528 OPTICS LETTERS / Vol. 30, No. 5 / March 1, 2005
10-GHz dispersion-managed soliton fiber-optical parametricoscillator using regenerative mode locking
Preetpaul S. Devgan, Jacob Lasri, Renyong Tang, Vladimir S. Grigoryan, William L. Kath, and Prem Kumar
Center for Photonic Communication and Computing, Department of Electrical and Computer Engineering,Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208-3118
Received June 28, 2004
We demonstrate a regeneratively mode-locked fiber-optical parametric oscillator that utilizes intracavity dis-persion compensation to generate pulses at a 10-GHz repetition rate in both soliton and nonsoliton regimes.At the threshold pump power the generated pulses are close to fundamental solitons. At higher pump powerswe found a significant deviation of the pulses from the sech2 shape. In addition, the use of an ultralow-jitter self-starting pump–pulse source in a regenerative feedback loop allows for a significant reduction of thesignal’s timing jitter and amplitude noise. © 2005 Optical Society of America
OCIS codes: 190.4970, 190.5530, 190.4380.
Regeneratively mode-locked (RML) erbium-fiberlasers have often been proposed for generating short,transform-limited soliton pulses in the 1550-nm wave-length band. Demonstrations of regenerative modelocking include feeding the signal back to the modula-tor1,2 and using the feedback to change the length ofthe cavity by means of a piezoelectric transducer.3 Asignificant benefit of regenerative mode locking hasbeen the reduction of noise in the generated pulsesrelative to those obtained from nonregenerative lasers.Improvement in the noise properties of these lasershas been demonstrated.4 – 6 In addition, experimentaland theoretical treatment of mode-locked (ML) lasershas utilized intracavity dispersion management for ob-taining soliton pulses.7,8 However, these lasers havelimitations that prevent broader use. Foremost is thefact that the tunability of these f iber lasers is limitedby the erbium gain bandwidth. A ML fiber-opticalparametric oscillator (FOPO), however, can facilitatethe generation of short optical pulses at wavelengthsnot available otherwise by utilizing the nonlinearprocess of four-wave mixing.9,10 A combination of re-generative mode locking and dispersion managementin a FOPO allows for the generation of widely tunablesoliton pulses with reduced noise characteristics.Recently we demonstrated, for the f irst time to ourknowledge, a regeneratively mode-locked (RML) FOPOfor obtaining soliton pulses in both the C and the Lbands simultaneously.11 Unlike other regenerativemode-locking techniques, our approach does not utilizean intracavity modulator or a piezoelectric transducer.Instead, we use an ultralow-jitter 10-GHz pump–pulsesource that is RML to the FOPO cavity by feedingback a portion of the oscillating signal to injectionlock the pump. Such regenerative feedback allowsfor stable operation with significant reduction of thepulses’ timing jitter and amplitude noise. In thisLetter we present results of our investigations intothe soliton properties of the generated pulses and thenoise characteristics of the RML FOPO.
A schematic of the experimental setup for a RMLFOPO is shown in Fig. 1. It is similar to what werecently used for generating pulses at two wave-
0146-9592/05/050528-03$15.00/0 ©
lengths simultaneously, one each in the C and Lbands.11 The FOPO consists of a 1-km-long piece ofhighly nonlinear f iber (HNLF), a 3-nm-wide tunableoptical bandpass f ilter, an adjustable optical-delayline, and an isolator, all connected in a ring-cavityconfiguration. The HNLF has a zero-dispersionwavelength �l0� of 1558 nm and a nonlinear coeffi-cient of �9 W21 km21. The remaining 15-m of fiberin the loop is standard SMF-28 fiber, providing adispersion-managed system to allow for soliton-pulsepropagation in the ring cavity. A 5% coupler is placedin the middle of the 15-m-long single-mode fiber toextract the oscillating-signal soliton pulses. Thepump pulses, which are injected into the FOPO viaa 50�50 coupler, are obtained from a self-startingelectroabsorption modulator-based optoelectronic os-cillator (EAM-OEO).12,13 We accomplish regenerativemode locking by feeding back a portion of the outputpulse stream from the FOPO to the high-speed pho-todetector in the optoelectronic oscillator. With thisarrangement, any change in the ring-cavity resonancefrequency causes a shift in the repetition frequencyof the pump pulses, providing automatic harmonicsynchronization between the pump–pulse train andthe pulses oscillating in the FOPO cavity.
Fig. 1. Experimental setup of the RML-FOPO: EDFAs,erbium-doped f iber amplifiers; OBFs, optical bandpass fil-ters; PDs, photodiodes; FPC, fiber polarization controller;SMF, single-mode fiber; ODL, optical delay line; EBF, elec-trical bandpass f ilter; EAM, electroabsorption modulator.
2005 Optical Society of America
March 1, 2005 / Vol. 30, No. 5 / OPTICS LETTERS 529
When the pulsed pump is coupled into the FOPOcavity and the cavity length is adjusted (by means ofthe optical delay line) such that its free spectral rangeequals a subharmonic of the pump repetition rate, har-monic mode locking of the FOPO is established. Whenthe pump power is just above threshold, the outputsignal pulses fit a sech2 shape with a FWHM of 3.9 ps,and their optical spectrum gives a time–bandwidthproduct of 0.35.11 For a soliton in a dispersion-managed system, the energy of the pulse is relatedto the pulse’s width by Esol � E0�1 1 0.7S2�.14 Here,S is the strength of dispersion management, which isgiven by S � �b1
00 2 bav00�l1�t2 2 �b2
00 2 bav00�l2�t2,
and E0 is the energy of the fundamental soliton:
E0 � 2 3 1.76bav00�gt , (1)
where b100 and l1 are the dispersion and the length
of the HNLF and b200 and l2 are the dispersion and
the length of the single-mode fiber, respectively. bav00
is the average dispersion of the cavity, g is the non-linear coeff icient, and t is the FWHM of the pulse.Our calculations show that the strength of the dis-persion management is small �S , 1�. Thus the dis-persion-managed soliton in the cavity is similar to afundamental soliton, explaining the sech2 shape of theoutput pulses. The average pulse energy is 17.8 fJ,whereas the expected energy for the measured pulsewidth from Eq. (1) is 14.4 fJ. The discrepancy can beattributed to the f ilter inside the cavity. When thepump power crosses threshold, the peak gain wave-length shifts by �0.5 nm. This is due to the depen-dence on pump power of the parametric-gain spectralprofile. Because the f ilter is set while the pump isbelow threshold, the oscillating signal above thresholddoes not pass through the center of the filter and addsextra dispersion to the signal. When the wavelengthis tuned, the soliton continues to stay to one side of thefilter, as it needs this extra dispersion to remain stable.The measured pulse energy coincides with the solitonenergy predicted by Eq. (1) if the dispersion throughthe filter contributes an additional 20.03 ps2�km tothe f iber’s path-averaged dispersion in the loop.
As noted above, we observed that the centralwavelength of the oscillating pulses is shifted fromthe peak transmission wavelength of the filter. Thewavelength shift increases as the pump power isincreased. When the shift becomes significant,the f ilter asymmetrically cuts the spectrum of thepulses, resulting in pulse shapes that deviate fromthe sech shape. To verify this effect we measuredthe autocorrelation functions of the signal pulses as afunction of pump power. The results are plotted inFig. 2(a), which clearly shows the increasing deviationof the signal’s autocorrelation shape from a sech2
shape with rising pump power. To further analyzethese results we reconstructed the temporal profilesof the pulses from the autocorrelation data, usingthe procedure described in Ref. 15. The recoverednormalized temporal profiles of the pulses are plottedin Fig. 2(b). One can see that as the pump powerrises above threshold the pulses’ FWHM begins tonarrow significantly and side wings start to develop.
At high pump powers, the measured pulse energies arealso several times larger than that given by Eq. (1) forthe measured pulse widths. The filter, however, doesact to limit the bandwidth growth of the oscillatingsignal pulses at higher pump powers, so the narrowingpulses can no longer sustain the sech shape for thecorresponding fundamental soliton.
To measure the noise performance of the RML FOPOwe characterized the pulses by using a 50-GHz p–i–ndetector followed by a wideband microwave preampli-fier and an electrical spectrum analyzer. We deter-mined the timing jitter and the amplitude noise of thepulse train by analyzing the spectral content of the de-tected signal harmonics in both cases when regenera-tive mode locking was enabled (RML FOPO) and whenit was disabled (ML FOPO). We accomplished the lat-ter by simply disconnecting the fiber from the FOPOoutput to the photodetector in the EAM-OEO. Thenoise power at the nth tone can be described as16
sn2 �
Z vhigh,n
vlow ,n
Sn�v�dv � sA2 1 �vR
2sJ2�n2, (2)
where Sn�v� is the single-sideband noise-power spec-tral density, sA is the rms amplitude-noise power, sJis the rms timing jitter, vR � 2pfR , and fR is the rep-etition rate. At each harmonic we define Sn�v� as theratio of the noise power per unit bandwidth at an off-set v from nvR to the power of the carrier at that
Fig. 2. (a) Autocorrelation traces and (b) reconstructedtemporal shapes of the output signal pulses as a functionof the input pump power. Inset, optical spectrum of thegenerated pulses.
530 OPTICS LETTERS / Vol. 30, No. 5 / March 1, 2005
Fig. 3. (a) Parameter f its of the rms noise over four har-monics for the pump and signal pulses with RML FOPOand without (ML-FOPO). (b) Summary of the results fortiming and amplitude f luctuations.
harmonic. Spectral density Sn�v� is integrated over aspectral range vlow, n to vhigh,n, which is kept constantfor all the harmonics. We measured the noise spectraldensity of four harmonics (10, 20, 30, and 40 GHz) overa frequency offset range of 100–10 MHz in each case.By fitting the data to Eq. (2), we obtained both the am-plitude noise �n � 0� and the timing jitter. Note that,in all measurements, a wideband microwave preampli-fier was used to ensure that the phase noise was wellabove the noise f loor of the electrical spectrum ana-lyzer over the entire frequency range and for all fourharmonics.
Figure 3(a) shows the harmonic-number dependenceof sn
2 for the pump and the signal pulses with andwithout regenerative mode locking, together with f itsto curves of the form of Eq. (2). The measurement re-sults are summarized in Fig. 3(b). The timing jitterand the amplitude noise of the pump pulses from theEAM-OEO are 68 fs and 0.2%, respectively. For thenonregenerative ML FOPO the timing jitter is 0.71 pswith an amplitude noise of 6.1%, whereas introducingregenerative feedback reduced the timing jitter and theamplitude noise of the RML FOPO to 0.23 ps and 3.9%,respectively. Thus the use of the EAM-OEO as aninjection-locked pump source permits a significant re-duction in both the timing jitter (by more than a factorof 3) and the amplitude noise (by a factor of �1.5) ofthe RML FOPO.
In conclusion, we have demonstrated a regenera-tively mode-locked soliton f iber-optical parametricoscillator that is capable of producing picosecondpulses at a 10-GHz repetition rate in both solitonand nonsoliton regimes. We have verified that, whenthe f iber-optical parametric oscillator is just abovethreshold, the generated pulses are consistent withthe sech soliton. At higher pump powers we foundsignificant deviations of the pulses from the solitonshape. By injection locking the FOPO output into aself-starting optoelectronic oscillator that acts as thepump, we reduced the timing jitter of the signal pulsesfrom 710 to 230 fs and decreased the amplitude noise.
We thank C. Joergensen of OFS Denmark for loan ofthe HNLF. Some of the photonic devices used in thisresearch were provided by the Photonics TechnologyAccess Program funded by the National Science Foun-dation and the Defense Advanced Research ProjectsAgency. This research was supported by the NationalScience Foundation under grants ANI-0123495 andIGERT DGE-9987577. P. S. Devgan’s e-mail addressis [email protected].
References
1. M. Nakazawa, E. Yoshida, and Y. Kimura, Electron.Lett. 30, 1603 (1994).
2. K. S. Abedin and F. Kubota, Electron. Lett. 40, 58(2004).
3. M. Nakazawa, E. Yoshida, and K. Tamura, Electron.Lett. 33, 1318 (1997).
4. E. Yoshida and M. Nakazawa, IEEE Photon. Technol.Lett. 11, 548 (1999).
5. K. K. Gupta, D. Novak, and H. Liu, IEEE J. QuantumElectron. 36, 70 (2000).
6. T. R. Clark, T. F. Carruthers, P. J. Matthews, andI. N. Durling III, Electron. Lett. 35, 720 (1999).
7. T. F. Carruthers, I. N. Durling III, M. Horowitz, andC. R. Menyuk, Opt. Lett. 25, 153 (2000).
8. M. Horowitz, C. R. Menyuk, T. F. Carruthers, andI. N. Durling III, J. Lightwave Technol. 18, 1565 (2000).
9. J. E. Sharping, M. Fiorentino, P. Kumar, and R. S.Windeler, Opt. Lett. 27, 1675 (2002).
10. J. Lasri, P. Devgan, R. Tang, J. E. Sharping, andP. Kumar, IEEE Photon. Technol. Lett. 15, 1058 (2003).
11. J. Lasri, P. Devgan, R. Tang, V. S. Grigoryan, W. L.Kath, and P. Kumar, Electron. Lett. 40, 622 (2003).
12. J. Lasri, P. Devgan, R. Tang, and P. Kumar, Opt. Ex-press 11, 1430 (2003), http://www.opticsexpress.org.
13. P. Devgan, J. Lasri, R. Tang, and P. Kumar, Electron.Lett. 39, 1337 (2003).
14. N. J. Smith, N. J. Doran, F. M. Knox, and W. Forysiak,Opt. Lett. 21, 1981 (1996).
15. J. Peatross and A. Rundquist, J. Opt. Soc. Am. B 15,216 (1998).
16. D. van der Linde, Appl. Phys. B 39, 201 (1986).