1. Introduction

Pulsed lasers are currently experiencing widespread use in micromachining applications, where a pulsed operation can be achieved using various methods. In the nanosecond range, Q-switching is the most commonly used approach. In this case, pulses are usually generated by an acousto-optical modulator [1]; however, alternative approaches can also be used, for example techniques based on a photo-elastic modulator [2, 3]. Pulsed operation can also be achieved using modulation of the pump source, the so-called gain switching, and one of the first studies of this approach [4] was conducted using a Nd:YAG crystalline laser.

Gain switching is based on a relaxation oscillation which appears every time the pump power is switched on. It consists of a series of pulses with amplitudes that gradually decrease with time until a steady state value is reached. It is possible to isolate only the first pulse by switching off the pump power at an appropriate time. By doing so one can achieve, using the pump source alone, pulsed operation of the laser in the nanosecond range with no additional elements inside the oscillator [5]. The complexity of this approach for producing laser pulse is limited by the pump system, and can be kept quite simple if the system is designed thoughtfully. Applying this technique to high-power fiber lasers, which have a number of advantages when compared to solid state lasers such as good thermal power management, stability, and beam quality, a robust nanosecond laser system can be designed. The main building block of fiber lasers is active fiber doped with rare earth ions. The adequate laser properties have enabled several realizations of gain-switched laser designs based on dopants such as erbium-ytterbium [6], neodymium [7], holmium [8], and thulium [9]. The latter dopant has, through faster energy transition dynamics, enabled researchers to achieve short pulses of 10 ns, at 2-μm laser light [10]. Despite good results with regards to short laser pulses, holmium and thulium fibers lose the advantage of simplicity due to a lack of appropriate high-power diode pump sources, and are usually pumped by more complex multistage systems [11].

On the other hand, high-power laser diodes are readily available for the ytterbium-doped fibers, which are used for fiber lasers operating in the most common 1-µm wavelength range. Several reports have been published on methods to achieve appropriate peak power and short pulses in this wavelength range [12, 13].

To obtain an appropriate peak power, such systems usually employ dual staged MOPA (Master Oscillator Power Amplifier) designs [14–16], which provide pulse durations of over 100 ns. Using an all-fiber designed ytterbium-doped gain-switched MOPA [17], pulses as short as 66 ns, with peak powers of 700 W, were achieved from a single-stage laser (oscillator only) pump by a laser diode emitting at 915 nm. By adding an additional stage (amplifier), a peak power of 1.4 kW was reached. Another approach is to combine the Q switched and gain switched section of ytterbium doped fiber laser in an all fiber design [18]. In this case the the gain switched section of the fiber laser serves as a sort of saturable absorber for Q switching. The dynamics of operation at 1040 nm and 1081 nm laser wavelengths caused output laser pulses to be 45 ns long with peak power of 1.4 kW. However, even shorter pulses of 27 ns [19] and higher peak power [20, 21] were achieved using a diode-based pump source emitting at 976 nm.

A short oscillator length and high pump powers are usually required in order to produce short laser pulses from a gain-switched fiber laser. However, in the case of a very short active fiber (i.e. in a range of 1 m or less), a significant amount of pump light may not be absorbed by the oscillator’s active fiber. As the peak power is proportional to the pump power, this leads to a decrease in the laser pulse peak power as well as a decrease in the slope efficiency of such lasers. In this paper, we present a concept of unused pump light recovery using a simplified amplifier stage. Therefore, despite a short pulse duration maintained by a short oscillator, high slope efficiency and high peak powers can be achieved. Because the amplifier is pumped by the unabsorbed pump light from the oscillator, no additional pump laser diodes are required. This proves to be a simple and cost-effective solution that further improves the performance of gain-switched fiber lasers.

2. Experimental setup

The laser system consists of a gain-switched fiber oscillator and a fiber amplifier stage (Fig. 1). The fiber used both for the oscillator and the amplifier was a micro-structured polarization-maintaining single-mode Yb-doped fiber, with 15 μm core diameter, 135 μm cladding diameter and NA of 0.055. The measured M2 is in good agreement with the manufacturer specifications, i.e. M2<1.2. The lengths of the fibers were 53 cm for the oscillator and either 108 cm or 172 cm for the amplifier. The oscillator was pulse-pumped at a repetition rate of 50 kHz with 976-nm light from the temperature-stabilized diodes using a forward-pumping scheme. The pump light was coupled to the angle-cleaved oscillator fiber end. On the other side, the fiber was straight-cleaved to provide 3.5% Fresnel reflection for the laser light. On the pump-side of the oscillator fiber, the laser light was deflected by a dichroic mirror (HT at 976 nm and HR at 1030/1064 nm) onto a diffraction grating to provide selective reflectivity for the desired wavelength. A polarizer was used to maintain the linear polarization of the laser output. Some of the oscillator output laser light was deflected onto a fast photodiode to provide a signal for the control unit, which was used to switch the pump off when the pulse was detected. This enabled a stable pulse-to-pulse operation of the oscillator.

 

Fig. 1 Schematics of the experimental setup. The control unit monitors the signal from the photodiode and consists of a user interface, switching logic, and temperature control. Fresnel reflection from the straight-cleaved fiber end is marked as R2. The pump and laser light were split by a dichroic mirror with high transmissivity (HT) at 976 nm and high reflectivity (HR) at 1030/1064 nm. The output of the oscillator (consisting of the laser and unabsorbed pump light) were coupled into an Yb-doped fiber amplifier.

Download Full Size | PDF

Because a short oscillator fiber is required in order to achieve short pulses from the oscillator, only some of the pump light is absorbed by the oscillator fiber. Therefore, a significant amount of pump light at the oscillator output is coupled to the fiber amplifier, together with the laser light from the oscillator. Since the amplifier does not require any additional pumping diodes, this proved to be a simple and cost-effective technique that significantly improved the slope efficiency of the entire laser system, while keeping the pulse duration relatively short compared to a system with an active fiber for optimal pump absorption.

3. Results and discussion

The slope efficiency with respect to the incident pump power of the laser oscillator at 1030-nm and 1064-nm wavelength was 31% and 25%, respectively. The relatively poor slope efficiency can be attributed to the short active fiber used, where a significant amount of unabsorbed pump power exited the fiber. By coupling both the output laser and pump light into a 108-cm Yb-fiber amplifier, the slope efficiency of the entire laser system was increased to 42% and 41% at 1030 nm and 1064 nm, respectively (Fig. 2). No significant increase in slope efficiency was observed when using a longer amplifier fiber. It should be noted that this efficiency was achieved on free space configurations of the laser setups. For the case of a fully integrated design one can expect better efficiency. However, the current study allows for adequate assessment of the influence of the added amplifier, which operated on unabsorbed pump light from the first stage of the gain-switched laser system.

 

Fig. 2 Output power of the oscillator (blue squares) and a 108-cm fiber amplifier (green circles). The wavelength of the oscillator was set to 1064 nm (using a diffraction grating). The bars denote the fraction of the laser (blue) and unabsorbed pump (red) power at the oscillator output. The slope efficiency increased from 25% of the oscillator alone (blue line) to 41% of the entire laser system (green line).

Download Full Size | PDF

The oscillator alone produced pulses with a duration (FWHM) of 24.5 ns at 1030 nm wavelength and 8.6-W pump power. Here the peak power was 1.8 kW. Because of the specific dynamics of gain-switched lasers, the laser pulse duration ($t_{Lp}$) depends on the pump power (P_pa) and pump pulse duration ($t_p$), which was presented in [12]:

The 1030-nm laser pulse shapes at different pump powers can be seen in Fig. 3. The stimulated emission cross section in Yb-doped fibers was significantly lower at the 1064-nm wavelength, which led to slower laser pulse build-up dynamics and consequently to both longer laser and pump pulse durations. The longer pump pulses also led to a higher average pump power, which may have caused damage to the fiber since a free-space setup was used. Consequently, the measurements at this wavelength were conducted at pump powers limited to a maximum of 11 W. Working at this pump power limit, the pulse duration and peak power were 46 ns and 1.2 kW, respectively.

 

Fig. 3 Shape of the output laser pulse of the oscillator at 1030 nm for three different pump powers. The pulse durations for 3.5-W, 5.3-W, and 8.6-W pump powers are 48 ns, 33 ns, and 24.5 ns, respectively.

Download Full Size | PDF

After coupling the oscillator output to a 172-cm-long fiber amplifier, the laser pulse duration increased slightly (from 24.5 ns to 26 ns at 1030 nm and 8.6 W pump power). This is expected as the amplifier provides some spontaneous emission, which causes a laser pulse to appear earlier within the pump pulse (Fig. 4), meaning that the population inversion is somewhat lower at the time the pulse appears. Therefore the dynamics of stimulated emission are slower, which leads to longer pulses. The longer pulse duration at higher pump powers is in good agreement with our model [12], as shown in Fig. 5(a). At lower pump powers however, there is a small discrepancy between the model and the measurements. It should be noted that the laser with the amplifier was modelled as a whole, meaning that any discrepancies in the oscillator model were carried into the amplifier model. All free-space coupling losses were accounted for in the model. The coupling efficiencies were estimated at around 80% for the pump-to-laser coupling, 70% between the laser and amplifier, and 40% at the diffraction grating.

 

Fig. 4 Pump-enabled signal and pulse output of the oscillator alone (blue, green), and that of the oscillator after it is coupled to the amplifier (red, cyan). The laser pulse appears around 34 ns sooner if the oscillator is coupled to the 172-cm-long amplifier.

Download Full Size | PDF

 

Fig. 5 (a) Laser pulse duration at 1030 nm of the oscillator without amplifier (blue circles), pulse duration of the oscillator after it is coupled to a 172-cm-long amplifier (green squares), and numerical model results for the oscillator without amplifier (blue line) and the oscillator after it is coupled to a 172-cm-long amplifier (green line). (b) Pulse duration at the input (green squares) and output (red diamonds) of the amplifier. The red line shows the predicted pulse duration at the output of the amplifier using our numerical model.

Download Full Size | PDF

The amplifier itself is highly saturated and therefore also has an effect on the pulse shape and duration. This result was compared to a numerical model of a laser amplifier based on [22]. Again the model and the measurements are in good agreement for higher pump powers. The comparison of the pulse duration at the input and output of the amplifier stage, as well as the modeled pulse duration are shown in Fig. 5(b). The peak power using this amplifier stage was increased from the initial 1.8 kW to 2.3 kW at 1030 nm, and from 1.2 kW to 1.8 kW at 1064 nm. Comparing the results for 1030 nm laser wavelength with the previously reported ones from a gain switched laser [19] an increase of laser pulse power from 1.4 kW to 2.3 kW is noted using the similar pump powers while the laser pulse durations are comparable.

The amplifier provided good pulse-to-pulse stability, as the standard deviation of the pulse duration was 2% and the standard deviation of the pulse energy was only 0.8% (Fig. 6).

 

Fig. 6 Pulse train at the output of the 172-cm amplifier at 1030-nm wavelength and 8.6-W pump power. High pulse–to-pulse stability was achieved as the measured standard deviation of the pulse FWHM and pulse energy were 2% and 0.8%, respectively, on a sample of around 100 pulses.

Download Full Size | PDF

In order to find the optimal balance between the oscillator and amplifier fiber lengths, a simplified oscillator model was used [13]. The goal of the simulation was to estimate the optimal (in regard to pulse duration and peak power) oscillator to amplifier length ratio, while keeping the total active fiber length at 225 cm, i.e.:$L_{osc}+L_{amp}=225cm$, which ensures that approximately 98% of the pump light is absorbed in the laser system. The results are shown in Fig. 7. From Fig. 7(a) it can be seen that the oscillator length of around 50 cm is optimal regarding the pulse duration. This is due to the fact that there is always some length of passive fibers or free space involved in such gain switched lasers and therefore further shortening of the active fiber within the cavity is not beneficial. Figure 7(b) shows that adding an amplifier stage with no coupling losses (e.g. spliced to the oscillator) is always beneficial, however in a free space setup this is not always true because of the coupling losses. At a certain length of the oscillator fiber (around 100 cm in this case) there is not enough pump power for the amplifier left and consequently the amplification is lower than the coupling losses between the laser and amplifier. In both cases it can also be seen, that the amplifier peak power is reached at the oscillator length of around 60 cm (amplifier length around 165 cm). This can be explained as with increasing the oscillator length, there is less pump power left on one hand and more laser power on the other hand at the input of the amplifier. Consequently, the increase of the average power at the output of the amplifier is less prominent than the increase of the pulse duration due to the longer oscillator and therefore the peak power decreases.

 

Fig. 7 The output of the laser system with a total active fiber length (oscillator + amplifier) of 225 cm with regard to the length of the oscillators’ active fiber length. (a) Laser pulse duration of the oscillator (blue), amplifier with no coupling losses (green) and amplifier with 70% coupling efficiency. (b) Laser peak power of the oscillator (blue), amplifier with no coupling losses (green) and amplifier with 70% coupling efficiency.

Download Full Size | PDF

4. Conclusion

In this paper, a novel method to increase the slope efficiency and the peak power of a short pulsed gain-switched fiber laser is presented. It uses a short oscillator length to produce laser pulses with duration of 28 ns at 1030-nm lasing wavelength. By recovering the unabsorbed pump light from the oscillator using an amplifier stage, an increase in slope efficiency from 25% to 41% was achieved. The peak power thus obtained was 2.3 kW and the pulse-to-pulse energy stability was 0.8%. Because the amplifier was pumped by the unabsorbed pump light from the oscillator, its gain was also in sync with the output seed laser pulses. This arrangement also means that the amplified, stimulated emission is negligible. The presented setup is therefore an energy and cost-efficient approach to increasing the output power of gain-switched lasers, which usually use short oscillator lengths to produce the desired laser pulse durations and thus sacrifice optical-to-optical efficiency.