A protocol to detect and automate mode locking in a pre-adjusted nonlinear polarization rotation fiber laser is presented. The detection of a sudden change in the output polarization state when mode locking occurs is used to command the alignment of an intra-cavity polarization controller in order to find mode-locking conditions.
When a laser is mode-locked, it emits a train of ultra-short pulses at a repetition rate determined by the laser cavity length. This article outlines a new and inexpensive procedure to force mode locking in a pre-adjusted nonlinear polarization rotation fiber laser. This procedure is based on the detection of a sudden change in the output polarization state when mode locking occurs. This change is used to command the alignment of the intra-cavity polarization controller in order to find mode-locking conditions. More specifically, the value of the first Stokes parameter varies when the angle of the polarization controller is swept and, moreover, it undergoes an abrupt variation when the laser enters the mode-locked state. Monitoring this abrupt variation provides a practical easy-to-detect signal that can be used to command the alignment of the polarization controller and drive the laser towards mode locking. This monitoring is achieved by feeding a small portion of the signal to a polarization analyzer measuring the first Stokes parameter. A sudden change in the read out of this parameter from the analyzer will occur when the laser enters the mode-locked state. At this moment, the required angle of the polarization controller is kept fixed. The alignment is completed. This procedure provides an alternate way to existing automating procedures that use equipment such as an optical spectrum analyzer, an RF spectrum analyzer, a photodiode connected to an electronic pulse-counter or a nonlinear detecting scheme based on two-photon absorption or second harmonic generation. It is suitable for lasers mode locked by nonlinear polarization rotation. It is relatively easy to implement, it requires inexpensive means, especially at a wavelength of 1550 nm, and it lowers the production and operation costs incurred in comparison to the above-mentioned techniques.
The purpose of this article is to present an automation alignment procedure to get mode locking (ML) in nonlinear polarization rotation fiber lasers. This procedure is based on two essential steps: detecting the ML regime by measuring the polarization of the output signal of the laser and then setting-up a self-start control system to get to ML.
Fiber lasers have become an important tool in optics nowadays. They are an efficient source of coherent near-infrared light and they are now extending into the mid-infrared portion of the electromagnetic spectrum. Their low cost and ease of use have made them an attractive alternative to other sources of coherent light such as solid-state lasers. Fiber lasers can also provide ultrashort pulses (100 fsec or less) when a ML mechanism is inserted in the fiber cavity. There are many ways to design this ML mechanism such as nonlinear loop mirrors and saturable absorbers. One of these, widely used for its simplicity, is based on nonlinear polarization rotation (NPR) of the signal1,2. It uses the fact that the polarization ellipse of the signal undergoes a rotation proportional to its intensity as it propagates in the fibers of the laser cavity. By inserting a polarizer in the cavity, this NPR leads to intensity-dependent losses during a roundtrip of the signal.
The laser can then be forced to ML by controlling the polarization state. Effectively, the high-power portions of the signal will be subjected to lower losses (Figure 1) and this will eventually lead to the formation of ultrashort pulses of light when the laser is turned on and starts from a low-power noisy signal. However, the drawback of this method is that the polarization state controller (PSC) must be properly aligned to get ML. Usually, an operator finds the ML manually by varying the position of the PSC and analyzing the output signal of the laser with a fast photodiode, an optical spectrum analyzer or a nonlinear optical auto-correlator. As soon as the emission of pulses is detected, the operator stops varying the position of the PSC since the laser is ML. Obviously getting the laser to self-start automatically leads to an important gain in efficiency. This is especially true when the laser is subject to perturbations changing the alignment or the cavity configuration since the operator has to go through the alignment procedure again and again. In the last decade, different methods have been proposed to achieve this automation. Hellwig et al.3 used piezo-electric squeezers to control polarization in combination with a full analysis of the polarization state of the signal with an all-fiber division-of-amplitude polarimeter to detect ML. Radnarotov et al.4 used liquid-crystal plate PSCs with an analysis based on the RF spectrum to detect ML. Shen et al.5 used piezo-electric squeezers to control polarization and a photodiode/high-speed counter system to detect ML. More recently, a strategy based on an evolutionary algorithm was presented in which the detection is provided by a high-bandwidth photodiode in combination with an intensimetric second-order autocorrelator and an optical spectrum analyzer. The control is then performed with two electronically driven PSCs inside the cavity6.
This article describes an innovative way of detecting ML and its application to an automation technique forcing the fiber laser to ML. The detection of ML of the laser is achieved by analyzing how the output polarization state of the signal varies as the angle of the PSC is swept. As will be shown, the transition to ML is associated with a sudden change in the polarization state detectable by measuring one of the Stokes parameters of the output signal. The fact that a pulse is more intense than a CW signal and will undergo a more important NPR explains this change. Since the output of the laser is immediately located before the polarizer in the cavity, the polarization state of a pulse at this location is different from the polarization state of a CW signal (Figure 2) and will be used to discriminate the ML state. Theoretical aspects of this procedure and its first experimental implementation were presented in Olivier et al.7. In this article, the emphasis will be on the technical aspects of the procedure, its limitations and its advantages.
This technique is relatively simple to implement and does not require sophisticated measuring instruments to detect the ML state and automate the alignment of the laser to get ML. A PSC adjustable externally through a programmable interface is required. Different PSCs could be used in principle: piezo-electric squeezers, liquid crystal, wave-plates rotated by a motor, magneto-optic crystals or a motorized all-fiber PSC based on squeezing and twisting the fiber8. In this article, the latter is used, an all-fiber motorized Yao-type PSC. To detect the polarization state an expensive commercial polarimeter can be used. However, since only the value of the first Stokes parameter is required, a polarizing beam splitter in combination with two photodiodes will be sufficient as shown in this article.
All these components are inexpensive for the widely used erbium-doped fiber lasers. A feedback loop based on this procedure can find ML in a few minutes. This response time is suitable for most applications of fiber lasers and is comparable to the other existing techniques. In fact, the response time is limited by the electronics used to analyze the polarization of the signal. Finally, although the procedure is applied here to a similariton9 erbium-doped fiber laser, it could be used for any NPR based fiber laser as soon as the above mentioned equipment or its equivalent becomes available at the wavelength of interest.
1. Setting Up a Fiber ML Fiber Laser Including a Motorized PSC
2. Analyzing the Polarization of the Output Signal
3. Setting Up a Feedback Loop to Automate the Alignment of the PSC Using the Commercial Polarimeter Measurements
4. Building a Rudimentary Homemade Polarization Analyzer
5. Replacing the Commercial Polarimeter by the Homemade Polarization Analyzer in the Automation Process
NPR mode-locked fiber lasers are known to provide a large variety of pulsing regimes such as Q-switched pulses10, coherent ML pulses, noise-like pulses, bound states of ML pulses, harmonic ML and complex structures of interacting ML pulses11. In the laser described here, after the birefringence of the PSC was fixed to be able to get ML, the pump power was adjusted to be relatively near the threshold of single-pulse ML. In doing so, the number of competing regimes was reduced to a minimum. At this pump power and depending on the angle of the PSC, the laser presented different regimes (Figure 5) but no multi-pulse regime. Noise-like pulses12,13 were avoided due to the pre-adjustment of the cavity fibers that were kept fixed once a standard single ML pulse was found. In fact, the cavity design is probably important in that respect too but this aspect was not investigated thoroughly here. Consequently, the only remaining regimes were continuous-wave emission (CW), Q-switched emission and a stable ML with a single coherent pulse. In the continuous-wave (CW) and Q-switched regimes, narrow lines (1 nm or so, sometimes limited by the optical spectrum analyzer resolution) are seen. These spectra are to be compared with the wide spectrum of the ML regime with a full width at half maximum of the order of 30 nm or even more. On the fast photodiode, CW shows almost no variations, Q-switching shows a pulse train with a repetition rate of the order of a few microseconds (3.5 µsec here) and ML appears as a much faster pulse train with a repetition rate of a few tens of nanoseconds (12.2 nsec here) corresponding to the roundtrip time of the laser cavity. When an autocorrelation trace is used, only the ML regime shows the presence of pulses because the Q-switched regime generates pulses that have a much longer duration and a much lower peak power. The autocorrelation trace in the ML regime shows a single peak with a width of 156 fsec from which we deduced that only a single coherent ML pulse is present with a FWHM duration close to 100 fsec (110 fsec assuming a Gaussian pulse shape and 101 fsec assuming hyperbolic secant squared pulse shape).
The measurement of the Stokes parameters as a function of the angle of the intra-cavity PSC (Figure 6) yielded a typical result as expected in theory7. Notice that each Stokes parameter changes abruptly when ML is reached. Consequently, a measurement of only one of them, say S1, is required to detect ML. Note that a discontinuity in the value of a given parameter that does not coincide with stable ML is sometimes observed. In fact, the laser could sometimes reach an unstable regime where it shifts really quickly between CW, Q-switched and ML regimes in a chaotic manner. In these situations, the values of the Stokes parameters might vary substantially in time. These variations appear as error bars on the graph. It can be seen that the variations are more important in some regions than others. However, in the stable ML regimes, the variations are really small. This suggests that the temporal variation of the Stokes parameters could be used as a complementary criterion to verify if ML is really reached or not after a discontinuous jump has been detected.
The previous analysis leads to the conclusion that the automation of the laser can be based on the search for a discontinuity of a given Stokes parameter. S1 was chosen here. The variation of S1 that is defined as a "discontinuity" is a priori arbitrary. Based on the measurements (Figure 6), it is found that S1 usually varies by steps smaller than 0.1 as the angle is varied by 1°. The only exception is when ML is reached where it varies by 0.6. It was thus decided to fix the threshold of discontinuity to 0.3. The automation procedure presented here (Figure 7) is based on that condition. The laser must not be in a ML condition when the routine starts otherwise the routine will stop when the discontinuity leading from ML to CW will be found and the laser will end up emitting CW. This constraint is not problematic because the range of angles giving ML is small compared to the full range of the PSC. It is thus easy to position the PSC at an angle really far from ML when the routine is engaged. Here, the PSC was brought to its minimum angle where a mechanical stop prevents it from moving further. At this position, the laser was not ML. Under these conditions, the routine works really well. It finds ML within a few minutes. In this case, speed is mostly limited by the communication time required between the commercial polarimeter and the computer as the angle is swept.
When measured with the homemade polarization analyzer (Figure 10), the curve of S1 as a function of the angle of the PSC is different from the curve measured with the commercial polarimeter (Figure 6). This is due to the fact that the x- and y- axes of both instruments do not necessarily coincide. However, the abrupt transition in S1 when ML is reached is clearly seen in both cases. In fact, the behavior of S1, S2 and S3 measured with the commercial polarimeter showed that the three parameters did not undergo the same discontinuity when ML was reached. It suggests that a change in the orientation of the polarizing beam splitter or, equivalently, the insertion of a manual PSC just before the polarization analyzer could help in making the transition more abrupt and easier to detect. In fact, this is exactly what happened here, the transition to ML is easier to see with the homemade polarization analyzer because the manual PSC was adjusted to make the transition appear more clearly. The automation procedure is then easier to achieve.
The automation with the homemade polarization analyzer works really well. ML is found within a few minutes. In fact, because the readings of the photodiode voltages are faster than the readings of the commercial polarimeter, the homemade polarization analyzer performs better.
Figure 1: ML based on nonlinear polarization rotation. The signal is first linearly polarized by the polarizer and then transformed into an elliptic polarization state by the PSC. Due to the Kerr nonlinearity of the fiber in the laser cavity, the polarization ellipse undergoes a rotation of its main axis proportional to the signal's power. Since the polarizer at the end transmits only the vertical component of the polarization, the transmission will depend on the power of the signal and could favor the formation of a pulse from noise if the PSC angle is adjusted correctly. Please click here to view a larger version of this figure.
Figure 2: Position of the polarization analyzer. For a given average power, a pulse will have a peak power larger than a continuous wave (CW) signal and will undergo a larger nonlinear polarization rotation. By positioning the analyzer just before the polarizer, discrimination between the polarization states will allow detection of the presence of a pulse in the cavity. Please click here to view a larger version of this figure.
Figure 3: The fiber laser ring cavity. The laser must be a ring cavity including single-mode optical fibers (blue), a gain fiber (green), an isolator, a polarizer, a PSC adjustable through a computer interface. The output coupler must be located just before the polarizer. Finally, 1% of the output signal is tapped in order to monitor the state of polarization of the signal and 99% of the output signal remains available. The polarization analyzer provides feedback to a control loop programmed in a computer that adjusts the angle of the motorized PSC (light red) via an electric cable (black). Please click here to view a larger version of this figure.
Figure 4: A motorized fiber-squeezer PSC. The birefringence of the PSC is fixed by the pressure of the screws on the left. The angle of the PSC is adjusted with the electronically controlled motor which is on the right. The electric cable connects the system to a computer interface. Please click here to view a larger version of this figure.
Figure 5: Detecting ML with an optical spectrum analyzer. Different regimes of the laser observed on the optical spectrum analyzer on the left, on a fast photodiode in the middle and on an autocorrelator on the right (when applicable): quasi-CW with multiple wavelengths (blue), Q-switched CW (green) and ML (red). The spectrum in the ML regime is much wider than the others and its dechirped autocorrelation trace shows a single peak with FWHM of 156 fsec and a relatively narrow pedestal. Please click here to view a larger version of this figure.
Figure 6: The value of the Stokes parameters as functions of the PSC angle and ML regions. The blue curves are the average value of each Stokes parameter over 5 measurements taken at intervals of 0.2 sec for a typical case. The error bars represent the standard deviation of the measurements and demonstrate the stability of the laser for a given PSC angle. As the angle of the PSC is varied, the values of the Stokes parameters change in a continuous fashion except when ML is reached (red areas on the figure). In this situation, their values undergo an abrupt variation that can be used to detect the ML. Please click here to view a larger version of this figure.
Figure 7: A routine to automatically align the PSC to get ML. This flowchart shows the simple routine used to automate the alignment of the polarization state controller (PSC) to get ML. Please click here to view a larger version of this figure.
Figure 8: Homemade polarization analyzer measuring S1. A free-space polarizing beam splitter splits the x- and y-polarization components of the signal. These components are sent separately to two photodiodes thus measuring the powers Px and Py in each polarization, allowing to compute the first Stokes parameter S1 = (Px–Py)/(Px+Py). Please click here to view a larger version of this figure.
Figure 9: Trans-impedance amplifier circuit for each photodiode. The InGaAs photodiode detects the 1,550 nm signal. It is connected to an operational amplifier, a resistance and a capacitor. The role of the capacitor is to reduce the bandwidth of the circuit thus reducing the risk of getting an electrical oscillation from the circuit itself. The voltage value will be averaged out by the oscilloscope as the mean value will be read from it and transformed into an optical average power through calibration with a commercial optical power-meter. Please click here to view a larger version of this figure.
Figure 10: The value of the first Stokes parameter as a function of the PSC angle using the homemade polarization analyzer. The behavior of S1 shows the typical abrupt transition at the angle where the laser reaches ML for a typical case. This was also seen with the commercial polarimeter. Please click here to view a larger version of this figure.
It has been shown that it is possible to automate the ML of NPR fiber ring lasers by using a feedback loop based on output polarization measurements. To realize this task it is crucial to insert an adjustable PSC in the cavity. The output coupler of the cavity must be located just before the polarizer in order to see a difference between the polarization state of a CW signal and a pulse signal (Figure 2). The birefringence of the PSC must be pre-adjusted so that ML can be found and the pump power must be set near single pulse ML the threshold in order to get a single pulse in the cavity and minimize the number of competing regimes that can occur. This explains why the ML regime found automatically by sweeping the angle in a certain direction was always the same during the experiment. The parameter measured at the output to detect ML is S1. This parameter changes continuously as the angle of the intra-cavity PSC is swept. The only exception to this is when ML is reached, the value of S1 then undergoes a discontinuity. The possibility to make small angle increments is important here. If large increments are used it might become difficult to discriminate between a sudden jump and a "normal" variation. The small range of angles leading to ML might also be stepped over without noticing it. The small increment also ensures that the ML state is always the same because the system does not fall anywhere in the ML range but always detect the edge of this region where the pulses have always the same optical spectrum. This is the only obvious way of ensuring the repeatability of the procedure and the parameters of pulses generated.
Assuming the above critical points have been considered, it is possible to build a homemade polarization analyzer that provides a value of S1 and allow the detection and automation of the ML. The setup proposed here was made up of a free-space polarizing beam splitter cube in combination with two photodiodes. An alternative would be to use a fiber-based polarization beam splitter. No alignment would be required and it would be an all-fiber setup. Note also that an oscilloscope was used to get the voltages of the photodiodes in order to communicate with it easily via a GPIB port. The use of an USB voltmeter or a homemade electronic circuit could reduce the cost of the apparatus.
The technique presented here is intended to work for NPR fiber mode-locked lasers. To apply it, one needs to work with a relatively stable cavity design that was pre-adjusted to be able to get ML. The fact that only a single parameter is varied to search for ML limits the generality of the technique. If the cavity is perturbed by, for instance, introducing a birefringence in the fibers, the system will be able to compensate and find ML when the perturbation is small. However, the PSC will not be able to compensate for a large modification of the birefringence of the cavity because its birefringence is fixed7. In that sense, this technique cannot be considered as general as the one presented in Hellwig et al.3. Also, the simple characterization of S1 at the output used here in combination with the control of a unique PSC angle does not allow the exploration of all the possible regimes of emission of the laser as discussed by Andral et al.6 for instance. Moreover, the ML detection technique presented here cannot discriminate between noise-like pulses11, coherent ML pulses and multiple-pulses regimes. The pre-adjustment of the cavity fibers, the pump power and the PSC birefringence must thus be carefully done to ensure that single coherent ML pulses will form instead of noise-like pulses or multiple-pulses regimes.
As mentioned in the introduction, other ML mechanisms exist and some of them do not require alignment. They all have some pros and cons. ML based on nonlinear loop mirrors14 requires an extra length of fiber inside the cavity and might not be suitable for high-repetition rate lasers15. ML based on saturable absorbers mirrors16 requires the design of custom mirrors appropriate for the power and spectral characteristics of the laser under consideration. The NPR ML mechanism remains the most widely used because of its simplicity, its effectiveness and low-cost implementation.
The automation of its alignment makes NPR an even more interesting option because it can now be used in commercial systems without requiring the intervention of the user to ensure ML occurs. The technique to automate its alignment presented here is sufficient to get ML in normal conditions and is simple to implement. It requires a few low-cost components and no expensive instruments such as an optical spectrum analyzer or an RF-spectrum analyzer. The cavity design does not have to be modified since it relies on output polarization measurements. In fact, only a fraction of the output is tapped for monitoring and the remaining portion can be used for the ongoing application.
In other words, the laser does not need to be disconnected to proceed with the alignment procedure. Secondly, the required average power is so small that a 1% monitoring tap is sufficient. This is to be contrasted with ML detection techniques based on nonlinear processes such as second -harmonic generation or two-photon absorption that would require a significantly larger fraction for the monitoring to be efficient. Finally, since this technique requires only the first Stokes parameter S1 to be measured, there is no need for a complete characterization of the polarization state and this renders the system much simpler and cheaper to design and build.
This technique is well-suited for commercial fiber lasers and, with that goal in mind, could be developed further to improve its performance. It will be interesting also to apply it to fiber lasers at different wavelengths. Here it was used in an erbium-doped fiber laser but it is easily transferable to ytterbium-doped fiber lasers since all the required equipment is readily available. It might become more challenging for lasers operating at non-conventional wavelengths but it is certainly feasible. More testing is required to verify its applicability to different dispersion regimes such as the soliton laser, the stretched-pulse laser, the similariton laser and the dissipative soliton laser.
The authors have nothing to disclose.
The authors would like to thank Christian Olivier and Philippe Chrétien for valuable help concerning electronics, Éric Girard at GiGa Concept Inc. for support with the motorized polarization controller, professor Réal Vallée for the loan of the commercial polarimeter and professor Michel Piché for many fruitful discussions.
This work was supported by the Fonds de recherche du Québec – Nature et technologies (FRQNT), the Natural Sciences and Engineering Research Council of Canada (NSERC) and Canada Summer Jobs.
Bare-Fiber adaptor | Bullet | NGB-14 | |
Drop-in polarization controller | General Photonics Corp. | Polarite PLC-006 | Manual polarization controller. |
DSP In-line polarimeter | General Photonics Corp. | POD-101D PolaDetect | Polarimeter with USB/serial computer connectivity. |
Fiber Cleaver | Fitel | S323 | |
FiberPort | Thorlabs Inc. | PAF-X-2-C | |
Fixed Fiber-to-Fiber Coupler Bench | Thorlabs Inc. | FBC-1550-APC | Any optical bench could be used. A 3-way bench would even be better. |
Fusion Splicer | Fujikura | FSM-40PM | |
High resolution all fiber polarization controller | Giga Concept Inc. | GIG-2201-1300 | All-fiber motorized polarization controller with USB computer connectivity. |
InGaAs PIN PD module | Optoway | PD-1310 | Pigtailed photodiode. |
Instrument communication interface | National Instruments | NI MAX | It comes packaged with National Instruments drivers (NI-VISA, NI-DAQmx, etc.) |
Operational amplifier | Texas Instruments | TLO81ACP | |
Optical Powermeter | Newport | 818-IS-1 with 1835-C | |
Optical spectrum analyzer | Anritsu | MS9710C | |
Oscilloscope | Tektronix | TDS2022 | Oscilloscope with GPIB computer connectivity. |
Polarizing beamsplitter module | Thorlabs Inc. | PSCLB-VL-1550 | |
Polyimide Film Tape | 3M | 5413 | Tape to fix the components on the table without damaging the fibers. |
Graphical programming language interface (GPLI) | National Instruments | LabVIEW | Interface to program in G Programming Language and communicate with laboratory instruments. |
Polarimeter controlling software | General Photonics Corp. | PolaView | Comes with the polarimeter General Photonics POD-101D. |