DCR-CC based on the parameters same with them in experiment is shown in Fig. 5(d), and the theoretical combined filtering effect of DCR-CC and SI-HBFBG is given in Fig. 5(f). Figures 5(e) and 5(f) show a good consistence between theoretical and experimental results. As can be seen in Fig. 5(e), four pure ultra-narrow passbands with SRs of <0.5, centered at 1555.863 nm, 1556.114 nm, 1556.354 nm and 1556.604 nm, are finally achieved for filtering, and, as shown in inset of Fig. 5(e), the FWHM of the four filtering passbands is ∼8.20 MHz, which is narrow enough for the SLM selection of an EDFL with a MRC length of ≤20 m. Note that, in the inset, slight fluctuation is seen in the enlarged spectrum, and that was from the laser frequency jitter during the low speed wavelength-sweeping, which has been verified experimentally.

Since the reflectivities of the four reflecting channels of the SI-HBFBG are highly polarization-dependent, with the fabricated DCR-CC filter and the SI-HBFBG inside the 4WS-EDFL in Fig. 1, the minimum single-pass cold cavity-losses of the four lasing wavelength channels were measured to be 26.80 dB, 28.76 dB, 27.05 dB and 27.50 dB respectively after turning off the pump LD and disconnecting the port 1 of the circulator and the WDM. In the measurement we used the T100S-HP wavelength-tunable laser connected with a DI-PC to control the SOP into port 1 and a hand-held power meter to monitor the remainder power.

3.2 Single-wavelength operation

Based on the excellent mode selecting capability of the DCR-CC with the SI-HBFBG, using a pump power of 200 mW for demonstration, through adjusting the PCs inside the 4WS-EDFL carefully, stable single-wavelength operation lasing at $\lambda 1$, $\lambda 2$, $\lambda 3$ and $\lambda 4$ respectively were obtained as shown in Fig. 6 and the switchable-operation among four wavelengths was implemented easily. Figures 6(a)–6(d) show the medium-term running stability of the 4WS-EDFL at all lasing wavelengths in a measurement time span of ∼150 min respectively, measured by the repeated scanning of OSA using a resolution of 0.02 nm and a data sampling interval of 0.001 nm. As can be seen, the four lasers are all with little to no wavelength fluctuation ${f_{\lambda i}}{\kern 1pt} {\kern 1pt} {\kern 1pt} (i = 1,2,3,4)$ (Maximum: 0.003 nm) and extremely low power fluctuation ${f_{Pi}}{\kern 1pt} {\kern 1pt} {\kern 1pt} (i = 1,2,3,4)$ (Maximum: 0.093 dB), and their OSNRs are all higher than 80 dB. Besides, the four wavelengths lasing at ∼1555.879 nm, ∼1556.145 nm, ∼1556.359 nm and ∼1556.605 nm are basically consistency with the center wavelengths of the four passbands in Fig. 5(e). We believe the little instability and wavelength deviation were mainly induced by the fluctuation of ambient temperature, the RIN of pump LD and the inevitable vibrations of surroundings.

 

Fig. 6. Spectra of single-wavelength switchable operations lasing at (a) $\lambda 1$, (b) $\lambda 2$, (c) $\lambda 3$ and (d) $\lambda 4$ respectively measured in a time span of ∼150 min. ${f_{\lambda i}}{\kern 1pt} {\kern 1pt} {\kern 1pt} (i = 1,2,3,4)$: fluctuation of lasing wavelength at $\lambda i$; ${f_{Pi}}{\kern 1pt} ({\kern 1pt} i{\kern 1pt} = 1,2,3,4)$: fluctuation of power lasing at $\lambda i$. OSNR: optical signal to noise ratio. Note that in each figure 15 repeated spectra measured by OSA with an interval of ∼ 10 min.

Download Full Size | PDF

The longitudinal-mode characteristics of the 4WS-EDFL at each lasing wavelength were investigated by a scanning Fabry–Pérot interferometer (Thorlabs, SA200-12B) with a FSR of 1.5 GHz and a resolution of 7.5 MHz, as shown in Fig. 7. As can be seen from Figs. 7(a)–7(d), the 4WS-EDFL was operating in a stable SLM state at every lasing wavelength. The SLM operation of the 4WS-EDFL was further confirmed using the self-homodyne method with a 400 MHz PD (Thorlabs PDB470C) and a radio frequency (RF) electrical spectrum analyzer (ESA, Keysight N9010A), as shown in Fig. 8(a), and we see that in a ∼10 min measurement using the maximum-hold (MH) mode of the ESA there is no any beating signal captured for every lasing wavelength. In order to verify the capability of DCR-CC filter, the RF beating spectra of laser output for every single-wavelength operation were measured when the DCR-CC was replaced by a section of SMF to maintain the original MRC length as shown in Fig. 8(b), and numerous spikes are seen, indicating that the 4WS-EDFL was lasing with dense longitudinal-modes. The spacing of adjacent peaks is ∼11.20 MHz corresponding to a length of MRC of ∼18.25 m, which is high consistent with the measured MRC length of ∼18.20 m. We also investigated the mode-hop characteristics of the 4WS-EDFL in every one of four switchable single-wavelength operations using a delayed self-heterodyne measurement system (DSHMS) composed of the 400 MHz PD, a Mach-Zehnder interferometer (MZI) with a 200 MHz acoustic optical modulator (AOM) and 100 km long SMF in two arms respectively, and the RF ESA. Using the MH mode of ESA in a ∼30 min measurement, the results in ranges of 0∼250 MHz and 175∼225 MHz are shown in Figs. 8(c) and 8(d) respectively. Due to the use of 100 km SMF delay-line in the MZI, any mode-hopping could be captured. However, only the strong beating signal at ∼200 MHz introduced by the AOM was captured in both situations, and considering that the longitudinal-mode spacing of MRC is ∼11.20 MHz, we believe that the 4WS-EDFL has the potential to work in stable SLM without mode-hop for a long time.

 

Fig. 7. Longitudinal-mode characteristics, measured by a scanning Fabry–Pérot interferometer, in single-wavelength operation of the 4WS-EDFL lasing at (a) $\lambda 1$, (b) $\lambda 2$, (c) $\lambda 3$ and (d) $\lambda 4$ respectively.

Download Full Size | PDF

 

Fig. 8. RF spectra measured by ESA for each single-wavelength operation of the 4WS-EDFL. (a) Self-homodyne RF spectra measured in a range of 0∼400 MHz using maximum-hold (MH) mode of ESA in ∼10 min with resolution bandwidth (RBW) of 51 kHz. (b) Self-homodyne RF spectra measured in a range of 0∼400 MHz in single-scan-mode of ESA with RBW of 51 kHz when DCR-CC replaced by SMF. Delayed self-heterodyne RF spectra measured in ranges of (c) 0∼250 MHz using MH mode of ESA in ∼30 min with RBW of 51 kHz and (d) 175∼225 MHz using MH mode of ESA in ∼30 min with RBW of 30 kHz. AOM: acoustic optical modulator.

Download Full Size | PDF

Using the DSHMS, the laser linewidth of 4WS-EDFL’s output at each lasing wavelength was studied. Using the average-mode of ESA for 100 times, Figs. 9(a)–9(d) show the measured delayed self-heterodyne RF beating spectra of four lasing wavelengths respectively and the Lorentz fitted curves of the experimental data with high adjusted R-Square (Adj. R-Square) of 0.99353, 0.99432, 0.99220 and 0.99168 respectively. As marked in four figures, the linewidths of four switchable single-wavelength lasers are 690 Hz, 680 Hz, 688 Hz and 685 Hz respectively, obtained by calculating the 1/20 times of the 20-dB bandwidth of the fitted curves. It is worth noting that, the exact linewidth for each laser should be measured using a SMF delay-line over 1000 km to achieve completely incoherent mixing of two arms of the MZI theoretically [26,28], since the interference effect can lead to a broadening of self-heterodyne spectrum. However, due to the limitation of output laser power and serious 1/f frequency noise from the ultra-long delay-line [43], it is impossible to obtain a pure Lorentzian linewidth spectrum. Consequently, considering the measured results contain the unavoidable broadening effects aforementioned and are all well fitted with Adj. R-Square values close to one, we believe that the linewidths obtained must be larger than the real values of lasers and can be regarded as conservative characterizations of the natural linewidths of our 4WS-EDFL for future practical applications.

 

Fig. 9. Measurements of laser linewidths for each single-wavelength operation of the 4WS-EDFL using the DSHMS. Delayed self-heterodyne RF beating spectra of output lasing at (a) $\lambda 1$, (b) $\lambda 2$, (c) $\lambda 3$ and (d) $\lambda 4$ respectively in 199.950∼200.050 MHz using average mode of ESA with 100 Hz RBW.

Download Full Size | PDF

The output power stability of the 4WS-EDFL was studied by measuring the laser output power fluctuation in a 30 min time span using a power meter with a sampling rate of 1 Hz. The measurement results of four lasing wavelengths are shown respectively in Figs. 10(a)–10(d). As can be seen, the maximum of laser output fluctuation is as low as 0.555 dB, indicating a good output stability of our 4WS-EDFL. Additionally, we also give the average output powers for the four operations respectively, which are slightly different, mainly due to the different total cavity-losses resulting from the polarization adjustments of light, for the four lasing operations. Besides, to investigate the high-speed fluctuation of laser output, we measured the RIN spectrum of output at each lasing wavelength using the 400 MHz PD, an oscilloscope (Tektronix, TDS2024C) and the N9010A ESA, as shown in Figs. 11(a)–11(d) respectively. In each figure, for comparison, the RIN spectrum of a commercial DFB laser (Han’s Raypro Sensing product, Model: RP-MP-10-0100-02, a single-frequency laser at 1550 nm with ultra-low RIN) is also plotted. We can see that the RIN of our 4WS-EDFL is less than that of the commercial low-noise DFB laser at every lasing wavelength and, when the frequency is larger than 3 MHz, the RIN of our fiber laser is ≤−156.7 dB/Hz for all four lasers. Meanwhile, we had measured the relaxation oscillation peaks for the lasers as shown in the inset of each figure and all of them are close to 38 kHz, lower than that of 60 kHz measured in our former work [28] since in this work the MRC length of ∼18.25 m is longer than that of ∼11.20 m in [28], which is consistent with the theory [44,45].

 

Fig. 10. Output power stabilities of four switchable single-wavelength operations of the 4WS-EDFL measured by a laser power meter using a data sampling rate of 1 Hz. ${f_{o\_\lambda i}}{\kern 1pt} {\kern 1pt} {\kern 1pt} (i = 1,2,3,4)$: fluctuation of output power lasing at $\lambda i$. Additionally, the average output powers also given for all of the laser operations.

Download Full Size | PDF

 

Fig. 11. RIN spectra of 4WS-EDFL lasing at (a) $\lambda 1$, (b) $\lambda 2$, (c) $\lambda 3$ and (d) $\lambda 4$ respectively in 0∼5 MHz with 10 kHz RBW of ESA; for comparison, in each figure, RIN spectrum of a commercial low-noise DFB laser also plotted; insets showing the same measurements in 0∼500 kHz with 100 Hz RBW of ESA.

Download Full Size | PDF

Finally, we also investigated the SOP characteristics of the laser output at $\lambda 1$, $\lambda 2$, $\lambda 3$ and $\lambda 4$ respectively using a polarization analyzer (General Photonics Cor., PSY-201) as shown in Figs. 12(a)–12(d) and each measurement was accumulated continuously for 5 min. Note that during the measurements the laser output fiber jumper connecting to the PSY-201 was fixed very well using adhesive tape in order for avoiding the external disturbance to the SOPs. As can be seen, the degree of polarization (DOP) of each laser is stable and larger than 99.8%, and the SOPs of lasers $\lambda 1$ and $\lambda 2$/lasers $\lambda 3$ and $\lambda 4$ are orthogonal respectively while the SOPs of lasers $\lambda 1$ and $\lambda 3$/lasers $\lambda 2$ and $\lambda 4$ are almost parallel respectively.

 

Fig. 12. SOP measurements of 4WS-EDFL lasing at (a) $\lambda 1$, (b) $\lambda 2$, (c) $\lambda 3$ and (d) $\lambda 4$ respectively. As can be seen, SOPs of lasers $\lambda 1$ and $\lambda 2$/lasers $\lambda 3$ and $\lambda 4$ are orthogonal respectively while SOPs of lasers $\lambda 1$ and $\lambda 3$/lasers $\lambda 2$ and $\lambda 4$ are parallel respectively.

Download Full Size | PDF

3.3 Multi-wavelength operation

By adjusting the two PCs, still using the pump power of 200 mW, we obtained six operations of dual-wavelength lasing respectively at $\lambda 1$&$\lambda 2$, $\lambda 1$&$\lambda 3$, $\lambda 1$&$\lambda 4$, $\lambda 2$&$\lambda 3$, $\lambda 2$&$\lambda 4$, and $\lambda 3$&$\lambda 4$ for our 4WS-EDFL, as seen in Figs. 13(a)–13(f). In each one, the measurement in a time span of ∼150 min was carried out by repeatedly scanning the OSA. As can be seen, in every situation, the fluctuations of lasing wavelength (${f_{\lambda i}}{\kern 1pt} {\kern 1pt} {\kern 1pt} (i = 1,2,3,4)$) are less than the resolution of 0.02 nm of the OSA and the OSNR of every laser is higher than 79 dB. Besides, we can see that the fluctuations of powers (${f_{Pi}}{\kern 1pt} ({\kern 1pt} i{\kern 1pt} = 1,2,3,4)$) in Figs. 13(b) and 13(e) are relatively higher than those in other figures, because the lasers $\lambda 1$&$\lambda 3$ and lasers $\lambda 2$&$\lambda 4$ are respectively with parallel polarizations, as displayed in Fig. 12, while in other dual-wavelength operations the two lasing wavelengths are all with orthogonal polarizations. That indicates the PHB effect was stronger in the EDF for the input lights with orthogonal polarizations than those with parallel polarizations. However, despite that, based on the data in Fig. 13, the stability of our 4WS-EDFL in all dual-wavelength operations is satisfactory compared with others in publications [22,46]. Besides, based on the excellent mode selection capability of the DCR-CC filter combing with the SI-HBFBG, the SLM lasing with a linewidth of <1 kHz for each laser in dual-wavelength operation was easily obtained and had been verified experimentally, similar with that in Fig. 9, but for saving space we decided not to show more similar figures.

 

Fig. 13. Spectra of dual-wavelength switchable operations lasing at (a) $\lambda 1$&$\lambda 2$, (b) $\lambda 1$&$\lambda 3$, (c) $\lambda 1$&$\lambda 4$, (d) $\lambda 2$&$\lambda 3$, (e) $\lambda 2$&$\lambda 4$, and (f) $\lambda 3$&$\lambda 4$ respectively measured in a time span of ∼150 min. The definitions of ${f_{\lambda i}}{\kern 1pt} {\kern 1pt} {\kern 1pt} (i = 1,2,3,4)$, ${f_{Pi}}{\kern 1pt} ({\kern 1pt} i{\kern 1pt} = 1,2,3,4)$ and OSNR are same with them in Fig. 7. Note that in each figure 15 repeated spectra measured by OSA with an interval of ∼ 10 min.

Download Full Size | PDF

Based on the principles of wavelength-switchable operation and the polarization-dependent cavity-loss adjustment described in Section 2.2, the switchable operations of power-equalized three-wavelength and four-wavelength lasing were also verified and characterized experimentally. Figure 14 shows the typical spectra of four operations of three-wavelength lasing respectively at $\lambda 1$&$\lambda 2$&$\lambda 3$, $\lambda 1$&$\lambda 2$&$\lambda 4$, $\lambda 1$&$\lambda 3$&$\lambda 4$ and $\lambda 2$&$\lambda 3$&$\lambda 4$, and the typical spectrum of four-wavelength lasing at $\lambda 1$&$\lambda 2$&$\lambda 3$&$\lambda 4$ for the 4WS-EDFL. As can be seen, the OSNR in every situation is higher than 79 dB, indicating the 4WS-EDFL in multi-wavelength operation was still with high output quality. But, in experiments, we found that the switching among five multi-wavelength operations was relatively harder than that among dual-wavelength and single-wavelength operations, since the extremely strong wavelength competition in multi-wavelength lasing was very difficult to suppress even though the enhanced PHB effect was working.

 

Fig. 14. Spectra of three-wavelength and four-wavelength switchable operations of the proposed 4WS-EDFL.

Download Full Size | PDF

4. Conclusions

We reported what we believe to be the first 4WS-EDFL using a SI-HBFBG as a four-wavelength-channel reflecting filter and a cascaded DCR-CC filter as a SLM selector. Thanks to the use of the SI-HBFBG, by adjusting the two PCs carefully, enhanced PHB effect in the 2.9 m EDF coiled inside the TL-PC and polarization-mismatch-induced losses at the fiber grating were introduced into the laser cavity to achieve wavelength-switching operation. The procedure of designing, fabricating and characterizing of the DCR-CC, including theoretical analysis and experimental results, was presented in detail, and we believe that is the first systematic approach for making a compound-cavity based filter used for selecting SLM in a fiber laser. By combining the SI-HBFBG and the DCR-CC, four filtering passbands all with a SR <0.5 and a FWHM of ∼8.20 MHz were obtained, which could ensure the 4WS-EDFL with the MRC length of ∼18.2 m output a laser in SLM operation regardless of the lasing wavelength. The switching among fifteen states including four single-wavelength operations, six dual-wavelength operations, four three-wavelength operations and one four-wavelength operation, lasing at any combination of the four oscillating wavelengths, was validated experimentally. In switchable single-wavelength operations, for demonstrating the high performance of the 4WS-EDFL, the four SLM lasing outputs were characterized in detail, and, for all four lasers, an OSNR of >80 dB, a linewidth of <700 Hz, a RIN of ≤−156.7 dB/Hz at frequencies over 3 MHz, an output power fluctuation of ≤0.555 dB and excellent polarization characteristics were obtained. We believe the performances of our 4WS-EDFL can be improved further if good temperature compensation and vibration isolation packaging are employed in future for practical applications.