1. Introduction

As a versatile laser source that can be engineered to produce pulses with duration ranging from the femtosecond to the nanosecond regime, Tm-doped fiber lasers (TDFLs) operating in the mode-locked state have been intensively investigated and widely employed in for example materials processing, surgical procedures, LIDAR, and environmental monitoring [1–3]. In general cases, the output of a TDFL is a convention soliton pulse owing to the anomalous dispersion of the single mode fiber at 2 µm [4–6]. However, the obtainable single pulse energy is highly limited with the governing of the soliton area theorem, and the output pulse is prone to split when the energy reaching a specific level (commonly sub-100 pJ) [7,8]. Fortunately, a particular condition called the noise-like pulse (NLP) generation from the mode-locked oscillator can circumvent this problem, and the multi-pulses that resulted from the soliton fission are regulated into a periodic pulse bunch with a stable envelope that has a duration/energy much wider/higher than the transform limited soliton [9,10], and even the NLP with shorter pedestal and spike can be obtained through the nonlinear amplification in recent years [11]. In addition, the further energy scaling of the NLP through fiber amplification is straightforward without any pulse shaping, rendering it attractive for applications such as materials processing, supercontinuum generation and coherence spectral interferometry that do not require high coherence of the laser [12–14].

At present, with the leveraging of new mode-locking regime [15–17], the saturable absorbers such as semiconductor saturable absorption mirrors (SESAM), carbon nanotubes, graphene [18–21], or the artificial saturable absorbers based on fiber nonlinearities such as nonlinear polarization rotation (NPR) and NALM [22–29], a series of demonstrations concerning NLP generation from TDFL has been reported. As the artificial saturable absorbers are more resistant to damage under high energy laser operation, they are preferable to construct mode-locked lasers that emit energetic pulses with high reliability. Nevertheless, most of the literatures choose to implement a non-PM fiber cavity, which inevitably compromises the stability and repeatability of the laser system. Alternatively, an all-PM oscillator has the high system stability due to the merit of insensitive to the external perturbations, and can also ensure a single polarization output, which however has been relatively less investigated for NLP generation in TDFLs [30,31]. For an all-PM laser cavity, the NALM has been recognized as a priority choice for initiating and maintaining the mode-locked operation, as the NPR can be hardly exploited as the intra-cavity polarization cannot be manipulated. Nevertheless, the NALM mode-locked PM fiber oscillators with the conventional figure-of-8 structure have a well-known disadvantage of difficult to self-start operation. This problem could be remedied by adding a phase shifter into the NALM to provide an extra phase offset for assisting the taking place of mode-locking, and it has been successively applied in mode-locked TDFLs with a figure-of-9 structure in recent years [32–34]. In addition to the robustness of an all-PM mode-locked fiber laser, it is also confirmed that the NALM has an inherent amplitude noise suppression mechanism that originate from the interaction between the transmission curve of the NALM and the peak power fluctuations of the laser pulse [35], enabling the realization of impressively low noise laser output [36–38]. However, at present there are no substantial investigations about the noise properties of NLPs, let alone that generated from a NALM mode-locked fiber laser. Although the intensity noise of an NLP has been measured before [39], a systematic examination of the overall noise properties of NLPs is still lacking and is important for understanding the intrinsic performances of this type of laser source.

In this work, we demonstrate a stable NLP generation from an all-PM NALM TDFL. By utilizing a phase shifter to enhance the self-starting ability of the mode-locking, and a chirped fiber Bragg grating (CFBG) with positive dispersion as the output coupler and filter, an NLP output has been obtained with a fundamental repetition rate of 3.25 MHz. Moreover, through measuring the phase noise and intensity noise of the NLP under different output power, and calculating the corresponding integrated time jitter and relatively intensity noise (RIN), the noise evolution of an NLP has been quantitative examined for the first time. Associated with monitoring the stability of the pulse, spectrum and output power, the results consistently show that the overall performance of the NLP laser is improved with the increase of output power.

2. Experimental setup

The experimental setup of the NLP operation all-PM TDFL based on NALM is depicted in Fig. 1. In the NALM loop as shown in the left ring of the setup, a 4.5 m long PM Tm-doped fiber with a core diameter of 10.5 µm was used to provide sufficient gain, with the pumping of a 1550 nm laser through a wavelength division multiplexer. A 50 m long PM passive fiber and a phase shifter which provides a linear π/4 phase bias were employed to provide sufficient nonlinear phase shift difference of the counter-propagated laser signal for enhancing the self-starting ability. A 2 × 2 PM fiber coupler with the splitting ratio of 50:50 was utilized to connect the reflection loop with the NALM loop. A PM CFBG with center wavelength of 1950nm, peak reflectivity of 10%, 3-dB bandwidth of 35 nm, and positive dispersion of −0.2 ps/nm was utilized as a partial reflector to output the generated NLPs through the port 3 of a PM circulator. It is noted that the dispersion compensation effect of the CFBG to the cavity is trivial since it is operated in the transmitting mode which can hardly affects the chirping of the signal. Nevertheless, another important function of the CFBG is that it can suppress the unwanted amplified spontaneous emission (ASE) at the short wavelength side, which is a common problem for a core pumped TDFL [40]. Although there are commercially available tunable filters in the 2 µm band, the corresponding narrow bandwidth (∼2 nm) is unfavorable for the generation of NLPs with a relatively wide spectral bandwidth. In addition, the significant insertion loss (∼4 dB) of commercial filters is also disadvantageous for constructing an efficient and compact fiber laser. The group velocity dispersion (GVD) of the passive fiber and the PM Tm-doped fiber are −0.068 ps²/m and −0.076 ps²/m, respectively [30]. As such, with the total length of the cavity of around 65 m, the calculated net cavity group delay dispersion (GDD) is −4.8 ps². In view of the low reflectivity of the CFBG, the reflected part was utilized as the cavity output through the port 3 of a PM circulator, while the 90% transmitted part was returned to the NALM loop through a PM isolator to form the oscillation.

 

Fig. 1. Experimental setup of the NLP operation NALM-based all-PM TDFL. WDM: wavelength division multiplexer, CIR: Circulator, CFBG: chirped fiber Bragg grating

Download Full Size | PDF

3. Results and discussion

In the experiment, the output power increased linearly with the enhancement of the pump power, as shown in Fig. 2(a). Under the pump power of 1.6 W, stable NLP operation was established with the output power of 5.5 mW, and the maximum output power was 13.6 mW at the pump power of 3 W. Figure 2(b) demonstrates the spectra evolution with the increase of the output power. At the starting point of the mode-locking, the spectrum is centered at 1950 nm with symmetrical Kelly sidebands, which is a typical feature of all-anomalous cavity. With further increasing the pump power, the spectrum spread slowly to the long wavelength side with the 3 dB bandwidth increased from 12 nm to 25 nm, probably resulting from the interaction of the intensified intra-cavity nonlinearity and the dynamical gain spectrum of the gain fiber. In addition, owing to the abundant nonlinear processes inside a single pulse, the spectrum profile tends to become disorganized at higher output power, with the long wavelength components more conspicuous aided by the stronger optical gain of the relatively long gain fiber. A similar evolution of the temporal profile was also observed and illustrated in Fig. 2(c), in which the pulse falling edge was gradually broadened from 3.1 ns to 6.6 ns at half maximum. It is therefore found that the calculated pulse peak power stays around 585 W with the increase of the pump power, attributing to the soliton peak power clamping effect that induces the formation of a pulse bunch. It is noted that at the falling edge of the pulse there is a slight dip, which is identified as a dark pulse that has been experimentally and theoretically verified to be induced by the intra-cavity nonlinear process [41,42]. The recorded pulse train at the maximum power is depicted in the inset of Fig. 2(c), with a pulse-to-pulse interval of 307 ns, corresponding to a fundamental repetition rate of 3.25 MHz. The measured autocorrelation trace at the output power of 5.5 mW with a large measurement range is shown in Fig. 2(d), in which a narrow coherent peak ride on a broad pedestal with a peak to pedestal intensity ratio of 1.37, manifesting an exemplary feature of the NLP lasers. Meanwhile, the gaussian fitting width is gradually narrowed from 424 fs to 303 fs with the increasing of the output power, as shown in Fig. 2(e), owing to the corresponding broadened optical spectrum [26,43].

 

Fig. 2. Output characteristics: (a) Output power. (b) Evolution of spectra (c) Evolution of pulse waveform, inset: the pulse train. (d) Autocorrelation trace measured in the range of 12 ps. (e) Autocorrelation traces of the coherent peak.

Download Full Size | PDF

To characterize the pulse stability, the PN and RIN of the generated NLP were respectively measured in the RF domain with a range of 10 Hz to 1 MHz by using a 12.5 GHz high-speed biased photodetector and the phase noise (PN) analyzer, as demonstrated in Fig. 3. Specifically, the PN was examined at the fundamental repetition frequency to reflect the timing jitter of a periodic pulse train, while the RIN represents the average power fluctuations of a laser pulse over a certain measurement time interval [44]. Although the NLP is a pulse bunch that formed by random ultrafast pulses, it is still a periodic pulse train with specific repetition rate, of which the fundamental frequency can be observed in the RF domain with significant signal to noise ratio (SNR). Therefore, it is reasonable to treat the NLP as a single pulse in terms of measuring its PN and timing jitter. It is observed in Fig. 3(a) that the PN is monotonically decreased with the frequency increasing from 10 Hz to 10 kHz, in which the noise spectra can be hardly separated from each other through changing the pump power. Whereas in the frequency range of 10 kHz to 1 MHz, the PN gradually approaches to an asymptotic level of around −120 dBc/Hz, and it manifests a decreasing trend (a maximum of ∼5 dB) with the enhancement of the output power, as shown in the inset of Fig. 3(a). The pronounced peak at ∼20 kHz is attributed to the relaxation oscillation (RO) of the laser, and with the increase of the output power it tends to be flattened and shifted to higher frequencies, as can be also observed for the RIN spectra in Fig. 3(b). In addition, the weak peak locates at 165 kHz in Fig. 3(a) might be related to the slight dark pulsing behavior, which however, needs to be further investigated in future work. As for the RIN spectra, it roughly keeps unchanged at frequency range smaller than the RO frequency, while at higher frequency range the spectra descend quickly. In the measured full frequency band, the intensity noise was reduced by more than 5 dB as the output power was increased from 5.5 mW to 13.6 mW, owing to the overdriven transmission function of the NALM [35]. However, this noise reduction effect is not effective for the PN at the low frequency range, where the origin of the PN is most probably predominated by the random ultrafast pulses that form the pulse bunch. The integrated timing jitter and rms RIN versus the output power in the frequency range from 10 Hz to 1 MHz are respectively shown in Fig. 4(a) and Fig. 4(b). With an integration frequency range of 1 kHz to 1 MHz, the obtained timing jitter and power instability are consistently decreased with increasing the output power, except that the timing jitter at the output power of 13.6 mW slightly elevated at the integration range of 1 kHz to 4 kHz, where the PN is higher compared with that at lower output powers, although this is trivial in Fig. 3(a). As a result, a minimum integrated timing jitter of 139 ps and RIN of 0.58% are respectively obtained at the output power of 11.5 mW and 13.6 mW.

 

Fig. 3. Noise characteristics of the laser pulse under different output powers. (a) Measured PN spectra, inset: PN spectra in the frequency range of 10 kHz to 1 MHz. (b) Measured RIN spectra.

Download Full Size | PDF

 

Fig. 4. Integrated timing jitter (a) and integrated rms RIN (b) under different output powers.

Download Full Size | PDF

Figure 5(a) depicts the RF spectrum of the laser pulse at 3.254 MHz with a 1.5 MHz range at the output power of 11.5 mW, manifesting a signal-to-noise ratio (SNR) of 60 dB. Meanwhile, the RF spectrum with a 100 MHz range shown in the inset of Fig. 5(a) demonstrates the absent of spurious peaks and further verifies a stable NLP mode-locking operation. To characterize the long-term stability of the laser, its output spectrum was recorded every 15 seconds with a total monitoring time of 2 h, and the results are shown in Fig. 5(b), in which any spectral variation can be hardly observed. Moreover, the output power stability measured in 2 h is depicted in Fig. 6(a), and the corresponding rms power fluctuation was calculated to be 0.45%. In addition, the amplitude of 10000 pulses were recorded and normalized with regard to the maximum, resulting in a calculated histogram exhibiting a normal distribution in a narrow range of the pulse intensities as shown in Fig. 6(b). The ratio of standard deviation to the mean was evaluated to be 1.25%, verifying the stable operation of the laser pulse likewise.

 

Fig. 5. (a) RF spectrum at the output power of 11.5 mW, inset: RF spectrums in a large frequency range of 100 MHz. (b) Spectrum evolution for a monitoring time of 2 h.

Download Full Size | PDF

 

Fig. 6. (a) Output power stability measured in 2 h. (b) The corresponding histogram of 10000 pulses amplitude.

Download Full Size | PDF

4. Conclusion

In conclusion, by exploiting a phase shifter and a CFBG, we have demonstrated a stable NLP generation from an all-PM Tm-doped fiber oscillator based on NALM. The wavelength of the laser pulse was centered at 1950 nm, and the corresponding 3 dB bandwidth was 25 nm. Owing to the limited reflection of the CFBG, the obtained maximum output average power was 13.6 mW with the repetition rate of 3.25 MHz. Meanwhile, the noise characteristics have firstly been systematically examined for the NLP laser, of which a minimum timing jitter and RIN was respectively calculated to be 139 ps and 0.58% with an integration frequency range from 1 kHz to 1 MHz. In addition, the experimental results also shown that the noise performances of the laser tend to be improved with the increasing of the output power. Through monitoring the optical spectrum and average power, the long-term stable operation of the NLP laser operation was further verified.