Characterization of 1.55 µm VCSELs using high-resolution and highdynamic range measurements of the CW optical spectrum

Transcription

1 Characterization o.55 µm VCSs using high-resolution and highdynamic range measurements o the CW optical spectrum Asier VillarancaF*F, Javier asobras and Ignacio arcés TOYBA ab., University o Zaragoza, Walqa Technology Park, d., 97 Huesca Spain ABSTACT We present a methodology or the characterization o the main parameters o VCSs or its use in direct modulation. Power response, chirp parameter alpha, linewidth, side-mode suppression ratio, relaxation oscillations peak requency, damping rate and relative intensity noise IN are obtained rom measurements o the emitted optical spectrum in continuous wave CW operation by means o a high resolution MHz and high dynamic range 8 db optical spectrum analyzer. Many o the main static and dynamic parameters o VCS lasers can be obtained rom the analysis o the optical spectrum when emitting in CW operation, but traditional spectrum analysis techniques do not achieve enough high resolution and dynamic range and high signal to noise ratio to perorm it. ecent developments in high resolution optical spectrum analyzer OSA technology allow a deeper characterization o the main properties o VCSs or its applications in optical communication systems by analyzing their CW emitted spectrum. Keywords: VCS, semiconductor laser, linewidth enhancement actor, relative intensity noise, relaxation oscillations, high-resolution optical spectroscopy. INTODUCTION Vertical-cavity surace-emitting lasers VCSs emitting at.55 µm are very promising sources or access and interconnection in telecommunications systems due to their technologically attractive properties such as very low threshold current and narrow spectral linewidth due to nearly single longitudinal mode operation. Common techniques to carry out a complete characterization o this type o lasers require the laser to be operated both in CW and under modulation. The method we propose in this work only requires measurements in CW mode or the complete experimental characterization o the working parameters or VCSs. The advantages o our method are its simplicity and the act that it does not require the laser to be modulated, thus reducing the experimental setup to use less resources and giving the possibility to use the same setup or the measurement o all the parameters. To carry out the complete characterization enough resolution, dynamic range, and specially a good combination o both is necessary, and using a commercial high-resolution optical spectrum analyzer, a BOSA rom Aragon Photonics abs, both resolution and dynamic range are reached. Main parameters we have characterized are power response, alpha parameter, linewidth, side-mode suppression ratio, relaxation requency, damping rate and IN. With traditional methods, to measure the power response, a power detector or a optical spectrum analyzer OSA is to be used; to measure the linewidth, either intererometric techniques or homodyne analysis should be perormed; or the linewidth enhancement actor, commonly a network analyzer and some hundred kilometers o ibers are needed 3 ; and or the measurement o the relative intensity noise 4 IN a photodetector and an electrical spectrum analyzer are normally used. All these setups needed to ully characterize a laser are very resource-demanding and time-consuming, while the analysis we propose can be carried out in a ew minutes when the methods are automated with a script or the analyzer. In this work, main static and dynamic parameters o VCS lasers have been obtained rom the analysis o the optical spectrum when emitting in CW operation mode 5.. THOTICA BACKOUND Semiconductor lasers can be modelled as non-ideal oscillators that have a narrow but inite width caused by noise-driven modulation. Noise in semiconductor lasers is dominated by spontaneous emission. ach spontaneously emitted photon adds to the coherent ield a small ield component whose phase is random, and thus perturbs both amplitude and phase, causing intensity and phase noise. The eect o these two sources o noise on the optical spectrum o a single mode *asier.villaranca@unizar.es;

2 semiconductor laser operating in CW ar rom threshold can be analyzed by an expression 6 or the power spectral density PSD: [ ] = r 4 / 4 / 4 / 4 / 4 / where is the ull width at hal maximum o the lasing mode, given by the phase noise, is the relaxation oscillation peak requency, is the damping rate o the relaxation oscillations, and is the linewidth enhancement actor. This expression can be separated in two components. The irst term is a orentzian curve centered at the lasing wavelength, related to the phase noise that gives the main contribution to the total power and linewidth and the second term arises a ew Hz rom the peak and has an asymmetric orentzian shape, given by the interaction between the relaxation oscillations and the intensity noise, which orm a pair o shoulders at both sides o the main peak. Thereore, this second term in can be considered as the output o the laser modulated with a noise source. The oscillation relaxation peak requency and damping rate can be obtained directly rom itting to, but also the relative intensity noise IN can be calculated rom the optical CW spectrum. The IN is deined as the ratio o the electrical power spectral density at a given requency to the average power. It is normally measured in the electrical domain, so that the power spectral density is the sum o the detected intensity noise and the contribution o the shot noise. To calculate the IN rom the electrical spectrum ater the photodiode the ollowing expression is used: p s p I I I IN = where I p is the intensity detected by a photodiode and I s is the shot noise. The optical spectrum is not aected by shot noise, but it has contributions o the phase noise that will not produce intensity in a photodetector. We know that the phase noise has a orentzian distribution that we can measure and subtract rom the optical spectrum so we can calculate the electrical spectrum o the signal without the contribution o the shot noise: [ ] = d IN / 3 where is the part o the optical power spectral density corresponding to the phase noise, given by the irst term o. When the laser is operated ar rom saturation, a simpler orm can be used: [ ] = d IN / 4 The laser spectrum can be itted to or most o the operating range o the laser, obtaining the values or the most oten used working parameters or the use o semiconductor lasers in telecommunications with a precise knowledge o the linewidth enhancement actor.

3 The method used or the measurement o is based on the measurement o the linewidth o the lasing mode as a unction o the emitted power 7. This relation predicted by the Henry law 8 is well known ar enough above threshold, and is described by: > = 4πP 6 I we look at the linewidth ar enough below threshold, the relation with emitted power ollows Schawlow-Townes equation or gas lasers 9 : < = 7 πp We use the values o these constants to obtain the value o applying the ollowing expression: > = 8 This equation provides a value o or the laser diode, which is constant as long as there is no power saturation. < 3. MTHODOOY The optical intensity spectrum o a laser emission in CW contains very signiicant inormation. To achieve a good characterization o the works parameters o semiconductor lasers rom it, enough resolution, dynamic range, and specially a good combination o both is necessary. nough dynamic range is achieved by means o traditional optical spectrum analyzers OSA based on gratings and enough high resolution by other spectrometry techniques such as tunable Fabry-Perot iltering or heterodyne detection, but none o them could obtain both enough dynamic range and resolution at the same time. However a new generation o high-resolution spectrum analyzers, based on the stimulated Brillouin scattering nonlinear eect, is able to obtain a resolution o MHz along with a dynamic range o around 8 db. When measuring the optical spectrum with this scope we can see details that cannot be measured by other means, as are normally not resolved either in requency or in dynamic range. This analyzer has some interesting properties, associated to the properties o the Brillouin scattering and to the act that the processing is perormed in the optical domain, in opposition to traditional methods, that perorm the analysis in the electrical domain, which makes the measurement process very ast. Two sets o measurements can be distinguished: a rough sweep or the whole range o operating currents o the laser, and a detailed sweep at shorter steps in the threshold region, which has a particular interest or power and linewidth. Above laser threshold region, an important narrowing o the laser linewidth is produced and it gets closer to the ilter bandwidth o the analyzer and it must be taken into account, however below threshold no correction due to the analyzer ilter is needed, as the spontaneous emission has linewidth values in the order o the Hz. To extract inormation rom the measured spectra, we should take into account that resulting measurement come rom a convolution o the proile o analyzer ilter with VCS proile. So, we have the convolution o aussian proile o this ilter with a very well deined orentzian shape o the phase noise o VCS lasers, resulting a well known Voigt unction: t = b e K a, b a t b dt π 9 with a = ln λ λ / and b = ln /, where λ is the central wavelength o the laser, is the width o the analysis ilter assuming aussian proile and is the linewidth o the laser.

4 4. MASUMNTS For this work, two commercial long-wavelength vertical cavity surace emitting lasers VCS which have similar characteristics have been characterized. The main dierence between them is that one has an anti-relection coating to reduce optical eedback aser VCS_AC and the other has not aser VCS. We operate the lasers in CW conditions using a bias controller orm Thorabs Inc. Due to strong sensitivity o this type o laser to eedback, we have used an additional isolator but even thus eedback issues are ound measuring the optical spectrum. All the data needed or our analysis is a series o measurements o the laser under test in continuous wave CW emission mode under dierent biasing conditions. The is swept rom the lowest the analyzer can measure, that is typically around 5- % below threshold or tested devices, to some arbitrary current in the operating range given by the laser manuacturer. The region around threshold must be measured in very small steps, as changes are very ast in this region or most o the properties o the laser. Measured spectra or the lasers around the threshold region can be seen in Fig. a-b. Measured Optical Power [dbm] Frequency elative to Maximum [Hz] Measured Optical Power [dbm] Frequency elative to Maximum [Hz] a b Fig. : Measured spectra o lasers or dierent s around threshold. Curve labels reer to the normalized I bias /I th. a VCS with anti-relection coating b VCS without anti-relection coating The lower curves correspond to spontaneous emission below threshold and they have a typical aussian lineshape, with a width in the order o the Hz. As is increased, stimulated emission becomes dominant and the width o the spectrum decreases. Also the lineshape suers some changes, resulting in a very well itted orentzian unction around threshold. At this point, only the phase noise o the laser is considered. I we look at the linewidth o the laser, we can see how it decreases rapidly when approximating to threshold, and then this decreasing continues, but at a soter rate. I we continue rising the, an additional phenomenon can be observed as a certain deormation in the orentzian shape o the spectrum. At higher s, this eect becomes quite more recognizable, as its behavior ollows clearly the theory o laser intensity noise 4, as can be seen in Fig. where a higher range o values is represented. The laser intensity noise appears as a sot increase o the power spectral density at requencies ar rom the peak o power where the spectral density due to phase noise has allen very low. The sum o phase noise and intensity noise orms a certain PSD plateau that continues until a maximum is reached and then decreases sharply.

5 Measured Optical Power[dBm] Frequency relative to Maximum [Hz] Measured Optical Power[dBm] Frequency relative to Maximum [Hz] a b Fig. : Measured spectra o lasers or dierent s above threshold. a VCS with anti-relection coating b VCS without anti-relection coating The requency separation between the peak o power and the maximum o the intensity noise varies with the value o the, ollowing the well-known linear relationship between the squared requency and the. The other consequence o the increasing is relected in the decreasing o the intensity noise PSD that is coherent with the typical behavior o the relative intensity noise IN. To measure ull emission power we must take into account VCS sensitivity to undesired optical eedback. The external optical eedback eects on the spectra o a semiconductor laser have been characterized in terms o the eedback strength 3. Optical eedback causes many eects such as line broadening, mode hopping and coherence collapse 4. In our case, VCS mode instability is observed 5, resulting in mode hopping that does not allow a good measurement o the laser proile above dbm. So, we can measure ull emission power around threshold as the integral o the calculated PSD but due to this instability we can not measure emission power or current bias above threshold this way and we must use the integrated photodiode o the BOSA as SMS is large enough, no signiicant error is added measuring the ull emission power above threshold. 5. SUTS Until now, spectra o the VCS lasers have been measured and analyzed qualitatively. In this section, the measurements will be evaluated by itting them to theoretical expressions. We will begin with the power response. 5.. Power response As explained in previous section, we can calculate the power contained in the lasing mode by integrating the PSD in the wavelength region where it is located or current bias in threshold region but it is not possible or current bias above threshold because optical eedback causes a distorted spectrum. The power measurement in threshold region was carried out with high-resolution optical spectrum analyzer. The irst order approximation 6 o the power response is quite inaccurate in the threshold region, which is critical or the measurement o. The itting is made to a more complex relation between the and the emitted power precise expression : FP I Ith I s FP I si = where I s is a spontaneous emission term, I th is the threshold current and F the inverse o the power to ratio in linear regime, given in A/W.

6 Power responses are shown in next igure. Fig.3 shows the power response transition in the threshold region o both lasers. Optical Power [dbm] aser VCS_AC aser VCS -, -, -,,5,,5,,5 I bias /I th - Fig. 3: Measured power o the lasing mode by integration o the PSD or lasers. As can be seen in Fig. 3, measurements it very well to the theoretical behavior given by. Measuring this with a photodetector normally leads to an overestimation o the mode power below threshold, as all modes contribute to the power and does not give a good itting. Above threshold, measuring with a photodetector does not add substantial error. 5.. aser linewidth and linewidth enhancement actor inewidth can be measured or low powers, around threshold region, due to in this zone optical eedback does not distort the laser spectrum. The linewidth o the VCS laser can be calculated using an easier expression to calculate laser linewidth 7 : = M where M is the measured linewidth, is the linewidth o the laser and is the width o the aussian analysis ilter. From we can deduce that the overestimation in the measured linewidth value is only important when the laser linewidth is comparable to the bandwidth o the analysis ilter. inewidth values corrected with are shown in next igure. The dependence o the linewidth with emitted power is quite complex 8. We can see how linewidth is much greater below threshold, as the contribution rom spontaneous emission is dominant. The threshold region can be identiied as a local maximum o the linewidth, and the linewidth continues decreasing.

7 inewidth [MHz] aser VCS_AC aser VCS -,6 -,,,6, I bias /I th - inewidth [MHz] aser VCS_AC aser VCS Inverse Optical Power [mw - ] a b Fig. 4: Measured linewidth in the threshold region, corrected with 3 and represented versus a normalized, and b inverse optical power. Asymptotic behaviors or b are represented with dashed lines. In Fig. 4b, asymptotic behaviors o the linewidth versus the inverse optical power are shown, ollowing the Schawlow- Townes and Henry laws. esulting alpha values applying 8 are 4.5 or VCS and 3.4 or VCS_AC. inewidth enhancement actor can also be measured using a network analyzer and using the transer unction ater propagation through some hundreds o kilometers o iber. This widespread method was proposed by Devaux et al. 3 and it requires typically around km o iber and a network analyzer with Hz bandwidth. Measurements with this technique give good agreement elaxation oscillations The optical spectrum o the lasers should exhibit a maximum coinciding with the relaxation oscillation peak requency. This maximum is observed in Fig., and can be easily measured in the optical spectrum. The common measurement method involves measuring the laser transer unction with a network analyzer 9. Squared requency [Hz^] aser VCS_AC aser VCS,,4,6,8,,4,6 I bias /I th - Fig. 5: Measured relaxation oscillations peak requency or the lasers under test. In Fig. 5, the values o the relaxation oscillation requency measured rom the optical spectra vs. the current are shown, presenting the usual square relation with above threshold.

8 5.4. elative intensity noise The measurement o IN rom the high-resolution optical spectrum is also very direct or requencies ar enough rom the peak o emission, normally above Hz, as or lower requencies the itting error to the Voigt unction is comparable or even higher than the intensity noise itsel. Applying 5 we obtain the desired IN curves as unction o requency, equivalent to those that we could measure in the electrical domain ater detection, but with the advantage that shot noise is not present. These curves, along with a it to the second term on are shown in Fig. 6. IN [db/hz] IN [db/hz] Frequency relative to maximum [Hz] Frequency relative to maximum [Hz] a b Fig. 6: elative intensity noise as a unction o requency rom the peak o emission or several. a VCS with antirelection coating. b VCS without anti-relection coating. IN measurements or the VCS without anti-relection coating are only valid over a small range o currents, due to the instability in the measured optical spectrum. Nevertheless, measurements made in this range are good enough or extrapolating IN values at higher currents. The peak IN and the damping rate value as a unction o are given in Fig. 7. IN [db/hz] aser VCS_AC aser VCS -35, I bias /I th - Damping rate [MHz] aser VCS_AC aser VCS,5,5 I bias /I th - a b Fig. 7: Peak IN a and damping rate b as a unction o or both lasers

9 5.4. Side-mode suppression ratio The single-mode nature o VCS lasers is quantiied through the side-mode suppression ratio SMS, deined as the ratio o the main-mode power to the power o the most dominant side mode. The SMS should exceed 3 db or a good single longitude mode laser. Thanks to the high dynamic range o the analyzer, we can measure the SMS o the VCS with anti-relection coating. We can see it in Fig. 8. The SMS is approximately 55 db. Optical power [dbm] Frequency relative to maximum [Hz] Fig. 8: Measured spectra o VCS with anti-relection coating. 6. CONCUSSION We have presented results rom the characterization o the main work parameters o two VCS lasers rom measurements o its optical spectrum at continuous emission mode, measured with a high-resolution optical spectrum analyzer BOSA-C rom Aragon Photonics. The main advantage o the proposed methods is their inherent simplicity: the lasers have to be manipulated minimally, as only a must be applied, and only equipment is used or all the measurements. Furthermore, results have proved to be very detailed and are consistent with other more traditional methods o measurement. The measurement o the power response, the linewidth and linewidth enhancement actor, the relaxation oscillations and thus the maximum modulation requency, the relative intensity noise, the damping rate and the ree spectral range have been accomplished. This combined with some more direct measurements that can be perormed also in CW operation, such as the emission wavelength and its shit with, give a very precise characterization o the laser properties or its use in communications systems or to perorm accurate simulations. AKNOWDMNT This work was supported by the Diputación eneral de Aragón DA, by means o project PIP8/5 and the unds provided or the I3A esearch aboratories in Walqa Technology Park and by Spanish Ministerio de ducación y Ciencia MC under project TC5-36 FNCS. K. Iga, Vertical-cavity surace-emitting laser: Introduction and review, Vertical-Cavity Surace-mitting aser Devices, Berlin, 3... Bjerkan, A. øyset,. Haskjaer and D. Myhre Measurement o aser Parameters or Simulation o High-Speed Fiberoptic Systems Journal o ightwave Technology, 4, , 996.

10 3. F. Devaux, Y. Sorel, and J.F. Kerdiles Simple measurement o iber dispersion and o chirp parameter o intensity modulated light emitter Journal o ightwave Technology,, , I. Joindoint Measurement o relative intensity noise IN in semiconductor lasers Journal o Physics III,, 59-63, A. Villaranca, J. asobras, J.A. ázaro and I. arcés Characterization o the main semiconductor laser static and dynamic working parameters rom CW optical spectrum measurements I Journal o Quantum lectronics. To be published. 6. M. P. van xter, W. A. Hamel, J. P. Woerdman, and B.. P. Zeijlmans Spectral Signature o elaxation Oscillations in Semiconductor asers I Journal o Quantum lectronics, vol. 8, no. 6, pp , A. Villaranca, J.A. ázaro, I. Salinas, and I. arcés Measurement o the linewidth enhancement actor o DFB lasers using a high-resolution optical spectrum analyzer I Photonics Technology etters, 7, pp C. H. Henry Theory o the inewidth o Semiconductor asers J. o Quantum lectronics, vol. Q-8, No., pp 59-64, A.. Schawlow and C. H. Townes Inrared and optical masers Phyical eview, vol., No. 6, pp , P. Agrawal Intensity Dependence o the inewidth nhancement Factor and Its Implications or Semiconductor asers I Photonics Technol. ett., vol., no. 8, pp. -4, 989. A. Villaranca, J. A. ázaro, I. Salinas, and I. arcés Stimulated Brillouin scattering gain proile characterization by interaction between two narrow-linewidth optical sources Optics xpress, vol. 3, pp , 5. K. Peterman aser diode modulation and noise. Kluwer Academic, W. Tkach egimes o Feedback ects in.5-µm Distributed Feedback asers Journal o ightwave Technology, vol. T-4, no., pp , Y.C. Chung, Y.H. ee Spectral Characteristics o Vertical-Cavity Surace-mitting asers with xternal Optical Feedback I Photonics Technol. ett., vol. 3, no. 7, pp , J.Y. aw,.p. Agrawal ects o Optical Feedback on Static and Dynamic Characteristics o Vertical-Cavity Surace-mitting asers I Journal o Selected Topics In Quantum lectronics, vol. 3, no.,pp , P. Agrawal Fiber-Optic Communication Systems John Wiley & Sons, New York, J.J. Olivero and.. ongbothum mpirical Fits to the Voigt inewidth: a brie eview J. o Quant. Spectrosc. adiat.transer, 7 p. 33, Z. Toano Investigation o Threshold Transition in Semiconductor asers J. o Selec. Topics Quant. lectron., vol. 3, no., , S. indgren, H. Ahleldt,. Backlin, et al., 4-Hz Modulation Bandwidth and Passive Alignment o Flip-Chip Mounted DFB aser Diodes I Photonics Technology etters, 9, 36-38, 997