1. INTRODUCTION

Visible micro light-emitting diodes ($\mathrm{\mu }\text{-}\mathrm{LEDs}$) based on III-nitride materials have emerged since 2000 and were successively proposed for versatile applications, including micro-display image processing, biomedical microscopic sensing and detection, optical communications, and so on [1]. In particular, the urgent demand for lighting fidelity within hospital facilities has recently emerged due to the unexpected pandemic situation, which relies strictly on using the optical wireless link as a supplementary candidate for the current wireless fidelity. Gallium nitride (GaN)-based $\mathrm{\mu }\text{-}\mathrm{LED}$ arrays with an elementary size of 20 µm and pitch of 30 µm were reported in 2003, with adequate emission power and minimal turn-on voltage [2]. Other than the fact that the current flow across the $\mathrm{\mu }\text{-}\mathrm{LED}$ array was measured as low as $5-10\mathrm{mA}$ under a direct current (DC) bias of 8 V at the time, no other characteristics were revealed until the realization of the indium GaN (InGaN) $\mathrm{\mu }\text{-}\mathrm{LED}$ array with blue, green, and ultraviolet colors (370–510 nm) in 2007 [3]. In 2009, the first demonstration of the ultraviolet AlInGaN $8\times 8\mathrm{\mu }\text{-}\mathrm{LED}$ array integrated with complementary metal–oxide-semiconductor (CMOS) circuits for square and pulsed data processing was reported for controllable optical pulse width at 0.3–40 ns [4]. Later on, the direct digital encoding of the III-nitride-based $16\times 16\mathrm{\mu }\text{-}\mathrm{LED}$ array with 72 µm diameter at wavelengths of 370–450 nm was reported with a maximal analog bandwidth of 245 MHz and potential digital data rate approaching 1 Gbit/s at 450 nm [5]. The permissible modulated bandwidth and its correlation with the emission aperture size of an $8\times 8\mathrm{\mu }\text{-}\mathrm{LED}$ array at 370–520 nm were characterized in 2012, given the significant bandwidth degradation of 1/2 when the aperture size was four times larger [6]. The green $\mathrm{\mu }\text{-}\mathrm{LED}$ array, which has an aperture size of 50 µm and a $-3\mathrm{dB}$ bandwidth of just 325 MHz, is among the colored lighting LED arrays that have been reported at an early stage. As most commercially available GaN LEDs are grown on “polar” $c$-plane sapphire substrates with (0001)-orientation, the polar $c$-plane sapphire substrate often results in a significant quantum-confined Stark effect (QCSE) to reduce emission efficiency under high injected current densities. Moreover, the wave functions of the electron and hole will be spatially isolated to cause a low carrier radiative recombination rate and internal quantum efficiency. However, the semipolar GaN surface exhibits a lower repulsive interaction with indium atoms than both polar and nonpolar surfaces. As the chemical potential needed for incorporating indium atoms into the semipolar plane is reduced significantly, semipolar devices can provide excellent output performance with a high modulation bandwidth and long emission wavelength. Johar et al. synthesized the (112-2)-oriented semipolar GaN/InGaN multiple quantum wells (MQWs) with tunable emission wavelength between 430 nm and 590 nm via changing the InGaN QW thickness [7]. In addition, the radiative lifetime can be detuned between 14 ps and 26 ps [7]. Parbrook et al. fabricated the semipolar $\mathrm{\mu }\text{-}\mathrm{LED}$ with its modulation bandwidth of 800 MHz to perform 1.5 Gbit/s non-return-to-zero on–off keying (NRZ-OOK) transmission [8]. Moreover, Haggar and co-workers utilized a green semipolar $\mathrm{\mu }\text{-}\mathrm{LED}$ to demonstrate the device modulation bandwidth to 540 MHz [9].

Recently, the combination of plastic optical fiber (POF) and $\mathrm{\mu }\text{-}\mathrm{LED}$ array has also emerged as an alternative approach for implementing a guided-wave visible light communication (VLC) link. For a single-directional data transmission record of 6.25 Gbit/s at 450 nm over 10 m of POF, advanced data formats such as pulse amplitude modulation (PAM) and quadrature amplitude modulation (QAM) data formats have been used [10]. A more advanced data format such as orthogonal frequency-division multiplexing (OFDM) with its QAM level ranging from four to 256 has also been demonstrated on the $\mathrm{\mu }\text{-}\mathrm{LED}$ array to examine its improved bandwidth despite substantial nonlinear modulation distortion [11]. A $\mathrm{\mu }\text{-}\mathrm{LED}$ with a size of $40\mathrm{\mu m}\times 40\mathrm{\mu m}$ and $-3\mathrm{dB}$ bandwidth of 230 MHz was directly encoded to provide 1.3 Gbit/s over 3 m in free space and 0.87 Gbit/s with transmission distance lengthening to 16 m [12]. As there is a trade-off set between the aperture size and modulation bandwidth of the $\mathrm{\mu }\text{-}\mathrm{LED}$, various subsequent experimental results have emphasized the optimized design of the $\mathrm{\mu }\text{-}\mathrm{LED}$ element or array to improve its communication performance. By using a $1\times 3\mathrm{\mu }\text{-}\mathrm{LED}$ array with 340 µm aperture size to obtain 18 mW power and 285 MHz bandwidth, the highest data rate under OOK, PAM, and QAM-OFDM data formats can, respectively, approach 2, 2.4, and 4.8 Gbit/s under a bias current density as high as $3\mathrm{kA}/{\mathrm{cm}}^2$ [13]. In contrast, a similar experiment demonstrated on a $14\mathrm{\mu m}2\times 5\mathrm{\mu }\text{-}\mathrm{LED}$ array with a current density as high as $13\mathrm{kA}/{\mathrm{cm}}^2$ allows only 1 Gbit/s OOK data transmission [14]. Rashidi et al. observed that the nonpolar InGaN/GaN $\mathrm{\mu }\text{-}\mathrm{LED}$ exhibits a 3 dB modulation bandwidth of 1.5 GHz under $1\mathrm{kA}/{\mathrm{cm}}^2$ operation [15]. In addition, Dinh and co-workers utilized the semipolar InGaN/GaN $\mathrm{\mu }\text{-}\mathrm{LED}$ with 1 GHz bandwidth to demonstrate back-to-back (BtB) data transmission at 2.4 Gbit/s [16]. Wun et al. further demonstrated 1.07 Gbit/s communication with a cyan $\mathrm{\mu }\text{-}\mathrm{LED}$ through 50 m step-index POF [17]. Nevertheless, the droop effect of the $\mathrm{\mu }\text{-}\mathrm{LED}$ inevitably causes a degradation in output power, signal-to-noise ratio (SNR), and data rate under high-bias and long-term operation [18]. This phenomenon essentially limits the high-power operation of the $\mathrm{\mu }\text{-}\mathrm{LED}$ array when demanding a high SNR during the data transmission [19,20]. Even though the dual-color or tri-color $\mathrm{\mu }\text{-}\mathrm{LED}$ array has recently been extensively studied to enable wavelength multiplexing communication [21,22], very few studies have focused on the data transmission performance of colored $\mathrm{\mu }\text{-}\mathrm{LED}$ components other than blue or ultraviolet ones. Because of its lower quantum efficiency and stronger drooping effect, the communication performance of the green $\mathrm{\mu }\text{-}\mathrm{LED}$ array has received less attention than that of other colored $\mathrm{\mu }\text{-}\mathrm{LED}$ arrays in previous studies. Particularly, the data communication efficiency of the $\mathrm{\mu }\text{-}\mathrm{LED}$ array should be taken into account for improving the energy/bit ratio under low power consumption.

For comparison, Table 1 summarizes the results of studies on the green $\mathrm{\mu }\text{-}\mathrm{LED}$ showing data rates in terms of NRZ-OOK and OFDM [23–28]. Figure 1 summarizes the benchmark of data rate versus current density for green $\mathrm{\mu }\text{-}\mathrm{LEDs}$. In 2012, McKendry et al. introduced a green $\mathrm{\mu }\text{-}\mathrm{LED}$ with a peak wavelength of 525 nm and diameter of 34 µm showing a data rate of 1.1 Gbit/s under $2.2{\mathrm{kA}/\mathrm{cm}}^2$ [23]. In 2018, Carreira et al. reported that an LED with a peak wavelength of 505 nm and 20 µm diameter obtained a data rate of 613 Mbit/s under $3.5{\mathrm{kA}/\mathrm{cm}}^2$ [24]. In the same year, Chen et al. developed a 520 nm green GaN-based LED, which achieved a data rate of 2.16 Gbit/s using OFDM modulation at a low current density of $679\mathrm{A}/{\mathrm{cm}}^2$ [25]. In 2020, Chen et al. demonstrated a long-wavelength (525 nm) semipolar $\mathrm{\mu }\text{-}\mathrm{LED}$ with a diameter of 50 μm achieving a data rate of 1.5 Gbit/s under $2{\mathrm{kA}/\mathrm{cm}}^2$ [26].

Table 1. Overview of Data Rate Performance in Terms of NRZ-OOK and OFDM for Green LED

View Table | View all tables in this article

 

Fig. 1. Benchmark of data rate versus current density for green-light LED.

Download Full Size | PDF

In 2021, Lin et al. demonstrated a green $\mathrm{\mu }\text{-}\mathrm{LED}$ in an underwater optical communication (UWOC) system and achieved a 660 Mbit/s data rate [27]. In the same year, Liu et al. reported a green mini-LED with a high data rate of 2.65 Gbit/s using OFDM under $357\mathrm{A}/{\mathrm{cm}}^2$ [28]. To date, this study presents the highest OFDM data rate of 5.02 Gbit/s achieved by a long-wavelength GaN-based $\mathrm{\mu }\text{-}\mathrm{LED}$. In comparison with previous works on the $\mathrm{\mu }\text{-}\mathrm{LED}$ array [26,29], the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array in this work exhibits operation of the lower current density to effectively decrease the power consumption in the transmission system. Moreover, the green $\mathrm{\mu }\text{-}\mathrm{LED}$ array in this work is directly packaged on the commercial transistor-outline can (TO-can) to demonstrate optical wireless communication. Therefore, this TO-can packaged green $\mathrm{\mu }\text{-}\mathrm{LED}$ array can be regarded as one of the candidates for the commercial transmitter of the VLC network in the future.

In this work, a green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array with an elementary aperture diameter of 50 µm is designed and fabricated for free-space visible-light optical wireless communication. The InGaN-based semipolar green $\mathrm{\mu }\text{-}\mathrm{LED}$ array is fabricated by employing nanostructured grating patterns, which improves the polarization contrast ratio. The linear power-to-current ($P-I$) and current-to-voltage ($I-V$) performances of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array under continuous-wave operation are measured, and their differential curves of $\mathrm{d}{P}_{\mathrm{out}}/\mathrm{d}{I}_{\mathrm{bias}}$ and $\mathrm{d}V/\mathrm{d}I$ are analyzed to understand the optimized modulation parameters. By employing different data formats for directly encoding the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array packed on the TO-can sub-mount module with a commercial sub-miniature-A (SMA) connector, the free-space back-to-back optical wireless transmission performance of OOK data streams without or with pre-compensation is discussed in detail. The QAM-OFDM data formats with and without the pre-emphasis algorithm are used to improve the spectral usage efficiency within the limited modulation bandwidth to fulfill the demand of network coverage with other wireless electrical networks. The short-reach access of the digitally encoded green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array reveals allowable data rates beyond 1.5 Gbit/s for the NRZ-OOK format and beyond 5 Gbit/s for the bit-loaded discrete multitone (BL-DMT) format. Such improved encoding and transmission bit rates are attributed not only to the size and electrode shrinkage of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array but also to a specifically designed TO-can+SMA connector module. These optimized designs allow the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array to have a minimal amount of resistance-inductance-capacitance (RLC) time added to its analog modulation response for improved communication performance.

2. EXPERIMENTAL SETUP

A. Design, Fabrication, and High-Speed Package of Green $2\times 2$ μ-LED Array with Emission Aperture Size of 50 μm

For the design and fabrication of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array, the material synthesis for the green LED structure is introduced at the beginning. A semipolar (20-21)-oriented GaN layer was synthesized on a patterned sapphire substrate with a surface normal orientation of (22-43) using the metalorganic chemical vapor deposition (MOCVD) method. Because the angle between the (22-43)-oriented sapphire and the $c$-plane sapphire is 74.64° and the angle between the (20-21)-oriented GaN and the $c$-plane GaN is 75.09°, the (20-21)-oriented GaN surface can be made to match the (22-43)-oriented sapphire substrate surface by adjusting the sapphire substrate offset angle by 0.45°. The surface patterning of the patterned (22-43)-oriented sapphire substrate is done using lithography and inductively coupled plasma reactive ion etching (ICP-RIE) techniques to meet this demand. The widths of trench and land are, respectively, designed as 3 µm and 6 µm, and the depth of the gap between trench and land is etched as deep as 1 µm. The silicon oxide layer was deposited on the surface of the substrate using a self-aligning angular evaporation technology to establish the sidewalls of the specific $c$-plane sapphire substrate. Thereafter, the semipolar GaN film with (20-21)-orientation can be grown on the patterned sapphire substrate. The Ge dopant was added early in the epitaxy process to prevent the development of stacking faults (SFs) during GaN growth. After the adjacent Ge-doped GaN stripes are merged with each other, an 8 µm thick undoped GaN is grown to form an approximately 10 µm thick GaN layer. Before the epitaxy of the green $\mathrm{\mu }\text{-}\mathrm{LED}$ structure, the wafer was taken out of the MOCVD chamber for surface smoothening by using the chemical mechanical polishing method, which assists the planarization of a smoothened (20-21)-oriented GaN surface for the InGaN/GaN LED growth. The green $\mathrm{\mu }\text{-}\mathrm{LED}$ structure was composed of 1.5 μm thick n-type GaN as the bottom layer, three pairs of InGaN/GaN MQWs as the active layer, and 100 nm thick p-type GaN as the top layer. With a $2\times 2$ array design, such a green $\mathrm{\mu }\text{-}\mathrm{LED}$ array is particularly developed with the intent of increasing the optical output power, which is beneficial to extend the transmission distance of optical wireless communication in free space. The $\mathrm{\mu }\text{-}\mathrm{LED}$ contact was created by depositing a transparent conductive layer of 200 nm thick indium tin oxide (ITO) over the $\mathrm{p}\text{-}\mathrm{GaN}$ layer, etching the transparent conductive layer with hydrogen chloride (HCl), and then dry-etching the mesa with a depth of roughly 1 µm using ICP-RIE. Following contact patterning, the entire sample was put in a nitrogen environment for quick thermal annealing at 450°C to generate an ohmic connection between $\mathrm{p}\text{-}\mathrm{GaN}$ and ITO. To reduce the impact of sidewall defects caused by dry etching, atomic-layer-deposition (ALD) technology was introduced to deposit a sidewall passivation layer, as formed by the 30 nm thick aluminum oxide (${\mathrm{Al}}_2{\mathrm{O}}_3$) deposited in an argon atmosphere at 300°C via the circulation of trimethylaluminum (TMA) and oxidane (${\mathrm{H}}_2\mathrm{O}$). Then, using plasma-enhanced chemical vapor deposition (PECVD), a 200 nm thick silicon dioxide (${\mathrm{SiO}}_2$) layer was created, and the entire passivation layer was dry-etched using ICP-RIE to open the contact hole. Finally, to complete the configuration of the $\mathrm{\mu }\text{-}\mathrm{LED}$ fabrication process, Ti, Al, Ni, and Au metallic films with corresponding thicknesses of 20, 150, 10, and 100 nm were deposited to create the contact electrode, which also serves as the ground–signal–ground bonding pad metal. The schematic of the $\mathrm{\mu }\text{-}\mathrm{LED}$ sample is shown in Fig. 2(a) with a detailed illustration of the device with nanostructured grating patterns on top of the $\mathrm{\mu }\text{-}\mathrm{LED}$ structure shown in Fig. 2(b). The top-view microscope images of the $\mathrm{\mu }\text{-}\mathrm{LED}$ and its emission pattern after lighting are shown in Figs. 2(c) and 2(d). In this work, all elements in the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array with an aperture diameter of 50 μm are connected in parallel.

 

Fig. 2. (a) Schematic of $\mathrm{\mu }\text{-}\mathrm{LED}$ sample. (b) Illustration of a device structure with nanostructured grating patterns on the top of $\mathrm{\mu }\text{-}\mathrm{LED}$ structure. (c) Optical image of the device before illumination. (d) Lighting of the $\mathrm{\mu }\text{-}\mathrm{LED}$. (e), (f) Nanostructured grating patterns on $\mathrm{\mu }\text{-}\mathrm{LED}$.

Download Full Size | PDF

The current fabrication technology benefits from the advantage that semipolar $\mathrm{\mu }\text{-}\mathrm{LEDs}$ have higher polarization ratios (PRs) when compared with the same $\mathrm{\mu }\text{-}\mathrm{LEDs}$ grown on a $c$-plane sapphire substrate. Because of the optical transition between the conduction band and the highest valence band in InGaN-based semipolar $\mathrm{\mu }\text{-}\mathrm{LEDs}$, a longer wavelength (i.e., under the higher indium content) shows a greater PR. In previous research, semipolar $\mathrm{\mu }\text{-}\mathrm{LEDs}$ by combining an ITO grating with Al-coated surface plasmons (SPs) have advantages of better electrical properties. Also, the periodic structures exhibited a significant enhancement of PR [30]. The PR can be obtained by $\rho =(I_{{\chi }^{\prime }}-I_{y^{\prime }})/(I_{{\chi }^{\prime }}+I_{y^{\prime }})$, where $I_{{\chi }^{\prime }}$ and $I_{y^{\prime }}$ are the maximal and minimal integrated intensities of the emission spectrum along [1–210] and [10-1-4] orientations, respectively, when the polarizer is aligned along the (20–21)-oriented $\mathrm{\mu }\text{-}\mathrm{LED}$. Therefore, in this semipolar $\mathrm{\mu }\text{-}\mathrm{LED}$, the ITO grating structure with an Al-coated surface was additionally fabricated upon the semipolar $\mathrm{\mu }\text{-}\mathrm{LED}$ device using e-beam lithography, which can be seen in Fig. 2(b). The pitch (width and spacing) of the ITO grating was set as 200 nm, as shown in Figs. 2(e) and 2(f). Unpolarized light from LEDs always suffers from a significant power loss due to absorption and reflection effects. The nanostructured grating patterns will not cause optical losses but will improve the output PR to result in better-polarized light emission.

With the optimal design through simulations, Fig. 3 reveals that the semipolar green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ enhances its output degree of polarization (DOP) by 0.71 to favor liquid-crystal display applications. Because of the dual-color SPs (DSPs) induced by the nanostructured grating patterns, the output power of the nanostructured grating-patterns-added green $\mathrm{\mu }\text{-}\mathrm{LED}$ can be enhanced by 78% with increasing radiative recombination efficiency [30]. By mounting the $\mathrm{\mu }\text{-}\mathrm{LED}$ on the TO-can package, the Au-coated pins of the TO-can sub-mount are connected using multiple Au bonding wires. The emission areas of four $\mathrm{\mu }\text{-}\mathrm{LED}$ elements were designed with identical diameters of 50 µm, and the areas of the coplanar waveguide ground-signal-ground (GSG) pads were $150\mathrm{\mu m}\times 150\mathrm{\mu m}$.

 

Fig. 3. Polarization ratio of semipolar green $\mathrm{\mu }\text{-}\mathrm{LED}$ (525 nm).

Download Full Size | PDF

B. Setup and Analysis of Directly Encoded Green $2\times 2\mathbf{\mu }\text{-}\mathrm{LED}$ Array for Optical Wireless Communication with Different Data Formats

After wire-bonding the adhered green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array to the TO-can and compact mounting to the specifically designed SMA connector shown in the upper left of Fig. 4(a), the small-form copper heat sink was employed to rigidly pack the device for communication testing, as shown in the upper right of Fig. 4(a). Figure 4(b) illustrates the experimental setup for continuous-wave emission, small-signal modulation, and digital data encoding characterizations. The fundamental emission properties of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array including $I-V$, $P-I$, and current-dependent spectrum analyses were investigated. A programmable $I-V$ source meter, power sensor, and charge-coupled device (CCD) spectrometer were used as the DC driver to simplify the testing geometry and accelerate the fast scanning within 10 s for each response curve of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array without executing temperature control during diagnosis. Later, the packaged green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array was standing along the optical axis of the testing environment and aligned the optical beam collimator constructed with a pair of plano–convex lenses. The transmission distance of the NRZ-OOK and QAM-OFDM transmissions carried by the TO-can-packaged $\mathrm{\mu }\text{-}\mathrm{LED}$ array with a collimated output can be set between 0 cm and 50 cm without degradation at the current stage. After beam expansion, propagation, and refocusing back to the receiver formed by an avalanche photodiode (APD) with analog receiving bandwidth of 1.5 GHz, the optically delivered data encoded onto the optical carrier output from the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array were inserted into the APD for optoelectronic conversion back to the electrical data stream for decoding. For data generation, an arbitrary waveform synthesizer (AWS, Tektronix 7122B) with a corresponding analog bandwidth of 5.3 GHz and sampling rate of 12 GSa/s was used to generate the NRZ-OOK, QAM-OFDM, and BL-DMT data streams. A pre-amplifier with a 26 dB unsaturated gain was optionally connected to the output port of the AWS to further amplify the peak-to-peak voltage of the data stream before encoding onto the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array. Indeed, the amplifier is added only for the initial optical alignment between the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array and the APD at the very beginning. As the receiving power of the optical data stream is quite small under initial alignment, the amplifier is necessary for magnifying the data power to optimize the positions and angles of all-optical components for light collimation. Afterwards, the amplifier is no longer needed for successive receiving and decoding procedures. A bias-tee was used to combine the DC and radio frequency (RF) biases for inserting into the input contact pad of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array. For data receiving, another bias-tee was used to separate the DC voltage bias for driving the APD and the alternating current (AC)-converted signal from the APD. Subsequently, the APD received output was fed into a post-amplifier for conditionally retrieving the peak-to-peak amplitude of the delivered data stream. Normally, the negative DC bias voltage was set at 90–100 V, the current limitation was set at 20 mA, and the amplifying gain was set at 20 dB during the receiving and decoding.

 

Fig. 4. (a) Photographs of the TO-can-packaged green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array on the copper cooling stage. (b) Experimental setup of the NRZ-OOK and QAM-OFDM transmissions carried by the TO-can-packaged green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array.

Download Full Size | PDF

Finally, the received data streams were analyzed with a digital serial analyzer (DSA, Tektronix 8200) with a 20 GHz (Tektronix, 80E03) module integrating the decoding software. To suppress the output fluctuation and drooping by the heat accumulation in the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array, the mount temperature was controlled at 20°C using a thermoelectric cooler and water-pipe cooled heat sink during operation. Without cooling, the temperature of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array gradually increases with a slope of about 0.25°C/min, which concurrently decreases the output power by $-0.4\mathrm{\mu W}/\min$. In contrast, the power-decaying slope of the same device under water cooling is only $0.033\mathrm{\mu W}/\min$. The water cooling procedure effectively stabilizes the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array to provide an extreme transmission data rate beyond 5 Gbit/s, but it is acceptable for obtaining the 80%–90% transmission performance at the uncooled condition.

3. RESULTS AND DISCUSSION

A. Continuous-Wave Emission and Analog Modulation Responses of the Green $2\times 2$ μ-LED Array

Figure 5 demonstrates the optical and electrical measurements of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array grown upon the semipolar (20-21)-oriented GaN buffered layer. Figure 5 shows the output $I-V$ curves, $P-I$ curves, and optical spectra for the $2\times 2$ green μ-LED array at different biases. The $I-V$ relationship can be described as

 

Fig. 5. (a) $I-V$ curve, (b) $P-I$ curve, and (c) optical spectra of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array. (d) Peak wavelength and FWHM of the EL spectra for the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array.

Download Full Size | PDF

From the bias-dependent emission spectral analysis, the main spectral peak centered in the green region exhibits a slight blueshift from 543 nm to 534 nm as the forward bias current increases from 5 mA to 40 mA. However, the green emission peak conversely redshifts back to 537 nm with the persistently enlarging forward bias current up to 100 mA. The shoulder emission peak at around 410 nm is attributed to the Mg-doped GaN barrier between QWs, which effectively improves the hole transport as well as radiative efficiency, as discussed in previous works [26,31]. The overall peak wavelength shift of the LED is only 7 nm, as shown in Fig. 5(d), and the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array retains good long-term stability as the injection current approaches 100 mA. Even while such a small wavelength shift is a better result than conventional devices produced on $c$-plane buffered GaN substrates, it can be attributed to a phenomenon associated with the lower polarization-related electric field and flattened energy band of the QW. As a result, the QCSE limiting device performance in $c$-plane devices will be inhibited. Furthermore, the FWHM of the emission spectrum shown in Fig. 5(d) varies only about 10 nm (including the maximum change) during operation with DC current ranging from 5 mA to 100 mA, indicating the strong uniformity of indium atom distribution and low defect density. Under the DC bias at 23 mA, the $-3\mathrm{dB}$ and $-6\mathrm{dB}$ modulation bandwidths of green $2\times 2\mathrm{\mu }\text{-}\mathrm{LEDs}$ are, respectively, obtained as 800 MHz and 1.02 GHz, as shown in Fig. 6. Owing to the limited bandwidth of the APD device, the $-3\mathrm{dB}$ and $-6\mathrm{dB}$ modulation bandwidths of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}+\mathrm{APD}$ set are, respectively, suppressed to 610 MHz and 830 MHz.

 

Fig. 6. Frequency responses of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array chip and green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array + APD set under the DC bias of 23 mA.

Download Full Size | PDF

B. NRZ-OOK Encoding Analysis of the Green $2\times 2$ μ-LED Array at the Optimized DC Forward Bias

The forward bias is set to 6 V to linearly encode the NRZ-OOK data stream onto the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array with enhanced modulation efficiency, with the forward bias current of $400\mathrm{A}/{\mathrm{cm}}^2$ located around the middle of the linear $P-I$ response. For directly encoded communication with the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array, such a forward bias selection is required to achieve maximum modulation efficiency. When DC biasing in such conditions, the transient differential resistance of approximately 191 Ω is nearly four times higher than the characteristic impedance of 50 Ω for all microwave instruments used in the testing environment. When considering impedance matching with the calculated reflection coefficient ($\mathrm{\Gamma }$) of $\mathrm{\Gamma }=(Z_{\mathrm{VCSEL}}-Z_{\mathrm{RF}})/(Z_{\mathrm{VCSEL}}+Z_{\mathrm{RF}})=0.585$ using the load impedances of the $\mathrm{\mu }\text{-}\mathrm{LED}$ and the microwave circuit ($Z_{\mathrm{VCSEL}}$ and $Z_{\mathrm{RF}}$), the return loss (${\eta }_{\mathrm{loss}}$) is determined as ${\eta }_{\mathrm{loss}}=-20\log |\mathrm{\Gamma }|=4.66\mathrm{dB}$, and the voltage standing wave ratio (VSWR) is determined as $\mathrm{VSWR}=(1+\mathrm{\Gamma })/(1-\mathrm{\Gamma })=3.82$. When performing direct modulation onto a green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array, such a large differential resistance would result in a significant amount of reflection of the high-frequency modulation signal or high-bit-rate data streams. To compensate for reflection loss, the peak-to-peak amplitude ($V_{\mathrm{pp}}$) of the encoding data amplitude must be increased from 0.8 V to 2.5 V to maintain receiving sensitivity and SNR after transmission. Figure 7 shows the received eye diagrams after direct encoding of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array via the amplified NRZ-OOK data stream, which is post-amplified and monitored by a real-time digital oscilloscope with sufficient bandwidth of 20 GHz. Without executing the pre-emphasis algorithm on the generated NRZ-OOK waveform with the AWS at the transmitting front-end, the blurred eye diagrams of the received data stream are easily observed even with its bit rate of 0.5 Gbit/s in Fig. 7(a). At 0.5 Gbit/s, the eye pattern in Fig. 7(a) exhibits SNR of 24.34 dB via a formula of $(V_{\mathrm{on}}-V_{\mathrm{off}})/(\delta V_{\mathrm{on}}+\delta V_{\mathrm{off}})$, with $V_{\mathrm{on}}$, $V_{\mathrm{off}}$, $\delta V_{\mathrm{on}}$, and $\delta V_{\mathrm{off}}$, respectively, denoting the on-level signal, off-level signal, on-level noise, and off-level noise amplitudes. In addition, an eye-crossing percentage [cross (%)] can be obtained as 80% by using a formula of $(V_{\mathrm{cross}}-V_{\mathrm{off}})/(V_{\mathrm{on}}-V_{\mathrm{off}})$ with $V_{\mathrm{cross}}$ representing the amplitude at the eye-crossing point.

 

Fig. 7. Eye diagrams of the NRZ-OOK transmission without pre-emphasis at data rates of (a) 0.5 Gbit/s, (b) 0.75 Gbit/s, and (c) 1 Gbit/s, and with pre-emphasis at data rates of (d) 1 Gbit/s, (e) 1.25 Gbit/s, and (f) 1.5 Gbit/s.

Download Full Size | PDF

Such a high level of the eye-crossing percentage denotes the mean value of the vertical histogram window, which indicates both the amplitude and the duty-cycle distortion owing to the insufficient modulation bandwidth of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array. As a result of the lower modulation throughput at higher frequencies, the data-bit symmetry deteriorated. Moreover, the rising time, falling time, and root-mean-square timing jitter were determined as 753.08 ps, 1.02 ns, and 300 ps, respectively, when operating at 0.5 Gbit/s. The timing jitter abruptly enlarges to distort the received eye diagram beyond 0.5 Gbit/s, and the eye diagram is entirely closed under 1 Gbit/s encoding of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array, as shown in Figs. 7(b) and 7(c). By performing the pre-emphasis algorithm on the generated NRZ-OOK data stream, Figs. 7(d)–7(f) reveal a clear opening eye diagram even when encoding the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array beyond 1 Gbit/s. After receiving the pre-compensated data waveform, the eye diagram reveals a slightly reduced bit amplitude owing to the transfer of the encoding power from low to high frequencies. Nevertheless, the received SNR is greatly improved from 6.52 dB to 20.578 dB as the data rate is fixed at 1 Gbit/s. In particular, the timing jitter is significantly suppressed to 75 ps at 1 Gbit/s, which somewhat enlarges from 125 ps to 200 ps when promoting the data rate from 1.25 Gbit/s to 1.5 Gbit/s, as shown in Figs. 7(e) and 7(f). More importantly, in all pre-emphasis scenarios, the crossing point of the received eye diagrams remains nearly constant at 55%, showing that both amplitude and duty-cycle distortions have been fully restored to the perfect condition. This is mainly attributed to the phase recovery on the encoded high-frequency component in addition to the amplitude compensation. Because of these eye diagrams without and with performing the pre-emphasis algorithm, the ultimate NRZ-OOK encoding data rate is recorded as 1.5 Gbit/s for such a green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array.

C. Broadband QAM-OFDM Encoding Analysis of the Green $2\times 2$ μ-LED Array at the Optimized DC Forward Bias

For the QAM OFDM data, the fast Fourier transform (FFT) size ($N_{\mathrm{FFT}}$) is set as 512. In addition, the inter symbol interference (ISI), cyclic prefix (CP), and training symbol ratio are also fixed as 1/32. Based on the formula of $N_{\text{subcarrier}}=\mathrm{BW}\times N_{\mathrm{FFT}}/R_{\text{sample}}$, with $N_{\text{subcarrier}}$, BW, and $R_{\text{sample}}$, respectively, denoting the subcarrier number, modulation bandwidth, and sampling rate, the subcarrier number under the modulation bandwidth of 1.5 GHz can be decreased from 192 to 48 with the increasing sampling rate from 4 GSa/s to 16 GSa/s. To eliminate low-frequency noise, the low-frequency cutoff is set as 15.625 MHz in our case based on the formula of $R_{\text{sample}}/N_{\mathrm{FFT}}$. For the broadband QAM-OFDM encoded transmission, the data generated from the AWS was characterized to optimize the oversampling rate for obtaining the best SNR with the least bit-error ratio (BER) at the very beginning, as shown in Fig. 8. For synthesizing the electrical 8-QAM OFDM data stream, the data bandwidth is set to 1.5 GBaud, and the BtB oversampling rate is set to 8 GSa/s, which includes the entire channel response, as shown in Fig. 8(a). The SNR remains around 15.4 dB with the corresponding BER slightly reducing to $2.7\times {10}^{-3}$, as shown in Fig. 8(b). Conversely, the 10-fold oversampling of the 8-QAM OFDM data stream with 16 GSa/s apparently degrades the SNR as much as $-3\mathrm{dB}$, which enlarges the decoded BER to $3.3\times {10}^{-3}$. The transmission characteristics of the received data stream with different bandwidths are compared to discover the greatest encodable bit rate of data using a five-fold oversampled 8-QAM OFDM data stream encoding to the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array.

 

Fig. 8. (a) RF spectra of the 4.5 Gbit/s 8-QAM OFDM data with different sampling rates. (b) Left: BERs and SNRs of the 4.5 Gbit/s 8-QAM OFDM data with different sampling rates. Right: constellation plots of the 4.5 Gbit/s 8-QAM OFDM data with sampling rates of 4 GSa/s and 16 GSa/s.

Download Full Size | PDF

First of all, the DC bias of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array was set as 6 V with a compliance current of 23 mA. To operate the APD in the linear detection region, the DC bias of the APD was stabilized at 94 V for obtaining the voltage gain of 100 and the responsivity of 15 V/W in the green wavelength region. Encoding a 2 GBaud 4-QAM OFDM data format with its subcarrier number of 256 and $V_{\mathrm{pp}}$ of 250 mV onto the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array provides an almost symmetric $V_{\mathrm{pp}}$ of about $\pm 0.2\mathrm{V}$ in the time domain, as shown in Figs. 9(a)–9(e). In general, the RF spectrum is the superposition of all QAM spectra encoded to all OFDM subcarriers. The comb-like spectrum with deep notches between adjacent subcarriers reveals that fewer subcarriers are employed in this case. A similar result obtained by using only 16 subcarriers to deliver the 4 Gbit/s QAM OFDM data was also presented previously [32].

 

Fig. 9. (a) Waveform, (b) RF spectrum, (c) CCDF of PAPR, (d) constellation plots, and (e) SNR spectrum of the electrical 4-QAM OFDM data with a modulation bandwidth of 2 GHz and a sampling rate of 8 GSa/s. (f) Waveform, (g) RF spectrum, (h) CCDF of PAPR, (i) constellation plots, and (j) SNR spectrum of the optical 4-QAM OFDM data with a modulation bandwidth of 2 GHz and a sampling rate of 8 GSa/s.

Download Full Size | PDF

After APD receiving and offline decoding, the received data stream in Figs. 9(f)–9(j) reveals a waveform with slightly asymmetric amplitude at positive and negative parts of $+0.4$ and $-0.6$ V after optoelectronic conversion and amplification in the APD with an output photocurrent of 0.32 mA on average. The asymmetric positive and negative parts reveal the output power saturation of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array due to the drooping effect under highly biased operation. After conducting the FFT of the time-domain waveform, the encoding throughput response was severely reduced by 36 dBm over a frequency range of 2 GHz, with a decaying ratio of $-18\mathrm{dB}/\mathrm{GBaud}$, due to the finite modulation bandwidth of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array. The received 4-QAM OFDM constellation plot was significantly blurred, resulting in a degraded probability at the same peak-to-average power ratio (PAPR), which causes at least a 1 dB linear increase in PAPR at the same complementary cumulative distribution function (CCDF), indicating that either time-domain clipping or frequency-domain filtering could be conditionally applied after receiving to reduce PAPR. After decoding, the error vector magnitude (EVM) is obtained as 31.6%, the PAPR at the probability of 0.1 is 6 dB, and the SNR linearly decreases from 20 dB to 4 dB with the corresponding BER determined as $7.8\times {10}^{-4}$. In contrast, when the data stream is transformed from 4-QAM to 8-QAM, the allowed data bandwidth is reduced to 1.5 GBaud due to the allocation of 96 OFDM subcarriers, while the oversampling rate stays at 8 GSa/s, as shown in Figs. 10(a)–10(e). The received data waveform can maintain its $V_{\mathrm{pp}}$ the same as that detected at the 4-QAM OFDM encoding case, as shown in Figs. 10(f)–10(j). Within the 1.5 GHz band, the 8-QAM modulation throughput spectrum appears to decline much more from $-20\mathrm{dBm}$ to $-50\mathrm{dBm}$, resulting in a decaying ratio of $-20\mathrm{dB}/\mathrm{GBaud}$, which is also larger than the 4-QAM encoding scenario. With the received peak-to-peak amplitude of 1.22 V, the 8-QAM OFDM constellation plot reveals an EVM of 24%, PAPR of 9.3 dB at CCDF of 0.1, and SNR decaying from 19 dB to 9 dB within the 1.5 GHz data bandwidth such that the decoded BER of $2.7\times {10}^{-3}$ is below the forward error correction (FEC) criterion of $3.8\times {10}^{-3}$.

 

Fig. 10. (a) Waveform, (b) RF spectrum, (c) CCDF of PAPR, (d) constellation plots, and (e) SNR spectrum of the electrical 8-QAM OFDM data with a modulation bandwidth of 1.5 GHz and a sampling rate of 8 GSa/s. (f) Waveform, (g) RF spectrum, (h) CCDF of PAPR, (i) constellation plots, and (j) SNR spectrum of the optical 8-QAM OFDM data with a modulation bandwidth of 1.5 GHz and a sampling rate of 8 GSa/s.

Download Full Size | PDF

For comparison, the decoding performances between the 4- and 8-QAM data streams delivered by the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array are summarized in Table 2. Note that the received 8-QAM OFDM data stream with its raw data rate of 4.5 Gbit/s (1.5 GBaud) also exhibits a less blurred constellation plot than the 4-QAM OFDM with its raw data rate of 4 Gbit/s (2 GBaud), which is mainly attributed to overestimation on the bandwidth of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array used for the 4-QAM case. After receiving and decoding, the data bits encoded beyond the bandwidth cause greater amplitude and phase errors than data bits encoded inside the bandwidth. This phenomenon necessitates the use of a more complex data structure, such as the BL-DMT algorithm, to make better use of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array’s finite bandwidth.

Table 2. Parameters for Broadband QAM OFDM Transmission Performance

View Table | View all tables in this article

D. Bit-Loaded DMT Encoding Analysis of the Green $2\times 2$ μ-LED Array at the Optimized DC Forward Bias

To enhance the usage efficiency of the full modulation band provided by the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array under direct encoding operation, the BL-DMT algorithm is employed to manage the maximal allowable QAM level of the data mapped to each OFDM subcarrier. At the same DC bias current, the allowable modulation bandwidth of 1.8 GHz was sliced into seven regions for encoding with M-ray QAM (M-QAM) OFDM data. For example, the low-frequency region extended from DC to 0.23 GHz can approach an SNR of 18 dB, which is assigned to be encoded by the 32-QAM OFDM data. As the SNR declines with the increasing subcarrier frequency, the assigned QAM level monotonically decays from 32 to four when assigning the QAM data from the first to the seventh sliced band (1.47–1.80 GHz) for directly encoding the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array. After managing the seven sub-bands for encoding different M-QAM OFDM data, the combined BL-DMT waveform synthesized from the AWG and detected at the receiving end is illustrated in Fig. 11(a), and the FFT spectrum of the data stream after off-line decoding with MATLAB is shown in Fig. 11(b). The corresponding power CCDF probability from a time-domain BL-DMT waveform is shown in Fig. 11(c), which reveals the percentage of probability for the time-domain data at different PAPRs. After decoding the received waveform of the BL-DMT data stream, the sliced frequency bands were slightly adjusted to adapt all assigned M-QAM data within each band for error-free receiving and decoding, as shown in Fig. 11(d).

 

Fig. 11. (a) Waveform, (b) RF spectrum, (c) CCDF of PAPR, (d) SNR spectrum, and (e) constellation plots of the optical BL-DMT data with a data rate above 5 Gbit/s.

Download Full Size | PDF

As a result, the first sub-band with a bandwidth of 0.23 GBaud allows 16-QAM encoding with a raw data rate of $0.23\mathrm{GBaud}\times {\log }_{2}16=0.92\mathrm{Gbit}/\mathrm{s}$, which can obtain the decoded EVM of 14.3%, average SNR of 16.9 dB, and received BER of $6.37\times {10}^{-4}$. The second sub-band located from 0.23 GHz to 0.45 GHz also allows 16-QAM encoding with a raw data rate of $0.22\mathrm{GBaud}\times {\log }_{2}16=0.88\mathrm{Gbit}/\mathrm{s}$, and the corresponding EVM, SNR, and BER are 16.3%, 15.8 dB, and $2.23\times {10}^{-3}$, respectively. The third sub-band located between 0.45 GHz and 0.7 GHz allows only 8-QAM OFDM with a raw data rate of $0.27{\text{GBaud}\times \text{log}}_{2}8=0.75\mathrm{Gbit}/\mathrm{s}$ with respective EVM, SNR, and BER of 19.3%, 14.3 dB, and $3.06\times {10}^{-4}$. The fourth sub-band located between 0.7 GHz and 0.97 GHz allows only 8-QAM OFDM with a raw data rate of $0.27\text{GBaud}\times {\log }_{2}8=0.81\mathrm{Gbit}/\mathrm{s}$ with respective EVM, SNR, and BER of 23.7%, 12.5 dB, and $2.44\times {10}^{-4}$. For the last three sub-bands covering the frequency range from 0.97 GHz to 1.8 GHz, the highest M-QAM level is up to four such that the encodable raw data rate is $0.83\text{GBaud}\times {\log }_{2}4=1.66\mathrm{Gbit}/\mathrm{s}$ with the respective EVM, average SNR, and receiving BER of 23.7%, 12.5 dB, and $2.442\times {10}^{-4}$.

Table 3 summarizes and compares the important parameters for directly encoding the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array with broadband 8-QAM OFDM and BL-DMT data streams. Overall, the total raw data rate of the BL-DMT data stream encoded onto the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array is 5.02 Gbit/s. By employing the BL-DMT algorithm, the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array has shown its superior potential on carrying the data beyond 5 Gbit/s for future applications in fields of VLC or optical wireless communication when packaging with handed mobile devices.

Table 3. Related Parameters of Optical Bit-Loaded DMT Data Carried by the 2 x 2 Green μ-LED Array

View Table | View all tables in this article

4. CONCLUSION

To facilitate high-speed data transmission, the specific design and package are demonstrated for the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array with nanostructured grating patterns grown on a semipolar (20-21)-oriented GaN buffered layer on (22-43)-oriented sapphire substrate. The unique structural design minimizes the polarization-related electric field and flattens the QW band diagram to suppress the QCSE in the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ with its turn-on voltage of 2.5 V and emission power of 0.3 mW at $1\mathrm{A}/{\mathrm{cm}}^2$. Such a green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array with the inhibition of QCSE shows a very small wavelength shift from 543 nm to 537 nm as compared to conventional devices fabricated on $c$-plane buffered GaN substrates. The 50 µm emission aperture of the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array ensures a lower capacitance for a larger $-3\mathrm{dB}$ modulation bandwidth, which is less affected by the high resistance of the single $\mathrm{\mu }\text{-}\mathrm{LED}$ element. The device also features a larger bias at its $-1\mathrm{dB}$ power compression point because of its low power consumption. Moreover, the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array also exhibits the transient differential resistance of 191 Ω to induce the reflection coefficient of 0.585, return loss of 4.66 dB, and VSWR of 3.82. These characteristics can sufficiently affect the transmission performance.

With a specific TO-can+SMA package, the green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array exhibits maximal data rates exceeding 1.5 Gbit/s for the NRZ-OOK format and beyond 5.02 Gbit/s for the BL-DMT format, which is very promising for optical wireless communication. As the sampling rate increases from 4 GSa/s to 16 GSa/s, the $\mathrm{\mu }\text{-}\mathrm{LED}$ array’s received SNR improves dramatically from 15.4 dB to 12.2 dB. The SNR remains about 15.4 dB, with a matching BER of $2.7\times {10}^{-3}$, whereas the 10-fold oversampling of the 8-QAM OFDM data stream with 16 GSa/s appears to reduce the SNR by $-3\mathrm{dB}$, resulting in a decoded BER of $3.3\times {10}^{-3}$. The green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array can deliver 8-QAM OFDM data with an EVM of 24%, PAPR of 9.3 dB at CCDF of 0.1, SNR from 19 dB to 9 dB, and decoded BER of $2.7\times {10}^{-3}$ smaller than the FEC criterion of $3.8\times {10}^{-3}$. Moreover, this device can deliver 16-QAM-OFDM to 4-QAM-OFDM data via the BL-DMT algorithm to achieve a data rate of 5.02 Gbit/s under the FEC criterion. The green $2\times 2\mathrm{\mu }\text{-}\mathrm{LED}$ array has demonstrated its greater potential in transmitting data beyond 5 Gbit/s using the BL-DMT algorithm for future applications in domains of VLC or optical wireless communication when packaged with handed mobile devices.