Open Access
Add to BibSonomy
Abstract
With the advent of the sixth-generation mobile communication standard (6 G), the visible light communication (VLC) technology based on wavelength division multiplexing (WDM) technology can effectively solve the problem of shortage of spectrum resources and insufficient channel capacity. This paper introduces one of our technical achievements, namely the construction of a near-real-time visible light laser communication (VLLC) system based on WDM, which includes a self-designed 10-λ fully-packaged visible light laser emission module, 1 m multimode fiber – 0.175 m free space – 1 m multimode fiber optical transmission link, and receiver array. In the transmitter system, we adopt adaptive discrete multitone (DMT) modulation technique combined with Quadrature Amplitude Modulation (QAM) modulation scheme to obtain maximum spectral efficiency (SE). In the receiving system, we utilize the sparse-structured reservoir computing post-equalization algorithm to achieve superior equalization performance on the basis of the traditional post-equalization algorithm. The experimental results indicate that this quasi-real-time communication system has achieved a signal transmission rate of 113.175Gbps. To the best of our knowledge, this work has set a record in the field of high-speed visible light laser communication. Therefore, the laser communication system constructed by this work, with its flexibility in deployment and high-speed performance, demonstrates the significant potential application of visible light laser communication in data center interconnection and high-speed indoor access networks.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
With the emergence of the sixth-generation mobile communication standard, the shortage of spectrum resources and the shortage of channel capacity greatly stimulate the development of new spectral resources and the research of higher spectral efficiency communication system. In view of these urgent scientific problems, visible light communication technology based on WDM has obvious advantages and application potential. [1]
In order to fully utilize the advantage of VLLC to make up for the serious shortage of spectrum resources, there are still abundant spectrum resources to be applied to WDM in the visible light band (approximately 400 nm to 780 nm), even when the research and development of visible light communication devices is not yet sufficient. The application of VLLC based on WDM is currently developing towards the characteristics of higher data rates in a single channel and multiplexing over different spectra. This development trend is presented in the Table 1. Currently, in known research, the transmission rate of a single visible laser in free space links can reach 14.590Gbps [6]. WDM-based VLLC systems can achieve 21 channels transmitting information simultaneously [5]. However, currently, achieving the visible light spectrum reuse across multiple wavelengths is a pressing challenge, both in terms of engineering and research. Therefore, recent advancements in free-space visible light communication systems based on WDM have predominantly relied on three to four different laser wavelengths [6]. This limitation poses difficulties in realizing visible light communication systems with higher channel capacities. Consequently, employing multiple wavelength laser sources in visible light communication systems can be advantageous for increasing channel capacity and making more efficient use of the spectral resources available in the visible light spectrum, which spans approximately 400 nm.
Table 1. Recent achievements of WDM-based free-space VLC systems
Affected by the severe nonlinear interference of visible light band devices and fiber channels, in order to achieve error-free transmission of visible light in complex fiber-free space-fiber channels, and impact higher SE, digital signal processing algorithms need to be used in the modulation and demodulation processes of the signal. Table 1 also shows the modulation formats in related research in recent five years [2–11]. It can be found that without using various multiplexing forms such as space division multiplexing (SDM) and orbital angular momentum (OAM) multiplexing [4], the highest single channel communication link rate is generally obtained using QAM-DMT modulation format [5]. For the visible light channel response, there is severe attenuation in the high frequency region, and effects such as severe noise in the low frequency region [12]. Therefore, the signal-to-noise ratio (SNR) within the visible light signal bandwidth has significant differences at different frequencies. In order to make full use of the in-band spectral resources of the signal, the DMT modulation format can allocate suitable QAM modulation order for each subcarrier according to its SNR, so as to maximize the spectral efficiency of the channel [13].
Reservoir Computing (RC) has found extensive applications and research in optical communication and sensing. In the context of optical fiber communication using intensity modulation-direct detection (IM-DD) systems, RC-based post-equalization algorithms have achieved remarkable results, with a bitrate-distance product of up to 11,200 Gbps·km in the C-band [14]. Moreover, in short-haul interconnections, the RC post-equalization algorithm has also achieved high rates, reaching up to 224 Gbps in IM-DD systems [15]. These achievements underscore the significant applicability of RC in addressing nonlinear signal response challenges [16].
In this paper, through the 10-λ fully packaged visible laser emitting module and QAM-DMT modulation format and RC equalization algorithm, a WDM-based VLLC system with a total transmission rate up to 113.175Gbps is achieved in a 1 m multimode fiber (MMF) - 0.175 m free space - 1 m MMF channel, with a spectral utilization range of 488nm-689 nm. Within the scope of our knowledge in related research, this achievement represents the most optimal combination of data rate and wavelength utilization. It holds significant implications for expanding new spectrum resources within the framework of the 6 G standard and serves as a valuable reference for high-speed communication applications in data centers.
2. Principle
2.1 Design of 10-λ WDM fully-packaged visible light emission module
The 10-λ fully packaged visible laser emitting module is shown in Fig. 1. The transmitter uses ten lasers with different wavelengths within visible light spectrum, packaged in the same heat sink. The heat sink uses an aluminum alloy cover for heat conduction of the lasers. The heat is removed by thermoelectric cooler (TEC), and the heat generated by the system is dissipated through a water-cooled radiator. TEC completes closed-loop control of the temperature by detecting the temperature of the lasers.
Fig. 1. The detail structure of high-speed WDM-VLC system with transmission module, cage system and receiver array.
In addition, the laser module is driven by a driver board, driving the lasers through the laser sockets integrated on the driver board. At the same time, the driver board integrates a bias-Tee chip, so only the RF signal and DC signal need to be input to complete internal modulation of the lasers. Since the operating current of the lasers is generally between 100mA-700 mA, the output power requirement for the laser driver is not high. Providing a maximum voltage of 10 V and current of 1A can meet the driving power requirements of all lasers.
The ten adjustable current sources can provide DC bias for the ten lasers. When starting the transmitting module, the output signals of the current sources need to be adjusted, and the output voltage and current limits need to be adjusted to prevent overvoltage and overcurrent. The ten lasers output through multimode fiber coupling, and are collimated through collimator lenses to minimize propagation loss in free space.
2.2 Reservoir computing algorithm
The input signal ${\mathbf u}(t )\in {R^M}$ where $t = 1, \ldots ,T$ is the discrete time and T the number of data points in the training set. In order to project this time onto the reservoir we will need an input to reservoir coupler, identified by the matrix ${{\mathbf W}_{\textrm{in}}} \in {R^{N \times M}}$.
The reservoir is constituted by N neurons connected in a Watts-Strogatz graph configuration and it is represented by an adjacency matrix ${\mathbf W}$ of size $N \times N$ with values drawn from a uniform random distribution on the interval [−1,1]. Understanding all of its components is a prerequisite for efficiently constructing an Echo State Network (ESN), as this is the most pivotal aspect of the network.
The sparse nature of the reservoir is a common observation in the literature. In the majority of research papers, it is evident that each node within the reservoir is typically linked to a limited number of other nodes, typically ranging from 5 to 12 connections. The sparsity, in addition to its theoretical implications, serves the practical purpose of accelerating computational processes.
After generating a random sparse reservoir matrix, the spectral radius $\rho ({\mathbf W} )$ is calculated, and then ${\mathbf W}$ is divided by this value. This procedure yields a matrix with a unit spectral radius, which can subsequently be scaled to a more appropriate value. It's worth noting that there are exceptions, particularly when the inputs ${\mathbf u}(t )$ are non-zero. However, maintaining a spectral radius below unity, $\rho ({\mathbf W} )< 1$, is a crucial condition for ensuring the echo state property. In a broader context, this parameter should be chosen with the aim of maximizing system performance, while still considering the unitary value as a valuable benchmark.
After the construction of the input layer and the reservoir we can focus on harvesting the states. The update equations of the ESN are:
- ${\mathbf v}(t )\in {R^M}$ is the predicted output.
- ${\mathbf x}(t )\in {R^N}$ is the state vector.
- ${{\mathbf W}_{\textrm{out}}} \in {R^{M \times N}}$ is the output layer.
- $f$ and g are two activation functions, most commonly the hyperbolic tangent and identity respectively.
$\alpha $ is the leaking rate.
The computation of ${{\mathbf W}_{\textrm{out}}}$ can be expressed in terms of solving a system of linear equations
After the training of the ESN, the prediction phase uses the same update equations showed above, but the input ${\mathbf u}(t )$ is represented by the computed output ${\mathbf v}({t - t} )$ of the preceding step.
In this paper, the RC and second-order Volterra LMS algorithm were employed as post-equalization schemes for quasi-real-time communication. Regarding the algorithm complexity, we use big-O notation to quantitatively express the time complexity of the RC. We have further supplemented pseudocode for RC to explicitly outline the core code for calculating complexity.
According to the pseudocode, it is evident that the time complexity of the reservoir computing algorithm is O(n), which is lower compared to second-order-Volterra LMS algorithm and traditional DNN networks. From the pseudocode, it is apparent that RC, in form, can be likened to an RNN with only one hidden layer. Moreover, it does not employ mean squared error as a loss term but completes biased estimation using the ridge regression algorithm. Consequently, it lacks the traditional backpropagation process found in NN networks. This significantly reduces the complexity of the algorithm and explains why RC has a lower time complexity compared to NN and the second-order Volterra LMS algorithm.
3. Experimental setup
Figure 2(c) shows the experimental system constructed with a 10-λ WDM-based fully-packaged visible light emission module and receiver array through a transmission link of 1 m MMF - 0.175 m free-space - 1 m MMF. We utilized an 8GSa/s arbitrary waveform generator (AWG, Keysight, M8190A) along with MATLAB program and host software to generate DMT signals. The signal is power amplified by an electronic amplifier (EA, MiniCircuits, ZHL-1042J+) with 4.2 GHz bandwidth and further adjusted for laser modulation depth through a tunable electronic attenuator (ATT.). The signal is then transmitted into the emission module through the SMA signal input port sequentially. Direct modulation of the lasers occurred on the driver board inside the module, which is designed with RF circuit impedance optimization and electromagnetic interference (EMI) shielding for signals above 2 GHz. Electronic switches control the working status of the current sources driving each laser, turning on the current sources to output preset optimized bias currents. The driver board integrates 6 GHz bandwidth bias-Tee chips (MiniCircuits, RCBT-63+) to couple the externally input signal and direct current, and the output modulated signal is loaded on the lasers to drive the lasers and output optical signals. The optical signals needed to be coupled from the multimode fiber of the lasers to the 1 m MMF (105um core diameter) outside the server cabinet and input into a fiber collimator optimized for the visible light band. After collimation, the laser transmits over a 0.175 m free space channel constructed in a cage system, and is coupled into the fiber collimator, then through another 1 m MMF, and finally received by the receiver array. The receiver array consists of ten high-speed fiber coupled photodetectors (PD, Thorlabs, DET025AFC) with 2 GHz bandwidth. The electrical signal output by the photodetectors goes through amplifiers and attenuators with the same specifications as the transmitting end before being displayed on an oscilloscope and processed by MATLAB on a PC.
Fig. 2. Experimental setup. (a) Bit-power loading algorithm and reservoir computing post-equalization algorithm applied to experimental setup system. (b) The spectrum of ten LDs. (c) The photograph of the experimental system.
As shown in Fig. 2(b), it is the spectral distribution of the 10 lasers used in the experiment, with wavelengths of 486.028 nm, 504.597 nm, 521.581 nm, 636.943 nm, 638.662 nm, 639.421 nm, 642.192 nm, 665.516 nm, 686.341 nm, and 687.095 nm respectively. The spectra of the lasers at their optimum operating points were measured multiple times using a spectrometer (Ocean Insight, HR4Pro). It is verified multiple times that the wavelengths of the ten lasers had significant resolvability within the measurement accuracy of the spectrometer. Therefore, it can be ensured that the inter-channel crosstalk in the collimated parallel light paths is minimized as much as possible. The whole experiment process was carried out sequentially under the test conditions that the 10 laser communication links were turned on and optimized.
The process of digital signal processing in the experimental system is shown in Fig. 2(a). Since the bit-power-loading algorithm and DMT modulation format signal is used to obtain high spectral efficiency, the communication and rate testing needs to be completed in two steps. The first step uses 4QAM signals as the transmission signals to estimate the SNR in the communication link. Therefore, only QAM mapping and DMT modulation need to be completed at the transmitting end. At the receiving end, waveform-to-waveform equalization is first performed through the RC algorithm, and the SNR is estimated after DMT demodulation. The role of the equalization algorithm here is to obtain higher SNR by recovering the signal, so as to achieve an increase in the transmission signal rate.
The second test of transmitting and receiving signals is the final bit rate and bit error rate (BER) testing process. The prerequisite for the bit-power-loading algorithm is the Levin-Campello (LC) allocation algorithm. Based on the SNR estimation results, the order of QAM signals loaded on each subcarrier and power are allocated according to the SNR requirements, so that the spectral efficiency of each subcarrier can be maximized. After receiving the signal from the oscilloscope, the receiving end also needs to go through the RC equalization algorithm to complete data recovery, so as to obtain a lower BER.
In the experiment, in order to obtain the conclusion about the superiority of the RC equalization algorithm, we will change various equalization algorithms or not add equalization algorithms at the stage of receiving end equalization digital signal processing, such as the part marked as “Reservoir Computing” in the blue box of the Rx part. Therefore, we define the abbreviation for no equalization algorithm as “w/o post-equalization”, and define the abbreviation for using cascaded least mean square (LMS) and second order Volterra equalization algorithms as “LMS-Volterra”.
4. Results and discussion
First, as shown in Fig. 3, we tested the optimal operating points of the ten lasers without using an equalization algorithm. By traversing the DC bias current loaded to the lasers and the peak-to-peak voltage (Vpp) of the modulation signal within a certain range, the maximum communication rate that the lasers could achieve in the communication link and their operating status could be obtained, and detailed contour maps were plotted. In each contour map, there is an irregular area outlined by black lines and marked with numbers. This curve indicates the operating point status of the lasers when the communication link enters a higher communication rate state. In addition, in each contour map, the optimal operating point is also marked with a star, and the optimal rate under no equalization algorithm is labeled in the bottom right corner of each contour map.
Fig. 3. The data rate ${R_b}$ of ten channels at all operating points. The ${R_b}$ contours of (a) 665.5 nm, (b) 521.5 nm, (c) 504.5 nm, (d) 639.4 nm, (e) 638.6 nm, (f) 637 nm, (g) 642.1 nm, (h) 687 nm, (i) 686.3 nm, (j) 486 nm.
For visible light wavelength laser diodes, extensive research and experimentation have been conducted prior to this. Lasers of different wavelengths exhibit variations in both emission power and modulation depth. Visible lasers of different wavelengths demonstrate differences in the effects of attenuation and scattering in fiber-optic-free space-fiber communication links. And meanwhile, photodetectors exhibit variations in sensitivity to different wavelengths. The variations in the performance observed in experimental laser links at different wavelengths can be attributed to these reasons.
After completing the test for each laser, we added the reservoir computing equalization algorithm again for each laser communication link to attempt higher communication rates. One of the hyperparameters in reservoir computing that has higher requirements for computational complexity and accuracy is the number of nodes in the reservoir. We take the ninth, third and tenth channels with red, green and blue lasers as typical examples to present the impact of the number of reservoir nodes on transmission rate. The related results presented and analyzed with three communication links as examples will also be used in the rest of this paper.
In our exploration of using the Reservoir Computing (RC) algorithm for channel equalization applications, we once attempted to employ a fully connected deep neural network (DNN) with two hidden layers. In each hidden layer, there are 49 nodes. The last dense linear layer serves as a regressor. We compared the post-equalization capabilities of RC and DNN in this system, as shown in Fig. 4. In the experiments where we varied the DC bias, we introduced the use of the DNN algorithm for post-equalization. In the contour plot depicted in Fig. 3, we can achieve this by fixing Vpp and iterating through various DC bias values. Consistent with the trend in the contour plot in Fig. 3, bias current, as the major variable affecting the emitted optical power, increasing the optical power appropriately is beneficial for improving the Signal-to-Noise Ratio (SNR) of the laser in the channel. However, excessively high bias current can cause the laser's operating point to enter the nonlinear region, and the marginal effect of increasing optical power can be compounded by the attenuation in the optical fiber. Therefore, it is evident that in the green and blue laser communication links, nonlinear effects are more pronounced at high power levels, leading to a rapid decline in transmission speed. The results indicate that, at the highest communication speed, DNN has limited effectiveness in improving communication rates, but it demonstrates certain advantages in the nonlinear working region. We attribute this outcome to the adaptive bit-power-loading algorithm, which optimally utilizes the spectral resources of the channel, leading to efficient optimization of spectral efficiency in the mid-frequency range where the signal-to-noise ratio is relatively good and nonlinear effects are less pronounced. Consequently, using DNN might not significantly enhance communication rates at the general optimal operating point. Considering both algorithm time complexity and post-equalization capabilities, we primarily discuss the comparative results between LMS-Volterra and RC algorithms for the realization of a quasi-real-time visible light communication system.
Fig. 4. The effect of without utilizing the post-equalization algorithm, utilizing the LMS-Volterra algorithm, and utilizing the reservoir computing on the system rate under different DC bias. (a) Red laser communication link. (b) Green laser communication link. (c) Blue laser communication link.
As shown in Fig. 5, the number of nodes was increased in a formulaic way as traversal nodes. It was found that when the number of nodes was 32, the communication rate of the third laser communication link reached the maximum, and then gradually decreased as the number of nodes increased. The reasons for the change on both sides of the optimal node number are different. In the case where the number of nodes increases from 4 to 32, the number of nodes in the reservoir is too small. When extracting features and generating output matrices through ridge regression algorithms, the data dimension of ridge regression is too small, resulting in the regression analysis results not being able to effectively represent all the data features. When the number of nodes is greater than 32, as the matrix dimension increases, the effect of reservoir computing on channel equalization gradually decreases, and even shows a deteriorating trend in communication rate compared to without equalization algorithms.
Fig. 5. The influence of the number of reservoir nodes on the bit rate of the RC equalization algorithm (Typical application of red, green and blue laser)
Through the above optimization and parameter adjustment of the RC equalization algorithm, we obtained the highest communication rates in visible light communication links under RC equalization. To compare the performance of the RC equalization algorithm, we contrasted the results with no equalization algorithm and using the LMS-Volterra equalization algorithm. As shown in Fig. 6, it is the comparison results in the red, green, and blue three typical communication links. Figure 6(a) shows the test results for the 686.3 nm laser communication link. By traversing the modulation depth of the laser to change the nonlinearity of the output optical signal, it can be seen that in the case of lower nonlinearity, the link communication rate using the RC algorithm for equalization has some advantage over the LMS-Volterra algorithm. However, as the Vpp increases, the optical signal nonlinearity enhances, the bitrate without using equalization algorithms decreases significantly, while the test rate result using the RC algorithm has a more gradual decreasing slope. At the operating point with the strongest nonlinearity (marked by the black elliptical curve, @Vpp = 400 mV), the bitrate obtained by using the RC algorithm is about 0.3Gbps higher than that of the LMS-Volterra algorithm.
Fig. 6. The effect of without utilizing the post-equalization algorithm, utilizing the LMS-Volterra algorithm, and utilizing the reservoir computing on the system rate under different Vpp, and corresponding AMAM image. (a)-(b) Red laser communication link. (c)-(d) Green laser communication link. (e)-(f) Blue laser communication link.
The experimental results marked by the black elliptical curve in Fig. 6(a) correspond by color to the AMAM curves in Figs. 6(b)(i)-(iii) below. Gray represents the results without using equalization algorithms, blue represents using LMS-Volterra equalization algorithms, and red represents using RC equalization algorithms. It can be seen that under the same operating point, using RC compared to using LMS-Volterra, the noise is greatly suppressed. Also, compared to the results without using algorithms, the influence of the laser's operating point on the nonlinearity of the signal is sufficiently filtered out. Overall, it presents very good linearity and a very high signal-to-noise ratio.
The same trend appears in the green laser link and blue laser link, as shown in Figs. 6(c)-(d) and Fig. 6(e)-(f). In the green and blue laser links, since the gallium nitride-based lasers have higher operating power, the effect of the algorithms is more significant: under high bias current, the signal in the green and blue laser communication links is strongly interfered by nonlinear effects, resulting in a serious decrease in communication rate within the corresponding laser operating range. Under the operating points marked by the black ellipse dotted line in Fig. 6(c) and Fig. 6(e), when the equalization algorithm is not used, the communication rate decreases obviously, which reflects the nonlinear effect of the system. While using the RC algorithm compensates for the signal distortion to some extent, compared to without equalization, the rates of the blue and green optical communication links can be improved by 1.1Gbps and 0.89Gbps respectively. Also, the equalization effect of RC is slightly better than that of LMS-Volterra at all operating points.
From the time domain perspective, the AMAM curves have fully demonstrated the excellent performance of the RC algorithm in resolving the nonlinearity issues caused by the channel and laser operating point. This can also be reflected in the frequency domain. As shown in Fig. 7, it is the SNR response curve and corresponding LC bit-loading allocation results and transmitted signal power distribution in the red, green and blue typical communication links. As shown in Figs. 7(a)-(c), they are the corresponding results for the ninth red laser communication link. Compared to using the RC algorithm, using the LMS-Volterra algorithm and not using the algorithm, the SNR curve has an obvious depression between the 100th to 200th subcarriers. However, the RC algorithm can effectively compensate for the SNR sag. Without using equalization algorithms, the highest SNR of the communication link can reach 25 dB. Using the LMS-Volterra algorithm, it can reach up to 26.3 dB, while using the RC algorithm, it can reach 28.2 dB. At the same time, based on the SNR curve, the LC allocation algorithm can allocate the highest possible signal modulation order for each subcarrier. With the help of the RC algorithm, signal transmission up to 256QAM can be achieved, and overall higher spectral efficiency can be realized. The results of normalized power ratio distribution are illustrated in Fig. 7(a)-(c) below. The gray dashed line represents the initial allocation power value of 1. Through the LC allocation algorithm, subcarriers with redundant power can redistribute a certain amount of power to subcarriers that require minimum signal power to improve one bit. The power ratio distribution of the red laser link is between 0.7 and 1.1.
Fig. 7. SNR response and LC bit-loading allocation results and the distribution of transmitted signal power for (a)-(c) red LD channel, (d)-(e) green LD channel, (g)-(i) blue LD channel.
The data rate of ten visible light laser communication supported by different equalization algorithms is showed in the Table 2. The ten channel communication rates and corresponding bit error rates of the 10-λ WDM fully packaged laser visible light communication system are shown in Fig. 8. It shows the rate comparison results of not using equalization algorithms, using LMS-Volterra algorithms, and RC algorithms, represented by light gray, blue, and red bar charts, respectively. Using the RC algorithm can achieve about 1Gbps/channel rate increase compared to not using equalization algorithms, and about 0.6Gbps/channel rate increase compared to the LMS-Volterra algorithm, while keeping the bit error rate of the communication system below the error-free threshold of 0.0038.
Fig. 8. The data rate and BER of ten optical-fiber channels, comparing the effect of without post-equalization algorithm, LMS-Volterra algorithm, and reservoir computing.
Table 2. Data rate of ten visible laser communication supported by different equalization algorithms
5. Conclusion
The paper is based on hardware equipment comprising a 10-λ WDM-based visible laser transmitter module, employing DMT modulation format, bit-power-loading algorithm, and RC post-equalization algorithm. It achieved a total communication rate of 113.175 Gbps with a BER below 0.0038, demonstrating the significant advantages of the RC algorithm in the communication system when compared to no-equalization and LMS-Volterra classic algorithms. To the best of our knowledge and understanding of related research, this work represents the optimal combination of rate and wavelength utilization.
In this work, considerations regarding the trade-off between single-channel communication speed and multi-channel communication speed were considered during system design and experimentation.
In fact, this work leans more towards engineering applications. We have simplified the approach to constructing the optical link, deviating from the use of complex optical systems seen in other projects that achieve more precise laser collimation, coupling, and alignment. Instead, we have opted for a set of non-adjustable, non-spherical collimating lenses based on the cage system for alignment. This significantly streamlines operational procedures during system deployment and testing. However, it unavoidably reduces the system's accuracy, making it challenging to enhance the single channel transmission rate.
Additionally, the transmitter in this system is designed as a 10-λ laser module. When all channels are activated, there may be issues such as electromagnetic interference (EMI) and temperature imbalances between laser drivers, potentially leading to unstable operating conditions of the lasers influenced by each other. This instability can result in challenges achieving the ideal optimal communication speed for each laser communication link. This is a notable design difficulty in the related work. In subsequent iterations of the design, we may introduce more refined driving and temperature control loops to ensure precise control for each channel when multiple channels are active. This aims to achieve synchronous optimization for single-channel speed and total transmission speed. In the future, our intention is to utilize our expertise in design to develop a series of highly integrated transmitter modules, with a focus on enhancing integration levels and minimizing power consumption.
Funding
National Natural Science Foundation of China; National Key Research and Development Program of China.
Disclosures
The authors declare no conflicts of interest.
Data availability
No data were generated or analyzed in the presented research.
References
1. N. Chi, Y. Zhou, Y. Wei, et al., “Visible Light Communication in 6G: Advances, Challenges, and Prospects,” IEEE Veh. Technol. Mag. 15(4), 93–102 (2020). [CrossRef]
2. Z. Luo, X. Lin, Z. Lu, et al., “Over 100Gbps Free-space Laser-based Visible Light Communication System Based on 10-λ WDM Module,” 2023 European Conference on Optical Communication (ECOC), Glasgow, UK, 2023.
3. Y. Di, C. Sun, S. Chen, et al., “Capacity Enhancement of VLC by Blue-green Wavelength Division Multiplexing Using Optical Phased Array,” 2023 Optical Fiber Communications Conference and Exhibition (OFC), San Diego, CA, USA, 2023, pp. 1–3.
4. S. Singh, S. Singh, G. Kaur, et al., “Tricolor lasers and optical angular momentum based visible light communication system,” Opt. Quantum Electron. 55(10), 851 (2023). [CrossRef]
5. W. H. Gunawan, Y. H. Chang, C. W. Chow, et al., “High Speed RGB Visible Light Communication (VLC) Using Digital Power-Domain Multiplexing (DPDM) of Orthogonal Frequency Division Multiplexed (OFDM) Signals,” 2022 Optical Fiber Communications Conference and Exhibition (OFC), San Diego, CA, USA, 2022, pp. 1–3.
6. J. Hu, F. Hu, J. Jia, et al., “46.4 Gbps visible light communication system utilizing a compact tricolor laser transmitter,” Opt. Express 30(3), 4365–4373 (2022). [CrossRef]
7. C. W. Chow, Y. H. Chang, L. Y. Wei, et al., “26.228Gbit/s RGBV Visible Light Communication (VLC) with 2 m Free Space Transmission,” 2020 Opto-Electronics and Communications Conference (OECC), Taipei, Taiwan, 2020, pp. 1–3.
8. L. Issaoui, S. Cho, H. Chun, et al., “High CRI RGB Laser Lighting With 11Gb/s WDM Link Using Off-the-Shelf Phosphor Plate,” IEEE Photonics Technol. Lett. 34(2), 97–100 (2022). [CrossRef]
9. W. C. Wang, C. H. Cheng, H. Y. Wang, et al., “White-light color conversion with red/green/violet laser diodes and yellow light-emitting diode mixing for 34.8Gbit/s visible lighting communication,” Photonics Res. 8(8), 1398–1408 (2020). [CrossRef]
10. Y. C. Wu, C. Y. Su, W. C. Wang, et al., “30 Gbit/s Optical Wireless Communication with Red/Green/Violet LD Mixed White Light,” 2019 24th OptoElectronics and Communications Conference (OECC) and 2019 International Conference on Photonics in Switching and Computing (PSC), Fukuoka, Japan, 2019, pp. 1–3.
11. T. C. Wu, Y. C. Chi, H. Y. Wang, et al., “Tricolor R/G/B Laser Diode Based Eye-Safe White Lighting Communication Beyond 8 Gbit/s,” Sci. Rep. 7(1), 11 (2017). [CrossRef]
12. Z. Xu, W. Niu, Y. Liu, et al., “31.38 Gb/s GaN-based LED array visible light communication system enhanced with V-pit and sidewall quantum well structure,” Opto-Electron. Sci. 2(5), 230005 (2023). [CrossRef]
13. W. Niu, Z. Xu, X. Lin, et al., “Key Technologies for High-Speed Si-Substrate LED Based Visible Light Communication,” J. Lightwave Technol. 41(11), 3316–3331 (2023). [CrossRef]
14. A. Yan, P. Luo, A. Sun, et al., “An Optoelectronic Joint Equalization Scheme Based on Photonic Reservoir Computing and TDNN for CD and Nonlinearity Compensation in IM/DD Link,” 2022 20th International Conference on Optical Communications and Networks (ICOCN), Shenzhen, China, 2022, pp. 1–3.
15. P. Luo, A. Yan, A. Sun, et al., “Study of Filter-based Neuromorphic Photonic Reservoir Computing for Signal Equalization in 224Gbps Sub-carrier Modulation IM-DD Short Reach Optical Fiber Communication System,” 2022 IEEE 7th Optoelectronics Global Conference (OGC), Shenzhen, China, 2022, pp. 41–44.
16. A. Sun, A. Yan, P. Luo, et al., “Silicon Photonic Integrated Reservoir Computing Processor with Ultra-high Tunability for High-speed IM/DD Equalization,” 2022 IEEE 7th Optoelectronics Global Conference (OGC), Shenzhen, China, 2022, pp. 227–230.
