- Email: [email protected]
Characterisation and optimisation of a dual-channel picosecond gain-switched DFB laser system for use as a pump–probe source
Optics Communications 248 (2005) 229–239 www.elsevier.com/locate/optcom
Characterisation and optimisation of a dual-channel picosecond gain-switched DFB laser system for use as a pump–probe sourceq Henry T.G. Bookey *, Ajoy K. Kar Department of Physics, School of Engineering and Physical Sciences, Heriot-Watt University, David Brewster Building, Edinburgh EH14 4AS, UK Received 29 July 2004; received in revised form 2 December 2004; accepted 3 December 2004
Abstract The output characteristics of a dual-channel gain-switched laser system are studied with respect to the gain-switch frequency and power. InGaAsP laser diodes operating at 1548 and 1549 nm are gain-switched at 1 GHz and the spectral and temporal profiles of the pulses are investigated. The development of a pump–probe system with the use of in-line amplifiers and amplified spontaneous emission filtering is detailed. 2004 Elsevier B.V. All rights reserved. PACS: 42.55; 42.81; 06.60J Keywords: DFB; Gain-switching; Picosecond; Pump–probe system
1. Introduction In order to increase the amount of data transmitted through a telecommunications system that employs WDM and TDM either the number of wavelength channels or the individual channel q
This work was supported in part by the Engineering and Physical Sciences Research Council. * Corresponding author. Tel.: +44 131 451 3047; fax: +44 131 451 3136. E-mail address: [email protected] (H.T.G. Bookey).
rates need to be increased. Increasingly both approaches need to be used, as there is a finite bandwidth available in silica fibre. High-speed sources are obviously crucial for systems running at 10 Gbit/s rates and there are a variety of solutions available. Also at high channel rates it is important to have short pulses as in TDM the pulsewidth of the signal pulses sets the upper limit on the bit rate. From the viewpoint of nonlinear optical switching, short pulses are advantageous as higher peak powers lead to shorter interaction length requirements and a reduction of the size of the switch.
0030-4018/$ - see front matter 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.optcom.2004.12.009
230
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
This sets out a challenge for a dual-channel system that could provide a pump and signal channel output at different wavelengths with the shortest possible pulse widths at high repetition rates. The pulse to pulse jitter also has to be minimal to enable complete switching as the overlap between a signal pulse and a pump (or gate) pulse determines the proportion of the signal that is switched. In order to assess the performance of new and novel optical fibres for nonlinear optical switching applications a high-speed dual-wavelength short pulse source was required. The design chosen was a dual-channel gain-switched distributed feedback (DFB) laser system. DFBs were chosen for their high side-mode suppression and wavelength stability and fibre amplifiers were used to achieve high peak powers. A description is now given of the system design and the principles of operation of each of the stages used. Finally the characteristics of this system is described for the suitability of nonlinear optical experiments. The dual picosecond source consisted of two synchronised RF generators driving two InGaAsP DFB laser units. The overall system design is shown in Fig. 1. Temperature controls on the
DFB unit allow the peak wavelength to be finely tuned. The HP8648B and Agilent 8648C RF signal generators are synchronised using a 10-MHz signal that also triggers the detection system. Gain-switching involves the modulation of a laser diode held just below the threshold current. The initial DC level, the applied AC frequency and magnitude determine the form of the optical pulse produced. The optical pulse shape can be broadened or compressed and the pulse can also be delayed by a suitable selection of these parameters [1]. Any semiconductor laser is sensitive to back reflections coupled into the cavity and particularly when there are multiple inline amplifiers in the system. Counter propagating optical power can be significant. To eliminate this a polarisation insensitive optical isolator is used. The process of gainswitching produces pulses that are considerably chirped because of temporal refractive index variations occurring during pulse generation. The chirp induced is negative which corresponds to a frequency shift of the leading and trailing edges to the blue and red, respectively. Since red-shifted components travel faster than blue-shifted components in the normal (or positive) dispersion regime,
Fig. 1. Configuration used in the pump–probe experiments.
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
231
the trailing edge is advanced and the leading edge retarded leading to pulse compression. Appropriate selection of the length of the normally dispersive fibre allows the production of pulses that are below 10 ps in duration [2]. In the 1.55-lm region dispersion shifted fibre (DSF) is required. In the system being described the pulse compressor consists of one 100-m length and one 200-m length of DSF both were of the type Lucent (DK-SM) with a dispersion of 99 ps nm1 km1 and a core diameter of 9.5 lm. Different lengths were used in order to observe different amounts of compression and also to spread the compression between in-line amplification. A simple relation can be used to estimate the optimum length of fibre for maximum pulse compression [3] DL ¼
Dt ; Dk
ð1Þ
where D is the dispersion parameter for the fibre, Dt is the initial full-width half-maximum pulse width and Dk is the initial spectral width. Given the D value for the fibre used and an initial pulse width of 35 ps and spectral width of 80 pm, then the optimum length of fibre will be of the order of 4 km. Clearly we will not be seeing optimum pulse compression in our system but some compression is expected. In the dual picosecond laser system amplifiers are required to increase the pump channel peak power (Channel 1 in Fig. 1). The signal channel (Channel 2) requires only one EDFA stage to increase the average power from approximately 100 lW to 3–4 mW. In Fig. 2 the types of amplifiers used in the amplification of the pump are shown. Initial amplification is given by WDMAMP7, an EDFA supplied by Nortel Networks. The output is then coupled into the first pulse compression stage (DSF1) consisting of 200 m of DSF before being amplified by FGM1 (a commercial EDFA from Corning, model FGM-D-080-03). Further pulse compression in DSF2 is followed by amplification at FGM2 (same model as FGM1). A band-pass filter with a bandwidth of 2 nm tuned to the peak of the laser emission is then used to remove the accumulated ASE before amplification in the high power EDFA (HP-
Fig. 2. Amplifier chain used for high power pump channel.
EDFA). The band-pass filter is required to improve the contrast of the output pulses which would be degraded if the ASE were allowed to propagate.
2. Experiment The following sections will describe the characterisation and optimisation of the output pulse trains from the laser system described previously. A description of the various detection and measurement techniques is provided followed by the presentation of results which characterise each stage of the system and also show how the performance is optimised. In order to observe the temporal pulse shape, measure the pulse width and repetition rate of the output from the system fast photodiodes were used. The New Focus ultrafast photodetector 1014 module was chosen. This module consists of an InGaAs Schottky photodiode of diameter 12 lm with a self-contained bias circuit and battery. This type of detector is suitable for the detection of light from 950 to 1650 nm. The optical signal is
232
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
delivered to the device using a single mode optical fibre. The bandwidth of the detector was quoted as 45 GHz [4]. The modules were connected directly to a high bandwidth oscilloscope thus eliminating the need for coaxial cables that can distort the signal. The oscilloscope used was a high bandwidth digital-sampling oscilloscope consisting of the Agilent Infiniium 86100A DCA with an 83484 Dual Channel 50 GHz electrical plug-in module. The unit is triggered externally using the 10-MHz signal from one of the RF generators driving the DFB unit. The output spectrum of the system is measured using an Advantest Q8384 optical spectrum analyser (OSA). This OSA uses a diffraction grating monochromator with a wavelength range extending from 600 to 1600 nm. The resolution is 10 pm with a dynamic range of 60 dB [5]. Average optical power measurements were made using a KD Optics Limited Device Alignment Test Set (DATS) with a high power detector head calibrated for 1550 nm. The unmodulated spectra for the two lasers were recorded both above and below threshold showing the characteristic suppressed side modes of diode lasers. It can be seen in Fig. 3 that below threshold the emission peaks at a different wavelength to that above threshold. This is due to the wavelength selective feedback that permits lasing only at the wavelength given by the grating structure.
Below threshold the broad spectrum of the spontaneous emission shows a kink near the fundamental lasing wavelength. This dip is due to the mode gap at the Bragg wavelength that results from the asymmetry of the grating as seen by waves travelling in different directions. This means that in a real DFB the fundamental output wavelength will be at the modes on either side of the Bragg wavelength. Single wavelength operation is achieved by either using an antireflection coating on one side of the device or the introduction of a 1/4 wave phase shift in the centre of the grating that enables lasing at the centre of the mode gap [6]. The threshold currents for the two DFB lasers were found to be 10 and 7 mA for DFB1 and DFB2, respectively. The wavelength stability with respect to temperature is important however it is also desirable to be able to tune the wavelength of the gainswitched DFB lasers by changing the temperature. The two lasers were gain-switched at 1 GHz with an applied RF power of 19 dBm and the peak wavelengths were tracked as the temperature was changed. In Fig. 4 the results can be seen where wavelength increase with temperature is evident. It should be noted that although there is only an 11 and 10 pm per C increase for DFB1 and DFB2, respectively, the two laser wavelengths are separated by 920 pm at similar temperatures
Fig. 3. Unmodulated DFB output spectra for: (a) DFB1 and (b) DFB2 below and above Ith.
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
233
Fig. 4. Effects of temperature changes on DFB output wavelength. The available tuning range of the peak wavelength can be seen.
and so a separation of over 1 nm can be achieved between the two peak wavelengths. The side mode suppression of the two DFB lasers was measured to be almost 40 dB.
3. Results To optimise the pulse width and pulse shape investigations were carried out on the influence of each of the applied current parameters on the temporal and spectral output. If the applied RF
modulation is kept at a constant frequency the effects of the magnitude of this modulation and the DC current on the DFB output can be observed. Fig. 5 shows the effects of increasing RF power with a constant DC current and the effects of increasing the DC current with a constant RF power on the temporal profile of the output pulse from DFB1. The RF modulation was kept at 1 GHz. From these results it can be seen how the pulse width can be minimised and the profile pedestal reduced by selecting appropriate current settings. The pedestal is due to the onset of the
Fig. 5. Effects of RF power (a) and DC current (b) on the temporal profile at constant repetition rate.
234
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
second relaxation oscillation. A delay can be induced on the output pulse by changing the DC current, that is the position below threshold that the diode is kept before the RF pulse is applied. Indeed the pulse is delayed by up to 100 ps by changing the DC current level. In Fig. 6 the effects of the DC current level and the RF power on the output spectrum can be seen. In Fig. 6(a) it can be seen that at low RF powers the spectrum is broad and the magnitude decreases rapidly. This corresponds to the situation where the unmodulated diode is below threshold and the magnitude of the applied current pulse is not high enough to bring the diode above threshold to start lasing. At gradually higher RF powers the lasing process begins and the lasing peak rises along with the side modes. In Fig. 6(b) the effects at a constant RF power of 19 dBm show that even at very low DC current levels the applied AC magnitude is high enough to bring the diode above threshold for lasing to occur. The repetition rate also determines the properties of an output pulse from a gain-switched laser. The optimum settings of RF power and DC current needed to obtain the shortest pulse with minimal pedestal are shown in Fig. 7. The corresponding pulse width and peak powers are shown in Fig. 8. It is evident that there exists a range of frequencies outside of which the pulse width becomes
broad. In particular outside of the range 500– 1500 MHz the pulse width becomes greater than 50 ps. Also outside this range the peak power decreases, this is not solely due to the increased pulse width but also due to a decrease in the output pulse energy. The overlap between the pump and signal was found to be unchanged when measured after the fused fibre coupler and then after transmission through the fibre, this confirms that there is no walk-off between the two pulses over the length of fibre used. The difference in group-velocity between the signal and the pump wavelengths is small enough such that the walk-off length is far greater than the length of the fibre tested. Using the trends observed in Figs. 6–8 the output from the DFB lasers could be optimised in terms of repetition rate, RF power and DC level. Once these parameters are set the temperatures of the two DFBs can be tuned to give the greatest separation of the two channels. This separation is needed because there is a need to be able to resolve the pump and signal channel light in an optical switch. However the two wavelength channels have been designed to be close together and the normal separation is of the order of 1 nm. This is useful for the study of fibres and devices where dispersion is high and cannot be controlled, for example in chalcogenide or telluride fibres so that any pump–probe walk-off is minimised.
Fig. 6. Effects of RF power (a) and DC current (b) on the output spectrum at constant repetition rate.
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
235
Fig. 7. RF power and drive current required to obtain the shortest pedestal free pulse from DFB1.
Fig. 8. Measured values of the pulse width and peak power from DFB1 using the current parameters in Fig. 7.
To examine the performance of the pulse compression unit an input pulse from DFB1 was coupled into the three different lengths of DSF. The pulse was produced with a 1 GHz, 20 dBm RF modulation with a DC drive current of 8.0 mA. The temporal profile taken using the high band-
width oscilloscope set to a large averaging mode can be seen in Fig. 9. It is clear that pulse compression does take place and the amount of compression increases with length of DSF. This indicates that the length of optimum compression has not been reached. The initial 37 ps pulse is compressed
236
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
Fig. 9. Pulse profile after propagation through three different lengths of DSF.
to 24 ps after 300 m of DSF. The transform limited pulse width (the smallest possible pulse width given the initial spectral and temporal profile) was found to be of the order of 10 ps if a Gaussian pulse shape and spectrum is assumed. However achieving these ideal pulse characteristics is not straightforward with gain-switching in this way. Average power measurements were taken to assess the performance of each of the EDFA units in the pump channel. An input pulse of 110 lW at 1 GHz was coupled into the amplifier chain. The initial EDFA denoted WDMAMP7 had a fixed pump drive current and so could not be adjusted to give different amounts of gain. The gain provided by this unit was 13 dB. Now coupling the input pulse from the DFB into the FGM1 the output from FGM2 was measured as a function of the pump drive current supplied to FGM1 and FGM2. It can be seen in Fig. 10 that the output from the coupled EDFA units saturates as FGM1 drive current is increased. That is with FGM2 held at a maximum pump current and the input signal from FGM1 increased, the gain of FGM2 quickly saturates. This means that in operation there is no need to run FGM1 at the maximum pump current, as there is little increase in output power from FGM2. The high power EDFA unit (HPEDFA) was also
characterised using a 110-lW input signal from the DFB1. As can be seen from Fig. 11, the average output power was seen to saturate at 16.8 dBm, which corresponds to a small signal gain of 26 dB. The use of the backwards pump laser gives an additional 1.3 dB gain before signal induced saturation occurs. As there are multiple EDFA units in series in the system the ASE effects were expected to be large. In order to reduce the ASE light propagating from the system a band-pass filter was used. This was in the form of a JDS-Fitel tuneable grating filter with a bandwidth of 2 nm. The effects of the filtering on the output from the amplifier chain can be seen in Fig. 12. The filter was placed between FGM2 and HP-EDFA. As can be seen the ASE peak has been reduced by more than 25 dB. The effective insertion loss of the filter, defined as the average power penalty due to the filter being placed in the system was found to be 0.3 dB. In Table 1 the average optical power through the system is listed.
4. Discussion The preceding sections have described the design, characterisation and optimisation of a dual-channel picosecond laser system that can
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
237
Fig. 10. Output power from FGM1 and FGM2 in series. Shows the output from the pair for the cases where the FGM1 drive current is increased with a maximised FGM2 drive current (m) and for vice versa (h).
Fig. 11. Average optical output power from HP-EDFA with a 110 lW input signal. The arrow indicates the use of dual pumping.
operate at GHz repetition rates. The output from the system consists of a signal pulse with an average power of the order of 5 mW and a pump pulse
with an average power of 90 mW. It is desired that the peak power of the pump pulse, Ppk, be as high as possible as this determines the magnitude of the
238
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
Fig. 12. Improvement of pump channel output spectrum due to insertion of tuneable filter before high-power amplifier.
Table 1 Average power at the main stages of the system Stage
Description
Average optical power (mW)
Average optical power (dBm)
DFB1 WDMAMP FGM2 BPF HPEDFA (no BPF) HPEDFA (with BPF)
Output from DFB after isolator Constant gain
0.11 2.39 48 11.5 91.24 84.75
9.56 3.78 16.8 10.6 19.6 19.28
Band-pass filter System output System output
nonlinear phase change D/, in a nonlinear all-optical switch of length L. This can be seen from the equation for the phase change in a Kerr shutter given by, D/
2p P pk n2 L; k A
ð2Þ
where n2 is the nonlinear refractive index and A is the effective mode area. So as Ppk increases the interaction length required is reduced. Assuming a pulse width tp, of 24 ps, as measured for the case of optimum pulse compression, and an average power, Pav of 90 mW at 1 GHz repetition rate the relation, P av 1 P pk ¼ ; ð3Þ tp R
where R is the repetition rate of the pulse train, can be used to determine the maximum peak power from the system. In this case this corresponds to a peak power of 3.75 W. This pump peak power enables a p phase change in single mode tellurite fibre possessing an n2 of 4 · 1019 cm2 W1 [7] after a length of 5 m or a p phase change in silica fibre after a length of 60 m. This is assuming an effective mode area of 9 lm2.
5. Conclusions We have demonstrated the operation of a pump–probe laser system for use in the study of nonlinear optical effects. An investigation
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
was made of the pulse spectrum and temporal profile and their dependence on the applied driving current characteristics and DFB temperature. It was shown that it is possible to tune the delay of the output pulses by varying either the magnitude of the applied ac modulation or the dc level. The peak wavelengths of the DFB output could finely tuned by changing the DFB temperature. Pulse compression through normally dispersive fibre was demonstrated. Pulses as short as 24 ps and with peak powers up to 3.75 W have been obtained using inline amplification with ASE filtering. This system allows the user to accurately vary the repetition rate, pulse-to-pulse delay, peak wavelength and pulse width through adjustment of the driving current characteristics, temperature and fibre compressor length making it an ideal tool for the study of picosecond pulse propagation in optical fibres and devices.
239
Acknowledgment The authors acknowledge the support and advice given by Nortel Networks Limited including the donation of equipment used in this work.
References [1] A.J. Vickers, S. Feivan, P. Yan, SPIE Proc. 3944 (2000) 219. [2] A. Takada, T. Sugie, M. Saruwatarii, Electron. Lett. 22 (25) (1986) 1347. [3] H.F. Liu, Y. Ogawa, S. Oshiba, T. Tonaka, IEEE J. Quantum Electron. 27 (6) (1991) 1655. [4] Models 100x, 101X, 102X, Users Manual, New Focus Inc., Santa Clara, CA. [5] Optical Spectrum Analyzer Operation Manual, FOE8335041C00, Q8384, Advantest Corporation, 1999. [6] A. Yariv, Optical Electronics in Modern Communications, fifth ed., Oxford University Press, 1997 (Chapter 16). [7] K.S. Bindra, H.T. Bookey, A.K. Kar, B.S. Wherrett, X. Liu, A. Jha, Appl. Phys. Lett. 79 (13) (2001) 1939.
Characterisation and optimisation of a dual-channel picosecond gain-switched DFB laser system for use as a pump–probe sourceq Henry T.G. Bookey *, Ajoy K. Kar Department of Physics, School of Engineering and Physical Sciences, Heriot-Watt University, David Brewster Building, Edinburgh EH14 4AS, UK Received 29 July 2004; received in revised form 2 December 2004; accepted 3 December 2004
Abstract The output characteristics of a dual-channel gain-switched laser system are studied with respect to the gain-switch frequency and power. InGaAsP laser diodes operating at 1548 and 1549 nm are gain-switched at 1 GHz and the spectral and temporal profiles of the pulses are investigated. The development of a pump–probe system with the use of in-line amplifiers and amplified spontaneous emission filtering is detailed. 2004 Elsevier B.V. All rights reserved. PACS: 42.55; 42.81; 06.60J Keywords: DFB; Gain-switching; Picosecond; Pump–probe system
1. Introduction In order to increase the amount of data transmitted through a telecommunications system that employs WDM and TDM either the number of wavelength channels or the individual channel q
This work was supported in part by the Engineering and Physical Sciences Research Council. * Corresponding author. Tel.: +44 131 451 3047; fax: +44 131 451 3136. E-mail address: [email protected] (H.T.G. Bookey).
rates need to be increased. Increasingly both approaches need to be used, as there is a finite bandwidth available in silica fibre. High-speed sources are obviously crucial for systems running at 10 Gbit/s rates and there are a variety of solutions available. Also at high channel rates it is important to have short pulses as in TDM the pulsewidth of the signal pulses sets the upper limit on the bit rate. From the viewpoint of nonlinear optical switching, short pulses are advantageous as higher peak powers lead to shorter interaction length requirements and a reduction of the size of the switch.
0030-4018/$ - see front matter 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.optcom.2004.12.009
230
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
This sets out a challenge for a dual-channel system that could provide a pump and signal channel output at different wavelengths with the shortest possible pulse widths at high repetition rates. The pulse to pulse jitter also has to be minimal to enable complete switching as the overlap between a signal pulse and a pump (or gate) pulse determines the proportion of the signal that is switched. In order to assess the performance of new and novel optical fibres for nonlinear optical switching applications a high-speed dual-wavelength short pulse source was required. The design chosen was a dual-channel gain-switched distributed feedback (DFB) laser system. DFBs were chosen for their high side-mode suppression and wavelength stability and fibre amplifiers were used to achieve high peak powers. A description is now given of the system design and the principles of operation of each of the stages used. Finally the characteristics of this system is described for the suitability of nonlinear optical experiments. The dual picosecond source consisted of two synchronised RF generators driving two InGaAsP DFB laser units. The overall system design is shown in Fig. 1. Temperature controls on the
DFB unit allow the peak wavelength to be finely tuned. The HP8648B and Agilent 8648C RF signal generators are synchronised using a 10-MHz signal that also triggers the detection system. Gain-switching involves the modulation of a laser diode held just below the threshold current. The initial DC level, the applied AC frequency and magnitude determine the form of the optical pulse produced. The optical pulse shape can be broadened or compressed and the pulse can also be delayed by a suitable selection of these parameters [1]. Any semiconductor laser is sensitive to back reflections coupled into the cavity and particularly when there are multiple inline amplifiers in the system. Counter propagating optical power can be significant. To eliminate this a polarisation insensitive optical isolator is used. The process of gainswitching produces pulses that are considerably chirped because of temporal refractive index variations occurring during pulse generation. The chirp induced is negative which corresponds to a frequency shift of the leading and trailing edges to the blue and red, respectively. Since red-shifted components travel faster than blue-shifted components in the normal (or positive) dispersion regime,
Fig. 1. Configuration used in the pump–probe experiments.
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
231
the trailing edge is advanced and the leading edge retarded leading to pulse compression. Appropriate selection of the length of the normally dispersive fibre allows the production of pulses that are below 10 ps in duration [2]. In the 1.55-lm region dispersion shifted fibre (DSF) is required. In the system being described the pulse compressor consists of one 100-m length and one 200-m length of DSF both were of the type Lucent (DK-SM) with a dispersion of 99 ps nm1 km1 and a core diameter of 9.5 lm. Different lengths were used in order to observe different amounts of compression and also to spread the compression between in-line amplification. A simple relation can be used to estimate the optimum length of fibre for maximum pulse compression [3] DL ¼
Dt ; Dk
ð1Þ
where D is the dispersion parameter for the fibre, Dt is the initial full-width half-maximum pulse width and Dk is the initial spectral width. Given the D value for the fibre used and an initial pulse width of 35 ps and spectral width of 80 pm, then the optimum length of fibre will be of the order of 4 km. Clearly we will not be seeing optimum pulse compression in our system but some compression is expected. In the dual picosecond laser system amplifiers are required to increase the pump channel peak power (Channel 1 in Fig. 1). The signal channel (Channel 2) requires only one EDFA stage to increase the average power from approximately 100 lW to 3–4 mW. In Fig. 2 the types of amplifiers used in the amplification of the pump are shown. Initial amplification is given by WDMAMP7, an EDFA supplied by Nortel Networks. The output is then coupled into the first pulse compression stage (DSF1) consisting of 200 m of DSF before being amplified by FGM1 (a commercial EDFA from Corning, model FGM-D-080-03). Further pulse compression in DSF2 is followed by amplification at FGM2 (same model as FGM1). A band-pass filter with a bandwidth of 2 nm tuned to the peak of the laser emission is then used to remove the accumulated ASE before amplification in the high power EDFA (HP-
Fig. 2. Amplifier chain used for high power pump channel.
EDFA). The band-pass filter is required to improve the contrast of the output pulses which would be degraded if the ASE were allowed to propagate.
2. Experiment The following sections will describe the characterisation and optimisation of the output pulse trains from the laser system described previously. A description of the various detection and measurement techniques is provided followed by the presentation of results which characterise each stage of the system and also show how the performance is optimised. In order to observe the temporal pulse shape, measure the pulse width and repetition rate of the output from the system fast photodiodes were used. The New Focus ultrafast photodetector 1014 module was chosen. This module consists of an InGaAs Schottky photodiode of diameter 12 lm with a self-contained bias circuit and battery. This type of detector is suitable for the detection of light from 950 to 1650 nm. The optical signal is
232
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
delivered to the device using a single mode optical fibre. The bandwidth of the detector was quoted as 45 GHz [4]. The modules were connected directly to a high bandwidth oscilloscope thus eliminating the need for coaxial cables that can distort the signal. The oscilloscope used was a high bandwidth digital-sampling oscilloscope consisting of the Agilent Infiniium 86100A DCA with an 83484 Dual Channel 50 GHz electrical plug-in module. The unit is triggered externally using the 10-MHz signal from one of the RF generators driving the DFB unit. The output spectrum of the system is measured using an Advantest Q8384 optical spectrum analyser (OSA). This OSA uses a diffraction grating monochromator with a wavelength range extending from 600 to 1600 nm. The resolution is 10 pm with a dynamic range of 60 dB [5]. Average optical power measurements were made using a KD Optics Limited Device Alignment Test Set (DATS) with a high power detector head calibrated for 1550 nm. The unmodulated spectra for the two lasers were recorded both above and below threshold showing the characteristic suppressed side modes of diode lasers. It can be seen in Fig. 3 that below threshold the emission peaks at a different wavelength to that above threshold. This is due to the wavelength selective feedback that permits lasing only at the wavelength given by the grating structure.
Below threshold the broad spectrum of the spontaneous emission shows a kink near the fundamental lasing wavelength. This dip is due to the mode gap at the Bragg wavelength that results from the asymmetry of the grating as seen by waves travelling in different directions. This means that in a real DFB the fundamental output wavelength will be at the modes on either side of the Bragg wavelength. Single wavelength operation is achieved by either using an antireflection coating on one side of the device or the introduction of a 1/4 wave phase shift in the centre of the grating that enables lasing at the centre of the mode gap [6]. The threshold currents for the two DFB lasers were found to be 10 and 7 mA for DFB1 and DFB2, respectively. The wavelength stability with respect to temperature is important however it is also desirable to be able to tune the wavelength of the gainswitched DFB lasers by changing the temperature. The two lasers were gain-switched at 1 GHz with an applied RF power of 19 dBm and the peak wavelengths were tracked as the temperature was changed. In Fig. 4 the results can be seen where wavelength increase with temperature is evident. It should be noted that although there is only an 11 and 10 pm per C increase for DFB1 and DFB2, respectively, the two laser wavelengths are separated by 920 pm at similar temperatures
Fig. 3. Unmodulated DFB output spectra for: (a) DFB1 and (b) DFB2 below and above Ith.
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
233
Fig. 4. Effects of temperature changes on DFB output wavelength. The available tuning range of the peak wavelength can be seen.
and so a separation of over 1 nm can be achieved between the two peak wavelengths. The side mode suppression of the two DFB lasers was measured to be almost 40 dB.
3. Results To optimise the pulse width and pulse shape investigations were carried out on the influence of each of the applied current parameters on the temporal and spectral output. If the applied RF
modulation is kept at a constant frequency the effects of the magnitude of this modulation and the DC current on the DFB output can be observed. Fig. 5 shows the effects of increasing RF power with a constant DC current and the effects of increasing the DC current with a constant RF power on the temporal profile of the output pulse from DFB1. The RF modulation was kept at 1 GHz. From these results it can be seen how the pulse width can be minimised and the profile pedestal reduced by selecting appropriate current settings. The pedestal is due to the onset of the
Fig. 5. Effects of RF power (a) and DC current (b) on the temporal profile at constant repetition rate.
234
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
second relaxation oscillation. A delay can be induced on the output pulse by changing the DC current, that is the position below threshold that the diode is kept before the RF pulse is applied. Indeed the pulse is delayed by up to 100 ps by changing the DC current level. In Fig. 6 the effects of the DC current level and the RF power on the output spectrum can be seen. In Fig. 6(a) it can be seen that at low RF powers the spectrum is broad and the magnitude decreases rapidly. This corresponds to the situation where the unmodulated diode is below threshold and the magnitude of the applied current pulse is not high enough to bring the diode above threshold to start lasing. At gradually higher RF powers the lasing process begins and the lasing peak rises along with the side modes. In Fig. 6(b) the effects at a constant RF power of 19 dBm show that even at very low DC current levels the applied AC magnitude is high enough to bring the diode above threshold for lasing to occur. The repetition rate also determines the properties of an output pulse from a gain-switched laser. The optimum settings of RF power and DC current needed to obtain the shortest pulse with minimal pedestal are shown in Fig. 7. The corresponding pulse width and peak powers are shown in Fig. 8. It is evident that there exists a range of frequencies outside of which the pulse width becomes
broad. In particular outside of the range 500– 1500 MHz the pulse width becomes greater than 50 ps. Also outside this range the peak power decreases, this is not solely due to the increased pulse width but also due to a decrease in the output pulse energy. The overlap between the pump and signal was found to be unchanged when measured after the fused fibre coupler and then after transmission through the fibre, this confirms that there is no walk-off between the two pulses over the length of fibre used. The difference in group-velocity between the signal and the pump wavelengths is small enough such that the walk-off length is far greater than the length of the fibre tested. Using the trends observed in Figs. 6–8 the output from the DFB lasers could be optimised in terms of repetition rate, RF power and DC level. Once these parameters are set the temperatures of the two DFBs can be tuned to give the greatest separation of the two channels. This separation is needed because there is a need to be able to resolve the pump and signal channel light in an optical switch. However the two wavelength channels have been designed to be close together and the normal separation is of the order of 1 nm. This is useful for the study of fibres and devices where dispersion is high and cannot be controlled, for example in chalcogenide or telluride fibres so that any pump–probe walk-off is minimised.
Fig. 6. Effects of RF power (a) and DC current (b) on the output spectrum at constant repetition rate.
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
235
Fig. 7. RF power and drive current required to obtain the shortest pedestal free pulse from DFB1.
Fig. 8. Measured values of the pulse width and peak power from DFB1 using the current parameters in Fig. 7.
To examine the performance of the pulse compression unit an input pulse from DFB1 was coupled into the three different lengths of DSF. The pulse was produced with a 1 GHz, 20 dBm RF modulation with a DC drive current of 8.0 mA. The temporal profile taken using the high band-
width oscilloscope set to a large averaging mode can be seen in Fig. 9. It is clear that pulse compression does take place and the amount of compression increases with length of DSF. This indicates that the length of optimum compression has not been reached. The initial 37 ps pulse is compressed
236
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
Fig. 9. Pulse profile after propagation through three different lengths of DSF.
to 24 ps after 300 m of DSF. The transform limited pulse width (the smallest possible pulse width given the initial spectral and temporal profile) was found to be of the order of 10 ps if a Gaussian pulse shape and spectrum is assumed. However achieving these ideal pulse characteristics is not straightforward with gain-switching in this way. Average power measurements were taken to assess the performance of each of the EDFA units in the pump channel. An input pulse of 110 lW at 1 GHz was coupled into the amplifier chain. The initial EDFA denoted WDMAMP7 had a fixed pump drive current and so could not be adjusted to give different amounts of gain. The gain provided by this unit was 13 dB. Now coupling the input pulse from the DFB into the FGM1 the output from FGM2 was measured as a function of the pump drive current supplied to FGM1 and FGM2. It can be seen in Fig. 10 that the output from the coupled EDFA units saturates as FGM1 drive current is increased. That is with FGM2 held at a maximum pump current and the input signal from FGM1 increased, the gain of FGM2 quickly saturates. This means that in operation there is no need to run FGM1 at the maximum pump current, as there is little increase in output power from FGM2. The high power EDFA unit (HPEDFA) was also
characterised using a 110-lW input signal from the DFB1. As can be seen from Fig. 11, the average output power was seen to saturate at 16.8 dBm, which corresponds to a small signal gain of 26 dB. The use of the backwards pump laser gives an additional 1.3 dB gain before signal induced saturation occurs. As there are multiple EDFA units in series in the system the ASE effects were expected to be large. In order to reduce the ASE light propagating from the system a band-pass filter was used. This was in the form of a JDS-Fitel tuneable grating filter with a bandwidth of 2 nm. The effects of the filtering on the output from the amplifier chain can be seen in Fig. 12. The filter was placed between FGM2 and HP-EDFA. As can be seen the ASE peak has been reduced by more than 25 dB. The effective insertion loss of the filter, defined as the average power penalty due to the filter being placed in the system was found to be 0.3 dB. In Table 1 the average optical power through the system is listed.
4. Discussion The preceding sections have described the design, characterisation and optimisation of a dual-channel picosecond laser system that can
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
237
Fig. 10. Output power from FGM1 and FGM2 in series. Shows the output from the pair for the cases where the FGM1 drive current is increased with a maximised FGM2 drive current (m) and for vice versa (h).
Fig. 11. Average optical output power from HP-EDFA with a 110 lW input signal. The arrow indicates the use of dual pumping.
operate at GHz repetition rates. The output from the system consists of a signal pulse with an average power of the order of 5 mW and a pump pulse
with an average power of 90 mW. It is desired that the peak power of the pump pulse, Ppk, be as high as possible as this determines the magnitude of the
238
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
Fig. 12. Improvement of pump channel output spectrum due to insertion of tuneable filter before high-power amplifier.
Table 1 Average power at the main stages of the system Stage
Description
Average optical power (mW)
Average optical power (dBm)
DFB1 WDMAMP FGM2 BPF HPEDFA (no BPF) HPEDFA (with BPF)
Output from DFB after isolator Constant gain
0.11 2.39 48 11.5 91.24 84.75
9.56 3.78 16.8 10.6 19.6 19.28
Band-pass filter System output System output
nonlinear phase change D/, in a nonlinear all-optical switch of length L. This can be seen from the equation for the phase change in a Kerr shutter given by, D/
2p P pk n2 L; k A
ð2Þ
where n2 is the nonlinear refractive index and A is the effective mode area. So as Ppk increases the interaction length required is reduced. Assuming a pulse width tp, of 24 ps, as measured for the case of optimum pulse compression, and an average power, Pav of 90 mW at 1 GHz repetition rate the relation, P av 1 P pk ¼ ; ð3Þ tp R
where R is the repetition rate of the pulse train, can be used to determine the maximum peak power from the system. In this case this corresponds to a peak power of 3.75 W. This pump peak power enables a p phase change in single mode tellurite fibre possessing an n2 of 4 · 1019 cm2 W1 [7] after a length of 5 m or a p phase change in silica fibre after a length of 60 m. This is assuming an effective mode area of 9 lm2.
5. Conclusions We have demonstrated the operation of a pump–probe laser system for use in the study of nonlinear optical effects. An investigation
H.T.G. Bookey, A.K. Kar / Optics Communications 248 (2005) 229–239
was made of the pulse spectrum and temporal profile and their dependence on the applied driving current characteristics and DFB temperature. It was shown that it is possible to tune the delay of the output pulses by varying either the magnitude of the applied ac modulation or the dc level. The peak wavelengths of the DFB output could finely tuned by changing the DFB temperature. Pulse compression through normally dispersive fibre was demonstrated. Pulses as short as 24 ps and with peak powers up to 3.75 W have been obtained using inline amplification with ASE filtering. This system allows the user to accurately vary the repetition rate, pulse-to-pulse delay, peak wavelength and pulse width through adjustment of the driving current characteristics, temperature and fibre compressor length making it an ideal tool for the study of picosecond pulse propagation in optical fibres and devices.
239
Acknowledgment The authors acknowledge the support and advice given by Nortel Networks Limited including the donation of equipment used in this work.
References [1] A.J. Vickers, S. Feivan, P. Yan, SPIE Proc. 3944 (2000) 219. [2] A. Takada, T. Sugie, M. Saruwatarii, Electron. Lett. 22 (25) (1986) 1347. [3] H.F. Liu, Y. Ogawa, S. Oshiba, T. Tonaka, IEEE J. Quantum Electron. 27 (6) (1991) 1655. [4] Models 100x, 101X, 102X, Users Manual, New Focus Inc., Santa Clara, CA. [5] Optical Spectrum Analyzer Operation Manual, FOE8335041C00, Q8384, Advantest Corporation, 1999. [6] A. Yariv, Optical Electronics in Modern Communications, fifth ed., Oxford University Press, 1997 (Chapter 16). [7] K.S. Bindra, H.T. Bookey, A.K. Kar, B.S. Wherrett, X. Liu, A. Jha, Appl. Phys. Lett. 79 (13) (2001) 1939.