Digital Color Shift Keying With Multicolor LED Array Volume 8, Number 4, August 2016 Naoya Murata Yusuke Kozawa Yohtaro Umeda
DOI: 10.1109/JPHOT.2016.2582645 1943-0655 Ó 2016 IEEE
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
Digital Color Shift Keying With Multicolor LED Array Naoya Murata, Yusuke Kozawa, and Yohtaro Umeda Department of Electrical Engineering, Graduate School of Science and Technology, Tokyo University of Science, Noda 278-8510, Japan DOI: 10.1109/JPHOT.2016.2582645 1943-0655 Ó 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Manuscript received June 7, 2016; accepted June 15, 2016. Date of publication June 20, 2016; date of current version July 7, 2016. This work was supported by Japan Society for the Promotion of Science (JSPS) KAKENHI under Grant 15K21397. This paper was presented in part at the 2015 IEEE Workshop on Visible Light Communications and Networking (VLCN) at the 2015 IEEE International Conference on Communications. Corresponding author: Y. Kozawa (e-mail:
[email protected]).
Abstract: Color shift keying (CSK) is a visible light communication (VLC) intensity modulation scheme, which is outlined in IEEE 802.15.7, that transmits data imperceptibly through the variation of the color emitted by red, green, and blue light-emitting diodes (RGB LEDs). In this paper, digital CSK (DCSK) is presented, which represents CSK symbols by digitally controlled multiple multicolor LEDs to avoid the effect of LED nonlinearity. By using this scheme, the DCSK system using nine RGB LEDs can transmit IEEE 802. 15.7 CSK constellations with an error-vector-magnitude (EVM) performance on an x y chromaticity of 3%. The bit-error-rate (BER) evaluation of both the conventional and the proposed DCSK system shows that the DCSK system achieves similar BER to that of CSK with a linear variable current driver including predistortion, with taking into account optical wireless channels including the LED response, undesired color shift, and channel crosstalk by the practice characteristics of RGB LEDs, RGB bandpass filters, and photodiodes. Index Terms: Visible light communications (VLC), light-emitting diode (LED), color shift keying (CSK), digital color shift keying (DCSK), LED non-linearity.
1. Introduction Visible light communication (VLC) is a wireless communication technology using light emitting diodes (LEDs) [1], [2]. For increasing attraction of VLCs, the modulation methods of VLCs should consider the effects of modulated light signal on the human eye, such as flicker mitigation, dimming control support, color temperature control support, high color rendering index (CRI), and so on. As a result, the first IEEE standard for VLCs in the form of IEEE 802.15.7 proposed three modulation methods for flicker mitigation and dimming support [3], [4]: On-Off Keying (OOK), Variable Pulse Position Modulation (VPPM), and Color Shift Keying (CSK). In particular, the PHY III level of IEEE 802.15.7 suggests the CSK system to increase the data speed by combined output of red, green, and blue in RGB LED (i.e., trichromatic LED: TLED). This CSK can also utilize LED original frequency response of tens MHz, while blue LED with phosphor is limited frequency response to several MHz by a phosphor [5], [6], although multiple color LEDs increase the cost of LED luminaire compared to using the blue LED with phosphor [7]. In CSK, M signal points are mapped inside the RGB constellation triangle on the CIE x y chromaticity diagram [8] formed by the center wavelength of the three RGB LEDs used in the TLED source. The signal points are transmitted by varying the intensities of the RGB LEDs. When the individual intensities of the RGB LEDs are PR , PG ,
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
and PB , the total intensity is constant, i.e., PR þ PG þ PB ¼ 1, for flicker mitigation, although each intensity may have a different instantaneous output power. For the dimming control, CSK employs amplitude dimming and controls the total intensity. Recently, Singh et al. [9] have presented enhanced CSK using a Quad LED (QLED) with four colors for improving the performance of the conventional CSK using TLED by employing a novel demodulation method and symbol mapping. This four color system can realize a 4-D signaling scheme and instantaneous intensities detection scheme in the 4-D signal space. This system enables gray mapping of symbols and increasing the signal point distance. Moreover, in [10], higher order modulations, such as 256-CSK, 1024-CSK, and 4096-CSK using QLED, have been evaluated over AWGN and diffuse optical channels. The uncoded and unequalized 4096-CSK using QLED can achieve data rates of up to 288 Mbit/s in 12 MHz bandwidth. However, these multi-level CSK systems suffer from LED non-linearity effects in adjusting to the desired optical intensity by changing the current driving with digital-to-analog converters (DACs) (i.e., analog controlled LED). This analog controlled LED causes the LED non-linearities of 1) current to voltage (I-V) characteristics and 2) intensity to current (-I) characteristics [11]. Moreover, the analog controlled LED also causes 3) undesired color shift, which is the peak wavelength shift with amplitude modulation (AM). In [12], the peak wavelengths of red AlGaInP shifts towards shorter wavelength with AM, while green and blue InGaN diodes shift towards longer wavelength with AM. Against problems 1) and 2), Monteiro and Hranilovic proposed a linear variable current driver including pre-distortion to linearizing LED output intensity [13]. This variable current driver allows desired driven current flows I into LED to linearly by open collector nature of OP AMP. In addition, due to the non-linear relation 2), a pre-distortion process is designed to operate with the only first 8 bits of the 10-bits DAC are assigned to drive each LED up to a 200 mA limit. However, these architectures increase the system complexity and cannot prevent the problem 3) because the undesired color shift is caused by the quasi-Fermi level away from the band edge with the increase in the injection current. In this work, a digital CSK (DCSK) system, based on digitally controlled multiple TLEDs, is presented [14], [15]. This DCSK system using multiple TLEDs can represent the conventional CSK symbol with the total amount of each non-varying “On-Off” intensity of multiple LEDs. This straightforward system can be considered as an application of digitally controlled LED concept in [16] to CSK. In this paper, our goal is to consider as a required number of TLEDs for representing the constellations designed using the IEEE 802.15.7 design rule (see Section 3) and, unlike previous work [15], clarify an effect of channel gain between multiple TLEDs (i.e., TLED array) and a color-sensor at the receiver (see Section 4). Additionally, we compare the performance of DSCK with those of the conventional analog controlled CSK (ACSK), ACSK with the linear variable current driver including pre-distortion (ACSK with undesired color-shift), and ACSK with linearly controlled LED (ideal ACSK) (see Section 5).
2. Proposed DCSK System With Multi-Color LED Array Fig. 1 depicts the system model of DCSK using multiple TLEDs. At the transmitter of DCSK system, NTx multi-color LEDs are used for representing the CSK symbol without DAC, where the multi-color LED is composed of Ncolor colors LEDs (e.g., TLED has Ncolor ¼ 3 color LEDs). At the receiver, NRx color-sensors are used, where the color-sensor has Ncolor photodiodes with narrowband optical filters of desired wavelengths, and the receiver is similar system to the conventional CSK receiver. Basically in this paper, we use a single ðNRx ¼ 1Þ color-sensor for fundamental analysis of DCSK. In [15], we confirmed the improvement of the symbol error rate (SER) performance of DCSK when NRx increases. In the DCSK system, the total number of LEDs and PDs is Nt ¼ Ncolor NTx , and Nr ¼ Ncolor NRx , respectively. DCSK represents the CSK symbol with the total amount of each “On-Off” intensity of multiple multi-color LEDs. There are Nt control points in the DCSK transmitter, and these Nt control points are controlled by only “on-off” signaling. The signal processing is therefore relatively easy to
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
Fig. 1. System structure of DCSK using NTx TLEDs (red, green, and blue) and a color sensor ðNTx ¼ 3Þ.
implement because the parallel signal processing is implementable by simple digital signal processing, and the “on-off” LED’s driven current can be managed by only current-controlled registers. DCSK with Nt LEDs can represent ðNTx þ 1ÞNcolor constellation points in Ncolor -dimensional signal power space because NTx þ 1 multi-level intensity can be represented by the “On-Off” signalings of NTx LEDs at each color. However, we don’t use the constellation points all because the total intensity of the constellation points is not constant, that causes the flicker effect. For the flicker mitigation, the proposed DCSK system turns on NTx at every data transmission, and keep the total intensity (i.e., optical intensity per symbol) constant. Thus, the number of DCSK constellation points can be written by [15] Mmax ¼
NTx þ Ncolor 1 NTx
(1)
indicates binomial coefficient. For the dimming control of DCSK, two kinds of scheme are considered [17]. One is the dimming control by pulse width modulation (PWM) which changes the duty cycle corresponding to desired dimming level. The other is the dimming control by digital pulse amplitude modulation (PAM) using collaborative multiple LEDs. For higher data rate, it is better to use the PAM dimming control. In the practical case, however, the number of LEDs in a TLED array is limited and the LEDs should be used for representation of the CSK constellations with maximum accuracy (see Section 3). Therefore, dimming control by PWM is better to employ for the DCSK system. Of course, when a TLED array has more TLEDs than desired constellation can be represented, PAM or hybrid dimming control method of PAM and PWM is applicable for achieving higher spectral efficiency. This DCSK system can also control the target color (e.g., color temperature) by adjustment of the constellation design rules using Billiards Algorithms [18] or interior point methods [19], as well as the conventional CSK systems. However, we need to consider the effect of which TLED array face send what color of a transmission signal to human eyes. One way of solving the issue is to apply a random mapping, or a cyclic mapping that cyclically rotate the active array elements in [14]. The active array mapping methods may improve not only the life span consideration but the illumination performance as well by averaging the emission color of each TLED element (i.e., the target color is averaged over TLED array face). Complete experimentally analyses of these techniques are subjects for future work. where
3. IEEE 802.15.7 CSK Constellation With DCSK The DCSK constellation points are discrete signal points on the CIE x y chromaticity diagram, and therefore, we must have an error margin for representing the conventional ACSK constellations. In this paper, we show the required number of TLEDs for representing the constellations designed using the IEEE 802.15.7 design rule within a permissible error range
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
Fig. 2. (a) EVM on x y chromaticity against the number of TLEDs for 4-DCSK, 8-DCSK, and 16DCSK. (b), (c) and (d) Comparison with IEEE 802.15.7 CSK constellation points and DCSK signal constellations points on x y chromaticity when the permissible error range is within 3%. (b) M = 4. (c) M = 8. (d) M =16.
and define the permissible error as Error Vector Magnitude (EVM) on x y chromaticity. EVM is given by rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi offi P n 2 2 ^ ^ ðx x Þ þðy y Þ 1=M M j j j j j¼1 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi E ðNTx ; MÞ ¼ 100 n o P 2 þy 2 x 1=M M j¼1 j j
(2)
where ðxj ; yj Þ is x y chromaticity value of the IEEE 802.15.7 CSK symbol, and (x^j , y^j ) is that of the DCSK symbol. Fig. 2(a) shows the EVM performance on x y chromaticity against the number of TLEDs, NTx , for 4-DCSK, 8-DCSK, and 16-DCSK. This figure shows that the EVM performance can be improved when the number of TLEDs increases. Specifically, the EVM performance becomes almost 0% when NTx is 3, 18, and 9 for 4-DCSK, 8-DCSK, and 16-DCSK, respectively. In this work, we consider the permissible error range against IEEE 802.15.7 CSK constellation points because the required number of TLEDs must be increased for achieving EVM is 0%. When the permissible error range is within 3%, the number of TLEDs can be reduced to 3, 8, and 9 for 4-DCSK, 8-DCSK, and 16-DCSK, respectively, and the EVM performance of 4-DCSK, 8-DCSK, and 16-DCSK is 0.03%, 2.27%, 0.06%, respectively (see Fig. 2(b), (c) and (d)). When the permissible error rate is 3%, nine TLEDs can dynamically change between the CSK constellations of M ¼ 4, 8, and 16. However, we also should consider the hybrid CSK method of DCSK and the linear variable current driver including pre-distortion for linearizing the higher-order CSK modulations and wider dimming resolution with high accuracy because the number of LEDs in a TLED array is limited. In the hybrid CSK system, the proposed system can offer the flexibility for linearizing methods, and relaxation of demands for higher resolutions of the linear variable current driver and DAC. The hybrid system allow greater flexibility for both lower order polynomial fitting I characteristics for pre-distortion and the lower number of LEDs for DCSK. In next section, we also evaluate the bit error rate (BER) performances with the DCSK constellation points when the permissible error range is 1%, 3%, 5%, and 15%.
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
4. Channel Gain Between TLED Array and Color-Sensor In this section, we analyze and evaluate the BER performances of the DCSK system to verify the effects of channel gain between TLED array and color-sensor. In the analysis, we consider generalized DCSK model has Nt LEDs and Nr PDs. We also evaluate the effects of the constellation points with the permissible error range against the BER performance.
4.1. BER Analysis of DCSK The DCSK signal is generated by on-off signalings of Nt LEDs. When the jth symbol ðj ¼ 1; . . . ; MÞ is selected in the M-DCSK system, the transmission signal xj ðt Þ is given with Nt -D vector as h i (3) xj ðt Þ ¼ x1j ðt Þ xkj ðt Þ xNj t ðt Þ 1; if k th LED turns on xkj ðt Þ ¼ 0; if k th LED turns off: (4) The on-off signals are converted to an optical signals at each LED. The optical signal sðt Þ is given as h i sj ðtÞ ¼ s1j ðt Þ skj ðt Þ sNj t ðt Þ (5) skj ðtÞ ¼Pt xkj ðt Þ gt k ðtÞ
(6)
where skj ðt Þ is optical signal of k th LED for representing jth DCSK symbol, Pt is transmission optical power of unit LED, and gt k ðt Þ is k th LED response with a 3-dB modulation bandwidth of ft k . In [20] and [21], blue LED and white LED responses are modeled by the first-order function, and the fitted response is quite representative of LEDs. However, the other color LED responses don’t be reported yet. In our model, this LED response therefore can be written as [11] gt k ðt Þ ¼ 2ft k expð2ft k t Þ:
(7)
This impulse response is defined by exponentially attenuated model, and the RGB LEDs have a generally 3-dB bandwidth up to 20 MHz [21]. In our DCSK system, for mitigating flicker effect, the P t 1 PNt 2 PNt M total optical power of TLED array is always constant, N k ¼1 sk ðt Þ ¼ k ¼1 sk ðt Þ ¼ ¼ k ¼1 sk ðt Þ. In this paper, we consider pure LOS channel, H, which is short distance aligned case, and is very close to an ideal impulse response ðt Þ [21]. The optical channel response from the k th LED to the k 0 th PD is given as [2] ðm þ 1ÞA dkk 0 m 0 0 t (8) cos ð Þgð Þcos ð Þrect hkk 0 ðt Þ ¼ kk c kk 2 c 2dkk 0 c 1 gð c Þ ¼ 2 (9) sin ð c Þ where m is the order of Lambertian radiation pattern, A is the receiving area, 1=2 is the semiangle at half illuminace of the LED, d is the distance between the transmitter and the receiver, is the angle of irradiance, is the angle of incidence, gð Þ is the gain of an optical concentrator, c is the width of the field of vision (FOV) at a receiver, and c is the speed of light [2]. The Lambertian radiation pattern and rectangular function rectðx Þ can be written by, respectively log2
m¼
(10)
log cos1
rectðx Þ ¼
Vol. 8, No. 4, August 2016
2
1; for jx j 1 0; for jx j > 1:
(11)
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
At the receiver, the received optical signal is converted to current signal in each PD with BPF, and the current signal is converted to voltage signal at the Trans-Impedance-Amplifiers (TIA). In our model, we assume the transimpedance of TIA is 1 . The voltage signal is fed into a correlator, and converted to digital signal by A/D convertor. The correlator outputs, rj , of jth DCSK symbol can be written as h i rj ¼ r1j ; . . . ; rkj 0 ; . . . ; rNj r (12) where rkj 0 is correlator output of k 0 th PD. The correlator output of k 0 th PD rk 0 is obtained by rkj 0
1 ¼ Ts
ZTs X Nt 0
k ¼1
Gkk 0 skj ðt Þ hkk 0 ðt Þ gr
k 0 ðt Þ
þ nk 0 ðt Þdt
¼skj 0 þ nk 0
(13)
where Ts is the symbol duration, Gkk 0 is the responsivity of the photodiode between the k th LED to the k 0 th PD, gr k 0 ðt Þ is an impulse response of receiver with the k 0 th PD, and nk 0 ðtÞ is receiver noise before correlator, and this is modeled as independent and identically distributed additive white Gaussian noise (AWGN) with the variance 2 ¼ N0 =2. Here, the responsivity Gkk 0 can be written by Z1 Gkk 0 ¼
0
Ptk ðÞkr ðÞd
(14)
0 0
where Ptk ðÞ is the emission spectrum per watt of the k th LED, and kr ðÞ is sensitivity of the k 0 th PD with Band Pass Filter (BPF). Finally, A/D converted signal is demodulated by Maximum Likelihood Estimation (MLE) that selects the symbol that has minimum Euclidean distance between received signal in Ncolor -D signal power space [9]. When symbol j is transmitted, the Euclidean distance between the received signal and the preamble signal, i.e., j 0 th DCSK symbol, is given as 8 2 < jnj ; j ¼ j0 0 j (15) Dðrj ; s0 Þ ¼ j 2 : s0 s0 j 0 þ n ; j 6¼ j 0 8 2 < jnj ; j ¼ j0 ¼ (16) 2 : P 2 hj hj 0 þ n ; j 6¼ j 0 t where hj is the received power with a unit transmission power in the case that symbol j is transmitted. In demodulation, the received signal is compared with preamble signals on the Ncolor -D signaling space [9], and the maximum likelihood signal is chosen. In the case when the jth symbol is transmitted, the pair wise error probability (PEP) that the signal is mistaken with the j 0 th symbol is written as j j0 PEPj!j 0 ¼ p D rj ; s0 G D rj ; s0 : (17) The symbol error performance union bound is therefore given as [30] 0sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 M 1 X M X 2 ðPt Þ2 hj hj 0 2 A Q@ SER F M j¼1 j 0 ¼jþ1 4N0
(18)
where k:kF is the Frobenius norm.
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
Fig. 3. (a) Relative intensity of TLED [22]–[24]. (b) Transmission of RGB BPF [27]–[29] and PD photosensitivity (in amperes per watt) [26].
4.2. Performance Evaluation In this paper, we evaluate the BER performances of the DCSK system using TLED array ðNcolor ¼ 3Þ and color-sensor ðNr ¼ 3Þ. In our simulation model, we model Red LED spectrum, Green LED spectrum, Blue LED spectrum, into LED spectrum modeling Ptk ðÞ by measuring OS5RKA5111P, OSG58A5111A, OSB56A5111A, respectively [22]–[24], with illuminance spectrometer IM-1000 of TOPCON Technique corporation [25], where wavelength resolution is 1nm (see Fig. 3(a)). In Fig. 3(a), LED spectrums are also presented when LED drive current is 0:1Imax , 0:5Imax , and Imax , where Imax is forward current of each LED. While, for model0 ing the sensitivity of PD with BPF kr ðÞ, we use PD characteristic of S6775 of Hamamatsu Photonics K.K. [26], and BPF characteristics of BP635, BP525, and BP470 of Midwest Optical Systems, Inc. [27]–[29] (see Fig. 3(b)). By the above LED spectrums and sensitivities of PD with BPF and (14), the sensitivity Gkk 0 is estimated to be 0 1 G11 G12 G13 B C G ¼@ G21 G22 G23 A 0
G31
G32
G33
0:381 B ¼@ 0:000
0:002 0:276
1 0:000 C 0:024 A
0:000
0:034
0:194
(19)
where G1k 0 , G2k 0 , and G3k 0 is the responsivity between Red LED and k 0 -th PD, Green LED and k 0 -th PD, and Blue LED and k 0 -th PD, respectively. Also, Gk 1 , Gk 2 , and Gk 3 is the responsivity between k -th LED and PD with BP635 (Red), k -th LED and PD with BP525 (Green), and k -th LED and PD with BP470 (Blue), respectively. Table 1 shows the channel response simulation parameters.
4.3. BER Performance of DCSK Constellation Within the Permissible Error Range In this subsection, the BER performances of DCSK with the permissible error range over a line of sight (LOS) are investigated. Table 2 shows the required minimum number of TLEDs used to represent EVM performances within the permissible error ranges (1%, 5%, 10%, and 15%,) when M is 4,8, and 16. Table 2 also shows the actual EVM performances on x y chromaticity. As expected, the EVM performance is degraded as NTx becomes small.
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array TABLE 1 Channel response simulation parameters
TABLE 2 Required number of TLEDs (NTx ) for representing IEEE 802.15.7 design rule within permissible error ranges (1%,3%,5%,15%)
Fig. 4. BER performances to ave using TLEDs in each EVM on x y chromaticity.
In Fig. 4, we show the BER performance of DCSK system versus signal-to-noise ratio ave which is defined by ave ¼
NTx Pt H N0
(20)
where NTx Pt H is the mean of received total optical power without LED non-linearity and undesired color-shift over the optical channel response H, and N0 is the white and Gaussian noise power per PD.
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
Fig. 5. (a) Geometry of NTx ¼ 7 TLEDs in a TLED array, and the TLED array and a color sensor within an indoor LOS environment. (b) Normalized ave to achieve at the BER of 10−3 against the distance between TLED elements for M ¼ 16, NTx ¼ 9, and Rx position (0, 0, 0). (c) Normalized ave to achieve at the BER of 10−3 against the distance between TLED elements for M ¼ 16, NTx ¼ 9, and D ¼ 1 m.
In this section, we also compare with the ACSK with linearly controlled LED (i.e., ideal ACSK). Here, for verifying the effect of the permissible range to the BER performance, we assume that there is no differences channel paths between the LED array and color-sensor, that means h1k 0 ¼ h2k 0 ¼ ¼ hNt k 0 . From this figure, it is found that the BER performances of DCSK with the permissible range is 1% and 3% can be achieved those of ideal ACSK. However, when the permissible range of 4-DCSK is 15%, the BER performance of DCSK is significantly better than that of ideal 4-ACSK system. This is because the minimum Euclidean distance of the DCSK system can be improved on the 3-D signal power space compared with that of ideal ACSK system, although the EVM performances are degraded on x y chromaticity. This is based on the effect of the instantaneous intensities detection scheme in the 3-D signal space [9].
4.4. Evaluation of the Impact of the Arrangement in DCSK In this subsection, we verify the effect of the difference channel responses between TLEDs and a color-sensor. The arrangement of TLEDs in a TLED array is distributed as a squared spiral as Fig. 5(a) where the distance between TLEDs is denoted by d , and the distance on the z-coordinate between the TLED array and the color-sensor is denoted by D. Fig. 5(b) and (c) show normalized ave required to achieve on BER ¼ 103 against the distance between TLEDs, i.e., d , for 16-DCSK using nine TLEDs (i.e., the permissible error range is 1%). From Fig. 5(b), it can been seen that the BER performance does not depend on the channel gain from any individual TLED to color-sensor when d G 0:02 m which means TLED array size is smaller than about 16 cm2. However, when d 9 0:02 m, the effect of the channel gain becomes large as D becomes small. For example, when d is 0.1 m, with D is 0.1 m will require an SNR of 15 dB to achieve an BER of 10−3 compared with d is 0.01 m. In fact, however, it is relatively easy to be reflow soldered on to a substrate at intervals of 0.02 m. Moreover, we show in Fig. 5(c) the effect of the receiver position on the error performance for the case of x-coordinate of the receiver position is 0, 0.2, 0.5, 1, and 3 m, and D is 1 m. When D is 1 m, the small difference of channel gain from each individual TLED element to color-sensor then lead to the resulting error performance degradation is almost constant in Fig. 5(c). These results therefore show the proposed TLED array architecture is never a practical issue for the error performance degradation.
5. Performances Comparison With Conventional ACSK Finally, we compare the BER performance of the proposed DCSK system with the conventional ACSK systems. For evaluating the BER performances of ACSK and ACSK with undesired color-shift on our simulation, we derive LED non-linearity effects of IV characteristics, -I
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
Fig. 6. (a) Forward voltage to optical intensity ðV Þ characteristics of three RGB LEDs [22]–[24]. (b) Effect of undesired color shift on Photosensitivity.
characteristics, and undesired color shift effects by measuring and modeling by polynomial curve fitting of “polyfit” function of MATLAB IðV Þ ¼
6 X
an V n
(21)
bn IðV Þn
(22)
cn IðV Þn
(23)
n
ðIÞ ¼
3 X n
Gkk 0 ðIÞ ¼
6 X n
where V is forward voltage of an LED by adjustment of DAC; IðV Þ is forward current of the LED; ðIÞ is optical intensity of the LED; Gkk 0 ðIÞ is sensitivity with undesired color-shift effect; and an , bn , and cn are the nth-order coefficients of polynomial model. In the measurement of LED characteristics, the maximum optical power of each color is unified to the optical power of green LED (OSG58A511A) at the rated forward current, and the input forward voltage and the output intensity are normalized as shown in Fig. 6(a). Fig. 6(b) shows the photosensitivity, Gkk 0 ðIÞ, versus normalized forward current I where we use the same BPFs and PD in Section 4.2. From this figure, it can be seen that the undesired shift of photosensitivity against the forward current is small although the center wavelength is shifted against the forward current (see Fig. 3(a)). Figs. 7(a) and 7(b) show the performances of DCSK, ACSK, ACSK with undesired color-shift, and ideal ACSK are compared when the LOS scenario has been considered. In the DCSK system, the number of TLEDs in a TLED array is determined when the permissible error range is 3%, and the distance between TLEDs in a TLED array, d , is 0.05 m, D is 3 m, and the receiver position is Rx(0,0,0). When Rb is 20 Mbps in Fig. 7(a), DCSK, ACSK with undesired color-shift and ideal ACSK present very similar BERs. In particular when M is 8 and 16, the BER performances of DCSK are slightly better than those of ACSK with undesired color-shift. The effect of undesired-color shift becomes more influential for higher order CSK systems, moreover in fact, the cost for a TLED array for DCSK is really low compared with ACSK with the linear variable current driver, including pre-distortion. Fig. 7(b) shows signal-to-noise ratio, ave , requirements for achieving on BER ¼ 103 against bit rate, Rb . It can be seen from Fig. 7(b) that DCSK modulation and ACSK with undesired color-shift present very similar SNR requirements. These three modulations give a data rate of
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
Fig. 7. (a) BER performances against ace for DCSK, ACSK, ACSK with undesired color shift, and ideal ACSK, when Rb is 20 Mb/s. (b) Signal-to-noise ratio ave to achieve at the BER of 10−3 against Rb for DCSK, ACSK, ACSK with undesired color shift, and ideal ACSK.
up to 62 Mbps for M ¼ 16 and 80 Mbps for M ¼ 4, 8. Fig. 7(b) also shows that 8-DCSK and 16DCSK outperform 8-ACSK and 16-ACSK. While, it can be found that the conventional 4-ACSK modulation can achieve the highest data rate of 100 Mbps. This is because the minimum Euclidean distance on the 3-D signal power space is fortuitously increased by LED non-linearity effect, so the ACSK with M ¼ 4 is significantly more robust against inter-symbol-interference (ISI) caused by the LED response. It is clear from the result shown in Fig. 7(b) that the optimum constellation on x y chromaticity leads to higher power requirements for the signal power space detection. Hence, the optimum constellation on signal power space will improve the maximum bit rate. The signal power space constellation will also be able to reduce the required number of TLEDs for representing symbol, and this approach is highly compatible with DCSK using TLED array. There are also nonlinearities in the CSK space on the CIE xyz color space by the photopic luminous efficiency function which is spectral sensitivity of human visual perception of brightness. Therefore, the nonlinearities will invalidate the gain of source linearity on the CIE xyz color space. While, there are no nonlinearities in the CSK space on the signal power space. The accuracy of CSK source therefore brings more potent effect on the communication quality than the illumination quality, when the CSK detection is the instantaneous intensities detection on the signal power space. Thus, we should consider the optimum constellation which achieves the minimum Euclidean distances in the signal power space or the most optimum constellation with a permissible error range for the human perception.
6. Conclusion A simplistic approach to linearizing multi-level CSK output by digitally controlled multiple TLEDs has been presented. In this paper, the effects of the permissible error rate caused by discretization of constellation points and the difference channel response between each TLED in a TLED array and color-sensor have also evaluated. From the BER evaluation with taking into account optical wireless channels including the LED response, undesired color-shift and channel crosstalk by the practice characteristics of TLED, RGB-BPFs and PD, there is no effect to BER when the permissible error rate is within 3%, and the permissible error on the IEEE 802.15.7 CSK constellation could be achieved by only nine TLEDs. The nine TLEDs can dynamically change between the CSK constellations of M ¼ 4, 8, and 16, and the difference channel response is never a practical issue for the BER degradation when the distance
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal
Digital CSK With Multicolor LED Array
between TLEDs is smaller than 2 cm. Additionally, the DCSK modulations can realize similar good BER performances of ACSK with linear variable current driver and ideal ACSK, and achieve a data rate of up to 62 Mbps for M ¼ 16 and 80 Mbps for M ¼ 4, 8 when 3-dB modulation bandwidth of RGB-LED is 20 MHz. Future works are to design optimum DCSK constellations for 3-D signal power space detection for achieving both higher data-rate and practical illumination performance and to evaluate experimentally of the target color performance over the TLED array face.
Acknowledgment The authors would like to thank Prof. H. Habuchi at Ibaraki University for valuable advice during manuscript preparation.
References [1] Y. Tanaka, S. Haruyama, and M. Nakagawa, “Wireless optical transmissions with white colored LED for wireless home links,” in Proc. IEEE PIMRC, 2000, vol. 2, pp. 325–1329. [2] T. Komine and M. Nakagawa, “Fundamental analysis for visible-light communication system using LED lights,” IEEE Trans. Consum. Electron., vol. 50, no. 1, pp. 100–107, Feb. 2004. [3] IEEE Standard for Local and Metropolitan Area Networks-Part 15.7: Short-Range Wireless Optical Communication Using Visible Light, IEEE Std. 802.15.7, 2011. [4] S. Rajagopal, R. D. Roberts, and S.-K. Lim, “IEEE 802.15.7 visible light communication: Modulation schemes and dimming support,” IEEE Commun. Mag., vol. 50, no. 3, pp. 72–82, Mar. 2012. [5] D. C. O’Brienl et al., “Home access networks using optical wireless transmission,” in Proc. IEEE PIMRC, 2008, pp. 1–5. [6] J. Grubor, O. C. Gaete Jamett, J. W. Walewski, and K.-D. Langer, “High-speed wireless indoor communication via visible light,” ITG Fachbericht, vol. 198, pp. 203–208, 2007. [7] P. H. Pathak, X. Feng, P. Hu, and P. Mohapatra, “Visible light communication, networking, and sensing: A survey, potential and challenges,” IEEE Commun. Surveys Tut., vol. 17, no. 4, pp. 2047–2077, 4th Quart. 2015. [8] CIE, Commission Internationale de lEclairage Proc., 1931. [9] R. Singh, T. O’Farrell, and J. P. R. David, “An enhanced color shift keying modulation scheme for high-speed wireless visible light communications,” J. Lightw. Technol., vol. 32, pp. 2852–2592, 2014. [10] R. Singh, T. O’Farrell, and J. P. R. David, “Higher order colour shift keying modulation formats for visible light communications,” in Proc. IEEE VTC—Spring, 2015, pp. 1–5. [11] H. Yonezu, Hikari Tusin Soshi Kougaku (Optical Communication Element Engineering). Tokyo, Japan: Kougaku Tosyo, 1984 (in Japanese). [12] S. Beczkowski and S. Munk-Nielsen, “LED spectral and power characteristics under hybrid PWM/AM dimming strategy,” in Proc. IEEE Energy Convers. Congr. Expo., 2010, pp. 731–735. [13] E. Monteiro and S. Hranilovic, “Design and implementation of color-shift keying for visible light communications,” J. Lightw. Technol., vol. 32, no. 10, pp. 2053–2060, 2014. [14] H. Shimamoto, Y. Kozawa, and Y. Umeda, “An experimental evaluation on EVM performance for 4-CSK (color shift keying) using visible light with multiple multi-color LEDs,” in Proc. IEEE RWS, Jan. 25–28, 2015, pp. 206–208. [15] N. Murata, H. Shimamoto, Y. Kozawa, and Y. Umeda, “Performance evaluation of digital colour shift keying for visible light communications,” in Proc. IEEE ICCW, 2015, pp. 1374–1379. [16] H. Qian, S. Zhao, S. Z. Cai, and T. Zhou, “Digitally controlled micro-LED array for linear visible light communication systems,” IEEE Photon. J., vol. 7, no. 3, Jun. 2015, Art. no. 7901508. [17] S. H. Lee, S.-Y. Jung, and J. K. Kwon, “Modulation and coding for dimmable visible light communication,” IEEE Commun. Mag., vol. 53, no. 2, pp. 136–143, Feb. 2015. [18] R. J. Drost and B. M. Sadler, “Constellation design for color-shift keying using billiards algorithms,” in Proc. IEEE GLOBECOM Workshop, 2010, pp. 980–984. [19] E. Monteiro and S. Hranilovic, “Constellation design for color-shift keying using interior point methods,” in Proc. IEEE GLOBECOM Workshop, 2012, pp. 1224–1228. [20] H. L. Minh et al., “100-Mb/s NRZ visible light communications using a postequalized white LED,” IEEE Photon. Technol. Lett., vol. 21, no. 15, pp. 1063–1065, Aug. 2009. [21] M. Biagi, T. Borogovac, and T. D. C. Little, “Adaptive receiver for indoor visible light communications,” J. Lightw. Technol., vol. 31, no. 23, pp. 3676–3686, Dec. 2013. [22] Optosupply Electronics “OS5RKA511P,” Jun. 2016. [Online]. Available: http://www.optosupply.com/UploadFile/PDF/ os5rka5111p.pdf [23] Optosupply Electronics “OSG58A511A,” Jun. 2016. [Online]. Available: http://www.peace-corp.co.jp/data/opto/OSG58A5111A-HPM.pdf [24] Optosupply Electronics “OSB56A511A,” Jun. 2016. [Online]. Available: http://www.peace-corp.co.jp/data/opto/OSB56A5111A-HPM.pdf [25] TOPCON “IM-1000,” Feb. 2016. [Online]. Available: http://www.topcon-techno.co.jp/en/products/op_meas/im-1000. html [26] Hamamatsu Photonics K.K. “S6775,” Jun. 2016. [Online]. Available: http://www.hamamatsu.com/jp/en/product/alpha/ S/4103/S6775/index.html
Vol. 8, No. 4, August 2016
1601313
IEEE Photonics Journal [27] [28] [29] [30]
Digital CSK With Multicolor LED Array
Midwest Optical Systems, Inc. “BP635,” Jun. 2016. [Online]. Available: http://midopt.com/filters/bp635/ Midwest Optical Systems, Inc. “BP525,” Jun. 2016. [Online]. Available: http://midopt.com/filters/bp525/ Midwest Optical Systems, Inc. “BP470,” Jun. 2016. [Online]. Available: http://midopt.com/filters/bp470/ W. O. Popoola, E. Poves, and H. Haas, “Error performance of generalised space shift keying for indoor visible light communications,” IEEE Trans. Commun., vol. 61, no. 5, pp. 1968–1976, May 2013.
Vol. 8, No. 4, August 2016
1601313
