Open Access
Add to BibSonomyAbstract
We propose a pulse-tail elimination and pulse shortening method using an optical interferometer, which is effective for picosecond chirped pulses from gain-switched multimode laser diodes. In a numerical simulation, when the delay distance between a chirped pulse and its replica in an optical interferometer matches two times the round-trip optical length of the laser cavity, the pulse-front and -rear tail parts are effectively eliminated from the input chirped pulse after passing through the optical interferometer. Using this method with a fiber Sagnac interferometer, a 33 ps pulse with a long-tail emitted from a gain-switched 1540 nm multimode laser diode was linearly transformed into a 20 ps pulse with a substantially reduced tail.
©2008 Optical Society of America
1. Introduction
Laser diodes are attractive as compact and stable picosecond pulse sources [1–4]. In the field of optical communication, high-end distributed feedback (DFB) laser diodes oscillating in a single longitudinal mode are already available as sources of low-chirp and tailless picosecond pulses at high repetition rates [5]. In contrast, general-purpose commercial Fabry-Perot (FP) laser diodes emit pulses of several tens of picoseconds with multimode spectra under gain-switching conditions [6]. Those pulses are often highly frequency down-chirped within the pulse durations [7, 8]. Although the applicability of methods for linearly or nonlinearly compensating for pulse chirping has been confirmed for single-mode laser diodes [9–12], they are not necessarily suitable for multimode laser diodes because different modes have slightly different chirp conditions and different group velocities in a medium used for pulse-chirp compensation [10]. In addition, the gain-switched pulses from commercial FP laser diodes are often accompanied by long tails resulting from their lower relaxation oscillation frequencies. The tail parts are generally difficult to eliminate and often form further pulse tails lasting in the nanosecond range when amplified [13]. In order to make efficient use of picosecond laser diodes in a wide range of applications, there is a strong demand for a method of eliminating the pulse tail and shortening the pulse duration that can be applied to gain-switched multimode laser diodes. To this end, we propose utilizing an optical interferometer as an optical frequency filter in order to eliminate pulse-front and -rear tail parts from gain-switched pulses emitted from multimode laser diodes. For a single-mode laser diode, a narrow-band FP etalon filter has been found to be effective for chirp reduction and pulse shortening of gain-switched pulses [14]. Here, in contrast, an optical filter used for multimode laser diodes is composed of a conventional Michelson (or Sagnac) interferometer with a constant optical delay between the two arms.
2. Numerical
We carried out numerical simulations of the proposed method using multimode laser diode rate equations [6]. A 1550 nm laser diode was chosen as a specific example in the simulation because it characteristically emits picosecond pulses with long tails due to nonlinear gain saturation effects [15, 16]. The characteristic parameters and pumping parameters in the rate equations were set to the same values used for a gain-switched 1550 nm laser diode in Reference [6] (Fig. 5), except for the linewidth enhancement factor α=5 and the gain compression factor ε=1.0×10-23 m3. The parameter values used in the numerical simulation are listed in Table 1. Numerical integration of the multimode rate equations was then carried out to obtain simulated gain-switched optical pulses.
Table 1. Notation of the parameters used in the numerical simulation.
Figure 1 shows an example of the simulated time-resolved spectrum (streak scope trace) of a gain-switched pulse, which is expressed as a time variation of the instantaneous frequency (time derivative of the phase of the complex pulse field) contrast-modulated by the corresponding pulse intensity. The corresponding pulse waveform and time-averaged spectrum are also shown in Fig. 1. A magnified view of part of the spectrum is shown at the bottom. For a realistic display of the simulated streak scope trace, a temporal resolution of 1 ps and a wavelength resolution of 0.05 nm, with Gaussian windows, were considered only for the time-resolved spectrum. The existence of a frequency down-chirp within the pulse duration of 37.8 ps is clearly represented in the time-resolved spectrum. It is found from the pulse chirp feature that the spectral components in the vicinity of shorter wavelengths in respective modes correspond mainly to the front part of the pulse tail, and in the same way, those in the vicinity of longer wavelengths correspond to the rear part of the pulse tail [6]. The distinct spectral peaks seen in respective modes, therefore, stem mainly from longer rear tails which are generated due to nonlinear gain saturation effects, as indicated in Fig. 1.
Those enhanced spectral peaks can be filtered out by passing them through a Michelson interferometer in which the spectral pitch of the sinusoidal energy transmittance function matches the mode spacing of the multimode spectrum (143 GHz). To achieve such a filter condition, the delay distance between the two arms of the interferometer (Ld) should be set to integral multiples of the round-trip optical length of the laser cavity (2mnLc, m: integer, n: refractive index of active layer, n=3.5, Lc: laser cavity length, Lc=300 µm), which corresponds to 143×m GHz of the spectral pitch. Figure 2 shows the time-resolved spectra of gain-switched pulses after passing through a Michelson interferometer when the delay distance is varied. Although the spectral position of the sinusoidal energy transmittance function of the interferometer is phase-sensitive, the phase problem between the input pulse and its replica could not be treated in the original numerical framework because the distance-step interval ΔL was longer than the wavelength assumed (1550 nm). The number of time-steps performed in the simulation was 215 for 1 ns, meaning that the corresponding ΔL was approximately 9.2 µm. To solve this problem, an arbitrary phase ϕint was then added to the replica pulse by assuming a delay distance variation of less than one wavelength. Thus, the sinusoidal energy transmittance function of the interferometer could be frequency-shifted by varying the value of ϕint, which enabled matching the sinusoidal function with the multimode spectrum. When the delay distance is not set to integral multiples of the round-trip optical length of the laser cavity (Ld≠2mnLc), the pulse tails are partially and asynchronously eliminated in respective modes, as shown in Fig. 2a. When the delay distance equals three times the optical length of the laser cavity (Ld=3nLc), the pulse tails are synchronously eliminated every two modes, as shown in Fig. 2b. The overall pulse waveforms in these cases, however, remain almost unchanged from the input pulse waveform in Fig. 1.
Fig. 1. Simulated time-resolved spectrum of gain-switched pulse from 1550 nm FP laser diode. The corresponding pulse waveform and time-averaged spectrum are also shown in Fig. . A magnified view of part of the time-averaged spectrum is shown at the bottom.
Figure 3 shows the variation of the pulse duration after passing through a Michelson interferometer when the delay distance Ld is varied and ϕint is set to a constant value of 14π/9. Pulse shortening is achieved only when the delay distance is set in the vicinity of integral multiples of the round-trip optical length of the laser cavity (Ld=2mnLc). Figure 4 shows the time-resolved spectra of gain-switched pulses after passing through a Michelson interferometer when the delay distance Ld is set to a) 2nLc (m=1), b) 6nLc (m=3), and c) 4nLc (m=2), respectively. In the case of a) m=1, because the sinusoidal filter function is a shallow curve, it is difficult to eliminate fully enhanced spectral peak components. In contrast, in the case of b) m=3, although the pulse duration is shortened to 10 ps, satellite pulse components are generated because of narrow transmittance windows of the sinusoidal filter function. As a result, under the condition of Ld=4nLc (m=2) in Fig. 4c, the long pulse tails are eliminated effectively from the input pulse without forming a distinct pulse pedestal, and the pulse duration is well shortened from 37.8 ps to 13.2 ps while partially maintaining the original frequency down-chirp. From a series of numerical simulations, the delay distance of Ld=4nLc provides, in many cases, the optimal condition for the above pulse-tail elimination problem for a wide range of characteristic parameter values and pumping parameter values suitable for multimode laser diodes. In the case of Fig. 4c, although almost 70 % of the pulse energy is lost from the other port of the interferometer, the output pulse peak intensity remains above the 70 % level of the input pulse.
Fig. 2. Simulated time-resolved spectra and the corresponding pulse waveforms, and time-averaged spectra of gain-switched pulses obtained after passing through Michelson interferometer when delay distance Ld is set to a) 147 ΔL (~1.35 mm) and b) 354 ΔL (=3nLc~3.24 mm), respectively. ϕint is set to a constant value of 14π/9 for both cases. Dotted sinusoidal functions in magnified views indicate energy transmittance spectra of the Michelson interferometer.
Fig. 3. Variation of pulse duration of gain-switched pulse after passing through Michelson interferometer when delay distance is varied. ϕint is set to a constant value of 14π/9.
Fig. 4. Simulated time-resolved spectra and the corresponding pulse waveforms, and time-averaged spectra of gain-switched pulses obtained after passing through Michelson interferometer when delay distance Ld is set to a) 236 ΔL (=2nLc~2.16 mm), b) 708 ΔL (=6nLc~6.48 mm), and c) 472 ΔL (=4nLc~4.32 mm), respectively. The spectral intensity of the time-averaged spectra is magnified to three times the original intensity corresponding to Fig. 1. ϕint is set to a constant value of 14π/9 for all cases. Dotted sinusoidal functions in magnified views indicate energy transmittance spectra of the Michelson interferometer.
3. Experimental
Figure 5 shows the experimental configuration. A commercial 1540 nm FP laser diode was dc-biased at a threshold current of 12 mA and gain-switched at a modulation frequency of 1 GHz. The gain-switched pulse was passed through an isolator and then made incident on a fiber Sagnac interferometer instead of the above Michelson interferometer. The fiber Sagnac interferometer was composed of a combination of a 50/50 fiber coupler, a fiber polarization controller operating as a half-wave plate, and a polarization maintaining fiber (PMF), allowing a required optical delay to be obtained in a confined optical fiber system. Because the refractive index difference between the fast and slow axes of the PMF was estimated to be around 4×10-4 and the delay distance required for pulse-tail elimination was about 4 mm, an approximately 10 m-long PMF was needed to realize the optimal filter effect mentioned above. To adjust the delay distance, several PMFs having different lengths (the shortest was 0.1 m) were prepared. Because the two arms of the interferometer were included in the same PMF loop, an arbitrary angle was allowed between the incident plane of the light and the polarization axis of the PMF, under the condition that the total number of pulse components passing along the slow axis of the PMF was equal to that passing along the fast axis. The polarization perturbations occurring in the PMF loop due to mechanical vibrations were therefore canceled out. Then, for spectral monitoring, a small portion (1%) of the pulse from the output port of the Sagnac interferometer passed out of the fiber through a 99/1 fiber coupler after passing through an isolator and an erbium-doped fiber amplifier (EDFA). The rest was incident on a second harmonic (SH) autocorrelation system, which was composed of a fiber Michelson interferometer with two external mirrors, a 1 mm-thick periodically poled lithium niobate nonlinear crystal (PPLN) and a photomultiplier; the latter two were used in generating and amplifying SH signals, respectively. The amplified SH signal was measured with a lock-in detection system. One of the external mirrors of the Michelson interferometer was moved until a complete SH autocorrelation trace was obtained. Typical values of the sweep distance and the sweep time of the external mirror were 90 mm and 30 minutes, respectively.
Fig. 6. Time-averaged spectrum (a) and the corresponding SH autocorrelation trace (b) of an input gain-switched pulse from a 1540 nm FP laser diode.
4. Results and discussion
Figures 6 and 7 show the experimental results of optical spectra (a) and the corresponding SH autocorrelation traces (b) of the gain-switched pulses before (Fig. 6) and after (Fig. 7) passing through the Sagnac interferometer. In Fig. 6a, distinct spectral peaks whose spectral pitch is 1.1 nm (corresponding to about 140 GHz) are seen in respective modes, as well as the above simulation results, which indicates the generation of long-lasting pulse tails. This is clearly reflected in the SH autocorrelation trace shown in Fig. 6b by the appearance of pulse pedestal components. The SH autocorrelation trace includes many coherent spikes appearing as the result of interference between a pulse having a multimode spectrum and its replica. The symmetrical shape of the SH autocorrelation trace indicates that the pulse oscillation was stable during the measurement time. The full width at half maximum (FWHM) of the SH autocorrelation waveform of the input pulse is measured to be 65.7 ps. The corresponding temporal pulse duration is then estimated to be 33 ps by assuming a Lorentzian pulse shape function (65.7/2~33), where the factor 2 is the ratio of the SH autocorrelation width to the pulse width for a Lorentzian pulse [17]. This is because the pulse shape function of gain-switched pulses with long tails can be regarded as Lorentzian-like in terms of the above-described factor in our simulations. To achieve the optimal condition for filtering, several PMFs were connected. The EDFA was temporarily used as an amplified spontaneous emission (ASE) source available in the vicinity of 1550 nm in adjusting the pitch of the sinusoidal filter function to that of the multimode spectrum. The connected PMFs were then temperature controlled for fine tuning to match the spectral position of the sinusoidal filter function with the multimode spectrum. When the total length of PMF was set to 12.2 m, finally, the enhanced spectral peaks almost disappeared, as shown in Fig. 7a. Thus the distinct pulse pedestals seen in Fig. 6b are largely reduced in the corresponding SH autocorrelation trace in Fig. 7b. The FWHM of the SH autocorrelation waveform is shortened to 30.9 ps, and the corresponding temporal pulse duration is estimated to be 20 ps by assuming a sech-squared pulse shape (30.9/1.55~20), where the factor 1.55 is for the sech-squared function. As with the case of the original gain-switched pulse with a reduced tail, the shape function of the output pulse can also be regarded as sech-squared in our simulations [18]. Although long pulse tails are thus successfully eliminated from the input pulse using this method, the obtained pulse shortening ratio of 1.65 (=33/20) is relatively small. According to our numerical estimations for this method, the shortened pulse duration was expected to reach around 15 ps or less. The discrepancy between the experimental and numerical results is thought to be due to overestimation of the frequency down-chirp inside the input pulse in the simulation, and/or low mismatch of the spectral pitch and position between the sinusoidal filter function and the multimode spectrum.
Fig. 7. Time-averaged spectrum (a) and the corresponding SH autocorrelation trace (b) of an output gain-switched pulse after passing through a Sagnac interferometer with a delay distance of two times the round-trip optical length of the laser cavity (~4 mm).
The averaged optical power of the pulses before and after passing through the Sagnac interferometer was 830 µW and 205 µW, respectively. It is found from the fact that almost 75 % of the pulse energy, corresponding mainly to the pulse tail parts, was lost from the other port of the interferometer and therefore the experimentally obtained pulses had longer tail parts than expected numerically. This means that the signal-to-noise (S/N) ratio of the output pulse was decreased by 3 dB from that of the input pulse. We need to pay attention to this decreasing in the S/N ratio in amplifying the shaped output pulses. In the above experiments, although no active feedback system for stabilization was employed, the frequency drift of the sinusoidal filter function or the multimode spectrum had only a limited influence on SH autocorrelation measurements, at least, for the measurement time of 30 minutes. When this system is used as a light source for long-time measurement (>30 minutes) in some applications, however, an active feedback system should be employed to compensate a mismatch between the sinusoidal filter function and the multimode spectrum due to the frequency drift. For the demonstration of the proposed pulse shaping method, we used a 1540 nm FP laser diode because gain-switched pulses from it had distinct long tail parts to be eliminated. In addition, the use of it made interferometric measurements simple because commercial optical fiber components (fiber couplers, polarization maintaining fiber cables, etc.) were available. The proposed method can also be applied to commercial FP laser diodes of different central wavelengths by using a spatial Michelson interferometer. The shaped pulses from those FP laser diodes may serve many uses in a wide range of applications by linearly amplifying them. Note that time-averaged measurements and interferometric measurements performed here were insensitive to the amplitude and temporal jitter of the pulses. We should pay attention to the jitter problem when the proposed system is used in some applications.
5. Conclusion
We have confirmed numerically that a conventional Michelson interferometer operates as an optical filter system effective for both pulse-tail elimination and pulse shortening of picosecond chirped pulses from gain-switched multimode laser diodes. The optimal condition for the pulse shaping is found that the delay distance between the two arms of the Michelson interferometer is set to two times of the round-trip optical length of the laser diode cavity. Using this method with an all-fiber Sagnac interferometer, a 33 ps pulse with a long-lasting tail emitted from a gain-switched commercial 1540 nm multimode laser diode was linearly transformed into a 20 ps pulse with a substantially reduced tail. Although attention needs to be paid to the degradation of the signal-to-noise ratio, jitter problems and the system stability, the proposed pulse shaping method is expected to be used in a wide range of applications because it can be applied to commercial Fabry-Perot laser diodes of different central wavelengths.
References and links
1. A. Aschwanden, D. Lorenser, H. J. Unold, R. Paschotta, E. Gini, and U. Keller, “2.1-W picosecond passively mode-locked external-cavity semiconductor laser,” Opt. Lett. 30, 272–274 (2005). [CrossRef] [PubMed]
2. A. Fragemann, V. Pasiskevicius, and F. Laurell, “Optical parametric amplification of a gain-switched picosecond laser diode,” Opt. Express13, 6482–6489 (2005), http://www.opticsinfobase.org/abstract.cfm?URI=oe-13-17-6482. [CrossRef] [PubMed]
3. K. Taira, T. Hashimoto, and H. Yokoyama, “Two-photon fluorescence imaging with a pulse source based on a 980-nm gain-switched laser diode,” Opt. Express15, 2454–2458 (2007), http://www.opticsinfobase.org/abstract.cfm?URI=oe-15-5-2454. [CrossRef] [PubMed]
4. H. Liu, C. Gao, J. Tao, W. Zhao, and Y. Wang, “Compact tunable high power picosecond source based on Yb-doped fiber amplification of gain switch laser diode,” Opt. Express16, 7888–7893 (2008), http://www.opticsinfobase.org/abstract.cfm?URI=oe-16-11-7888. [CrossRef] [PubMed]
5. K. Otsubo, M. Matsuda, K. Takada, S. Okumura, M. Ekawa, H. Tanaka, S. Ide, K. Mori, and T. Yamamoto, “Uncooled 25 Gbit/s direct modulation of semi-insulating buried-heterostructure 1.3 µm AlGaInAs quantum-well DFB lasers,” Electron. Lett. 44, 631–633 (2008). [CrossRef]
6. K. Wada, H. Sato, H. Yoshioka, T. Matsuyama, and H. Horinaka, “Suppression of side fringes in low-coherence interferometric measurements using gain- or loss-modulated multimode laser diodes” Jpn. J. Appl. Phys. 44, 8484–8490 (2005). [CrossRef]
7. K. Wada, Y. Akage, and Y. Cho, “α-parameter dependence of chirped pulse from injection-modulated semiconductor lasers,” Opt. Rev. 1, 217–219 (1994).
8. A. Bresson, N. Stelmakh, J. M. Lourtioz, A. Shen, and C. Froehly, “Chirp measurement of multimode Q-switched laser diode pulses by use of a streak camera and a grating monochromator,” Appl. Opt. 37, 1022–1025 (1998). [CrossRef]
9. A. Takada, T. Sugie, and M. Saruwatari, “Picosecond optical pulse compression from gain-switched 1.3 µm distributed-feedback laser diode through highly dispersive single-mode fiber,” Electron. Lett. 21, 969–971 (1985). [CrossRef]
10. R. Takahashi, H. F. Liu, M. Osinski, and T. Kamiya, “Picosecond single-mode pulse compression using a 1.3 µm Fabry-Perot laser diode, a dispersion-shifted fiber, a grating monochromator,” Appl. Phys. Lett. 55, 2377–2379 (1989). [CrossRef]
11. L. P. Barry, B. C. Thomsen, J. M. Dudley, and J. D. Harvey, “Optimised design of fibre-based pulse compressor for gain-switched DFB laser pulses at 1.5µm,” Electron. Lett. 35, 1166–1168 (1999). [CrossRef]
12. P. M. Anandarajah, C. Guignard, A. Clarke, D. Reid, M. Rensing, L. P. Barry, G. Edvell, and J. D. Harvey, “Optimised pulse source employing an externally injected gain-switched laser diode in conjunction with a nonlinearly chirped grating,” IEEE J. Slected Topics in Quantum Electron. 12, 255–264 (2006). [CrossRef]
13. A. Blter, R. Erdmann, and T. Schmitt, “Amplified picosecond diode lasers simplify biomedical research,” Bio-photonics International July , 38–41 (2005).
14. M. Nakazawa, K. Suzuki, and Y. Kimura, “Transform-limited pulse generation in the gigahertz region from a gain-switched distributed-feedback laser diode using spectral windowing,” Opt. Lett. 15, 715–717 (1990). [CrossRef] [PubMed]
15. R. Olshansky, P. Hill, V. Lanzisera, and W. Powazinik, “Frequency response of 1.3 µm InGaAsP high speed semiconductor laser,” IEEE J. Quantum Electron. QE-23, 1410–1418 (1987). [CrossRef]
16. K. Wada, Y. Akage, H. Marui, H. Horinaka, N. Yamamoto, and Y. Cho, “Simple method for determining the gain saturation coefficient of a distributed feedback semiconductor laser,” Opt. Commun. 130, 57–62 (1996). [CrossRef]
17. K. L. Sala, G. A. Kenny-Wallace, and G. E. Hall, “CW autocorrelation measurements of picosecond laser pulses,” IEEE J. Quantum Electron. QE-16, 990–996 (1980). [CrossRef]
18. K. Wada and Y. Cho, “Improved expression for the time-bandwidth product of picosecond optical pulses from gain-switched semiconductor lasers,” Opt. Lett. 19, 1633–1635 (1994). [CrossRef] [PubMed]
