3DS Max following the dimensions and shape of a Boeing 727-. 757 fuselage. For our ... dimensions of the section of the Boeing 737 fuselage [20]. Since.
SNIR Predictions for on-Aircraft VLC Systems Dario Tagliaferri, Carlo Capsoni Dipartimento di Elettronica, Informazione e Bioingegneria Politecnico di Milano Via Ponzio 32, 20133 Milano, Italy {dario.tagliaferri, carlo.capsoni}@polimi.it Abstract— This work reports a set of preliminary predictions of the Signal-to-Noise-plus-Interference Ratio (SNIR) of a Visible Light Communication (VLC) system inside an airplane cabin devoted to provide high-speed Internet access to air travelers. Different conditions of signal sources, noise and interference have been addressed, making use of an in-house developed Modified Monte Carlo (MMCA) ray tracer simulator. The results show that the SNIR is essentially affected only by the interfering sources from the side seat row, while solar noise and other lightings have a relatively low impact. Moreover, we observed that also VLC links using white phosphorescent LED sources are feasible, allowing systems with unity wavelength reuse factor. Keywords—VLC; airplane cabin; MMCA ray tracing; SNIR;
I. INTRODUCTION Visible Light Communications (VLC) are becoming a promising alternative to classical radio communications in many difficult environments, thanks to their intrinsic advantages, such as unlicensed use, non EM pollutant, very high capacity per unit area, and availability of high number of low cost, potential transmitters (LED lights). Among the various applications, inaircraft VLC systems are gaining large attention in recent years, for the possibility to provide low-cost, ubiquitous and green access to high speed Internet in E.M. sensitive spaces. For guaranteeing high QoS links, adequate values of signal-to-noiseplus-interference ratio (SNIR) are required. Inside an aircraft cabin, however, the large number of potential light transmitters (LED reading lamps) makes the communication challenging, due to the interference they also generate while lightings such as ceiling, sidewall or dome lights as well as solar radiation entering through the fuselage windows contribute to increase the noise at the receivers front-end. The possibility of in-flight VLC was already introduced and studied in [1], [2] and [3], while the specific problem of exploring the interference and the SIR was reported by [4], for a simple airplane cabin geometry and sources-receivers configuration. It is therefore of interest to simulate the optical propagation inside an airplane cabin of realistic shape and dimensions, in order to investigate how the environment impact of the VLC systems performance, and to explore the effects of different types of signal (or noise) sources and receiver configurations. In this paper we have addressed these topics with the use of our Modified Monte Carlo (MMCA) ray tracer simulator, that we developed for the characterization of the visible and infrared light propagation inside any arbitraryshaped space. The code is able to provide detailed output data of channel impulse responses and SNIR. We considered separately the effect of each disturbance (noise and interference) on the
VLC systems in order to identify the most critical aspects of the design. The paper is organized as follows: section II explains the main blocks of the software, giving an overview of the involved models and techniques, section III reports the description and the results relative to the selected case studies and finally section IV draw the conclusions. II. SIMULATOR DESCRIPTION A. Propagation algorithm We made reference to the Modified Monte Carlo ray tracing technique [5] to build up a software able to simulate any visible light or infrared propagation channel inside any kind of space, for arbitrary configurations of sources and receivers. The algorithm is conceptually simple, and proceeds as follows. The code preliminary checks whether the direct ray path is free from obstacles (LOS) and then the link budget is evaluated (at each wavelength under tests) by means of the general formula presented in [6]. The received powers and currents are stored together with their arrival times in order to build the impulse responses, if requested. Afterwards, a fixed number of rays are launched from each source position (in case of point sources) to any random direction, according to the radiation pattern. It is enough to generate a random couple of angles in local spherical coordinates (elevation and azimuth), for example with the inversion method in [7], and then rotate the orientation vector of the source (expressed in global coordinates) according to those angles. Each ray is therefore identified by a unit-vector expressed in the global coordinate system. The choice of the local reference system of the source is critical if the radiation pattern is not rotationally symmetric (dependence on the azimuth angle), but since in our study we are dealing only with Lambertian and Phong radiation patterns (which are independent on the azimuth angle), we simplified the implementation by randomly selecting the reference on the surface of the source. Since we modelled the space by using a CAD software (Autodesk 3DS Max), every object is composed of triangular facets each one with its orientation vector and we used the same approach also for the sources. Before launching the ray, the code selects the operating wavelength by randomly selecting (with uniform probability) of a wavelength sample from a given set. The power carried by each ray is equal to the power emitted by the source at the selected wavelength (power spectrum) divided by the total number of transmitted rays for that wavelength (in our case, the same number for each one). Once a ray is generated, the code simulates the multiple bounces propagation in the space by: 1) computing the impact point in the space and the travel time; 2) multiplying the power
associated to the ray for the reflection coefficient of the hit surface; 3) checking which paths from the impact point to the receivers are free from obstacles; 4) computing the link budgets from the impact point to all the visible receivers (free paths) and save the results and the arrival times; 5) generating a new random ray from the impact point according to the reflection pattern of the surface. The channel impulse response for each source-receiver couple is obtained by summing the power/current contributions from each bounce according to their arrival time and wavelength. In case of extended light sources, as commercial illumination lamps, the previous procedure must be modified to account for the effective size of the emitter. In our implementation, we solved this problem by randomly choosing (with uniform probability) the starting point of each ray on the surface of the emitter, and then we proceed as previously indicated. With this technique, emitters of any size (lamps and/or windows, from which the solar radiation can affect the receivers) can be considered. The code provides the impulse response for each source-receiver couple, for each bounce order and desired wavelengths. In this way the user can have a comprehensive view of the temporal spread of the signal and/or interference power. Moreover, the contributions of the artificial noise (lamps) and natural noise (solar radiation) are calculated separately to evidence their effects on the receivers. The simulator provides as output only the total received powers and currents, since the temporal information is of no importance in this case. All the parameters involved in the propagation, like the number of rays, the number of light bounces, the number of sources and receivers, the wavelengths to simulate, etc., are set at the beginning of the simulation. At the end of the propagation algorithm, the SNIR is calculated as indicated in subsection E. Output data are stored on different text files depending on whether they are relative to signal sources or noise sources, to facilitate post-processing operations. We wrote the code in ANSI C language to obtain high performance levels of simulation times, and we further employed the “OpenMP” library to parallelize the algorithm over the available CPUs [8]. Fig. 1 shows an example of wavelength dependent impulse response computed for the empty room space described in [9] (configuration A). Instead of taking scalar reflection coefficients for the objects, we used a combination of white plaster wall and granite to evidence the different behavior of each color (i.e. blue, green, red). B. Airplane cabin modeling We modeled the airplane cabin using the Autodesk 3DS Max software, which allows to create graphically CAD models of any desired object and space. By the use of this CAD software, we are able to model also other types of aircraft (or spaces) that could be of interest for the implementation of VLC systems. By default, each element in 3DS Max is composed of a mesh of triangular facets characterized by a large set of parameters, which can be exported as a text file where all the geometrical information of the mesh are summarized (vertexes coordinates, normals, etc.). The code retrieves not only the geometrical data from the output text file, but also specifies for each object in the scene (and even for each facet) the material, the correspondent spectral diffusion coefficient (taken from the NASA Aster Spectral library [10]) and the reflection pattern.
Fig. 1. Wavelength dependent impulse response of the system configuration A reported by Barry et al. in [9]. The three sample wavelengths are 460 nm (blue curve, wavelength 0), 532 nm (green curve, wavelength 1) and 650 nm (red curve, wavelength 2).
C. Sources and receivers Each light source (i.e. LED lamp) is characterized by the following features: position in space, orientation, extension (in case of extended source), emitted radiant (or luminous) power, emitted spectrum and radiation pattern. Since position, orientation and extension are set in 3DS Max software, and the input power is simply a scalar value specified at the beginning of the simulation, the only features to be modeled are the spectrum and the radiation pattern. VLC signal sources can be either phosphorescent white LED lamps and/or trichromatic RGB LED lamps. In the first case, the white light is obtained by combining the emission of a blue LED with the phosphorescence effect caused by a phosphor coating that turns part of the blue light to yellow light. In the second case (RGB) the white light comes out from the superposition of the emissions of three different LEDs, red, green and blue ones. To obtain an analytical model to use in the software, we approximated the spectra by a weighted sum of Gaussian functions of which the user specifies relative amplitudes and standard deviations. Fig. 2 shows an example of this approximation, reporting a white LED spectrum obtained by phosphorescence, where the blue peak is centered around 460 nm and the yellow one around 550 nm. For RGB sources, the blue peak is always centered around 460 nm, the green one around 532 nm and the red one around 650 nm. The radiation pattern of each source can be approximated by a Lambertian function of the form: R ( )
m 1 cos m ( ) 2
(1)
in which “θ” is the co-elevation angle, measured from the normal direction (orientation) and “m” is the mode number or Lambertian order. Typical beam angles of on-board reading lamps can be as small as 10°, as for the ones in [11]. The receivers usually have a greater number of parameters to be defined with respect to the sources: position, orientation, optical
concentrator parameters (if present, they are the refractive index “n” of the concentrator and the overall field of view, “FOV”), optical filter (if present, we modeled its spectral mask with a Butterworth curve as in [12]), spectral responsivity and noise parameters (electrical bandwidth of the receiver, ambient temperature of the receiver, dark current power, noise factor and gain of the photodiode in case of APDs). Refractive index of the concentrator and overall field of view determine the concentration gain (also referred as geometrical gain) of the receiver, that has the usual expression indicated by [6] which we also used in our software. Depending of the concentrator type, which could be hemispherical or compound parabolic, the filter mask could shift as the incident rays strike with a direction different from the normal one. To handle this case, we made use of the analytical form of the spectral shift reported in [12]. In Fig. 3 is depicted an example of RGB LED light with superposed an example of filter mask used to recover the green component. In this case, the filter is a 5-th order Butterworth mask centered around 532 nm and has a -3dB bandwidth of 100 nm [12].
D. Evalutation of the solar radiation outside the windows When evaluating the sun contribution to the receiver noise, the first step is the modeling of the solar radiation outside the aircraft. In this work, we have always considered the case of clear sky weather, not considering the presence of clouds. From the geometrical position of the Sun with respect to the airplane, we computed the direct and diffuse spectral irradiance in front of each cabin window. We combined the simple but effective models proposed by Bird and Riordan [13] and Gueymard [14] to compute the direct spectral irradiance normal to the direction of the rays and the diffuse spectral irradiance on an arbitrarytilted plane, over a desired wavelength range. For our purposes, the diffuse irradiance is evaluated only on a vertical plane, as for the common case of windows. The algorithm takes as inputs the geographical position of the airplane, the day of the year, the hour of the day, the altitude of the receivers (we set to 8000 m, as a common cruise flight altitude), the absolute temperature and pressure (retrieved from [15] for the standard atmosphere case) and the relative humidity at that height above sea level. In addition to these data, the model requires the knowledge of the ozone layer thickness (estimated through [16]) and the aerosols turbidity factor [17]. An example of direct and diffuse irradiance spectra is reported in Fig. 4, over the 200-1100 nm wavelength range (with a spacing of 1 nm). The effective noise power that propagates inside the cabin must also take into account the transmissivity of the window. We simulate this phenomenon by using the simple multi-layer model proposed in [18].
Fig. 2. Example of white LED spectrum obtained by phosphorescence, over the 380-780 nm range. Relative amplitudes of the blue and yellow (phosphorescent) peaks are equal to 1, while standard deviations are, respectively, 12 nm and 55 nm (curve fitting from average real spectra).
Fig. 4. Example of direct (red) and diffuse (blue, dashed) irradiance computed at 17:30 p.m. at the latitude and longitude of Milan, Italy. Elevation angle of the Sun: 32.13°, azimuth angle 260°. Total direct irradiance is 1000.61 W/m2, total diffuse irradiance 128.52 W/m2.
E. Signal-to-Noise-plus-Intereference Ratio computation After having simulated the signal and noise propagation inside the cabin, the last step is the computation of the SNIR at each receiver. The general expression is: Fig. 3. Example of white RGB LED spectrum (black, dashed) with a filter mask to recover the green component (green). Relative amplitudes of the spectrum are all equal to one and standard deviations of the blue, green and red peaks are, respectively, 12 nm, 18 nm and 28 nm (curve fitting from average real spectra).
SNIR
S NI
(2)
where “S” is the squared electrical current generated by the sum of the useful signals, “I” is the squared electrical current generated by the sum of the interfering signals by the other light sources and “N” is the sum of the shot noise power generated by all the received radiation (unmodulated background lights and signal sources) and the thermal noise. The latter is simply computed as σ2th = (4kBTB)/RL where “kB” is the Boltzmann constant, “B” is the electrical bandwidth of the whole receiver, “T” is the ambient temperature and “RL” is the load resistance of the photodetector, or the feedback resistor of a transimpedance amplifier. Depending on the type of the photodetector, the noise term indicated with “N” assumes two different expressions, [19]. We implemented the two formulas under the assumption that the spectrum of RGB consists of a background light component (unmodulated colors) and a signal or interfering light (modulated colors).
three receivers in figure (R1, R2, R3) to simulate a 3 separate downlink VLC systems, while the other three (S4, S5, S6) interfere with the previous. All the reading lamps emit a luminous power of 100 lumens. There are also four illumination lamps, labeled in the figure with ES7, ES8, ES9 and ES10 (E stands for “extended”), which has been created following the ceiling wash lights specifications in [11]. They are 70 cm by 8 cm length x width, oriented toward the ceiling and tilted of 30 degrees from the vertical, and emit a luminous power of 1926 lumens each (156 cd). The power spectrum of the ceiling lights is a warm white LED light (phosphorescent). The radiation pattern was assumed to be perfect Lambertian (m = 1). In addition to the noise generated by these lamps, we added the contribution of the solar radiation. In all the simulations the solar spectrum was the one correspondent to 12:30 p.m. in the Milan area on May 7. The elevation of the Sun was set to 55.13°, while the azimuth was 153.82°, measured from the North direction (y axis).
III. SIMULATIONS AND RESULTS We have studied various configurations of sources and receivers in two VLC systems configurations and the same airplane cabin configuration. The latter has been recreated in 3DS Max following the dimensions and shape of a Boeing 727757 fuselage. For our purposes (characterization of the VLC channel), we considered three seats rows only, with the receivers placed in the central one. This assumption implies that interfering lights from further reading lamps have no effect on the receivers, which has to be verified. Fig. 5 reports the main dimensions of the section of the Boeing 737 fuselage [20]. Since the computational complexity of the code (i.e. the simulation time) is linearly proportional to the total number of facets [5] [7], we simplified in some way some details of the cabin and of the seats in order to have more manageable simulations without radically change the geometry of the space. Fig. 6 and 7 show, respectively, the front and the side views of the airplane cabin as it appears in the 3DS Max workspace. As it can be seen, some details, especially some curved surfaces as the sidewalls, are modeled as planes to reduce the complexity of the scene. The maximum dimensions of the cabin are the following: 5 m along the x axis (which is taken from left to the right in the side view), 3.54 m along the y axis (which is taken from left to the right in the front view) and 2.19 m along the z axis (vertical direction). The latter two are in accordance with the maximum width and height indicated in Fig. 5. Airplane cabin walls are modeled with a high reflective material among the ones in [10] (white plaster wall), while seats are modeled with a less reflective material (60% of the reflectance of the white walls). Both the two have a perfectly diffusing reflection pattern (percentage of diffusely reflected power p = 1 and Lambertian order of the specular lobe m = 1, as specified in [21]). Windows are simply modeled as oriented square surfaces, of 25 cm side. We considered them as homogeneous glass multilayers of 1 cm thick, with the standard refractive index reported in [18] and a reflection pattern with p = 0 and m = 400 (strong specular reflection). A. System 1 The first configuration considered is depicted in Fig. 8. Three sources (labeled as S1, S2 and S3) are pointed towards the
Fig. 5. Section of the Boeing 737 fuselage cabin with the main dimensions [20].
Fig. 6. Front view of the airplane cabin model as it appears in 3DS Max. Black lines are the edges of the facets that constitute the objects.
Fig. 7. Side view of the airplane cabin as it appears in 3DS Max. Small blue squares represent the windows, while yellow and red planes, respectively, sources and receivers.
We simulated both the phosphorescent white sources and the RGB ones, comparing the results. All the simulations we carried out with the following common parameters: the wavelengths under test (460 nm, 532 nm and 650 nm), the number of rays launched for each source and for each wavelength (1000000) and the number of light reflections (5). In case of phosphorescent reading lamps, receivers cannot reject the interference through optical filters, which is governed by the emitting angle of the sources and the space geometry. Therefore, we tested four possible beam angles for the sources, with the receivers pointed vertically and equipped with glassy hemispherical lenses (FOV = 70°). Fig. 9 shows the spectra of the reading lamps (daylight white) and of the illumination lamps (warm white). The RGB sources allow the modulation of the three colors separately, thus we assigned a color to each source (with the scheme indicated in Table I) and the receivers with optical filters of the same kind as in section II. Fig. 10 shows the filters masks and reports their parameters in the caption. In this last case, however, we assumed the sources and the receivers to be broadcast (m = 1 for the sources and FOV = 85° for the receivers), in order to check if the SNIR was enough even for strong interference levels. In all the simulations, we assumed that the receivers had a fixed receiving area of 1 cm2.
Fig. 9. Daylight white spectrum (blue, solid line) assigned to the reading lamps vs. warm white spectrum (red, dashed line) assigned to ceiling lights.
Fig. 10. Transmissivity of the three filters employed together with RGB sources. The black curve (solid line) is the RGB sample spectrum; the blue curve (dashed-dotted line) is the blue filter, centered at 460 nm and with a 50 nm bandwidth; the green curve (dashed line) is the green filter, centered at 532 nm and with a 75 nm bandwidth; the red curve (dotted line) is the red filter, centered at 650 nm with a 100 nm bandwidth.
TABLE I. POSITION OF THE SIGNAL SOURCES AND THE RECEIVERS INSIDE THE CABIN, WITH ASSOCIATED COLORS.
Fig. 8. Distribution of sources and receivers inside the cabin for system 1.
Name
Position
Name
Position
(color)
[m]
(color)
[m]
S1 (R)
(2.8,3,1.55)
R1 (R)
(2.8,2.14,0.65)
S2 (B)
(2.8,3.1,1.55)
R2 (B)
(2.8,2.64,0.65)
S3 (G)
(2.8,3.2,1.55)
R3 (G)
(2.8,3.14,0.65)
S4 (R)
(2.8,0.3,1.55)
-
-
S5 (B)
(2.8,0.4,1.55)
-
-
S6 (G)
(2.8,0.5,1.55)
-
-
The results are summarized in Table II. For each kind of source, the SNIR results are reported at each receiver site. For the calculation of the thermal noise powers, we took the following values as reference: 100 MHz for the electrical bandwidths of the receivers, 27°C for the operating temperatures and 0.2 nA for the dark currents. In this system configuration, the receivers are not exposed to the direct sunlight (which can enter only from windows W2, W4 and W6), but they collect the diffuse sunlight entering from all the windows. As far as phosphorescent lamps are concerned, we can see that in case of highly directive sources (beam angles 13° [11] and 30°), the SNIR level is high enough to overcome the interfering lights and the background noise, even with the receivers unequipped with filters and not strongly directive (FOV = 70°). According to [22], when the SNIR values are in the order of 9 dB, by employing a DCO-OFDM modulation on a 10 MHz bandwidth, (available by equalizing the frequency response of the LED lamp), the achievable throughput for a target BER of 10-3 can reach 40 Mbit/s with a 16-QAM modulation and a DC bias of 7 dB. Considering the differences in the geometry of the cabin and in the sources-receivers configuration, these results are consistent with the ones reported in [4]. In case of wide-angle sources, the links are no more reliable, unless optical filtering is employed. It is worth to underline that we forced each receiver to have a wide field of view, because of the possibility that the user move or rotate the receiving equipment. With a wide field of view, the probability of maintaining the LOS path is maximized. In this regard, color filtering appears to be the only solution that allows to have both broadcast sources and broadcast receivers. In order to see the influence of the solar radiation on the system we show also the SNIR results obtained without the natural noise contribution (Table III). As it can be seen, for this kind of system configuration, the SNIR values are generally increased by fractions of decibels in most of the cases. The effect of the diffuse sunlight inside the cabin is therefore negligible with respect to the one of the adjacent signal sources.
TABLE II.
TABLE III.
SNIR RESULTS FOR DIFFERENT TYPES OF SOURCES. THE SOLAR RADIATION IS NOT CONSIDERED IN THIS CASE. SNIR [dB] (no solar radiation)
Source type (BA) R1
R2
R3
PHOS. (13°)
43.6
61.15
67.8
PHOS. (30°)
9.265
9.78
20.625
PHOS. (45°)
2.544
0.785
8.78
PHOS. (120°)
- 0.88
- 2.6
3.5
RGB (120°)
17.18
23.85
27.8
features as in the previous experiment, with the same color assignment scheme repeated for every row of seats. The receivers are placed exactly in the same position as before, and have the same characteristics. Table IV contains the SNIR values associated to each receiver, for the same experiments of the previous case. All The SNIR values are reduced by an amount that basically depends on the position of the receiver, but the robustness of the communication is substantially unaffected. This let us guess that for a rough estimation of the VLC system performance it is sufficient to consider the sources belonging to the same row of seats, that in most cases illuminate both their assigned receivers and the adjacent ones. The SNIR at the receiver is almost completely determined by the signal power and the LOS power of the interfering sources.
SNIR RESULTS FOR DIFFERENT TYPES OF SOURCES. SNIR [dB]
Source type (BA) R1
R2
R3
PHOS. (13°)
43.55
61.1
67.7
PHOS. (30°)
9.26
9.77
20.61
PHOS. (45°)
2.54
0.78
8.77
PHOS. (120°)
- 0.9
- 2.62
3.4
RGB (120°)
17.15
23.8
27.7
B. System 2 The second system we analyzed is an extended version of the previous one: we inserted, in addition to the main sources, all the reading lamps associated to the other seats (Fig. 11). The goal is to explore the possible detrimental effect of the interfering sources of adjacent seat rows, which for the considered geometry are not in line of sight with the receivers (the LOS path is blocked by the back of the seat). The sources have the same
Fig. 11. Distribution of sources and receivers inside the cabin for system 2. New interfering sources are added with respect to the previous case: they are placed in the same relative position with respect to their seats as before.
TABLE IV.
SNIR RESULTS FOR DIFFERENT TYPES OF SOURCES UNDER THE INFLUENCE OF ALL THE ADJACENT READING LAMPS. SNIR [dB]
Source type (BA) R1
R2
R3
PHOS. (13°)
42.22
57.08
63.09
PHOS. (30°)
9.12
9.70
20.48
RGB (120°)
14.38
22.05
24.60
IV. CONCLUSIONS The paper reports SNIR predictions for in-aircraft downlink VLC systems, obtained with our MMCA ray tracer software. We analyzed two different VLC system configurations inside a CAD model of the Boeing 737 airplane cabin, exploring various configuration of sources (reading lamps) and receivers in presence of solar background light and artificial lights (illumination fixtures). Under some hypothesis regarding the directivity of the sources and the geometry of the links (availability of the LOS path), these preliminary results show that there is the possibility to communicate with common phosphorescent reading lights as the one we reported as reference [11], without employing RGB sources and/or optical filters at the receivers. This opportunity is interesting because it would lead to a lower cost, unity wavelength reuse system. Moreover, we observed that the signal sources relative of the frontal and rear seat rows have a relatively low impact on the performances of the system, as far as their LOS path to the receivers is blocked. The last condition is realistic inside an airplane cabin, where seat backs and passengers often shield the receiver from adjacent lights. We also found that, under some hypothesis, the solar noise coming from the windows can be reasonably neglected for a rough estimation of the SNIR
REFERENCES [1]
[2]
[3]
[4]
[5]
S. Dimitrov, R. Mesleh, H. Haas, M. Cappitelli, M. Olbert and E. Bassow, "Path Loss Simulation of an Infrared Optical Wireless System for Aircrafts," Global Telecommunications Conference, 2009. GLOBECOM 2009. IEEE, Honolulu, HI, 2009, pp. 1-6. doi: 10.1109/GLOCOM.2009.5425552 D. Krichene, M. Sliti, W. Abdallah and N. Boudriga, "An aeronautical visible light communication system to enable in-flight connectivity," 2015 17th International Conference on Transparent Optical Networks (ICTON), Budapest, 2015, pp. 1-6. doi: 10.1109/ICTON.2015.7193336 R. Perez-Jimenez, J. Rufo, C. Quintana, J. Rabadan and F. J. LopezHernandez, "Visible light communication systems for passenger in-flight data networking," Consumer Electronics (ICCE), 2011 IEEE International Conference on, Las Vegas, NV, 2011, pp. 445-446. doi: 10.1109/ICCE.2011.5722675 C. Quintana, V. Guerra, J. Rufo, J. Rabadan and R. Perez-Jimenez, "Reading lamp-based visible light communication system for in-flight entertainment," in IEEE Transactions on Consumer Electronics, vol. 59, no. 1, pp. 31-37, February 2013. doi: 10.1109/TCE.2013.6490238 F.J Lopez Hernandez, R. Perez Jimenez and A. Santamaria. “Monte Carlo calculation of impulse response on diffuse IR wireless indoor channels”. Communications, IEEE transactions, Vol 43 1613-1623, 1998.
[6]
[7]
[8] [9]
[10]
[11]
[12] [13]
[14]
[15]
[16] [17]
[18] [19] [20] [21]
[22]
Kahn, J.M.; Barry, J.R., "Wireless infrared communications", Proceedings of the IEEE , vol.85, no.2, pp.265,298, Feb 1997 M. I. S. Chowdhury, W. Zhang and M. Kavehrad, "Combined Deterministic and Modified Monte Carlo Method for Calculating Impulse Responses of Indoor Optical Wireless Channels," in Journal of Lightwave Technology, vol. 32, no. 18, pp. 3132-3148, Sept.15, 15 2014. doi: 10.1109/JLT.2014.2339131 OpenMP Library, documetation available at: “http://openmp.org/wp”. J.R. Barry, J.M. Khan, W.J. Krause, E.A. Lee, D.G. Messerschmitt. “Simulation of multipath impulse response for indoor wireless optical channels”. Selected Areas in Communications. IEEE journal Vol 11, 367379, Apr. 1993. Baldridge, A. M., S.J. Hook, C.I. Grove and G. Rivera. “The ASTER Spectral Library Version 2.0”. Remote Sensing of Environment, vol 113, pp. 711-715, 2009, available at: “http://speclib.jpl.nasa.gov“. UTC Aerospace Systems, documentation available at: “http://utcaerospacesystems.com/cap/systems/Pages/Interiors-documentlibrary.aspx“. Ho K, Kahn JM. “Compound parabolic concentrators for narrowband wireless infrared receivers”. Opt. Eng. 0001;34(5):1385-1395. Bird, R. E., and Riordan, C. J. "Simple Simple Solar Spectral Model for Direct and Diffuse Irradiance on Horizontal and Tilted Planes at the Earth's Surface for Cloudless Atmospheres." Journal of Climate and Applied Meteorology. Vol. 25(1), January 1986; pp. 87-97. Gueymard Christian A. “Parametrized transmittance model for direct beam and circumsolar spectral irradiance”. Solar Energy, vol. 71, no. 5, pp. 325-346, 2001. G. S. Aglietti, S. Redi, A. R. Tatnall and T. Markvart, "Harnessing HighAltitude Solar Power," in IEEE Transactions on Energy Conversion, vol. 24, no. 2, pp. 442-451, June 2009. doi: 10.1109/TEC.2009.2016026. Van Heuklon T.K. “Estimating atmospheric ozone for solar radiation models”, Solar Energy, 22, 63- 68, 1979. Shettle, E. P., & Fenn, R. W. (1979). Models for the aerosols of the lower atmosphere and the effects of humidity variations on their optical properties (No. AFGL-TR-79-0214). AIR FORCE GEOPHYSICS LAB HANSCOM AFB MA. Furler, Reto A.. "Angular Dependence of Optical Properties of Homogeneous Glasses." ASHRAE Transactions 98, Part 1, 1991. S. Donati “Photodetectors: Devices, Circuits and Applications”, Prentice Hall, 1999. “http://www.satcom.guru/2016/03/reflecting-on-boeing-727.html”, Satcom website. S. Rodriguez Pérez, R. Perez-Jimenez, F.J Lopez Hernandez, O. B. Gonzalez Hernandez and A.J Ayala Alfonso, “Reflection model for calculation of the impulse response of IR-wireless indoor channel using ray tracing algorithms”, Microwave and optical technology letters, p. 296300, 2002. J. Armstrong and B. J. C. Schmidt, "Comparison of Asymmetrically Clipped Optical OFDM and DC-Biased Optical OFDM in AWGN," in IEEE Communications Letters, vol. 12, no. 5, pp. 343-345, May 2008. doi: 10.1109/LCOMM.2008.080193
