1. Introduction

Visible light communication (VLC) is an emerging technology which is being considered as a potentially significant part of future wireless networks [1,2]. An important characteristic of a VLC system is going to be its maximum data rate and this depends upon its bandwidth and the signal-to-noise ratio (SNR) at the output of the receiver [3]. In many VLC systems, the SNR is enhanced using an optical concentrator at the receiver. However, conventional concentrators, such as compound parabolic concentrators (CPCs), can be comparatively bulky and the conservation of étendue means that there is a trade-off between their optical gain and field of view (FOV) [4].

To build compact VLC receivers with wide field of views (FOVs), optical antennas incorporating fluorophores have been studied recently and some promising results have been obtained. In particular, it has been shown that because the fluorophores only absorb certain wavelengths they act as optical filters which reject ambient light [5]. In addition, these antennas are designed so that photons emitted by the fluorophore can be trapped within the antenna and guided to the photodiodes. This means that they can concentrate the light and hence increase the signal at the output of the photodiode [6–8]. Finally, since light is concentrated by fluorescence rather than reflection and refraction, étendue is not conserved and so it should be possible to achieve both a high light concentration and a wide FOV [6].

The working principles of a fluorescent antenna are identical to those of a luminescent solar concentrator (LSC) [9,10]. However, to be effective a LSC has to absorb a wide range of wavelengths and the emission of photons with a longer wavelength reduces the efficiency of the LSC. In contrast, a fluorescent antenna only has to absorb the wavelengths emitted by the transmitter and a large change in wavelength can improve efficiency by reducing re-absorption. The potential advantages of using fluorescent antennas mean that since 2014 several different antenna structures and fluorescent materials have been investigated in different VLC or optical wireless communication (OWC) systems [5,7,8,11–20]. Some of the antennas have been made using existing fluorescent optical fibres [8,12,15,19,20]. However, these are not optimized for VLC applications and this limits the choice of wavelengths that can be used to transmit data. This limitation can be overcome by fabricating antennas using commercially available fluorophores. Previously, fluorophores have been combined with microscope slides to make rectangular antennas [7,18]. Unfortunately, when rectangular structures are used, the inefficient coupling between the antenna and the photodetector means that the coupling between the antenna and the photodiode is inefficient [5].

In this paper, we introduce a new flexible method of creating fluorescent antennas. This method is then used to investigate the suitability of different fluorescent materials for an important example application, specifically white light-emitting diode (LED)-based VLC systems. The new antenna structure uses a glass capillary to contain a cured epoxy doped with a fluorophore. Compared to previous fluorescent antennas based on microscope slides, this structure can be easily coupled with typical photodiodes which leads to much lower photon escape rates at the antenna edge. Moreover, in this work, three different organic fluorescent materials are considered and compared when incorporated into the antenna. We show that using a previously unconsidered fluorophore, Coumarin 504 (Cm504) which is also known as Coumarin 314, results in a higher bandwidth for a white LED-based VLC system than the previously considered Cm6 and 4-(Dicyanomethylene)-2-methyl-6-(4-dimethylaminostyryl)-4H-pyran (DCM). The impact of the concentration of these fluorophores on performance factors including the FOV, the receiver output signal, the modulation bandwidth and the bit-error rate (BER) have been studied. Furthermore, the comparison has been undertaken at different illuminance levels at the receiver. Results are presented which show that when orthogonal frequency-division multiplexing (OFDM) modulation is used and the illuminance level is low, the antenna made of DCM can achieve lower BERs. In contrast, when the illuminance level is relatively high, the antenna made of Cm504 can achieve lower BERs.

2. White LED-based VLC

VLC using white LEDs for both indoor illumination and wireless data transmission is being considered for future wireless networks that support a wide range of applications, such as the connections of Internet of things (IoT) devices used in a smart home environment [21,22]. In these systems the white LED used to illuminate a space would also be used as a transmitter. As shown in Fig. 1, the light from a white LED has two components. The first component is blue light from the blue gallium nitride (GaN) LED within the white LED. The other component is the yellow light emitted by a phosphor excited by some of the blue LED light. The bandwidths of the transmitter and receiver are important characteristics of any communications system. For both the LED and the phosphor, their bandwidths are inversely proportional to the carrier recombination and photoluminescent (PL) lifetimes respectively. Typically the carrier recombination lifetime is shorter than the PL lifetime [3] and so the bandwidth of the blue LED is wider than the bandwidth of the yellow phosphor. To determine the bandwidth of these two components, the bandwidths of a cool white LED and a royal-blue LED have been measured. These measurements led to the conclusion that the 3 dB bandwidth of the royal-blue LED is 5 MHz. Whilst the bandwidth of the white LED and the yellow component of the output of the white LED were both found to be 2.3 MHz. The measured spectrum of the two LEDs used in these experiments are shown in Fig. 1.

 

Fig. 1. The spectra of a phosphor-coated white LED (LUXEON, Rebel LL-PWC2) and a royal-blue LED (LUXEON, Rebel LXML-PRO2-A900) measured using a photonic multichannel spectral analyzer (Hamamatsu, PMA-11 model C7374-36). Together with the responsivity of the APD used to create VLC links (Thorlabs, APD130A) obtained from its datasheet.

Download Full Size | PDF

Although the bandwidth is a very important parameter that affects the performance of a VLC link, it was found in [23,24] that blue filtering does not always improve the link’s performance. In particular, if the subcarrier frequencies are low enough for the yellow light to carry information then using a blue filter will reduce the system’s performance. In contrast, if higher subcarrier frequencies are used then the yellow can’t carry information. However, it does add shot noise which can be amplified when equalization is applied to the subcarrier. In this situation, by reducing the shot noise a blue filter can improve the link’s performance. The different consequences of adding blue filtering to systems mean that it is prudent to test the performance of alternative designs of VLC systems.

3. Fluorescent antennas in VLC

The yellow component from a white LED can be rejected using either interference or absorption filters [3]. However, the response of interference filters depends upon the angle of incidence. The problems that this can cause can be avoided by using absorption filters. Unfortunately, the proximity of the outputs from the blue LED and the phosphor will mean that the absorption filter will also attenuate the blue light. In contrast, with a fluorescent antenna, it should be possible to both reject the light from the phosphor and increase the photon flux reaching the receiver.

In [7], an optical antenna structure made of a thin fluorescent layer sandwiched between two glass microscope slides was described. These structures are a convenient way to create an optical antenna containing a fluorophore. However, a photodetector only occupies a small percentage of the length of the edges through which fluorescence escapes the antenna. Consequently, these antennas attenuate the signal reaching the photodetector. Despite this attenuation, these antennas improve the performance of VLC systems operating in ambient light [5].

A better match between the size and shape of photodetectors and the area through which fluorescence escapes an antenna can be achieved using a cylindrical rather than a rectangular antenna. Figure 2 illustrates the main physical processes within a fluorescent antenna. When the light arrives at the cylinder’s surface some light is reflected but most of the light is transmitted into the cylinder. Then, if the wavelength of the light is not within the absorption range of the fluorophore or the host matrix, the light will pass through the cylinder. In contrast, when the photon’s wavelength is within the absorption range of the fluorophore, it can be absorbed by the fluorophore. The absorbed energy might be dissipated by a non-radiative process. However, for a well-chosen fluorophore most absorbed photons will result in the emission of a longer wavelength photon. This photon can be emitted in any direction and so it may escape from the antenna. Alternatively, any overlap between the absorption spectrum and the emission spectrum of the fluorophore may result in re-absorption. If this doesn’t happen the photons can be retained in the antenna by total internal reflection until it reaches one end. The aim when designing a fluorescent concentrator is to increase the photon flux on a photodetector placed at the end of the cylindrical concentrator.

 

Fig. 2. The physical processes that occur within a fluorescent antenna.

Download Full Size | PDF

One consequence of the change from a rectangular concentrator to a cylindrical concentrator is that parallel light from a distant transmitter will have different angles of incidence on the curved surface of the cylinder. The probability of transmission into a cylinder with a refractive index of 1.5 for different angles of incident has been calculated using the Fresnel equations. The results are shown in Fig. 3. It can be seen that the probability of light being transmitted into the cylinder decreases as the angle of incidence increases. However, higher angles of incidence are associated with smaller projected areas. Taking this into account the average probability of transmission can be calculated using

 

Fig. 3. The probability that a photon is transmits into a glass cylinder plotted as a function of the incident angle with respect to the curved cylinder surface.

Download Full Size | PDF

The gain of a fluorescent antenna also depends upon the retention of the fluorescence within the antenna. This can be calculated using the probability that a photon will escape from a radius $r$ in a cylinder with a maximum radius $R$ and a refractive index $n$, which is given by [25]

Both of these probabilities are shown in Fig. 4(a) when $n=1.5$. It can be seen that, when $r=R$ and photon emission occurs at the surface, the probability that a photon is retained is, $75\%$. However, this reduces to $33\%$ if the photon is emitted from the centre of the cylinder. Fortunately, the probability of a fluorophore being at a particular radius increases as the radius increases and is proportional to the perimeter of the circle with the same radius. Therefore, the average probability of a photon escaping from the cylinder if the fluorophore is distributed evenly between the centre and a particular normalized radius, $r/R$, can be calculated using

Therefore, the average probability of a photon being retained is

Figure 4(b) shows that the average probability increases significantly if the fluorophore occupies most of the cylinder and raises to $55\%$ if the fluorophore occupies all the cylinder.

 

Fig. 4. (a) The escape probability and the retention probability of an emitted photon plotted as a function of $r/R$, (b) The average escape probability and the average retention probability for a fluorescent layer that occupies a cylinder up to a radius $r$, shown as a function of $r/R$.

Download Full Size | PDF

Less light will enter a cylinder than a rectangle of the same size and less fluorescence will be retained in the cylinder. However, these losses are insignificant compared to the amount of fluorescence which escaped from the previously described rectangular antennas before they reached the receiver. Cylindrical antennas with an area that matches the area of the photodetector in the receiver are therefore expected to give a better performance than existing rectangular antennas.

4. Fluorescent concentrator fabrication

The fabrication steps of a cylindrical fluorescent optical antenna are introduced in this section. The main steps include the selection of the host system, the preparation of the fluorescent samples and the design of the antenna structure.

4.1 Host system

Light is retained in an antenna by total internal reflection and so the outer surface of an antenna should be as smooth as possible. To achieve this a glass capillary has been chosen as the host system. To match the size of the APD (Thorlabs, APD130A), that will be coupled to the antenna, the antenna should have a diameter of 1 mm. This is the internal diameter of the selected cylindrical capillary (Hirschmann, 9600150). These capillaries are up to 125 mm long. As calculated in the previous section, the effective area of the capillary is $0.932{\times }2RL=116.5\;\text {mm}^2$ which is 148 times the area of the APD. However, the light escapes the capillary from both ends and only approximately $50\%$ of the light is retained and so the photon flux on the APD might increase by a factor of approximately 37. The quantum efficiency of the fluorophore and light scattering out of the antenna by unavoidable optical inhomogeneities mean that this is an optimistic estimate. However, even taking this into account we would expect to increase the photodetector’s output signal by significantly more than the ‘increase’ by a factor of 0.28 previously reported [5].

4.2 Antenna fabrication process

In the fluorescent antenna fabrication process, a selected fluorophore was first mixed with NOA61 (Norland Products Inc). NOA61 was selected because the cured epoxy has high transmission coefficients for the visible light range. More importantly, NOA61 has a relatively low viscosity of 300 cps and so is easily stirred when mixing with the fluorophore.

In the sample preparation stage, samples with different concentrations of the fluorophore, including 0.2 mg/ml, 0.5 mg/ml, and 1 mg/ml, were prepared. To reduce the amount of air which can cause bubbles in the antenna after the epoxy is cured, each mixture was placed in a vacuum chamber for approximately 20 mins immediately before being drawn into a capillary. As shown in Fig. 5(a), the liquid mixture was then drawn into the capillary using an injection tube. In the final step, the capillary was placed under a 365 nm UV light for curing. A photograph of three fabricated capillary antennas (after UV curing) under 400 nm blue light is shown in Fig. 5(c). A photograph of the prepared sample examples of 0.2 mg/ml under 400 nm blue light before UV curing is shown in Fig. 5(b).

 

Fig. 5. (a) The key steps of fabricating a fluorescent antenna, (b) a photo of the prepared samples under 400 nm blue light before UV curing, (c) a photo of three fabricated capillary antennas (after UV curing) under 400 nm blue light.

Download Full Size | PDF

 

Fig. 6. The absorption and emission spectrum of Cm504, Cm6 and DCM. All three different samples have the same concentration of 0.2 mg/ml. The dashed lines are the adsorption and emission spectra of the samples sandwiched between two glass substrates. The solid lines are the emission spectra of the florescence antennas.

Download Full Size | PDF

4.3 Selection of the fluorescent dyes

To investigate their relative performance in VLC applications, different fluorophores need to be incorporated into cylindrical antennas. Previously, receivers for use with white LED transmitters have used filters to increase the bandwidth of the link by blocking the yellow light from the LED. This past practice suggests that the best fluorophore for use in a receiver working with a white LED will absorb as much blue light as possible. The white LED spectrum shown in Fig. 1, then indicates that the fluorophore should absorb wavelengths shorter than 480 nm without absorbing a significant amount of light at longer wavelengths. Cm6 has previously been used in a VLC system which used a blue transmitter [7]. However, as shown in Fig. 6, it absorbs some wavelengths longer than 480 nm. This isn’t the case with Cm504 and so antennas containing Cm6 and Cm504 have been compared.

Another consideration when designing a receiver is the responsivity of the photodetector. Some of the data in Fig. 1 shows that the responsivity of the APD used to compare antennas increases rapidly as the wavelength of the light increases from 400 nm to 800 nm. Since the responsivity of the APD is expressed in Amperes per Watt some of this increase can be explained by the reduction in photon energy as wavelength decreases. However, the energy of an 800 nm photon is half the energy of a 400 nm photon. The change in photon energy would therefore explain a doubling of the responsivity of the APD as the photon wavelength changes from 400 nm to 800 nm. In contrast, over this range of wavelengths the responsivity increases by a factor of 10. This clearly shows that the APD is significantly better at detecting longer wavelength photons. When this APD is used in a receiver a fluorophore that absorbed 450 nm light and has a large Stokes shift might therefore be a good choice. Since DCM has a relatively large Stokes shift, despite the potential reduction in photoluminescence quantum yield (PLQY) associated with a large Stokes shift, it has been compared to Cm6.

The measured absorption and emission spectra of Cm504, Cm6 and DCM are shown in Fig. 6. In the absorbance measurements, the liquid sample in which the fluorescent dye was mixed into epoxy NOA61 with a concentration of 0.2 mg/ml was sandwiched between two glass substrates with a gap of 0.5 mm. A photo of the used samples is provided in Supplement 1. The absorbance of the sample was measured using a spectrophotometer (JASCO V-570, UV-Vis-NIR spectrophotometer) before and after UV curing. The absorbance of a sample with only NOA61 was also measured. Then, the absorbance of the fluorescent dye was obtained by subtracting the absorbance of the cured NOA61 from the absorbance of the sample containing the dye. The measurements showed that UV curing does not affect the absorbance within the visible light range. Next, the same samples used for the absorption measurements were also used for the emission spectra measurements. In this measurement, the samples sandwiched by two glass substrates were excited by 400 nm light and the emitted light from the samples was coupled into an optical fiber which is connected to a photonic multichannel analyzer (Hamamatsu, PMA-11 model C7473-36) for the spectra measurements. Then, the emission spectra of the antennas were measured by placing the optical fiber input at one antenna end when the excitation source was a 450 nm royal-blue LED. It can be seen from Fig. 6, the emission spectra of the samples are different from the emission spectra of the antenna. This is because, overlap between the samples absorption and emission spectra, means that when light propagates within the antenna self-absorption occurs. Since the shorter wavelengths suffer most self-absorption this changes the shape of the emission spectra, including a further shift of the emission peak.

Figure 1 shows the spectrum of the white LED used to compare the antennas. Comparing this figure with Fig. 6(b), confirms that light emitted from the phosphor between 480 nm and 510 nm can be absorbed by Cm6. However, Fig. 6(a) shows that, as expected, this isn’t the case for Cm504. A VLC system incorporating the Cm504 antenna is therefore expected to have a larger bandwidth than one containing a Cm6 antenna. Finally, Fig. 6(c) confirms that DCM has a larger Stokes shift than Cm6. When combined with the responsivity of the representative APD used to compare antennas this means that a DCM antenna is expected to generate a larger signal than a Cm6 antenna.

4.4 Influence of the fluorophore concentration

Increasing the concentration of the fluorophore from a low value will initially increase both the amount of incident light that is absorbed and re-absorption by the fluorophore. However, at high concentrations, the amount of incident light absorbed will saturate but the impact of self-absorption will continue to increase. This means that the concentration of the fluorophore mixed into the NOA61 will have an impact on the performance of the antenna.

Figure 7 shows the measured emission spectra of the antennas when different fluorophore concentrations, specifically 0.2 mg/ml, 0.5 mg/ml, 1 mg/ml, are exposed to light from a royal blue LED, whose measured spectrum is shown in Fig. 1. From Fig. 7(a)-(c), it can be seen that for these concentrations increasing the fluorophore concentration results in a lower light intensity level and a shift in the peak wavelength of the emitted light. Based upon these results, the antennas with the lowest concentration of fluorophore, 0.2 mg/ml, were used in communications experiments.

 

Fig. 7. The emission spectrum of the fluorescent antenna when three different concentrations of samples were used in the fabrication process. (a) the spectra of the emitted light from the Cm504 antennas, (b) the spectra of the emitted light from the Cm6 antennas, (c) the spectra of the emitted light from the DCM antennas.

Download Full Size | PDF

4.5 PL lifetime

Another important parameter of the optical antenna is the PL lifetime of the fluorophore. This is important because it creates a single pole response which might determine the bandwidth of the whole system. Fortunately, unlike the yellow phosphor coating on a white LED, many organic materials have a PL lifetime of several nanoseconds or less. Figure 8 shows the measured PL decay results of the three fluorophores that have been compared. It can be seen that the lifetimes of Cm504, Cm6 and DCM are 3.8 ns, 3.2 ns and 2.4 ns, respectively. This means the 3 dB bandwidths of these three antennas are at least 40 MHz. This is larger than the bandwidth of a white light LED and so if an LED is used with these materials the system’s bandwidth will be determined by the LED rather than the fluorophore. However, this may not be the case if the transmitter is a laser or micro LED.

 

Fig. 8. The measured PL decay plots of three considered materials. These results were measured using a streak camera (Hamamatsu, C5680) when the excitation source is a 408 nm laser.

Download Full Size | PDF

5. Communication performance

The important performance metrics for a VLC system are the FOV and the data rate at which the BER is less than an allowed maximum. These characteristics have therefore been measured, In addition, to understand the relative performance of different antennas, the VLC systems bandwidth and the signal at the output of the APD have been measured.

5.1 FOV and concentration gain

Usually, the gain of an optical concentrator is defined in the optical domain as the ratio of the collected optical power with the antenna to that without the antenna. However, for a VLC system, the important parameter is the signal at the receiver’s output. The peak-to-peak voltage at the APDs output has therefore been measured when the transmitted signal was a 1 MHz sine wave. Figure 9(a) shows the measured peak-to-peak voltage without an antenna and with three different antennas when the excitation light source is a royal blue LED. The peak-to-peak voltages when an antenna is used show that the fluorescent antennas provide signal gain. As described previously this arises from a combination of an increase in the number of photons reaching the APD and a change in wavelength that increases the effective quantum efficiency. Furthermore, the results show that the wide absorption band and large Stoke’s shift of DCM means that the DCM antenna gives the highest gain.

 

Fig. 9. (a) The measured output voltage for different antennas at different angles. (b) The normalized output voltage for different antennas at different angles, (c) the measured gain for different antennas at different angles. The symmetry of the antenna around its axis means that the angle that was varied in these experiments was measured from the axis of the antenna.

Download Full Size | PDF

To determine the FOV the measured output voltages have been normalized. The results in Fig. 9(b) show that the FOV is determined by the cosine law that arises from the projected area of the antenna. Finally, the gains of the three antennas are shown in Fig. 9(c). As expected these results show that the gain is independent of the incident angle. Furthermore, the gain of the Cm504 antenna is approximately 1.3, whilst the gain of the Cm6 antenna is 3.5 and the gain of the DCM antenna is 4.7. Although, none of these are as high as the maximum theoretical gain they are all larger than one and therefore these antennas are all acting as concentrators.

5.2 Measured bandwidth

The modulation bandwidth is an important characteristic of a communication link. When a fluorescent antenna is used with a white LED transmitter, the bandwidth depends on the overlap between the absorption spectrum of the fluorophore and the spectrum of the white LED. In particular, a fluorophore which only absorbs the light emitted from the blue LED within the white LED should result in a receiver with a higher bandwidth. The frequency responses of the three selected antennas were measured when each antenna was flood illuminated. The results in Fig. 10 include the results obtained without an antenna. The generation of light by recombination in an LED, phosphorescence or fluorescence all lead to an exponential temporal response and hence a single pole frequency response. Since the output voltage of a photodiode is proportional to the photon flux the output voltage from a photodiode then also has a single pole response. The frequency response of the photodiode with and without different antennas in Fig. 10 is therefore shown with an axis that emphasizes an interest in the output voltage rather than the output power. First, it can be seen that the response of the white LED and blue LED are comparable to the expected single pole response. However, although it contains a blue LED the bandwidth of a white LED is only 2.3 MHz. This is also the bandwidth when a long pass filter (GG495) was placed in front of the receiver. The GG495 filter only allows light emitted from the phosphor to be detected. This shows that, when all the white light emitted from a LED is used, the PL lifetime of the phosphor dominates the overall frequency response. Figure 10 also shows that all three antennas increase the bandwidth of the link. More importantly, as expected the selective absorption of light from the blue LED by Cm504 results in a higher bandwidth. In particular, using Cm504 results in a bandwidth of 4.7 MHz, compared to bandwidths of 3.1 MHz for Cm6 and 2.9 MHz for DCM. Also, the bandwidth achieved using a Cm504 antenna is very similar to the bandwidth of a blue LED. This means that the Cm504 only absorbs the blue light emitted from the GaN LED and does not absorb the yellow light emitted from the phosphor.

 

Fig. 10. The measured frequency responses of links with and without an antenna.

Download Full Size | PDF

5.3 Data transmission

The schematic diagram and a photograph of the VLC experimental setup are shown in Fig. 11. During the data transmission experiments, the bit-loaded OFDM transmitted signal was generated offline using MATLAB and then uploaded into an arbitrary waveform generator (AWG, Siglent SDG2082X). The OFDM signal was pre-clipped in MATLAB at both its top and bottom with a clipping ratio of 10 dB [26,27]. In this case, significant signal peaks can be removed so that the dynamic range of the voltage signal provided by the AWG can be well utilized for a given peak-to-peak voltage value whilst the signal distortion caused by the clipping noise is very minor. In all experiments, both the top and the bottom clipping levels were set relative to the standard deviation of the unclipped OFDM signal which has a Gaussian distribution. In particular, the 10 dB clipping level that was used meant that the clipping level was 3.16 times larger than the standard deviation of the Gaussian variable [26]. Moreover, by avoiding sudden extremely high peaks in the OFDM signal generated in different rounds transmissions, the performance comparison becomes fair for a fixed peak-to-peak voltage value. Then, the signal was amplified using an electrical amplifier (Mini-circuits, ZHL-32A-S) and superimposed onto a DC current using a bias-T (Mini-circuits, ZFBT-4R2GW). This combined signal was then used to drive a white LED (LXML-PWC2). At the receiver, a fluorescent antenna collected the light from the white LED and guided some of the resulting fluorescence to the selected APD (Thorlabs APD130A). An internal transimpedance amplifier then converted the output current from the APD into a voltage signal. Finally, the voltage signals were captured by a digital oscilloscope (LeCroy, 204Xi-A) and processed offline using MATLAB to recover the transmitted data.

 

Fig. 11. A schematic diagram and a photograph of the experimental setup used to character the performance of VLC links when data is transmitted using OFDM.

Download Full Size | PDF

Figure 12 shows the results of a typical data transmission experiment, in particular the results when 25 Mbps are transmitted to a Cm504 antenna. In particular, Fig. 12(a) shows the estimated SNRs as well as the number of loaded bits on different OFDM subcarriers. Then, Fig. 12(b) shows the measured BERs of different OFDM subcarriers and Fig. 12(c)-(f) show the received signal constellation with different sizes. From Fig. 12(b), it can be seen that, although different subcarriers have different SNRs, the use of bit loading can lead to similar BERs on different subcarriers which can result in overall higher transmission data rate.

 

Fig. 12. The transmission performance at 25 Mbps when a Cm504 antenna is used and the illuminance at the receiver is 1113 lx, (a) the estimated SNR and the loaded number of bits on different OFDM subcarriers, (b) the measured BER on different OFDM subcarriers, (c) the received BPSK constellation points, (d) the received 4-QAM constellation points, (e) the received 8-QAM constellation points, (e) the received 16-QAM constellation points.

Download Full Size | PDF

Next, the BER performance when the illuminance at the receiver is relatively high was investigated. In particular, Fig. 13(a) shows the measured BERs for different data rates when the distance between the LED transmitter and the antenna is 15 cm and the measured illuminance on the antenna is 1113 lx. These results show that, as usual, the BER for each antenna increases as the transmission data rate increases. A comparison of the performance of the three antennas shows that Cm504 gives a better performance than DCM and Cm6. Consequently, the Cm504 antenna can support data rates up to 35 Mbps with BERs that are less than the forward error correction (FEC) limit is $3.8{\times }10^{-3}$.

 

Fig. 13. The measured BER with and without antennas, (a) when the illuminance at the receiver is 1113 lx, (b) when the illuminance at the receiver is 285 lx.

Download Full Size | PDF

In addition, the BER performance when the illuminance at the receiver was relatively low was investigated. Figure 13(b) shows the BER results when the distance between the LED and the antenna is increased to 30 cm. Since the power of the LED was fixed, increasing the transmission distance reduced the illuminance on the antenna to 285 lx. The BER results in Fig. 13(b) show that both the DCM antenna and the system that doesn’t include an antenna can support the best data rate of 19 Mbps.

The results in Fig. 13 show that the best choice of the antenna depends upon the illuminance at the receiver. It appears that at higher illuminances the SNR when the antennas are used is high enough for the bandwidth to be an important factor. Consequently, the Cm504 antenna supports the highest data rate. At lower illuminances, it appears that the receiver output signal, and hence SNR, is more important. This means that under these conditions the best antenna contains DCM. However, since this antenna doesn’t absorb all the white light the data rate that it supports is comparable to the data rate supported by the APD alone. The advantage of using an antenna at all illuminances is that, unlike the APD, it will have a very wide FOV in two dimensions.

Finally, the achievable data rates at 285 lx, 409 lx, 627 lx and 1113 lx were measured. It can be seen from Fig. 14, that when the illuminance level is higher than 1000 lx, the Cm504 antenna supports the highest data rate. In addition, at irradiances less than approximately 280 lx the APD alone supports the highest data rate. However, the recommended illuminance level for most work environments is 300 lx or higher and at these illuminances the DCM antenna supports a higher data rate than the APD alone. Furthermore, the general illuminance level recommended for offices is 500 lx and at this illuminance the DCM antenna will support a significantly higher data rate that the APD alone.

 

Fig. 14. The achievable data rate at different illuminances for four different receivers.

Download Full Size | PDF

6. Conclusion

The use of fluorescent antennas in a VLC is a promising approach to improving the data rate supported by the VLC system. In particular, a fluorescent antenna is capable of filtering the incident light and then concentrating light on the photodetector in a VLC receiver. In addition, since this light concentration is based on fluorescence rather than reflection and refraction, it can exceed the étendue limit. This means that the receiver can have both a high concentration gain and a wide FOV.

In this paper, a method of making a fluorescent antenna that can be used to investigate the performance of different fluorescent materials has been described. This method is based upon a glass capillary which is filled with epoxy that has been mixed with a selected fluorophore. The resulting structure acts as a waveguide that concentrates light onto the receiver’s photodetector. Since a capillary can be chosen which is the same size and shape as the photodetector photon leakage at the antenna edge is avoided. Consequently, much higher concentration gains have been achieved compared to those previously achieved with rectangular antennas.

This proposed method has been used to create antennas containing three fluorophores, including one Cm504, which hasn’t previously been used in a VLC system. As expected, since it doesn’t absorb yellow light, the antenna that incorporated Cm504 resulted in a VLC system with the highest bandwidth. The performance of the Cm504 antenna was compared to the performance of the antenna containing two previously studied fluorophores, Cm6 and DCM. Results have been presented to show that the performance of the antenna is related to the illuminance at the receiver. When the illuminance is low, its higher SNR means that the DCM antenna can support the highest transmission data rate. In contrast, at high illuminances, the Cm504 antenna provides the best performance.

If fluorescent antennas are integrated into future systems, the fluorophore can be encapsulated in durable and flexible plastic fibers. Furthermore, the results in Fig. 4 show that the performance of these fibers could be improved by incorporating the fluorophore towards the outer edge of these fibers. In addition, the system’s performance could be improved by using a photodiode at both ends of the fibre or a photodiode at one end and a reflector at the other end [28].