Open Access
Add to BibSonomyAbstract
Optical background noise and second-order nonlinear distortions are two main challenges faced by indoor high-speed VLC system. In this paper, a novel phase shifted Manchester (PS-Manchester) coding based on PAM-8 is proposed and experimentally demonstrated to mitigate these noise and distortions. With the aid of PS-Manchester coding and WDM, a total data rate of 3.375-Gb/s can be successfully achieved in the RGB-LED based VLC system. The BER is under 7% HD-FEC limit of 3.8x10−3 after 1-m indoor free space transmission. To the best of our knowledge, this is the highest data rate ever achieved in PAM VLC systems.
© 2016 Optical Society of America
1. Introduction
Nowadays, light emitting diodes (LEDs) have been used in many places and considered to be a major candidate for future illumination. LED based visible light communication (VLC), which provides simultaneous illumination and communication, has become a promising and attracting technology [1]. VLC has been widely investigated due to its many advantages, such as license-free, high-speed, and immune to electromagnetic radiation and so on. The feasibility of VLC has been both demonstrated by utilizing blue LEDs in combination with phosphor, and red-green-blue (RGB) LEDs [2,3]. Compared with phosphor-based blue LEDs, RGB LEDs are more suitable for high-speed VLC systems, because three wavelengths can carry different data streams simultaneously. However, the limited bandwidth of LEDs is still one of the main challenges for high-speed VLC systems. To overcome the limitation, advanced high-order modulation techniques are widely investigated in [4–7]. Pulse amplitude modulation (PAM) has attracted great interest of researchers for its simpler structure, lower computational complexity and more flexible implementation. A 1.1-Gb/s white lighting LED-based visible light link with PAM-4 [6] and a 300-Mb/s VLC transmission with PAM-8 [7] have been successfully demonstrated by Grzegorz Stepniak.
On the other hand, the optical noise and distortions are another key challenge for high-speed VLC system. The noise mainly comes from the conventional fluorescent lamps, and sunlight. The second-order nonlinear distortions are introduced by the photodetector, which is a square-law detector [8], for the photon detection rate and signal current is proportional to the square of the electric field. The system performance is seriously deteriorated by the optical noise and nonlinear distortions, especially for PAM VLC system due to its vast direct current (DC) and low frequency (LF) components. In order to mitigate noise and distortions, controlling the spectral shape for modulated signal is generally utilized. By using specific coding schemes, the LF components of signal spectra can be reduced to suppress crosstalk noise and nonlinear distortions. Manchester coding, as a simple and easy coding scheme, has been widely investigated in optical wireless communication systems [9,10]. Besides, it has advantages of providing synchronization and improving the clock recovery. In [9], Alin-Mihai Cailean has presented the simulation and analysis results in the VLC system based on Manchester and the Miller coding. However, it lacks the relevant experimental verification. In [10], further experiments have been demonstrated to evaluate the performance of Manchester-coded optical wireless communication. But the transmission bit rate is only 2.5-Mb/s. Furthermore, the Manchester coding in this work just focuses on on-off keying (OOK) signal. However, for high-order PAM modulation, there is still absence of effective coding solution to eliminate optical background noise and nonlinear distortions.
In this paper, we propose a novel coding scheme for PAM modulation, called phase shifted Manchester (PS-Manchester) coding, to mitigate the optical noise and nonlinear distortions. By using opposite signals to odd and even symbols in coding process and subtraction operation in differential decoding process, the DC component, the second-order nonlinear distortion, and a part of optical background noise can be eliminated at the receiver. Furthermore, we experimentally demonstrate a high-speed RGB-LED based WDM VLC system employing PS-Manchester coding and scalar modified constant multi-modulus algorithm (S-MCMMA) based adaptive equalization [11]. An aggregate data rate of 3.375-Gb/s is successfully achieved in 1-m indoor free space transmission with the bit error rate (BER) under the 7% hard decision forward error correction (HD-FEC) limit of 3.8x10−3 [12]. As far as we know, this is the highest data rate ever achieved in PAM VLC systems at common indoor distance. The results clearly show the feasibility and benefit of PS-Manchester coding scheme for indoor high-speed VLC systems.
2. Principle
In this section, we will comprehensively discuss and analyze the encoding and decoding process of the PS-Manchester scheme. Figure 1 shows the schematic diagram of PS-Manchester coding and coded signal spectrum. Traditional Manchester coding assigns logic “1” to the signal transition from low to high, and logic “0” to the signal transition from high to low [13]. PS-Manchester coding, however, employs opposite signals to odd and even symbols. In the specific implementation process, the generated PAM signal can be divided into several blocks, and the second symbol is reverse from the first symbol in every block. Since encoded symbols will occur hopping in the duration of each clock, the LF component can be inhibited greatly and signal spectrum can be re-shaped. In the differential decoding process, the consequent odd and even symbols are separated into two sequences. Then the original sequence is subtracted by the opposite sequence to eliminate the DC component and second-order intermodulation distortion. Thus the clock signal can be easily recovered from the encoded data stream. Besides, a part of optical background noise can be reduced by the subtraction between adjacent symbols. However, double bandwidth is required as the expense, as shown in Fig. 1(b).
Another reason for spectrum shaping is because of the bias tee circuit in the transmitter. Generally the LED is a DC driven device. The signal is super-imposed onto the DC current by the bias tee so that the LED light intensity could be an information bearer. In our transmitter circuit the bias tee is a bandpass device, which will block the low-frequency component. Clearly for a single carrier signal with abundant low-frequency component, this band-pass feature will lead to signal distortion and result in failure detection. From this aspect, the input signal with DC-balanced pre-coding is preferred for our VLC system. And it should be noted that the spectral shaping technique is more efficient to mitigate distortion but at the cost of system complexity.
In this coding scheme, the signal can be divided into several blocks. The PAM signals in the kth block are given by
wherein, is the PAM symbol. And the PAM signals modulated on the LED light through bias-tee can be expressed aswhere and are respectively the bias voltage and reversal voltage of LED. The coefficient is used to describe the ratio of PAM band strength. is optical carrier frequency. After transmitting through the indoor environment, the received PAM signals can be approximated asThe is the Additive White Gaussian Noise (AWGN). After being received by the square-law detector, the photo current for the kth block can be approximately written as wherein, is the receiver current including signal current and noise current, is the noise induced by receiver, and is the optical background noise current including AWGN noise current and receiver noise current. In Eq. (5), the first term is DC component, the second term is signal component, the third term is second-order intermodulation distortion between signals, and the fourth term is noise current as shown in Eq. (6). It can be observed that the amount of the second-order term coefficient with respect to the linear (first-order) term coefficient is 1/2. In Eq. (6), the first term is the noise between AWGN to AWGN, and the second term is the noise between signals to AWGN. The photo current of 2nd in the kth block expressed in Eq. (7) is similar to Eq. (5). By conducting subtraction operation between Eq. (5) and Eq. (7), we can obtain the photo current in kth block as:where the first term is the signal and the second term is the noise. It can be observed from Eq. (8) that the DC component and the second-order term related to intermodulation distortion can be greatly eliminated. And the sensitivity of receiver can be improved by 3dB. In addition, a part of optical background noise can be mitigated by subtraction between adjacent symbols.Furthermore, we simulate the PAM-8 signal with/without PS-Manchester coding scheme. It can be observed from Fig. 1(b), compared with general PAM-8 signals, the bandwidth of PS-Manchester encoded signals is almost increased doubly, from 400MHz to about 800MHz. However, the LF component of coded spectrum is mitigated greatly, while its high-frequency (HF) component is preserved as much as possible.
The simulation results are shown in Fig. 2. Firstly, we fit a negative exponential decay curve according to the attenuation characteristic of the VLC channel, shown in Fig. 2(a). In particular, the DC and LF components are greatly attenuated in our fitting curve since bias tee used in our experiments cannot pass the DC component of signals. Through VLC channel added with AWGN (SNR = 20dB), the received spectra for PAM-8 signals with/without PS-Manchester coding are portrayed in Figs. 2(b) and 2(c), respectively. Since the PAM signals have vast DC and LF components, the attenuations seriously deteriorate the performance of received signal, which induces the failure of decoding original bit. Therefore, the generated PAM signals are usually up-converted to the sub-carrier for transmission in previous experiments [14], which greatly increases the system complexity. However, the DC and LF component can be inhibited effectively by using PS-Manchester coding, as shown in Fig. 2(c). Moreover, the performance of PS-Manchester encoded PAM-8 signals without/with differential decoding is investigated and demonstrated in Figs. 2(d) and 2(e), respectively. Compared with direct decoding, spectra of differential decoding is more flat because a part of background noise can be mitigated by subtraction. Therefore, the combination of PS-Manchester coding and differential decoding scheme can obtain the optimal performance, and it can be testified by the constellations presented in Fig. 2(f).
Fig. 2 Simulation results: (a) VLC channel curve; received spectrum for PAM-8 signals: (b) without coding; (c) with PS-Manchester coding; (d) with PS-Manchester coding but without differential decoding; (e) with PS-Manchester coding and differential decoding; (f) constellations for PAM-8 signals: (i) without coding; (ii) with PS-Manchester coding but without differential decoding; (iii) with PS-Manchester coding and differential decoding.
Besides, a post equalizer is also utilized at the receiver to mitigate inter-symbol interference (ISI). We utilize a novel blind equalizer suitable for PAM systems, called scalar modified cascaded multi-modulus algorithm (S-MCMMA) [11]. Different from traditional equalization algorithms, such as, least mean square (LMS) [13], and recursive least square (RLS) [14], the error function can be calculated without training sequences. With the aid of post-equalization, the ISI is effectively reduced, so the system performance is improved.
3. Experimental setup
Figure 3 shows the experimental setup of the RGB-LED based WDM VLC system with PAM-8 modulation, employing PS-Manchester coding and S-MCMMA equalization. At the transmitter, the original bit sequence is firstly generated in MATLAB and mapped into PAM-8 symbols. Then the PAM symbols are up-sampled by 2, and converted to two serial sequences through serial-to-parallel conversion. The second symbol sequence is opposite from the first one and delayed by half one-bit time. Followed by parallel-to-serial conversion to merge two sets of sequences, the coded signals are up-sampled with a factor of 4 to generate the transmitted signals.
Fig. 3 The experimental setup of WDM VLC system employing PS-Manchester coding and S-MCMMA equalization.
In the experiment, the PAM data is fed into an arbitrary waveform generator (AWG, Tektronix AWG710) to generate signals for the three color chips of the RGB-LED (LZ4-00MA00). The maximum output power of the LED is 10W, and the wavelengths of red, green, and blue chips are 621nm, 525nm and 460nm, respectively. Then, each generated PAM signal is pre-amplified by a self-designed bridged-T based pre-equalizer, to compensate for the attenuation at HF components [15]. The bandwidth of the pre-equalizer is 375MHz. After amplified by a mini-circuits electrical amplifier (EA) with 25-dB gain, the electrical signal and DC-bias voltage are combined by a bias tee and applied to the RGB-LED. In order to decrease the beam angle of the LED and achieve longer transmission distance, a reflection cup with 60° divergence angle is applied to the RGB-LED. Due to the limited experimental conditions that AWG 710 only has two independent outputs, the output of channel 1 is used for the red chip, while the output of channel 2 and its inverted copy are used for green and blue chips, respectively.
At the receiver, a commercially available PIN photodiode is used as the receiver. It is equipped with a lens (70-mm diameter and 100-mm focus length) to focus light and optically filter the background noise from other light sources than the communication source itself [16]. Then the optical R/G/B filters are employed to filter out different colors. The received signal is amplified by an EA and recorded by a digital storage oscilloscope (Agilent DSO54855A) for further offline signal processing. In offline signal processing, the received signal is firstly normalized with the theoretical average power determined by the level of encoded symbols, and down-sampled with a factor of 4. Then the S-MCMMA based adaptive post-equalization is utilized to eliminate the ISI and compensate for the channel loss. Finally, the data is decoded and demodulated to obtain the original bit sequence. In the differential decoding processing, we conduct serial-to-parallel conversion to separate the original and opposite symbol sequences. Then the original sequence is subtracted by the opposite sequence, which is delayed by half one-bit time, to eliminate the noise and distortion.
4. Experimental results and discussions
Firstly, the transmitted and received signal spectra are measured by spectrum analyzer, and the decoded signal spectra are analyzed in offline processing, as shown in Fig. 4. For simplicity, we just measure red LED chip and set the transmitted bandwidth as 800MHz and 400MHz for coded and original PAM-8 signals, respectively. It can be found that the received original PAM-8 signal spectrum has a large DC component induced by square-law detector so that it suffers serious distortion in LF component. Therefore, the original bit sequence without coding cannot be recovered directly. For the PAM-8 signals with PS-Manchester coding, the cut-off frequency of spectrum is about 500MHz, which is limited by receiver bandwidth. However, we can still recover signals because the effective data only occupies one-half of total data. Namely, the effective signal bandwidth is 400MHz. In offline processing, the spectra of coded signals with/without differential decoding scheme are also depicted. Compared with signals without differential decoding, the DC component and LF noise for differential decoded signals can be eliminated greatly. The constellations inserted in Figs. 4 (b), 4(e) and 4(f), clearly demonstrate that the signals with PS-Manchester coding and differential decoding can obtain the optimal performance, while the signals without coding completely cannot be recovered.
Fig. 4 Measured experimental results: PAM-8 signals without coding: (a) transmitted spectra; (b) received spectra; PAM-8 signals with PS-Manchester coding: (c) transmitted spectra; (d) received spectra; (e) received spectra without differential decoding; (f) received spectra with differential decoding.
It should be noted that, in the receiver we use the DC coupling and an integrated trans-impedance amplifier (TIA) chip to realize low noise small-signal pre-amplification. The PIN has to be set with an extremely large reverse bias voltage of 12V to achieve large bandwidth. The tradeoff is that from one hand the high bandwidth is obtained, from the other hand the common mode cancellation is no longer effective. That is the reason why the received signal has some common mode noise and the noise can be efficiently eliminated by phase-shift Manchester coding.
In previous works, some simple-designed circuits are used to reduce received noise at the receiver. In [17], two external capacitors of about 10 nF are used in the cancellation of received background DC noise. In [18], the optical receiver including integrated TIA chips can provide the function of DC cancellation, but a specific designed circuit is needed. These techniques can only eliminate DC noise, but not AC noise. Therefore, besides aforesaid hardware circuits, some digital signal processing (DSP) techniques should be considered to mitigate AC noise and distortions. One of the efficient idea is to using specific coding scheme. In this way the signal spectrum can be re-shaped to achieve a better fitting with the bias tee circuit. Moreover, the coding process introduces some connections between consecutive data sequence. The common mode noise can be diminished during differential decoding process. Thus, the PS-Manchester coding scheme can greatly improve performance in our system.
Besides, we investigate the influence of different bias voltages and input signal peak-to-peak value (Vpp) to render the RGB-LED work at the optimal condition. In this demonstration, the bandwidth of signal fed to red, green and blue chip is respectively set as 850MHz, 600MHz and 650MHz, according to the quality of transmission channel. And the distance between transmitter and receiver is fixed as 1-m. The measured results and constellations related to optimal working point are shown in Fig. 5. It can be found that the optimal working point of the red, green and blue chip is at (2.1V bias voltage, 0.3V input signal Vpp), (3.7V bias voltage, 0.35V input signal Vpp), and (3.6V bias voltage, 0.35V input signal Vpp), respectively. At the optimum point, all the BER of the three color chips can below the 7% FEC limit of 3.8x10−3. Besides, we can observe that red chip can work in the maximum region with BER under FEC threshold, and then green chip, while the blue chip has the minimum operation region.
Fig. 5 Measured BER versus bias voltages and input signal Vpp: (a) red chip; (b)green chip; (c) blue chip.
At the optimal work condition, we measure the BER performance versus different bitrates to investigate the highest transmitted bitrate achieved by three color chips, as shown in Fig. 6. The transmission distance is fixed at 1-m in this demonstration. With the aid of PS-Manchester coding, differential decoding, and S-MCMMA adaptive post-equalization, the aggregative bitrate of 3.375-Gb/s (red: 1.35-Gb/s; green: 0.975-Gb/s; blue: 1.05-Gb/s) is successfully achieved. We also demonstrate the BER performance for signals without differential decoding at the same condition, and the total bitrate of 1.8-Gb/s is obtained (the transmission bitrate for each color chip is 600-Mb/s). Therefore, the highest bitrate is increased by 1.575-Gb/s by utilizing differential decoding. In addition, the eye diagrams of three color chips at the bitrate of 1.05-Gb/s are inserted in Fig. 6. It shows that the performance of red chip is much better than green and blue chips at the same bitrate.
Fig. 6 Measured BER performance versus different bitrates: (a) red chip; (b) green chip; (c) blue chip; (d) eye diagrams.
The BER performances versus different distances employing PS-Manchester coding and differential decoding are also presented in Fig. 7. We measure the BER performance when the transmission distances vary from 1-m to 2-m with a step of 0.25-m. It can be observed that the aggregate data rate of 3.375-Gb/s, 3.225-Gb/s and 2.925-Gb/s are successfully obtained at the distances of 1-m, 1.5-m and 2-m, respectively. Thus, the results clearly demonstrate the feasibility and benefit of PAM VLC system based on PS-Manchester coding scheme for common indoor communication. The coding scheme requires extra computational resources at the transmitter and the receiver, but the complexity of the hardware circuits at the receiver is reduced. The phase-shift Manchester coding provides benefit for the system bandwidth improvement and the sensitivity enhancement.
Fig. 7 Measured BER performance versus different distances: (a) red chip; (b)green chip; (c) blue chip.
It should be noted that the luminance of the LED is one of key factors that limit the transmission distance in VLC system. In our experiment, the output optical power of the RGB LED is below 2.5W. And the luminance of different color chips at 1-m is measured as: red 129lux, green 210lux, blue 15lux. The measured results show that the luminance needed at the receiver is far below the standard value for brightness (500lux). The luminance decreasing with distance will reduce the system transmission capacity. Thus, the system performance and transmission distance can be further improved by deploying several LEDs together to increase luminance.
On the other hand, it is worth noting that the nonlinearity behavior seriously deteriorates the system performance and reduces the transmission distance. Hence a nonlinear post-equalizer at the receiver is also an effective method to compensate nonlinear distortion. A Volterra series based nonlinear equalizer is proposed and successfully demonstrated in carrierless amplitude and phase (CAP) based VLC system [19]. Similar scheme could be applied in PAM modulated VLC system to obtain additional improvement on the receiver sensitivity. The performance of the nonlinear post-equalization will be investigated in our future work.
5. Conclusion
In this paper, for the first time we propose a PS-Manchester coding scheme with differential decoding to mitigate the optical background noise and nonlinear distortions. By employing opposite signals to odd and even PAM symbols at the transmitter and differential subtraction operation at the receiver, the noise and distortion can be eliminated efficiently. Furthermore, we experimentally demonstrate a high-speed RGB-LED based WDM VLC system employing PAM-8 modulation. With the aid of PS-Manchester coding scheme and S-MCMMA equalization, a data rate of 3.375-Gbt/s for PAM-8 is successfully achieved in 1-m indoor free space transmission with the BER below 3.8x10−3. Moreover, we measure the transmission bitrates versus different distances, and compare the BER performance of coded signals with/without differential decoding. The measured results indicate that the bitrate can be increased by 1.575-Gb/s through using differential decoding. Thus, the investigated results clearly show the feasibility and benefit of PS-Manchester coding scheme for indoor high-speed VLC systems with PAM modulation.
Funding
National Natural Science Foundation of China (NSFC) (61571133); National “863” Program of China (2015AA016904), and State Grid Corporation of China (SGCC): Research on Key Technologies of High Reliability and Short Distance Wireless Communication in Power Complex Electromagnetic Environment.
Acknowledgments
We gratefully acknowledge the invaluable contribution of Yuanquan Wang, Yiguang Wang and Xinying Li in providing constructive suggestions and revising our manuscript.
References and links
1. D. O’Brien, H. L. Minh, L. Zeng, G. Faulkner, K. Lee, D. Jung, Y. Oh, and E. T. Won, “Indoor visible light communications: challenges and prospects,” Proc. SPIE 7091, 709106 (2008). [CrossRef]
2. Y. Wang, N. Chi, Y. Wang, R. Li, X. Huang, C. Yang, and Z. Zhang, “High-speed quasi-balanced detection OFDM in visible light communication,” Opt. Express 21(23), 27558–27564 (2013). [CrossRef] [PubMed]
3. Y. Wang, X. Huang, L. Tao, J. Shi, and N. Chi, “4.5-Gb/s RGB-LED based WDM visible light communication system employing CAP modulation and RLS based adaptive equalization,” Opt. Express 23(10), 13626–13633 (2015). [CrossRef] [PubMed]
4. Y. Wang, Y. Wang, N. Chi, J. Yu, and H. Shang, “Demonstration of 575-Mb/s downlink and 225-Mb/s uplink bi-directional SCM-WDM visible light communication using RGB LED and phosphor-based LED,” Opt. Express 21(1), 1203–1208 (2013). [CrossRef] [PubMed]
5. G. Cossu, A. Wajahat, R. Corsini, and E. Ciaramella, “5.6Gbit/s downlink and 1.5Gbit/s uplink optical wireless transmission at indoor distances (≥1.5m),” in European Conference on Optical Communication (ECOC), Amsterdam, Netherlands, 2014, We.3.6.4.
6. G. Stepniak, L. Maksymiuk, and J. Siuzdak, “1.1Gbit/s white lighting LED-based visible light link with pulse amplitude modulation and volterra DFE equalization,” Microw. Opt. Technol. Lett. 57(7), 1620–1622 (2015). [CrossRef]
7. G. Stepniak, J. Siuzdak, and P. Zwierko, “Compensation of a VLC phosphorescent white LED nonlinearity by means of Volterra DFE,” IEEE Photonics Technol. Lett. 25(16), 1597–1600 (2013). [CrossRef]
8. W. Martin, “Photodetector responsivity” in Computer Science and Communications Dictionary (Academic, 2001), pp. 1268.
9. A. M. Cailean, B. Cagneau, L. Chassagne, V. Popa, and M. Dimian, “Evaluation of the noise effects on Visible Light Communications using Manchester and Miller coding,” in IEEE International Conference on Development and Application Systems (DAS, 2014), pp. 85–89. [CrossRef]
10. C. Chow, C. Yeh, Y. Liu, and P. Huang, “‘Mitigation of optical background noise in light-emitting diode (LED) optical wireless communication systems,” IEEE Photonics J. 5(1), 7900307 (2013). [CrossRef]
11. M. Zhang, Y. Wang, Z. Wang, J. Zhao, and N. Chi, “A novel scalar MCMMA blind equalization utilized in 8-PAM LED based visible light communication system,” presented at IEEE International Conference on Communications, Kuala Lumpur, Malaysia, 23–27 May, 2016.
12. A. Leven, F. Vacondio, L. Schmalen, S. Brink, and W. Idler, “Estimation of soft FEC performance in optical transmission experiments,” IEEE Photonics Technol. Lett. 23(20), 1547–1549 (2011). [CrossRef]
13. K. Zhong, X. Zhou, T. Gui, L. Tao, Y. Gao, W. Chen, J. Man, L. Zeng, A. P. T. Lau, and C. Lu, “Experimental study of PAM-4, CAP-16, and DMT for 100 Gb/s short reach optical transmission systems,” Opt. Express 23(2), 1176–1189 (2015). [CrossRef] [PubMed]
14. G. Stepniak, L. Maksymiuk, and J. Siuzdak, “Experimental comparison of PAM, CAP and DMT modulations in phosphorescent white LED transmission link,” IEEE Photonics J. 7(3), 7901708 (2015). [CrossRef]
15. X. Huang, J. Shi, J. Li, Y. Wang, Y. Wang, and N. Chi, “750Mbit/s Visible Light Communications employing 64QAM-OFDM Based on Amplitude Equalization Circuit,” in Optical Fiber Communication Conference, OSA Technical Digest (online) (Optical Society of America, 2015), paper Tu2G.1.1385–1391 (2003). [CrossRef]
16. I. Takai, S. Ito, K. Yasutomi, K. Kagawa, M. Andoh, and S. Kawahito, “LED and CMOS image sensor based optical wireless communication system for automotive applications,” IEEE Photonics J. 5(5), 6801418 (2013). [CrossRef]
17. J. S. Youn, M. J. Lee, K. Y. Park, and W. Y. Choi, “10-Gb/s 850-nm CMOS OEIC receiver with a silicon avalanche photodetector,” IEEE J. Quantum. Elec. J. 48(2), 229–236 (2012). [CrossRef]
18. B. Analui, D. Guckenberger, D. Kucharski, and A. Narasimha, “A fully integrated 20-gb/s optoelectronic transceiver implemented in a standard 0.13-cmos soi technology,” IEEE J. Solid-St Circ 41(12), 2945–2955 (2006). [CrossRef]
19. Y. Wang, Y. Wang, N. Chi, J. Yu, and H. Shang, “Demonstration of 575-Mb/s downlink and 225-Mb/s uplink bi-directional SCM-WDM visible light communication using RGB LED and phosphor-based LED,” Opt. Express 21(1), 1203–1208 (2013). [CrossRef] [PubMed]
