Single-mode (SM) vertical-cavity surface-emitting lasers (VCSELs), especially those operating around 850 nm, have been studied intensively in recent years for short distance transmission. Despite the demonstrations of increased data rate and system distance, the impact of frequency chirp that is commonly present in directly modulated lasers is an area that needs more detailed studies for 850 nm VCSEL-based systems. In this paper, we explore the interaction between a laser chirp and fiber chromatic dispersion using an 850-nm SM VCSEL over a standard SM fiber that is two-mode at the operating wavelength. Our transmission experiments show that the system can enjoy a benefit from negative fiber dispersion instead of a penalty compared to the back-to-back case, due to the favorable chirp–dispersion interaction, which is also supported by our system bandwidth measurements. Furthermore, we measure the chirp value of the SM VCSEL and conduct modeling using the time-domain pulse concept to illustrate the impact on the chirp–dispersion interaction and explore the optimal chirp parameters for different transmission data rates. Our study indicates a significant system benefit of using 850-nm SM VCSELs with a high bandwidth single-mode fiber around 850 nm due to the favorable chirp–dispersion interaction. Such a benefit can enable high data rate and longer distance system transmission for modern data center applications.
I. INTRODUCTION
For short distance communications, 850 nm multimode (MM) vertical-cavity surface-emitting laser (VCSEL)-based transmission systems over multimode fibers (MMFs) have been a low-cost solution for many years. As the traffic increases over time, the transmission data rate is getting higher, which poses new challenges for short reach systems using VCSELs. 850 nm single-mode (SM) VCSELs have been actively studied alongside with multimode (MM) VCSELs to improve system performance. At high data rates of 25 Gb/s and above, the fiber chromatic dispersion (CD) becomes a limiting factor due to the large laser linewidths (∼0.6 nm) of MM VCSELs. SM VCSELs with much narrower linewidths of around 0.1 nm have the advantage over MM VCSELs with reduced CD effects and, therefore, improved system level bandwidth. Although SM VCSELs at longer wavelengths such as 980 and 1060 nm have been proposed to take advantage of lower fiber CD at these wavelengths, 850 nm SM VCSELs can utilize the matured 850 nm MM VCSEL technology by reducing the aperture size, and hence, the cost of such VCSELs can be potentially lower than that of the longer wavelength VCSELs. Significant progress has been made in developing 850 nm SM VCSELs for transmission over MMFs modulated at 25, 100 Gb/s, and beyond.1–6
In addition to the improved capability of transmission over MMFs, SM VCSELs can also launch into standard single-mode fibers due to their small spot size and low numerical aperture (NA).7 As reported in Ref. 7, a graded-index standard single-mode fiber, which supports two linearly polarized (LP) modes at 850 nm, was shown to have a high modal bandwidth and 25 Gb/s transmission over 1.5 km of such fiber has been demonstrated.
On the other hand, it is known that directly modulated lasers (DMLs) can possess frequency chirps. It was suggested that the impairment of negative CD can be compensated by the laser chirp, resulting in enhanced system performance for metropolitan area networks.8 Indeed, depending on the sign of the laser chirp parameter and fiber CD, the laser chirp and CD interaction can either enhance or worsen the transmission performance.9 Around 850 nm, the wavelength for most of VCSEL transmission, the optical fibers (both MMF and standard single-mode fiber) have high negative dispersion that limits system performance beyond 25 Gb/s. A question that can be raised is if an SM VCSEL has a meaningful chirp that interacts with the negative CD and impacts the transmission performance. To our knowledge, there have not been any confirmed reports in the literature of SM VCSELs at 850 nm showing a significant laser chirp and having favorable interaction in transmission despite relevant studies.10,11 In Ref. 10, a new technique to characterize the chirp parameters of SM VCSELs operating at 850 nm was proposed by taking advantage of the two-mode propagation through a G.652 fiber. However, no interaction between the VCSEL chirp and fiber chromatic dispersion was reported due to the large modal dispersion in the fiber. In Ref. 11, an 850-nm SM VCSEL was studied, and it was concluded that the laser has a low or negligible chirp. On the other hand, effort was made to tune the chirp of the 1550-nm VCSEL to work favorably with the fiber having positive dispersion.12 We note that if an SM VCSEL around 850 nm wavelength window could be designed with a right laser chirp to compensate fiber CD, it could significantly improve high-speed transmission systems by reducing the penalty due to the high dispersion at 850 nm. To date, the report of such a study is lacking.
In this paper, we report the observation of a chirp of an 850-nm SM VCSEL and present a detailed study on the interaction between the laser chirp and fiber CD using the SM VCSEL over a graded-index standard single-mode fiber. In particular, we show that a negative VCSEL chirp can compensate fiber CD that results in a power benefit for short fiber lengths of a few hundreds of meters at 850 nm. To the best of our knowledge, this is the first report of favorable interaction between the chirp of SM VCSEL and the negative dispersion of fiber at 850 nm, which could provide a new design parameter for optimizing VCSEL-based short reach transmission systems for data center applications. The transmission experiments discussed in Sec. II show that the system with fiber under test (FUT) in place can enjoy a benefit instead of a penalty compared to the back-to-back (B2B) case, which prompts us to pursue a further detailed study. In Sec. III, we show the system level transfer function with various lengths of the FUT, which supports the finding in Sec. II. To have a better understanding of the chirp–dispersion interaction, we measure the laser chirp value of the SM VCSEL in Sec. IV and present a simple picture of chirp–dispersion interaction using the time-domain pulse concept in Sec. V. Finally, we present conclusions in Sec. VI.
II. SYSTEM TRANSMISSION EXPERIMENTS
In a recent experiment using SM VCSELs for a mode division multiplexed transmission system,13 we observed that the system with a 1000-m fiber can enjoy a transmission benefit compared to the B2B case. Such a finding prompted us to conduct more detailed studies to understand the phenomena, which we believe is due to the interaction of the laser chirp and fiber CD.
In the current study, we use a graded-index standard single-mode fiber that has two LP modes around 850 nm with a high modal bandwidth as the transmission medium in order to observe effects between the VCSEL chirp and fiber CD. Figure 1 shows the experimental setup for the system transmission measurements at 25 Gb/s using an SM VCSEL. As reported in Ref. 7, the graded-index standard single-mode fiber has a modal bandwidth of 48.3 GHz km (referred to as Fiber1) at 850 nm, which leads to negligible modal dispersion penalty. The fiber attenuation is around 2 dB/km at 850 nm, and the cable cutoff wavelength of the fiber is below 1260 nm. The SM VCSEL we used in the experiments is from V-I-Systems, whose optical spectrum is shown in Fig. 2. The VCSEL has a center wavelength at 847.95 nm, similar to those in Ref. 1. The higher order mode suppression is about 45 dB, so the VCSEL is essentially single mode in the 850 nm window. The FWHM linewidth is 0.043 nm. The SM VCSEL is packaged with a V-connector and mounted on a plate followed by two sequential lenses in a cage system to enable the light coupling from the VCSEL to the fiber. An Agilent BERT system is used to measure the bit error rate (BER). A pattern generator (N4951B) is used to generate a non-return-to-zero (NRZ) signal with 0.54 Vpp at 25 Gb/s, and a bias-T (SHF 122C) is used to combine the RF modulation signal with the DC voltage to drive the VCSEL. With 2.32 V DC driving voltage (∼2.5 mA current), the optical power coupled into the single-mode fiber is around −2 dBm. After transmitting through the FUT, the optical signal is detected by a Discovery Semiconductor’s 850 nm Lab Buddy optical receiver (R409) with a 15 GHz bandwidth and an error detector (N4952A-E32).
The transmission performance was measured over different lengths of Fiber1: B2B (1 m), 300, 500, 1000, and 1500 m. The results of the bit error rate (BER) as a function of received optical power are shown in Fig. 3(a). The received optical power was controlled by detuning the distance between the receiving fiber and the two-lens coupling system. Since the fiber has a very high modal bandwidth as noted above, the possible slight change in the launch condition during the detuning would have very little effect on the system performance. Under the B2B condition, the system reaches error-free performance with −7.3 dBm power, while the error-free power threshold values with 300, 500, 1000, and 1500 m Fiber1 are −7.9, −8.2, −9.3, and −9.4 dBm, respectively. The system performance shows power benefits of 0.6, 0.9, 2.0, and 2.1 dB under a BER level of 10−10 for 300, 500, 1000, and 1500 m of Fiber1, respectively. We also measured the corresponding eye diagrams under the five configurations as shown in Fig. 3(b). Each eye diagram was measured with the maximum received optical power under each configuration. As shown, all the eyes are clearly open. The slightly noisier feature as the fiber length increases is due to the lower received optical power.
In the above study, we have chosen a fiber with a high modal bandwidth to highlight the chirp–CD interaction effect around 850 nm. In a practical system, the fiber modal bandwidth is another limiting factor for the system performance. As an example, a second graded-index standard single-mode fiber (Fiber2) with a worst-case modal bandwidth of 4.0 GHz km at 850 nm was used to illustrate the modal bandwidth effect. The system BER as a function of received optical power measured over 300-m Fiber2 is shown in Fig. 4, where the BER curves of the B2B and 300-m Fiber1 are also included in for comparison. With 300-m Fiber1, the system enjoys a benefit of around 0.6 dB compared to the B2B case, while the performance of 300-m Fiber2 is comparable to that of B2B. Nevertheless, for Fiber2, the system still enjoys a power benefit instead of power penalty that is expected from the lower bandwidth.
III. SYSTEM BANDWIDTH MEASUREMENTS
We expected that the observed performance benefit in the system transmission with a fiber in the link compared to the B2B case came from higher system bandwidth due to the cancellation of VCSEL chirp effects by fiber CD. To verify this, we measured the system bandwidth using the setup shown in Fig. 5(a). A vector network analyzer (VNA) Agilent N5230C is used to capture the frequency response of the system consisting of the SM VCSEL, the FUT, and the optical receiver. The RF signal output by the VNA serves as the modulation signal for the VCSEL, and the modulated light out of the VCSEL is coupled into the FUT using the same lens system in Fig. 1. The optical receiver converts the optical signal back into the electrical signal to be detected by the VNA. The system bandwidth measurement results are depicted in Fig. 5(b). As shown, the frequency response curve moves toward higher frequency as the length of the FUT increases, indicating higher system bandwidths. Specifically, the three optical dB (or six electrical dB) bandwidths for the B2B, 300, 500, 1000, and 1500 m of Fiber1 are 17.19, 17.76, 18.16, 18.89, and 18.76 GHz. These results are consistent with the system transmission experiments, clearly showing that the system has better performance with longer lengths of Fiber1 in the link. To show more clearly the system bandwidth improvement, Fig. 5(c) plots the system bandwidth as a function of fiber length. It can be found the system bandwidth increases linearly with fiber length up to about 1000 m, reaches a maximum at 1200 m, and starts to decrease for longer length. We believe that the system bandwidth increase is due to the compensation of fiber CD by the laser chirp. At about 1200 m length of fiber, the fiber CD is completely compensated by the laser chirp, corresponding to the maximum system bandwidth. After the optimal fiber length, the residual fiber CD starts to degrade the system bandwidth. From Fig. 5(c), it is estimated that for a fiber length up to 2000 m, a typical maximal link distance for hyperscale data centers, the SM VCSEL can offer benefit of compensating fiber CD to improve the system performance, which is attractive for VCSEL system applications in the hyperscale data centers.
IV. VCSEL CHIRP MEASUREMENT
The system performance improvement with the high bandwidth Fiber1 is presumably due to the interaction of the laser chirp and fiber dispersion. The laser chirp parameter that dominates the interaction with the fiber CD is the transient chirp or the linewidth enhancement factor α.14,15 To quantify the linewidth enhancement factor, we measured the chirp response of the SM VCSEL using the technique described in Ref. 16. We follow the chirp convention defined in Ref. 16, which is consistent with that in Ref. 9. The experimental setup is the same as that in Fig. 5(a), while in the chirp measurement experiments, a 15 km Hi780 fiber, single mode at 850 nm, is used to provide enough fiber CD to see the resonance dips in the system frequency response, as shown in Fig. 6. The fiber has attenuation around 2 dB/km at 850 nm and a fiber cutoff wavelength around 720 nm. Following Ref. 16, the resonance dips in the frequency response occur at frequencies fk determined by the following equation:
where c is the speed of light in the vacuum, D is the fiber dispersion, L is the fiber length, λ0 is the operation wavelength, and k = 0, 1, 2, … is an integer corresponding to the first, second, third dips, and so on. Therefore, the laser chirp can be calculated without the need to know the exact fiber dispersion as
In Fig. 6, the values for the first and second frequency dips are f0 = 15.825 GHz and f1 = 22.914 GHz. Using Eq. (2), the chirp parameter of the VCSEL is αlw = −3.52. The CD of the fiber is calculated to be −100.27 ps/(nm km) using Eq. (1), which agrees well with the CD value of −100.298 ps/(nm km) as calculated from the measured fiber refractive index profile. This confirms that the SM VCSEL has a significant negative chirp consistent with the study in Secs. II and III. We also measured the chirp parameter of the SM VCSEL used in Ref. 7, which is an earlier version of SM VCSEL from V-I-Systems. The SM VCSEL in Ref. 7 has a center operating wavelength of 842 nm and the measured chirp parameter is similar to the SM VCSEL in the current study. However, we did not notice the benefit of chirp–CD interaction in the 25 G/s system testing in Ref. 7 as Fiber1 has a lower modal bandwidth of 12.5 GHz km at 842 nm, which caused some transmission penalty that was convoluted with the benefit of chirp–CD interaction. As a result, we missed the earlier chance of identifying the transmission benefit from the favorable chirp–CD interaction. However, the benefit of chirp–CD interaction is shown in Fig. 4 of Ref. 7 as the system performance of 500 m fiber is similar to the B2B condition.
V. IMPLICATION OF CHIRP–CD INTERACTION FOR HIGH DATA RATE SYSTEMS
To better illustrate the impact of the interaction between the SM VCSEL chirp and fiber CD at 850 nm, we modeled the propagation of a chirped Gaussian pulse in a fiber following Chapter 3 in Ref. 9 using relevant parameters that were measured from and assumed for the SM VCSEL and FUT. Note that the laser chirp–CD interaction has often been studied for single-mode fiber at the O-band or the C-band. For standard single-mode fiber, the CD value is nearly zero around 1310 nm, a common wavelength for single-mode transmission in the O-band. At 1550 nm, another commonly used wavelength in the C-band, the CD is positive around 16 ps/(nm km). As the laser chirp–CD interaction requires a significant amount of CD, it often requires tens of kilometers of fiber in the O- or C-band to show observable effects. As the CD at 850 nm is much higher than that in the O- or C-band, a shorter fiber can have enough dispersion to interact with the laser chirp. The study here can shed light on how the chirp–CD interaction can affect the short distance transmission in the order of 100–2000 m at 850 nm. The assumed input chirped Gaussian field is
The output field is then described as
where T0 is the pulse width of the input Gaussian pulse, L is the fiber length, and the dispersion parameter , where D is the fiber dispersion. At 850 nm, all silica-based fibers have high negative dispersion, which means that the dispersion parameter β2 is positive. According to the pulse propagation theory, the negative chirp can interact with the positive dispersion parameter to reduce pulse broadening. For Fiber1 used in our experiments, the CD is −88.76 ps/(nm km). The amount of negative CD at 850 nm is significantly higher than those at 1310 or 1550 nm, which suggests that the chirp–CD interaction can have a significant observable effect at very short fiber lengths of a few hundred meters.
Using the measured chirp parameter αlw = −3.52 for the SM VCSEL from the experiment, we calculated pulse width changes assuming the input pulse width for three cases, A: T0 = 20 ps; B: T0 = 10 ps, and C: T0 = 5 ps. These pulse widths correspond to the half of the bit intervals for 25, 50, and 100 Gb/s data rates. Figure 7(a) plots the calculated pulse width changes for the three cases due to interaction between the laser chirp and the fiber CD for fiber length up to 2000 m, a typical fiber length range for data centers. For Case A, the pulse becomes narrower for all the fiber lengths, which is consistent with the system testing results in Sec. II. For Case B, the pulse width decreases initially, reaches a minimum at 750 m, and then starts to increase again, but the pulse width is below 10 ps for fiber length up to 1500 m. For Case C, the pulse width change has a similar behavior to Case B, except that the length corresponding to the minimum is reduced to about 200 m and the pulse width stays below 5 ps for a fiber length up to about 380 m. While the chirp parameter αlw = −3.52 works very well for Case A, it is not optimal for Cases B and C with shorter pulse widths. To improve the chirp–CD interaction, we changed the chirp parameter for each case to find the optimal chirp parameter. For the three cases, the optimal chirp parameters are −5.7, −1.5, and −0.4, respectively, corresponding to the laser pulse widths of 20, 10, and 5 ps. Using the optical chirp parameters, Fig. 7(b) plots the pulse width changes as a function of the fiber length for the three cases. For Cases A and B, the pulse width can stay below the input pulse widths for the whole range of fiber length below 2000 m. For Case C, the pulse width stays below 5 ps for a fiber length up to about 500 m, which can cover most link distances for data centers.
The chirp parameter of VCSELs depends on materials and design parameters of the active region. The VCSEL used in our study was a commercially available device and we do not have detailed information on its design. The optimization of the chirp parameter of VCSELs for directly modulated high-speed short-reach systems is subject to further study. Nevertheless, our study shows that designing VCSELs with a right amount of negative chirp can result in favorable chirp–CD interaction, which could potentially enable future higher data rate transmission developments beyond 100 Gb/s.
VI. CONCLUSIONS
In this paper, we report the observation of a negative laser chirp on an 850 nm SM VCSEL and study its interaction with fiber CD over a graded-index standard single-mode fiber, which is two-mode around 850 nm. The transmission experiments show that the system can enjoy a benefit compared to the back-to-back case, even for a fiber as short as 300 m, due to the favorable chirp–dispersion interaction. The improved system performance is supported by system bandwidth measurements, which show that the system bandwidth increases due to the compensation of fiber CD by the VCSEL chirp. Furthermore, we characterize the chirp value of the SM VCSEL and provide a simple illustration using the time-domain pulse concept for the laser chirp–dispersion interaction through which optimal chirp parameters for very high data rate transmission are identified. To the best of our knowledge, this is the first report of favorable interaction between the chirp of SM VCSEL and the negative dispersion of fiber at 850 nm, which illustrates its benefit in very short distance transmission systems. In the current study, we have chosen to use the graded-index single-mode fiber due to its high modal bandwidth so that the effect of laser chirp–CD can stand out. In principle, the mechanism of chirp–CD interaction could also apply to SM VCSEL over high bandwidth MMF systems. The understanding of favorable SM VCSEL chirp–CD interaction and its implication to VCSEL designs could enable higher data rate and longer distance transmission for 850 nm SM VCSEL-based systems in data center applications.
AUTHOR DECLARATIONS
Conflict of Interest
The author declares that they have no conflict of interest.
DATA AVAILABILITY
The data that support the findings of this study are available within the article.