Open Access
Add to BibSonomy
Abstract
A 850 nm single-mode vertical cavity surface emitting laser (VCSEL)-based data transmission is demonstrated, enabling four-level pulse amplitude modulation (PAM-4) at over 100–300 m in an OM4 multi-mode fiber (MMF). By optimizing the bias of the single-mode VCSEL with a differential resistance of 159 Ω, the related electrical return loss of is determined to provide an analog modulation bandwidth of 18.9 GHz. After pre-emphasizing the waveform of the PAM-4 format, the PAM-4 data stream can be successfully delivered by the single-mode VCSEL at under back-to-back and 100 m long OM4 MMF conditions. Lengthening the transmission distance worsens the signal-to-noise ratio (SNR) of PAM-4 data to the FEC criterion, as the waveform pre-emphasis of the PAM-4 data stream inevitably induces spectral power compensation from low to high frequencies. Therefore, increasing the OM4 MMF distance from 200 to 300 m significantly reduces the peak-to-peak amplitude of data to suppress the SNR and reduce the transmission capacity from 52 to .
© 2018 Chinese Laser Press
1. INTRODUCTION
With the explosive demand for data streaming in cloud computing, data switching, and real-time multimedia applications, high-speed and cost-effective intra-data centers for handling heavy data traffic in short-reach optical interconnects are mandatory nowadays [1]. Among all approaches for rack-to-rack transmission within data centers, multi-transverse-mode vertical cavity surface emitting laser (MM VCSEL)-based multi-mode fiber (MMF) transmission links have emerged as the best candidate at the current stage [2–4]. These VCSELs exhibit several advantages including low power consumption [5], high conversion efficiency [6], large modulation bandwidth, and fiberized packaging flexibility [7], which can be packaged into a standard “quad small form-factor pluggable” (QSFP) module to serve as a transmitter for intra-data centers. To study the ultimate transmission performance of the VCSEL, Kuchta et al. have employed a multi-mode VCSEL to demonstrate a transmission link with a non-return-to-zero on-off-keying (NRZ-OOK) format [2]. However, the transmission distance in the MMF is limited to 7 m because the data carried by the multi-mode VCSEL suffers from serious modal dispersion.
Going beyond 100-Gbit/s standard QSFP transmitter modules based on VCSEL array with conventional NRZ-OOK modulation, the larger data rate of approached with or VCSEL array transmitters is expected at the present stage, which pursues higher-capacity streaming and longer transmission distance according to the IEEE P802.bs standard [8,9]. On the other hand, the allowable OOK data format is limited by the finite modulation bandwidth of the MM VCSEL. Hence, several kinds of advanced data formats have been successively developed to provide larger data rate within the same bandwidth owing to their higher spectral usage efficiency, such as four-level pulse amplitude modulation (PAM-4) [10–15], discrete multi-tone (DMT) modulation, quadrature amplitude modulation, and orthogonal frequency-division multiplexing (QAM-OFDM) [16–18]. To study performance using these data formats, Tasi et al. employed an MM VCSEL to carry out pre-leveled 16-QAM OFDM data transmission at over 100 m long OM4 MMF [16]. In 2016, Bo et al. demonstrated a single-mode VCSEL-based DMT link at over 300 m long OM4 MMF [19]. Kao further compared VCSELs with different mode numbers for QAM-OFDM transmission at nearly [20]. Although both the QAM-OFDM and DMT data formats benefit from significantly high spectral usage efficiency [21], the need for a high-speed digital-to-analog converter (DAC) and analog-to-digital converter (ADC) with high sampling rate makes the system difficult to realize, and the complex signal processing inevitably increases the construction cost of the whole infrastructure at the current stage [22].
In contrast, the circuitry of the PAM- data link is easier to realize than the QAM-OFDM/DMT cases, and can support data rates twice as large as the OOK format within the same bandwidth. Moreover, the use of PAM-4 does not require high-speed DAC/ADC, which lowers the complexity and cost of encoding/decoding circuits [23,24]. With these advantages, Szczerba et al. demonstrated an 850 nm VCSEL based PAM-4 link at [9]. Sun et al. used short wavelength-division multiplexing technology to demonstrate 51.5625 Gbit/s PAM-4 transmission over 150 m with receiving bit error ratio (BER) of [25]. In addition, Motaghiannezam et al. employed the same technology to perform 180 Gbit/s transmission over 300 m long OM4 fiber with a BER as small as [15].
When considering the MM VCSEL carried data transmission in MMF, the modal dispersion induced among transverse modes inevitably limits the maximal allowable data rate owing to the distortion of transmitted bit shape [26]. Apparently, the use of single-mode VCSELs can essentially favor modal-dispersion-free transmission [27]. This feature has lately increased research interest in promoting the use of single-transverse-mode VCSELs for intra-data-center applications. In view of previous works, several methods have been proposed to suppress the high-order transverse mode of VCSELs, such as shrinking the aperture of emission [28–30], using spatial mode filters [31], and introducing photonic crystal caps [32]. Among the proposed techniques, shrinking the oxide-confined aperture is the most effective and economical way to obtain a single-mode VCSEL [29]. By employing an aperture size as small as 3 μm, Moser et al. successfully demonstrated an OOK transmission at over 500 m OM3 MMF in 2013 [30]. To compensate the signal-to-noise ratio (SNR) declination induced by the dispersion-related radio frequency (RF) power fading after MMF transmission, waveform pre-emphasis (also called pre-distortion) technology in the time domain is proposed [12,33]. With the assistance of pre-emphasis and off-line equalization, Lavrencik et al. used an unpackaged 850 nm VCSEL for PAM-4 data transmission at over 100 m in OM4 MMF in 2016 [14]. Except for this work, the use of single-mode VCSELs for delivering pre-distorted PAM-4 data transmission in OM4 MMF with distance beyond 100 m has seldom been discussed.
In this work, the transmission performance of a small-aperture-size single-mode VCSEL carrying pre-emphasis PAM-4 data over 100–300 m in OM4 MMF is characterized. The fundamental characteristics of the single-mode VCSEL, including power–current (P-I) curve, differential resistance, and modulation throughput, are analyzed for optimization. To perform PAM-4 data transmission in OM4 MMF with lengthening distance, weighting factors such as amplitude and phase parts within a channel of the whole link are evaluated so as to implement the waveform pre-emphasis of the PAM-4 data. For practical intra-data-center application, the transmitted data qualities are discussed and compared in detail after propagation through back-to-back (BtB), 100, 200, and 300 m OM4 MMF segments without and with PAM-4 waveform pre-emphasis. The characteristic parameters observed from RF spectra, eye diagrams, bathtub curves, and power penalties are analyzed for comparison.
2. EXPERIMENTAL SETUP
Figures 1(a) and 1(b) illustrate the block diagrams of the waveform pre-emphasis technology for the PAM-4 data stream and the on-chip testing bench for characterizing the single-mode VCSEL, which is directly encoded by the PAM-4 data for transmission over 100–300 m in the OM4 MMF, respectively. During the experiments, a homemade probe station with a 40 GHz ground-signal-ground (GSG) probe (GGB, 40A-GSG-100-DP) and a lensed OM4 MMF segment was established; this was used to seed the electrical signal into the single-mode VCSEL and collect its modulated output. To stabilize the output dynamics of the single-mode VCSEL chip, its outer-shell temperature was controlled at 22°C after adhering to a stainless-steel plate-based heat sink with cooling water circulation.
Fig. 1. (a) Concept of the pre-emphasis technology. (b) Experimental setup of the proposed VCSEL based on 16-QAM OFDM and PAM-4 over 100 m MMF.
To encode the single-mode VCSEL chip, an electrical PAM-4 data with peak-to-peak amplitude of 1 V and pattern length of was exported from an arbitrary waveform generator (AWG, Keysight M8195A) with a sampling rate of . After passing through a 35 GHz wideband microwave amplifier (AMP, Picosecond 5882) with a power gain of 16 dB and noise figure (NF) of 6 dB, the PAM-4 data was employed to directly encode the single-mode VCSEL through a bias-tee (Anritsu, V250) with its 3 dB bandwidth of 65 GHz. For intra-data-center application, the collected light with carried PAM-4 data stream was launched into the OM4 MMF (Corning, ClearCurve OM4 fiber with core/cladding diameter of 50-μm/125-μm). The transmittance distance in the OM4 MMF can be lengthened from 100 to 300 m at increments of 100 m. Then, the optical PAM-4 data was received by a high-speed photodetector (PD, New Focus 1484-A-50) with a 3 dB bandwidth of 22 GHz, and another microwave amplifier (AMP, Picosecond 5882) was employed to compensate the power attenuation induced during the OM4 MMF transmission. At last, the amplified PAM-4 data was captured by an electrical sampling module (Tektronix, 80E10) plugged inside a digital sampling oscilloscope (DSO, Tektronix 8300) for eye-diagram and BER analyses.
For data pre-emphasis, the pre-emphasizer is a software program constructed with the AWG, which is designed by measuring and compensating the throughput frequency response of the whole transmission system to be tested. Initially, both amplitude and phase parts of the delivered PAM-4 data stream are measured and recorded to retrieve the systematic response. Afterwards, the AWG can generate a modified PAM-4 data stream for pre-compensating the amplitude and phase parts of the throughput response of the whole transmission channel. To implement the broadband waveform pre-emphasis of the PAM-4 data for pre-compensating the distortion made after propagating through the whole testing system in our case, the AWG initially delivers a wideband PAM-4 data at specified bit rate to encode the single-mode VCSEL for checking the distortion-induced amplitude and phase variations after transmitting through the whole MMF link. Subsequently, the waveform of the receiving data was received by a real-time digital serial analyzer (DSA, Tektronix DPO77002SX) with a bandwidth of 33 GHz to determine the weighting factors for amplitude and phase parts. Afterwards, the retrieved weighting factors were employed to pre-distort the waveform of the PAM-4 data generated from the AWG. With this procedure, the pre-emphasized PAM-4 data can be regenerated to overcome the channel-response-induced data degradation and improve its transmitted eye diagram significantly. After the transmissions in the BtB case and in the OM4 MMF with distances of 100, 200, and 300 m, different amplitude/phase weighting factors were individually employed to pre-emphasize the PAM-4 data waveform before encoding the VCSEL chip. Finally, a commercial software program (Tektronix, 80SJNB) built up with the digital sampling oscilloscope was used to analyze the received BER and the related bathtub curve.
3. RESULTS AND DISCUSSION
Regarding the design and fabrication of the single-mode VCSEL, the detailed structure of the single-mode VCSEL configuration is shown in Fig. 2(a). The top distributed Bragg reflector (DBR) is made of 21 pairs of n-type layers, and the bottom DBR is made of 37 pairs of p-type layers. Between the DBRs, there are three strained quantum wells (QWs) grown to serve as the active region. Upon the active layer, an oxide-confined aperture with its size reduced to 3 μm is formed to achieve single-transverse-mode lasing. To improve the conductance, an additional zinc-diffusion process is introduced on the contact layer to decrease the contact resistance and enhance the modulation bandwidth of the single-mode VCSEL chip.
Fig. 2. (a) 3D structure, (b) P–I curve with corresponding differential resistance, and (c) optical spectrum of the single-mode VCSEL chip.
The fundamental characteristics of the single-mode VCSEL, including the P-I-V curve with corresponding differential resistance and lasing spectrum obtained at a bias of 3.7 mA, are shown in Figs. 2(b) and 2(c), respectively. Note that the single-mode VCSEL ensures high injection current density within an active region to exhibit a threshold current of only 0.37 mA with a P–I slope of and a differential quantum efficiency of 0.12 (including the coupling efficiency between VCSEL and lensed MMF). As the single-mode VCSEL with small aperture is prone to cause heat accumulation, the related Auger effect is easily induced to gradually saturate its optical output power when the bias is beyond 1.5 mA. The single-mode VCSEL biased at 3.7 mA exhibits a single-transverse mode with a side-mode suppression ratio as high as 31.3 dB and a full width at half maximum of 0.06 nm. Biasing at 3.7 mA leads to a differential resistance of 159 Ω for the single-mode VCSEL, which causes an electrical return loss of and a voltage standing wave ratio (VSWR) of 3.18. Note that a fluctuating trend can be observed at bias current of because of the unstable zinc dopants when operating at high compliance voltage and suffering from the heat-accumulation effect.
Subsequently, the frequency responses of the single-mode VCSEL under small-signal analog modulation at different bias currents are illustrated in Fig. 3. By increasing the bias current from 1.5 to 4 mA, the single-mode VCSEL further enlarges its 3 dB modulation bandwidth from 17.9 to 18.9 GHz. In the latter experiments, the bias current is therefore adjusted at 3–4 mA for pursuing the 3 dB modulation bandwidth as large as possible.
To optimize the bias current of the single-mode VCSEL for maximizing the data capacity, the BER responses and related eye diagrams of the PAM-4 data at 12 GBaud (24-Gbit/s) carried by the single-mode VCSEL biased at different currents are illustrated in Fig. 4.
Fig. 4. BER performance and corresponding eye diagrams of the PAM-4 data directly modulated onto the single-mode VCSEL at different biases.
When operating the single-mode VCSEL at , the received eye diagram reveals a significant waveform clipping effect on the top eye pattern, which seriously degrades its receiving BER to . This originates mainly from the turn-off of the VCSEL below threshold. It is necessary to enlarge the DC bias to 3.7 mA for up-offsetting the amplitude level of the modulated PAM-4 data, which effectively suppresses the waveform clipping effect to provide a clear PAM-4 eye diagram with improved BER of . Overly biasing the single-mode VCSEL up to 4.2 mA causes interior heat accumulation as well as Auger effect, which inevitably saturates the output to distort the PAM-4 data waveform with degrading BER of . Note that the enhancement on the 3 dB modulation bandwidth is trivial when enlarging the DC bias from 2.5 to 4 mA, and the optimization is mainly focused on excluding the power saturation and waveform clipping under direct modulation.
After bias optimization, the weighting factors with both amplitude and phase parts of the modulated throughputs for the BtB, 100, 200, and 300 m MMF transmissions are measured to perform the PAM-4 waveform pre-emphasis, as shown in Fig. 5. When compared with the BtB case, the 100 m MMF transmission case slightly weakens its high-frequency throughput by but retains a similar phase response. Lengthening the MMF distance from 200 to 300 m not only degrades the high-frequency throughput from 4 to 8 dB, but also induces distinct phase shift from 0° to 800° on the modulated signal at different frequencies, as the modulated frequency increases from 0 to 24 GHz. In more detail, Fig. 6 shows the RF spectra and related eye diagrams of the O-to-E converted electrical PAM-4 data at 24 GBaud with and without pre-emphasis after BtB, 100, 200, and 300 m MMF transmissions.
Fig. 5. Amplitude and phase responses of the single-mode VCSEL after BtB, 100, 200, and 300 m MMF transmissions.
Fig. 6. RF spectra and corresponding eye diagrams of single-mode VCSEL-carried PAM-4 data at 24 GBaud with and without pre-emphasis after 0, 100, 200, and 300 m MMF transmissions.
Under BtB transmission, the electrical PAM-4 data at 24 GBaud reveals the clearest eye diagram with BER when compared with other cases. Nevertheless, when encoding such data onto the single-mode VCSEL, the BtB transmitted eye diagram becomes blurred due to the declined throughput and limited bandwidth of the single-mode VCSEL. Throughput degradation beyond 10 GHz can be confirmed by comparing the RF spectra of electrical and optical PAM-4 data streams without pre-emphasis, as shown in the left and right parts of Fig. 6(a). To compensate such a declined throughput response with increasing modulation frequencies, waveform pre-emphasis technology is introduced for energy re-arrangement according to the weighting factors extracted from the amplitude and phase responses shown in Fig. 5. The pre-emphasis on the phase part effectively reduces the data-dependent jitter, and the pre-emphasis on the amplitude part compensates the throughput attenuation. As compared to the case without pre-emphasis, the RF spectra of the received PAM-4 data with pre-emphasis exhibit weaker throughput at lower frequency and stronger throughput at higher frequency. When encoding the pre-emphasized 24 GBaud PAM-4 data onto the single-mode VCSEL, the received data shows a flatter throughput and clearer eye diagram, with a smaller BER of .
After transmitting the PAM-4 data with pre-emphasis over 100 m MMF, a clear eye diagram can also be obtained with a BER of . When compared with the BtB case, it is found that more modulation energy transformation from lower to higher frequencies is required to compensate the dispersion and RF fading induced throughput power declination. A similar trend can be observed for the 200 and 300 m OM4 MMF cases. However, even with the assistance of the waveform pre-emphasis, the 200 and 300 m MMF transmitted BERs still deteriorate to and . As the transmission distance lengthens, more severe dispersion is induced to cause the lower SNR, which forces the low-frequency band to sacrifice more energy to compensate for the declination occurring at the high-frequency band. Such a low-to-high frequency energy rearrangement, accompanied with the PAM-4 data waveform pre-emphasis, inevitably decreases the signal amplitude and degrades the overall SNR of the received data. Therefore, the 300 m MMF transmission reveals the lowest throughput intensity with the highest BER of compared with other cases.
In more detail, the related parameters including BER, RMS jitter, eye width, RMS noise, average rising/falling time, and peak-to-peak amplitude of the pre-emphasized eye diagrams for the 24 GBaud PAM-4 data after 0, 100, 200, and 300 m MMF transmissions are summarized in Table 1. Among them, the smallest RMS jitter of 0.92 ps is observed for the BtB case. In addition, the rising and falling times of the top, middle, and bottom eyes are 15.7 and 15.6 ps, respectively. After transmitting over 300 m MMF, the low-to-high frequency spectral power transformation of the PAM-4 data stream seriously decreases the peak-to-peak signal amplitude from 420 to 152 mV, which increases the RMS jitter to 1.77 ps and the rising/falling time to .
Table 1. Parameters of the Received Eye Diagrams after 0, 100, 200, and 300 m MMF Transmissions
For detailed comparisons of the received BER tolerance in the time domain, the bathtub BER curves of BtB, 100, 200, and 300 m MMF transmitted 24-GBaud PAM-4 data carried by the single-mode VCSEL with data waveform pre-emphasis are shown in Fig. 7. To meet the KP4-FEC criterion, the BtB case allows the largest jitter tolerances of 14.7, 20.5, and 14.1 ps (or 0.35, 0.49, and 0.34 unit interval, U.I.) for the top, middle, and bottom eyes compared with other cases.
Fig. 7. Bathtub curves of 24 GBaud PAM-4 data with pre-emphasis after BtB, 100, 200, and 300 m MMF transmissions.
After 100 and 200 m transmissions in the OM4 MMF, the received top/middle/bottom eye widths shrink to suppress the received jitter tolerances to 13.2/16.4/12.2 ps (0.32/0.39/0.29 U.I.) and 13.6/14.7/11 ps (0.33/0.36/0.26 U.I.), respectively. As the transmission distance continuously increases, the induced group velocity delay reduces jitter tolerance. As a result, the jitter tolerances of the received top, middle, and bottom eyes significantly degrade to 10.7, 11.4, and 9.3 ps (0.26, 0.27, and 0.22 U.I.), respectively. The received PAM-4 waveform and eye diagram are reversed as the negatively biased photodetector with an inverse output is employed in this work. Therefore, the bottom eye of the received PAM-4 eye diagram shown in Fig. 7 is actually the top eye after receiving and reversing by the PD. As the top eye of the modulated PAM-4 data would be influenced by the saturated output of the VCSEL and the frequency response of the transmission system, the top eye exhibits the highest BER among all eye diagrams, where the modal dispersion degrades the high-frequency response and extends the rising/falling time to result in the worse waveform and BER for the middle and bottom eye diagrams, compared to the top one, after increasing the transmission distance.
To investigate the maximal allowable transmission capacity of the single-mode VCSEL when delivering the PAM-4 data stream, Fig. 8(a) shows the BERs of the received PAM-4 data without and with pre-emphasis at different bandwidths after BtB, 100, 200, and 300 m OM4 MMF transmissions. Without pre-emphasis, the BtB cases can only support the maximal allowable PAM-4 bandwidths of 16 GBaud at data rates of . Utilizing the data waveform pre-emphasis significantly increases the maximal allowable PAM-4 data bandwidth to 32 GBaud at a data rate of for both cases, and the corresponding BERs of and which pass the KP4-FEC criterion are observed for the BtB and 100 m MMF cases, respectively. Note that, due to the systematic limitation, the PAM-4 data bandwidth cannot be increased beyond 32 GBaud as the used AWG exhibits an inherent sampling rate of only . Further increasing the PAM-4 data bandwidth inevitably suffers from the aliasing effect to deteriorate the original data quality.
Fig. 8. (a) BERs of the single-mode VCSEL-carried and waveform pre-emphasized PAM-4 data at different bandwidths after BtB, 100, 200, and 300 m OM4 MMF transmissions. (b) Eye diagrams of the pre-emphasized PAM-4 data at 32, 32, 27, and 24 GBaud carried by the single-mode VCSEL over BtB, 100, 200, and 300 m OM4 MMF.
As the transmission distance increases, the required low-to-high-frequency energy transformation inevitably decreases the signal amplitude, which limits the transmission capacity. In contrast, for the carried PAM-4 data after 200 and 300 m MMF transmissions, the maximal allowable bandwidths seriously shrink to 27 and 24 GBaud with corresponding data rates of 54 and for receiving BERs of and , respectively. Obviously, the receiving performance after 200 and 300 m transmissions has already approached the margin of KP4-FEC criterion for qualified detection and decoding process. For a fair comparison at the same data rate, the BER curves of the single-mode VCSEL carrying 24 Gbaud PAM-4 data with pre-emphasis after BtB, 100, 200, and 300 m MMF transmissions are shown in Fig. 9. To meet the KP4-FEC required BER of , the BtB receiving requires a power sensitivity of only ; however, this is increased to with a power penalty of 0.84 dB after 100 m OM4 MMF transmission. By further lengthening the OM4 MMF distance to 200 and 300 m, the receiving power sensitivities gradually increase up to and with corresponding power penalties of 2.5 and 4.2 dB (as compared to the BtB case), respectively.
Fig. 9. Measured BER of BtB, 100, 200, and 300 m MMF transmitted pre-emphasized PAM-4 data carried by the single-mode VCSEL chip.
Note that the linear increments per power penalty of 1.7 dB obtained by lengthening the OM4 MMF distance to 100 m are confirmed from this experiment. As expected, the data waveform pre-emphasis technology successfully provides an adaptive compensation for the degradation of both amplitude and phase channel responses, which effectively overcomes the uneven frequency responses within the modulation bands caused by the finite frequency bandwidth of the used components and RF power fading induced SNR degradation. These results indicate that the single-mode VCSEL supports data rate over 200 m MMF and thus becomes the best candidate for intra-data-center application.
4. CONCLUSION
A single-mode VCSEL at a central wavelength of 843 nm is employed to carry pre-emphasized PAM-4 data at 32 GBaud and to achieve single-channel transmission at a bit rate as high as . The single-mode VCSEL-carried PAM-4 data stream guarantees error-free transmission in OM4 MMF over a distance of 100–300 m, which ensures the data capacity required for high-speed data-center applications. At an optimized bias current of 3.7 mA, the single-mode VCSEL exhibits a threshold current of 0.37 mA, power-to-current slope of 0.12 (measuring after coupling into the OM4-MMF), and differential resistance of . Such a high differential resistance inevitably leads to the electrical return loss of and VSWR of 3.18, which limits the analog modulation with a 3 dB bandwidth to only 18.9 GHz. The waveform pre-emphasis of PAM-4 data is performed by extracting the weighting factors of the amplitude and phase parts of the received PAM-4 data. After compensating the throughput declination of the encoded PAM-4 data propagating over 0, 100, 200, and 300 m OM4 MMF, error-free transmission can be obtained with such a PAM-4 waveform pre-distortion during data generation. As a result, both BtB and 100 m OM4 MMF cases successfully support up to PAM-4 transmission with KP4-FEC certified data qualities. When lengthening the OM4 MMF distance from 200 to 300 m, the waveform pre-emphasis causes low-to-high frequency spectral power transformation of the PAM-4 data, which further attenuates the peak-to-peak data amplitude to degrade the transmission capacities from 54 to . As compared to the BtB case, the receiving power penalties of the 100, 200, and 300 m OM4 MMF transmission cases are 0.84, 2.5, and 4.2 dB, respectively. The PAM-4 waveform pre-emphasis technique essentially helps ensure the ability of the single-mode VCSEL to carry PAM-4 data beyond over 200 m MMF, and shows great potential for intra-data-center applications.
Funding
Ministry of Science and Technology, Taiwan (MOST) (MOST 104-2221-E-002-117-MY3, MOST 106-2218-E-005-001, MOST 106-2221-E-002-152-MY3).
REFERENCES
1. P. Westbergh, E. P. Haglund, E. Haglund, R. Safaisini, J. S. Gustavsson, and A. Larsson, “High-speed 850 nm VCSELs operating error free up to 57 Gbit/s,” Electron. Lett. 49, 1021–1023 (2013). [CrossRef]
2. D. M. Kuchta, A. V. Rylyakov, F. E. Doany, C. L. Schow, J. E. Proesel, C. W. Baks, P. Westbergh, J. S. Gustavsson, and A. Larsson, “A 71-Gb/s NRZ modulated 850-nm VCSEL-based optical link,” IEEE Photon. Technol. Lett. 27, 577–580 (2015). [CrossRef]
3. P. Westbergh, J. S. Gustavsson, B. Kögel, A. Haglund, A. Larsson, A. Mutig, A. Nadtochiy, D. Bimberg, and A. Joel, “40 Gbit/s error-free operation of oxide-confined 850 nm VCSEL,” Electron. Lett. 46, 1014–1016 (2010). [CrossRef]
4. S. A. Bolkhin, J. A. Lott, A. Muti, G. Fiol, N. N. Ledentsov, M. V. Maximov, A. M. Nadtochiy, V. A. Shchunkin, and D. Bimberg, “Oxide-confined 850 nm VCSELs operating at bit rates up to 40 Gbit/s,” Electron. Lett. 45, 501–503 (2009). [CrossRef]
5. M. Lebby, C. A. Gaw, W. Jiang, P. A. Kiely, C. L. Shieh, P. R. Claisse, J. Ramdani, D. H. Hartman, D. B. Schwartz, and J. Grula, “Characteristics of VCSEL array for parallel optical interconnects,” in Proceedings of the Electronic Components and Technology Conference (1996), pp. 279–291.
6. P. Westbergh, R. Safaisini, E. Haglund, B. Kögel, J. S. Gustavsson, A. Larsson, M. Geen, R. Lawrence, and A. Joel, “High-speed 850 nm VCSELs with 28 GHz modulation bandwidth operating error-free up to 44 Gbit/s,” Electron. Lett. 48, 1145–1147 (2012). [CrossRef]
7. H. E. Li and K. Iga, Vertical-Cavity Surface-Emitting Laser Devices (Springer, 2003).
8. S. J. Trowbridge, “Ethernet and OTN - 400G and beyond,” in Optical Fiber Communication Conference, OSA Technical Digest Series (Optical Society of America, 2015), paper Th3H.1.
9. T.-T. Shih, Y.-C. Chi, R.-N. Wang, C.-H. Wu, J.-J. Huang, J.-J. Jou, T.-C. Lee, H.-C. Kuo, G.-R. Lin, and W.-H. Cheng, “Efficient heat dissipation of uncooled 400-Gbps (16 × 25-Gbps) optical transceiver employing multimode VCSEL and PD arrays,” Sci. Rep. 7, 46608 (2017). [CrossRef]
10. K. Szczerba, P. Westbergh, M. Karlsson, P. A. Andrekson, and A. Larsson, “60 Gbits error-free 4-PAM operation with 850 nm VCSEL,” Electron. Lett. 49, 953–955 (2013). [CrossRef]
11. H.-Y. Kao, Y.-C. Chi, C.-Y. Peng, S.-F. Leong, C.-K. Chang, Y.-C. Wu, T.-T. Shih, J. J. Huang, H.-C. Kuo, W.-H. Cheng, C.-H. Wu, and G.-R. Lin, “Modal linewidth dependent transmission performance of 850-nm VCSELs with encoding PAM-4 over 100-m MMF,” IEEE J. Quantum Electron. 53, 8000408 (2017).
12. J. M. Castro, R. Pimpinella, B. Kose, Y. Huang, B. Lane, A. A. Correa, M. Bigot, D. Molin, and P. Sillard, “200 m 2 × 50 Gbps PAM-4 SWDM transmission over wideband multimode fiber using VCSELs and pre-distortion signal,” in Optical Fiber Communication Conference, OSA Technical Digest Series (Optical Society of America, 2016), paper Tu2G.2.
13. J. M. Castro, R. Pimpinella, B. Kose, Y. Huang, B. Lane, K. Szczerba, P. Westbergh, T. Lengyel, J. S. Gustavsson, A. Larsson, and P. A. Andrekson, “50 Gb/s 4-PAM over 200 m of high bandwidth MMF using a 850 nm VCSEL,” in Optical Fiber Communication Conference, OSA Technical Digest Series (Optical Society of America, 2015), paper W1D.1.
14. J. Lavrencik, S. Varighese, A. Varghese, G. Landry, Y. Sun, R. Shubochkin, and K. Balemarthy, “100 Gbps PAM-4 transmission over 100 m OM4 and wideband fiber using 850 nm VCSELs,” in European Conference and Exhibition on Optical Communication Conference, OSA Technical Digest Series (Optical Society of America, 2016), paper Th.1.C.5.
15. S. M. R. Motaghiannezam, I. Lyubomirsky, H. Daghighian, C. Kocot, T. Gray, J. Tatum, A. AmezcuaCorrea, M. Bigot-Astruc, D. Molin, F. Achten, and P. Sillard, “180 Gbps PAM4 VCSEL transmission over 300 m wideband OM4 fibre,” in Optical Fiber Communication Conference, OSA Technical Digest Series (Optical Society of America, 2016), paper Th3G.2.
16. C.-T. Tsai, C.-Y. Peng, C.-Y. Wu, S.-F. Leong, H.-Y. Kao, H.-Y. Wang, Y.-W. Chen, Z.-K. Weng, Y.-C. Chi, H.-C. Kuo, J.-J. Huang, T.-C. Lee, T.-T. Shih, J.-J. Jou, W.-H. Cheng, C.-H. Wu, and G.-R. Lin, “Multi-mode VCSEL chip with high-indium-density InGaAs/AlGaAs quantum-well pairs for QAM-OFDM in multi-mode fiber,” IEEE J. Quantum Electron. 53, 2400608 (2017). [CrossRef]
17. R. B. Nunes, H. R. de O. Rocha, D. A. A. Mello, F. D. Simões, M. E. V. Segatto, and J. A. L. Silva, “Transmission of CE-OFDM signals over MMF links using directly modulated 850-nm VCSELs,” IEEE Photon. Technol. Lett. 27, 315–318 (2015). [CrossRef]
18. H.-Y. Kao, Y.-C. Chi, C.-T. Tsai, S.-F. Leong, C.-Y. Peng, H.-Y. Wang, J. J. Huang, J.-J. Jou, T.-T. Shih, H.-C. Kuo, W.-H. Cheng, C.-H. Wu, and G.-R. Lin, “Few-mode VCSEL chip for 100-Gb/s transmission over 100 m multimode fiber,” Photon. Res. 5, 507–515 (2017).
19. W. Bo, Z. Xian, M. Yanan, L. Jun, Z. Kangping, Q. Shaofeng, F. Zhiyong, L. Yazhi, M. Agustin, N. Ledentsov Jr., J. Kropp, V. Shchukin, N. N. Ledentsov, I. Eddie, and L. Chao, “Close to 100 Gbps discrete multitone transmission over 100 m of multimode fiber using a single transverse mode 850 nm VCSEL,” Proc. SPIE 9766, 97660K (2016). [CrossRef]
20. H.-Y. Kao, C.-T. Tsai, S.-F. Leong, C.-Y. Peng, Y.-C. Chi, J.-J. Huang, H.-C. Kuo, and G.-R. Lin, “Comparison of single-/few-/multi-mode 850 nm VCSELs for optical OFDM transmission,” Opt. Express 25, 16347–16363 (2017). [CrossRef]
21. Y.-C. Chi, Y.-C. Li, H.-Y. Wang, P.-C. Peng, H.-H. Lu, and G.-R. Lin, “Optical 16-QAM-52-OFDM transmission at 4 Gbit/s by directly modulating a coherently injection-locked colorless laser diode,” Opt. Express 20, 20071–20077 (2012). [CrossRef]
22. J.-J. Liu, K.-L. Chi, C.-C. Wei, T.-C. Lin, C.-Y. Chuang, X.-N. Chen, J.-W. Shi, and J. Chen, “High bit-rate distance product of 128 Gbps·km 4-PAM transmission over 2-km OM4 fiber using an 850-nm VCSEL and a Volterra nonlinear equalizer,” in Optical Fiber Communication Conference, OSA Technical Digest Series (Optical Society of America, 2017), paper W3G.5.
23. F. Breyer, S. C. J. Lee, S. Randel, and N. Hanik, “Comparison of OOK and PAM-4 modulation for 10 Gbit/s transmission over up to 300 m polymer optical fiber,” in Optical Fiber Communication Conference, OSA Technical Digest Series (Optical Society of America, 2008), paper OWB5.
24. K. Szczerba, P. Westbergh, E. Agrell, M. Karlsson, P. A. Andrekson, and A. Larsson, “Comparison of intersymbol interference power penalties for OOK and 4-PAM in short-range optical links,” J. Lightwave Technol. 31, 3525–3534 (2013). [CrossRef]
25. Y. Sun and R. Lingle Jr., R. Shubochkin, K. Balemarthy, D. Braganza, T. Gray, W. J. Fan, K. Wade, D. Gazula, and J. Tatum, “51.56 Gb/s SWDM PAM4 transmission over next generation wide band multimode optical fiber,” in Optical Fiber Communication Conference, OSA Technical Digest Series (Optical Society of America, 2016), paper Tu2G.3.
26. A. Gholami, D. Molin, and P. Sillard, “Compensation of chromatic dispersion by modal dispersion in MMF- and VCSEL-based gigabit ethernet transmissions,” IEEE Photon. Technol. Lett. 21, 645–647 (2009). [CrossRef]
27. G. Stepniak, A. Lewandowski, J. R. Kropp, N. N. Ledentsov, V. A. Shchukin, N. Ledentsov Jr., G. Schaefer, M. Agustin, and J. P. Turkiewicz, “54 Gbit/s OOK transmission using single mode VCSEL up to 2.2 km MMF,” Electron. Lett. 52, 633–635 (2016). [CrossRef]
28. P. Moser, J. A. Lott, and D. Bimberg, “Energy efficiency of directly modulated oxide-confined high bit rate 850 nm VCSELs for optical interconnects,” IEEE J. Sel. Top. Quantum Electron. 19, 1702212 (2013). [CrossRef]
29. Å. Haglund, J. S. Gustavsson, J. Vukusic, P. Modh, and A. Larsson, “Single fundamental-mode output power exceeding 6 mW from VCSELs with a shallow surface relief,” IEEE Photon. Technol. Lett. 16, 368–370 (2004). [CrossRef]
30. P. Moser, J. A. Lott, P. Wolf, G. Larisch, H. Li, and D. Bimberg, “85-fJ dissipated energy per bit at 30 Gb/s across 500-m multimode fiber using 850-nm VCSELs,” IEEE Photon. Technol. Lett. 25, 1638–1641 (2013). [CrossRef]
31. R. Safaisini, E. Haglund, P. Westbergh, J. S. Gustavsson, and A. Larsson, “20 Gbit/s data transmission over 2 km multimode fibre using 850 nm mode filter VCSEL,” Electron. Lett. 50, 40–42 (2014). [CrossRef]
32. M. P. Tan, S. T. M. Fryslie, J. A. Lott, N. N. Ledentsov, D. Bimberg, and K. D. Choquette, “Error-free transmission over 1-km OM4 multimode fiber at 25 Gb/s using a single mode photonic crystal vertical-cavity surface-emitting laser,” IEEE Photon. Technol. Lett. 25, 1823–1825 (2013). [CrossRef]
33. S. M. R. Motaghiannezam and C. Kocot, “104 Gbps PAM4 transmission over OM3 and OM4 fibers using 850 and 880 nm VCSELs,” in Lasers and Electro-Optics Conference, OSA Technical Digest Series (Optical Society of America, 2016), paper SW4F.8.
