1. Introduction

Random fiber lasers (RFL) were first introduced in 2007 [1], using a microstructured fiber with a colloidal solution in its empty core. Following the pioneer proposal of Letokhov in 1968 [2], random lasers, as well as random fiber lasers, differ from conventional lasers because the optical feedback is provided by scatterers rather than by two fixed mirrors. The first unequivocal demonstrations of a random laser (RL) was reported by Lawandy and co-workers in 1994 [3], whereby a colloid consisting of Rhodamine 6G as the gain medium and 250 nm TiO2 nanoparticles as the scatters were appropriately pumped by the second harmonic of a pulsed Nd:YAG laser. RLs can be optically or electrically pumped [4,5], and their characterizations and theoretical aspects have been thoroughly reviewed over the last years [4–7]. RFL are more than a decade younger than RL, but their development has been very fast, particularly after the demonstration of different schemes using random fiber gratings as the scattering media (and erbium ions as the gain) in 2009 [8,9] followed, in 2010, using stimulated Raman scattering (gain) and Rayleigh scattering (scattering mechanism), both in conventional telecom fibers [10,11]. Fiber lasers Q-switched by random phase distributed backscattering exploiting Brillouin gain have been reported, e.g. [12,13], as well as mode-locked hybrid femtosecond lasers [14]. The different architectures using conventional optical fibers triggered great developments and applications of RFL, as reviewed in [15,16]. Among novel architectures for RFL, recent demonstrations employed SOA as the gain media and kilometer-length conventional fibers as scatterers [17–20]. Random fiber lasers have recently been used for speckle-free imaging [21], for spectral super-resolution spectroscopy [22] in the near to mid-infrared spectral region, as platform for complex systems studies of photonic turbulence [23,24], photonic spin-glass behavior [25,26] and Lévy-like statistics of intensity fluctuations [27]. Generally, random fiber lasers incorporate one optical amplifier and a fixed distributed reflector.

We report here an innovative scheme relying on a flexible dual-gain system, formed by a hybrid SOA-EDF combination, potentially capable of several MHz repetition rate and electronically addressable operation. It allowed choosing with ∼1 m resolution the position along the gain fiber where the effective feedback takes place, within many meters. An optically biased erbium doped fiber (EDF) provided population inversion in the cw regime, and a pulsed SOA operating as a fast modulator with gain provided time-resolved pulses. Light oscillation of the amplified spontaneous emission was obtained by synchronizing the frequency of the modulated SOA within a loop mirror, with the arrival to the SOA of the backscattered light from a section of the EDF, thus providing the required scattering feedback mechanism characteristic of random lasers. Laser oscillation was obtained in a modulation frequency span from 570 kHz to 680 kHz and was inhibited by de-tuning the SOA pulses. By tuning the frequency, the random feedback section along the fiber could be chosen electronically in microseconds within 17 meters. We discuss the results in the spectral and temporal regime, explain the random laser mechanism and anticipate extension and applications of this hybrid electrically addressable random laser (HEAR laser).

2. Results

A schematic diagram of the HEAR laser is shown in Fig. 1. The random laser cavity consists of a relatively long (81 m) piece of cw-pumped Erbium doped fiber (EDF) coupled via a 3-dB coupler to a fiber loop mirror containing a SOA and a piece of passive standard single mode fiber (SMF). The SOA was used as a modulator that blocked light circulation in the cavity except for when it was driven by current pulses, when it provided gain. It consisted of a booster amplifier with 15 dBm nominal saturated gain, 27 dB small signal gain at -20 dBm input, 7.5 dB noise figure and 85 nm FWHM bandwidth at 1550 nm. The EDF was a GeO2-SiO2 fiber doped with 168 ppm of Er ions, with cut-off wavelength 1.15 µm, core diameter 4.2 µm, and 4.2 dB/m absorption at 1.53 µm, fabricated by CPqD, Telebras, Brazil. The EDF was optically pumped with cw radiation at 1480 nm and was angle-cut at the end. A polarization controller (PC) allowed adjusting the polarization of the signal light between the loop mirror and the EDF. The output port of the 3 dB coupler was used for simultaneous temporal (TDS 3032B, 300 MHz digital oscilloscope + PDB460C, 200 MHz, NIR photodetector) and spectral measurements (86142B Optical Spectral Analyzer, 0.06 nm spectral resolution). During the laser studies, a small fraction (10%) of the light from the EDF entering the loop mirror was tapped to the photodiode and added to a sample of the optical signal produced by the gated SOA, as shown in Fig. 1. This allowed us visualizing on the oscilloscope the relative delay between the pulses returning from the EDF and the (much larger) pulses newly gated and amplified by the SOA. This tap could be removed during normal laser operation and use. Except for the power level, the spectral characteristics of the tap and output ports are the same.

 

Fig. 1. Simplified schematic diagram for the HEAR laser. Gain is provided by the pulsed unidirectional semiconductor optical amplifier and by the cw pumped Erbium doped fiber. The EDF also provides the necessary backscattering for random laser action. Adjustable frequency of the driving pulses defines the addressed section of the EDF. The 10% fiber tap used during laser characterization could be removed during normal operation.

Download Full Size | PDF

The unidirectional SOA was excited by 16 ns pulses of up to 14 V amplitude (∼1 A current) through a 11 Ω series resistor, which was used to improve impedance matching of the 50 Ω electronics to the low resistance of the SOA alone. The pulsed peak current was twice the nominal CW current, so that we estimate the small signal pulsed gain to be around 30 dB. The drive pulses were produced by an HP 8011A pulse generator triggered externally by an RF signal source Agilent N9310A with nominal 1 Hz precision over 2.3 GHz, providing good timing jitter. Automated frequency modulation was available, so that the frequency of drive pulses to the SOA was varied in small steps (e.g., 50-100 Hz) while the laser performance was monitored. In some experiments, shorter duration pulses (10 ns, 10 V) were generated by a Picosecond Pulse Labs pulse generator (Picosecond Step Generator Model 4050). The shorter pulses improved the spatial resolution of the experiments but reduced the peak power. The frequency of the HEAR laser was brought to a convenient range (570 - 680 kHz) by adding to the loop mirror 230.5 m of passive single-mode fiber. A LABVIEW program was used to control data handling for automated frequency scanning and simultaneous spectral-temporal data acquisition.

During the early stages of the experiment, the exact length of SMF fiber, including the (1480/1550) WDM, the three waveplate polarization controller, the SOA and coupler pigtails, available in the spool was determined by replacing the EDF in the set-up by a 1-m long piece of standard fiber cleaved to provide a 4% end reflection at a known position. The conventional fiber laser thus constructed operated at the fundamental modulation frequency found to be 868 kHz, enabling an optical pulse to circulate along the cavity.

Once the resonance frequency was measured and the spool length determined, we replaced the 1-m long SMF by the EDF. Figure 2 shows the spectral-temporal results for a cw 1480 nm pump power of 110 mW, and for 16 ns duration 1A square current pulses applied to the SOA. The modulation frequency was scanned in the range from 608.1 kHz to 570 kHz. As the modulation frequency was reduced, sections further along the EDF were addressed periodically.

 

Fig. 2. (a) spectral and (b) temporal behavior of the HEAR laser at modulating frequencies 570 kHz (red trace, below threshold) and 608.1 kHz (blue trace, above threshold) under 110-mW cw pump power.

Download Full Size | PDF

The two examples illustrated in Fig. 2 clearly show the alternation between regimes of plain amplified spontaneous emission (ASE) (red traces, here at a frequency 570 kHz) and random lasing (blue traces, at f = 608.1 kHz). Figure 2(a) (red) shows two relatively broad spectral peaks at ∼1532 and ∼1550 nm, characteristic of Erbium-doped fiber amplified spontaneous emission, measured with the optical spectrum analyzer. Note that with the set-up used, the cw emission of the EDF does not reach the spectral analysis port of the coupler, except when the SOA is gated. Figure 2(a) then implies that the SOA’s output spectrum is dominated by the amplified EDF light input even when the laser is below threshold. The blue trace in Fig. 2(a) shows the spectral narrowing measured when laser action takes place, at a wavelength ∼1552 nm.

Figure 2(b) shows the temporal behavior of the output together with the backscattered light from the EDF measured with the photodiode and oscilloscope. The main (left) peak of red and blue traces, with a pulsewidth ∼16 ns, is the output generated by the modulated SOA. A lower amplitude ∼0.4 µs wide band at around 0.7 µs is also visible here (right peak). It is the result of the amplified backscattering of the SOA pulse as it travels along the forward direction in the EDF, detected via the 10% tap in the loop mirror. To bring both signals to a comparable scale, the SOA pulse was attenuated before detection.

Another clear evidence of the laser character of the system described can be seen in Fig. 3, where the cw pump power to the Erbium-doped fiber was varied and both the emission intensity [Fig. 3(a)] and spectral width [Fig. 3(b)] of the ASE (red traces) and laser (blue traces) were measured. The SOA drive pulse amplitude was 607 kHz and Vc was 9.9 V (measured on a 50 Ω oscilloscope), their best value for maximum RL emission. For cw pump powers above ∼102 mW, a clear inflection (indicated by the dashed vertical line) on the emission intensity was observed, together with a corresponding appearance of a sharp spectral line characteristic of the onset of laser action as from Fig. 2(a). Above threshold the linewidths of the two spectral components were separately calculated and plotted. To make the behavior of both components clear, the ASE peak intensity was interpolated and also plotted, together with the total intensity above threshold.

 

Fig. 3. (a) Emission peak intensity and (b) spectral width of HEAR laser output as a function of EDF pump power at 607 kHz modulation rate and 9.9 V bias pulses. The blue traces refer to the laser emission peak and the red traces to the ASE. The solid lines are a guide to the eye. A change in regime is observed at ∼102 mW, when the laser goes above threshold. Note that ASE remains as a pedestal also above threshold, and it is plotted in the graph.

Download Full Size | PDF

The temporal response of the HEAR laser as the frequency of the current pulses applied to the SOA is varied is shown in Fig. 4. Figure 4(a) displays a 3D plot with 1100 time-resolved traces. As in Fig. 2(b), (a) 16 ns peak corresponding to the SOA’s pulsed emission is followed by the broad (∼0.4 µs) bell-shape signal, which is measured via the tap port in the fiber loop. As mentioned above, it is found that strong laser action takes place when addressing 1.6-m long sections of the EDF between 43 m and 60 m, i.e., the EDF gain is too low for lengths below 43 m and the loss due to pump absorption too high beyond 60 m. The 17 m length that can be electronically addressed corresponds to the frequency range 580 kHz to 680 kHz, including all the fiber in the cavity (∼240 m). This length was limited by various parameters, including the 1480 nm pump power, the absorption of the pump and the level of back propagating Rayleigh scattering in the EDF. Closer inspection of the SOA pulses (peaks on the left-hand side) in Fig. 4(a) reveals that the amplitude measured varies with the addressed feedback section of the EDF and whether the laser is above or below threshold. Also, Fig. 4(b) shows a zoom within the backscattered signal at a shorter modulation frequency range, where we can clearly observe the lasing process. The black trace shows the onset of lasing, which becomes optimized at a given modulation frequency and then fades away as the modulation frequency varies beyond the optimum. The several peaks around the lasing one arise from Phase OTDR-like (Φ-OTDR) processes occurring simultaneously in our system, as discussed in the next section.

 

Fig. 4. (a) Temporal response of HEAR laser emission with frequency modulation under 140 mW pump power; (b) temporal response for selected modulation frequencies, (c) contour plot of temporal response shown in (a).

Download Full Size | PDF

Figure 4(c) allows visualizing the timing of the laser pulses as a function of modulation frequency, by triggering the oscilloscope synchronized to the peak of the tapped EDF signal. The top view displayed in Fig. 4(c) conveys similar information as in Fig. 4(a). By fixing the random laser pulses at the tap port as a vertical line at ∼1.18 µs, one now sees the effect of varying the frequency of the drive pulses, and the relative timing of the laser pulses and the bell-shaped pulse from the EDF. One can address the scattering from various segments of the EDF by simply adjusting the modulation frequency.

Figure 5 illustrates the spectral behavior of the light circulating inside the HEAR laser when the frequency of drive pulses to the SOA is varied. The 1100 measurements included in the 3D plot seen in Fig. 5(a) display the output intensity versus modulation frequency and wavelength. The measurement is accompanied by the peak optical power generated by the system, seen in Fig. 5(b). The same information is shown as a contour plot of the spectral response [Fig. 5(c)], where it is clear that the HEAR laser wavelength shifts randomly in the absence of a spectral filter. For clarity, the individual traces for five distinct frequency values and reduced wavelength range are shown in Fig. 5(d). The HEAR laser tends to remain above threshold in one of the two Erbium-related bands (1.53 µm and 1.55 µm), which can be selected by polarization control. Careful inspection of Fig. 5(d) shows that above threshold, the broadband EDF emission suffers depletion at the expense of the intense laser pulse.

 

Fig. 5. (a) Spectral response of HEAR laser emission with frequency modulation under 140 mW pump power; (b) variation of peak optical power with modulation frequency; (c) contour plot of spectral response shown in (a); (d) spectral response for selected modulation frequencies at reduced wavelength range.

Download Full Size | PDF

3. Discussion

The HEAR laser is distinct in various aspects from previously reported random fiber lasers [16], or even from other SOA based random lasers already reported [17–20]. It has two amplifiers in the laser cavity, both of which providing more than optical gain. The sub-mm long SOA guarantees nanosecond modulation and electrical control, while the EDFA, with tens-of-meters length, provides for distributed random-phase feedback. As explained above, electronic adjustment of the modulation frequency resulted in microsecond control of the cavity length of the HEAR laser and the choice of the fiber section providing random scattering. It is important to notice that the scattering fiber length used (∼1 m) was a couple of orders-of-magnitude shorter than previously needed in random fiber laser demonstrations using passive fibers [16]. This opens the possibility, with appropriate electronics and fibers, to electronically adjust the length of the cavity by up to 50% (>10% demonstrated here).

We estimate below the loss and gain conditions prevailing, starting by listing the passive loss of the cavity which adds to 9 dB: Coupler 6 dB (double pass loss); tap 0.5 dB; WDM and connectors in EDF arm (double pass) 1 dB; connectors in loop 0.5 dB. It is reasonable to add an extra 1 dB of excess loss unaccounted for, reaching 9 dB. The backscattering loss is estimated at -60 dB/m. The typical Rayleigh level expected from one meter of SMF fiber is ∼-70 dB [28], but in a high-NA Er-doped fiber, this can be much higher and indeed to reach one order of magnitude higher [28,29]. Therefore, the total loss of the cavity is estimated to be -69 dB. The small signal gain of the SOA is ∼30 dB. For the laser to reach threshold, it is then necessary that the gain in the EDF per pass reaches ∼20 dB. Even considering saturation when the EDF is probed by intense SOA pulses, this double-pass small signal gain level (40 dB) is quite reasonably achieved. The way the cavity is constructed leads to double-pass amplification in the EDF and once in the SOA, compensating the high loss. This is in contrast to most random fiber lasers described previously, which even in a double-pass amplification regime lack the extra 30 dB gain from an additional amplifier and thus require a long scattering fiber or a grating to reach threshold.

The interplay between the different noise mechanisms and level dynamics that semiconductor and rare-earth gain media exhibit raises interesting questions regarding the noise, the coherence and the statistics of the HEAR laser. For example, under weak EDF pumping conditions and high-current pulsed bias, the laser behavior may be well described by strong re-amplified SOA pulses scattered in a random medium. In contrast, for strong EDF pumping and low currents, the laser may behave as dominated by a rare-earth amplifier periodically gated by the SOA. The transition between such regimes deserves further investigation.

Simultaneous visualization of the output signal together with the signal which was fed back to the SOA provided rich information to the interpretation of the laser behavior. The tap port signal of the HEAR laser can also be seen as a time-domain reflectometer trace of SOA probe pulses. For low accumulated gain, for example, for a de-tuned SOA, a relatively smooth bell-shaped pulse is measured in the tap. It maps the time-resolved double-pass gain in the distributed fiber amplifier as in a standard OTDR [29]. Because of the limited detector sensitivity, the noise floor is high and only the more amplified section can be seen. The accumulated gain is always small near the entrance of the EDF (L < 43 m), where the propagation length is still short, and towards its end (> 60 m), where the attenuation is strong because of pump depletion. In the remaining region of the EDF, the feedback can be sufficient for random laser action. In such a case, a narrow temporal peak is formed over the broad bell-shaped pulse, concurrently with the spectrally narrow feature seen in Fig. 2(a). Above threshold, in a tuned condition, the spectral width sharpens down to ≤ 0.5 nm, narrow enough for the OTDR-like backscattered signal to present phase sensitive fluctuations as a Φ-OTDR. This is clearly shown in the backscattered intensity oscillations displayed in Fig. 4(b). In fact, 10-30% intensity fluctuations are expected in meter-length resolutions for linewidths around 0.4 nm [30]. This phenomenon also explains the intensity fluctuations as the pump power is increased above threshold in Fig. 3. Indeed, the lasing wavelength varies slightly as the pump power increases, thus changing the conditions of random interference in the addressed section.

It is noticeable in the spectral domain how energy is taken away from the ASE background and transferred to the lasing peak [refer to Fig. 5(d)] as it decreases (from black trace to blue trace). We believe that stronger pulses from the SOA would deplete the gain in the EDF more fully, increasing the laser peak at the expense of the ASE, therefore allowing a nearly background free HEAR laser emission. The random coalescing of the light emission, Fig. 5(b), around the peak of the EDF band shows the intensity fluctuation with the addressed fiber section. Indeed, the modulation frequency maps the position of the resonant lasing section so that the same fluctuations observed in a Φ-OTDR are directly seen here in the HEAR laser intensity. Hence, the HEAR laser can be directly used as a fast distributed sensor within the addressable section of the EDF, the laser intensity being sensitive to external effects over the fiber, such as stress or temperature, with the resolution given by half the modulation pulsewidth as in a Φ-OTDR. The big advantage here is that instead of a tiny backscattered signal measurement as it is in an OTDR, it is the full output of the HEAR laser that provides the sensing signal. Hence, no averaging is needed, and a fast response is available. Even though the addressable section is limited, remote pumping is possible, and all the active devices can be concentrated at the near end of the system. The highest resonance frequency here is limited by the bulk pigtailed devices such as the SOA, WDM and couplers, as well as the EDF itself to a few MHz. However, there is no restriction for overtone driving because there is no ambiguity concerning the position of the addressed section. Therefore, the EDF could be placed as far as a kilometer, the ∼20-m sensitive section still being addressable with ∼ 1 MHz pulses driving the 10th harmonic of the resonance frequency.

4. Conclusion

In conclusion, we have demonstrated a novel electronically addressable random fiber laser, operating in a pulsed mode and potentially capable of MHz range repetition rates. The light guidance inherent to the fiber geometry prevents the multidirectional emission typical of bulk random lasers. The experimental setup is such that it also works as amplified OTDR or phase-OTDR. Although it was demonstrated for EDF, it can also be reproduced using other media gain such as Raman, Brillouin or with Neodymium or Ytterbium doped fibers, with the SOA operating in appropriate spectral regions, therefore expanding the spectral range of the HEAR laser. Considering the mature technological development of EDFAs and SOAs, a packaged HEAR laser is also technologically feasible, making it a turn-key device.