Open Access
Add to BibSonomyAbstract
1.55-µm vertical cavity surface-emitting low-parasitic lasers show open eyes up to 22-Gb/s modulation speed. Uncooled error-free operation over a wide temperature range up to 85°C under constant bias conditions is demonstrated at 12.5-Gb/s data rate. At these fixed bias conditions the laser characteristics are practically invariant with temperature. These are the highest data-rates reported from a long-wavelength VCSEL structure to date.
©2009 Optical Society of America
1. Introduction
Vertical-cavity surface-emitting lasers (VCSELs) are cost-effective and energy-efficient light sources for optical data transmission. High-speed VCSELs with data-rates in excess of 20 Gb/s have been demonstrated recently in the 850-nm wavelength regime [1,2]. As the waveband around 850 nm is only suitable for short-reach interconnects, many efforts have been made to develop long-wavelength, high-speed VCSELs for metro-range links [3–7]. For the long-wavelength regime, approaches incorporating wafer-fusion [3], GaInNAs-based quantum-wells [4] or monolithic approaches on InP [5–7] have shown promising results with data-rates of 10-Gb/s. The incorporation of a buried tunnel-junction (BTJ), enabling the replacement of p- by n-conducting material with lower electrical and optical losses, seems to be mandatory to achieve reasonable high-speed or high-temperature performance. However, each approach has its trade-offs, making high data-rates in a wide operating temperature range at long wavelengths challenging.
In this paper, we present the first long-wavelength VCSEL with open eyes at 22 Gb/s and the first error-free data-transmission at 12.5 Gb/s up to 85°C. This was achieved by reducing device parasitics, a new highly strained active region and a well-tailored mode-gain offset in the VCSEL cavity.
2. Structure
The VCSEL under investigation was grown by molecular beam epitaxy on an InP substrate, and optimized for high-speed operation at high temperatures [5]. This was achieved by a low doped region in the overgrowth as shown in Fig. 1 , reducing the space-charge capacitance of the blocking diode around the BTJ [8,9]. Furthermore, this VCSEL incorporated very highly strained quantum wells with a large mode-gain detuning for more differential gain and higher relaxation oscillation frequencies with lower damping. The detuning results in lower threshold currents at high operation temperatures, enabling a temperature-invariant behavior [9]. This optimized structure was presented recently, expecting high potential data-rates from small-signal measurements [9].
Fig. 1 Schematic diagram of the high-speed vertical-cavity surface-emitting laser device. A reduction of parasitic response was achieved by a low-doped layer in the overgrowth. An equivalent circuit of laser parasitics is inset.
Reducing parasitics by lowering doping is not very straight-forward, as resistance rises linearly (assuming constant mobilities) and capacitance scales down with a square-root law. However, the RC-time constant τp is not a simple multiplication of resistance and capacitance. The simplest equivalent circuit describing this VCSEL structure well [8] is overlaid on Fig. 1 with the series resistance Rm of the mirrors and the overgrowth, and the elements Ra and Ca around the active region. This simple circuit can model the laser parasitics due to negligible contact-pad capacitances and low-resistive ohmic behavior of contacts and hetero-junctions. Above laser threshold, the active region is modeled in this case as a short-circuit Zacitve = 0 for small signals due to Fermi-level pinning.
Equation (1) gives the parasitic roll-off frequency fpar of this circuit with Z50Ω as impedance of the driver. Even with a slightly higher series resistance Rm, the reduction in Ca causes a much better parasitic roll-off [7,9], as Ra rather than Rm contributes directly to the parasitic time constant.
3. Results
The light output versus current versus voltage (L-I-V) characteristics and the single-mode spectrum are presented in Fig. 2 . Besides the continuous-wave (CW) characteristics we also present pulsed L-I curves with a 500-ns pulse and 50% duty cycle simulating a device under large-signal modulation. This is necessary to understand why quite high bias conditions, e.g. 8 mA at 85°C, which is above the thermal roll-over, yield best results. Since thermal resistance is an inherited challenge in long-wavelength VCSELs, the devices perform even better under large-signal modulation, as the effective duty-cycle is only 50%. The VCSEL emits single mode with very high side-mode suppression ratios greater than 30 dB even for the polarization mode. The CW output power is above 2 mW at room temperature, and around 1 mW at 85°C. The lasers work CW up to 120°C heat-sink temperature and still show reasonable powers up to 115°C. Fiber-coupled average powers under modulation of 0 dBm are achievable. Due to a designed slight increase in Rm, threshold voltage and differential resistance rise from 0.9 to 1.0V and from 35 to 50 Ω, respectively compared to the reference design with higher doping levels in the overgrowth.
Fig. 2 L-I-V characteristics of 1.55 µm VCSEL for 25°C, 85°C and 115°C. Dashed: CW under large-signal modulation (500 ns pulse, 50% duty cycle). Solid lines: CW. Laser operates CW up to 120°C. Single-mode spectrum inset. The polarization-mode is suppressed by more than 30 dB, and higher order transverse modes are suppressed more than 50 dB.
3.1 Small-signal modulation performance
In order to evaluate modulation capabilities of these redesigned lasers, small-signal experiments were carried out. The VCSELs could be directly probed without potential additional parasitics from bonding wires and sub-mounts due to their microstrip-layout and coplanar connectivity [5].
The small-signal modulation performance (S21) is presented in Fig. 3 . Superior modulation bandwidths in excess of 12 GHz can be stated. As the equivalent circuit presented in Fig. 1 models the parasitics well and can be mapped to a first-order low-pass filter function according to Eq. (1), the whole transfer-function can be fit to the following equation
a three-pole filter function including relaxation-oscillation frequency fR, intrinsic damping γ and parasitic roll-off fpar as defined in Eq. (1). The constant terms are the differential quantum efficiencies ηd of laser and detector. The solid lines in Fig. 3 represent curve-fits to this equation.Fig. 3 Small-signal modulation performance of a high-speed long-wavelength VCSEL at room-temperature for different bias-currents. The symbols represent the measured data. The solid-lines are fit to Eq. (2) squared for intrinsic parameter extraction.
The extracted results are presented in Fig. 4 . In (a) the modulation bandwidth and resonance-frequency fR are plotted versus the square-root of laser current above threshold. For comparison, the parasitics-limited response from the previous design is also given. Details can be found here [9]. The laser response is limited by intrinsic damping [1,5,6,8]. From rate-equation analysis the following widely accepted relationship can be derived
with the photon-lifetime τP, the confinement factor Γ, gain g, the gain derivates a and -ap versus carrier N and photon-density S, respectively, the carrier losses Jth and the spontaneously generated photon-rate Jsp [8]. This defines the K-factor and the damping offset γ0. In Fig. 4(b) we plot the damping rate versus the resonance-frequency fR squared extracting a K-factor of 0.30 ns. This is a clear improvement versus the previous design only yielding 0.41 ns [8] or 0.46 ns [5]. This is attributed to the enhancement of differential gain a of the highly strained quantum-wells [9]. However, as can be seen in Eq. (3), a reduction in photon-lifetime is expected to have an even stronger effect. Therefore, we very recently realized a radical redesign of this laser structure with a shortened cavity length [10], yielding a K-factor of only 0.15 ns and modulation bandwidths in excess of 15 GHz. These are very exciting results to be verified by large-signal experiments in the near future.Fig. 4 Extracted parameters from small-signal modulation characterization and parameter fitting in comparison to the previous design [6] with higher parasitics and less differential gain from the quantum-wells. In (a), modulation bandwidth and resonance frequency are plotted versus the square-root of laser-current above threshold, in (b) the K-factor is derived.
3.2 Large-signal experiments
High-speed edge-emitters require more than 13-GHz modulation bandwidth for 10-Gb/s data-rates. This is caused by the non-flat modulation response with a large relaxation overshoot. These VCSELs however, have a very flat modulation response with bandwidths exceeding 12 GHz [8] as shown in Fig. 3. Further, as can be seen in Fig. 4(a) not only the power is characterized by a thermal roll-over in long-wavelength VCSELs, but also the resonance-frequency. On the other hand, as shown in Fig. 2, the thermal budget is significantly larger for a 50% duty cycle pulsed operation, like under large-signal modulation. Therefore, much higher data-rates could be expected for these devices compared to distributed feed-back (DFB) edge-emitting lasers with similar bandwidth.
Large signal experiments were carried out with a setup originally designed to verify the performance of ultra-high-speed VCSELs in the 850 nm waveband [2]. This is the reason, why only back-to-back (BTB) data is available. The PRBS was generated by a SHF 12100 Bit Pattern Generator with reference signal produced by a PSG CW Signal Generator, Agilent Technologies, E8247C. The PRBS signal with 0.5 Vpp was attenuated by 6 dB with an Anritsu 41V-3 attenuator. For the bit-error-rate (BER) testing, one sub channel of Pattern Generator was used instead of MUX channel. The PRBS was mixed with CW current by a Wiltron V250 bias-T. The optical signal was transmitted with 62.5-µm fiber with PC or APC connectors. For the BER measurements, the optical power was attenuated by variable JDSU OLA-54 attenuator and measured with a JDSU OLP-55 power meter. The photo detector used was a New Focus 3547, DC-12-GHz IR receiver with option 1554-B-50 installed. For bit error rate analysis the SHF 11100B error analyzer was used and the eye-diagrams presented here were recorded by an Agilent Infinium DCA-J 86100C digital communication analyzer with the following installed modules: Agilent 86107A precision time-base module 10/20/40GHz and 86118A remote sampling module 70 GHz.
This modified setup had some influence on the quality of the driving eye, as can be seen in Fig. 5 and the detector was designed for use up to 12 GHz only. Nevertheless, wide open eyes with significantly low signal-to-noise (S/N) ratios could be demonstrated as shown in Fig. 5. A record-high data-rate of 22 Gb/s with clearly open (S/N of 3.9) eye-diagrams was demonstrated. These experiments were carried out at room temperature with a pseudo-random bit-sequence (PRBS) of 231-1 bit-pattern length. As these long bit-patterns are only required by long-haul applications, and to save experiment-time the BER curves were recorded at a PRBS of 27-1.
Fig. 5 Eye diagrams of a 1.55-µm high-speed VCSEL at 25°C and 231-1 PRBS sequence. The bit-rate is varied from 10 Gb/s up to 25 Gb/s. Clear open eyes up to 22 Gb/s. The electrical driver (upper left) showed significant rise and fall-times. Eye at 25 Gb/s may be limited from the detector-response.
3.3 Temperature-invariant characteristics
Edge-emitting lasers usually require monitor diodes for automatic power control for a constant and a consistent output from these devices, especially if the temperature is varied. These VCSELs, however have a natural power limit. With a well-tailored mode-gain offset, the power can even be kept at constant levels over a wide temperature range [9]. Figure 6 presents modulation bandwidth f3dB and pulsed laser power Popt (simulating large-signal modulation) at this bias condition together with bias current Ith versus temperature. The laser characteristics are not affected much by temperature. This enables driving the laser at constant biasing conditions uncooled over a very large operation range up to 85°C.
Fig. 6 Characteristics of 1.55-µm VCSEL versus temperature. Modulated output-power (pulsed, 50% duty cycle), threshold current and modulation bandwidth are practically constant at constant driving conditions (here: 8 mA). This unique VCSEL feature can redundantize closed-loop power monitoring and temperature control.
3.4 Uncooled data transmission at constant bias
As mentioned before, unlike edge-emitters, well-tailored long-wavelength VCSELs can be driven uncooled in optical communications systems without closed-loop monitoring. This makes monitor-diodes, Peltier cooling and thermistors redundant, dropping packaging cost and last but not least reducing system power consumption and overall expense. Large-signal experiments at constant driving conditions were carried out to underline this unique feature and are presented in Fig. 7(a) . A data-rate of 12.5 Gb/s was chosen, as this was the highest supported data-rate of the setup at time of experiment. No error-floor and negligible BER-penalties can be observed over a wide temperature range. The bias conditions were optimized in this case for 85°C operation temperature yielding 8 mA bias current and 0.25 Vpp voltage swing. Only the pulsed characteristics presented in Fig. 2 make it understandable why this biasing works. As bias conditions were optimized for this point, we can observe a negative BER-penalty at very high operation temperatures. In Fig. 7(b) the corresponding eye-diagram at 60°C is shown yielding a S/N of 6.0 at these fixed conditions. With optimized conditions for that temperature, an S/N of 7.0 would be possible, both sufficient for error-free operation. This demonstrates again the quite tolerant behaviour of these novel devices. Laser characteristics are not affected much by temperature and constant biasing conditions are feasible.
Fig. 7 Large-signal modulation performance from 25°C to 85°C: (a) BER-measurement at 12.5-Gb/s; constant driving conditions (IBias=8 mA). (b) Eye-diagrams at 60°C for optimum and at fixed driving conditions corresponding to BER-data in (a). All experiments were carried out back-to-back (BTB).
4. Conclusion
We presented a long-wavelength high-speed VCSEL operating at data-rates up to 22 Gb/s. Error-free data-transmission is demonstrated uncooled up to 85°C at 12.5 Gb/s without adjusting bias conditions. To the best of our knowledge, these are the highest values reported to date.
Acknowledgments
We gratefully acknowledge the German Academic Exchange Service (DAAD) for supporting W. Hofmann by a fellowship within the Postdoc-Program and the German Research Council (DFG) for funding via SFB 787.
References and links
1. P. Westbergh, J. Gustavsson, Å. Haglund, M. Sköld, A. Joel, and A. Larsson, “High-Speed, Low-Current-Density 850 nm VCSELs,” IEEE J. Sel. Top. Quantum Electron. 15, 694–703 (2009).
2. F. Hopfer, A. Mutig, G. Fiol, M. Kuntz, V. Shchukin, V. Haisler, T. Warming, E. Stock, S. Mikhrin, I. Krestnikov, D. Livshits, A. Kovsh, C. Bornholdt, A. Lenz, H. Eisele, M. Dähne, N. Ledentsov, and D. Bimberg, “20 Gb/s 85°C Error-Free Operation of VCSELs Based on Submonolayer Deposition of Quantum Dots,” IEEE J. Sel. Top. Quantum Electron. 13(5), 1302–1308 (2007).
3. A. Syrbu, A. Mereuta, V. Iakovlev, A. Caliman, P. Royo, and E. Kapon, “10 Gbps VCSELs with High Single Mode Output in 1310nm and 1550 nm Wavelength Bands”, Proc. OFC/NFOEC 2008, (San Diego, USA, 2008), OThS2, pp. 1–3.
4. Y. Onishi, N. Saga, K. Koyama, H. Doi, T. Ishizuka, T. Yamada, K. Fujii, H. Mori, J. Hashimoto, A. Yamaguchi, and T. Katsuyama, “Long-Wavelength GaInNAs Vertical-Cavity Surface-Emitting Laser With Buried Tunnel Junction,” IEEE J. Sel. Top. Quantum Electron. 15, 838–843 (2009).
5. W. Hofmann, N. H. Zhu, M. Ortsiefer, G. Böhm, J. Rosskopf, L. Chao, S. Zhang, M. Maute, and M.-C. Amann, “10-Gb/s Data Transmission Using BCB Passivated InGaAlAs–InP VCSELs,” IEEE Photon. Technol. Lett. 18(2), 424-426 (2006).
6. M.-C. Amann and W. Hofmann, “InP-Based Long-Wavelength VCSELs and VCSEL Arrays,” IEEE J. Sel. Top. Quantum Electron. 15, 861–868 (2009).
7. M. Ortsiefer, W. Hofmann, E. Rönneberg, A. Boletti, A. Gatto, P. Boffi, J. Rosskopf, R. Shau, C. Neumeyr, G. Böhm, M. Martinelli, and M.-C. Amann, “High speed 1.3 µm VCSELs for 12.5 Gbit/s optical interconnects,” Electron. Lett. 44(16), 974–975 (2008).
8. W. Hofmann, InP-based Long-Wavelength VCSELs and VCSEL Arrays for High-Speed Optical Communication, (Sel. Topics Semiconductor Phys. Technol., 99, Munich, 2009), http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:91-diss-20081119-679286-1-5.
9. W. Hofmann, M. Müller, G. Böhm, M. Ortsiefer, and M.-C. Amann, “1.55 µm VCSEL with Enhanced Modulation Bandwidth and Temperature Range,” IEEE Photon. Technol. Lett. 21(13), 923–925 (2009).
10. M. Müller, W. Hofmann, G. Böhm, and M.-C. Amann, “Short-Cavity Long-Wavelength VCSELs with Modulation-Bandwidth in Excess of 15 GHz,” IEEE Photon. Technol. Lett. (accepted for publication).
