Visible light communication (VLC) refers toshort-range optical wireless communicationusing the visible light spectrum from 380 to 780nm. VLC transmits data by intensity modulatingoptical sources, such as light emitting diodes(LEDs) and laser diodes, faster than the persis-tence of the human eye [1–3]. There has beenrenewed interest in visible light optical commu-nication due to the widespread deployment ofLEDs for energy efficiency and recent advance-ments in LED technology with fast nanosecondswitching times. Traditional radio frequency(RF) communication below 6 GHz is rapidlyrunning out of spectrum bandwidth for high-data-rate communication. With ~300 THz ofbandwidth available for VLC, multi-gigabit-per-second data rates could be provided over shortdistances, for example, using arrays of LEDs in amultiple-input multiple-output (MIMO) fashion[4]. In addition, communication is provided inconjunction with lighting providing gigabit-per-second data rates with only simple LEDs andphotodetectors (PDs) compared to expensive RFsolutions that require high power consumption(Watts) for transmitting, sampling, and process-ing gigabit-per-second data.
The two main challenges for communicationusing visible light spectrum are flicker mitigationand dimming support. Flicker refers to the fluc-tuation of the brightness of light. Any potential
flicker resulting from modulating the lightsources for communication must be mitigatedbecause flicker can cause noticeable,negative/harmful physiological changes inhumans. To avoid flicker, the changes in bright-ness must fall within the maximum flickeringtime period (MFTP). The MFTP is defined asthe maximum time period over which the lightintensity can change without the human eye per-ceiving it. While there is no widely acceptedoptimal flicker frequency number, a frequencygreater than 200 Hz (MFTP < 5 ms) is generallyconsidered safe [5]. Therefore, the modulationprocess in VLC must not introduce any notice-able flicker either during the data frame orbetween data frames. Dimming support is anoth-er important consideration for VLC for powersavings and energy efficiency. It is desirable tomaintain communication while a user arbitrarilydims the light source. The human eye respondsto low light levels by enlarging the pupil, whichallows more light to enter the eye. This responseresults in a difference between perceived andmeasured levels of light. The relation betweenperceived and measured light is given by [6]
(1)
As shown in Fig. 1, a lamp that is dimmed to10 percent of its measured light output is per-ceived as being dimmed to only 32 percent.Hence, communication support needs to be pro-vided when the light source is dimmed over alarge range, typically between 0.1–100 percent.
VLC is being investigated by a number ofuniversities, corporations, and organizationsworldwide. In 2007 the Japan Electronics andInformation Technology Industries Association’s(JEITA) established standards for a “visible lightID system.” In 2008, the Visible Light Commu-nications Consortium (VLCC) introduced aSpecification Standard. The Home GigabitAccess project (OMEGA) in Europe is alsodeveloping VLC for home networks [3]. Howev-
Perceived lightMeasured light
(%)(%)
= ×100100
ABSTRACT
Visible light communication refers to short-range optical wireless communication using visi-ble light spectrum from 380 to 780 nm. Enabledby recent advances in LED technology, IEEE802.15.7 supports high-data-rate visible lightcommunication up to 96 Mb/s by fast modula-tion of optical light sources which may bedimmed during their operation. IEEE 802.15.7provides dimming adaptable mechanisms forflicker-free high-data-rate visible light communi-cation.
TOPICS IN STANDARDS
Sridhar Rajagopal, Samsung Electronics
Richard D. Roberts, Intel
Sang-Kyu Lim, ETRI
IEEE 802.15.7 Visible LightCommunication: Modulation Schemesand Dimming Support
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 72
IEEE Communications Magazine • March 2012 73
er, none of the above standards focus on flickermitigation and dimming, which has been inte-grated into IEEE 802.15.7 and is the focus ofthis article. The IEEE 802.15.7 standard sup-ports multiple diverse topologies, such as peer-to-peer and star topologies, with data ratesranging from 11.67 kb/s to 96 Mb/s for indoorand outdoor applications. In the following sec-tions, the article describes the various modula-tion methods available in IEEE 802.15.7 andtheir benefits for flicker mitigation and dimming.
MODULATION METHODSINCORPORATED IN 802.15.7
The IEEE 802.15.7 standard offers three physi-cal (PHY) types for VLC. PHY I operates from11.67 to 266.6 kb/s, PHY II operates from 1.25to 96 Mb/s and PHY III operates between 12and 96 Mb/s. PHY I and PHY II are defined fora single light source, and support on-off keying(OOK) and variable pulse-position modulation(VPPM). PHY III uses multiple optical sourceswith different frequencies (colors) and uses aparticular modulation format called color shiftkeying (CSK). The different modulation schemesallow trade-offs between data rates and differentdimming ranges [7]. For example, under dim-ming conditions, modulation using OOK pro-vides constant range and variable data rate byinserting compensation time, while modulationusing VPPM provides constant data rate andvariable range by adjusting the pulse width. Allthree PHYs have been crafted to coexist witheach other while mitigating flicker and support-ing dimming.
Each PHY mode contains mechanisms formodulating the light source, run length limited(RLL) line coding, and channel coding for for-ward error correction (FEC). RLL line codesare used to avoid long runs of 1s and 0s thatcould potentially cause flicker and clock anddata recovery (CDR) detection problems. RLLline codes take in random data symbols at inputand guarantee DC balance with equal 1s and 0sat the output for every symbol. Various RLLline codes such as Manchester, 4B6B, and 8B10Bare defined in the standard, and provide trade-offs between coding overhead and ease of imple-mentation. IEEE 802.15.7 also supports variousFEC schemes to work reasonably well in thepresence of hard decisions that would be gener-ated by the CDR. The channel codes supportboth long and short data frames for high-data-rate indoor and low-data-rate outdoor applica-tions. For outdoor applications, stronger codesusing concatenated RS and CC codes are devel-oped to overcome the additional path loss dueto longer distance and potential interferenceintroduced by optical noise sources such as day-light and fluorescent lighting. Reed-Solomon(RS) and convolutional codes (CC) are pre-ferred over advanced coding schemes such aslow density parity check (LDPC) codes in orderto support short data frames, hard decisiondecoding, low complexity, and their ability tointerface well with RLL line codes. For indoorapplications, where the coding requirements areless stringent for short distances, RS codes are
used for FEC since they are better suited tohigh-data-rate implementations. RS codes alsointerface well in conjunction with the RLL linecodes, where the errors detected from the RLLline code at the receiver could be marked as era-sures to the RS decoder, providing performanceimprovements of around 1 dB. For PHY I, aninterleaver between the RS code and the CCcode provides an additional 1 dB of performanceimprovement.
Each PHY modulation mode has an associat-ed optical clock rate which is “divided down” bythe various coding schemes to obtain the finalresulting data rates, as shown in Tables 1, 2, and3. The optical clock rate for PHY I is chosen tobe ≤ 400 kHz to account for the fact that LEDsused in applications such as traffic lights requirehigh currents to drive the LEDs and thereforeswitch slowly. For PHY II, the optical clock rateis chosen to be ≤ 120 MHz to accommodate fastLEDs used in mobile and portable devices forcommunication. The optical clock rate is chosento be ≤ 24 MHz, which is the maximum clockrate supported by current infrastructure (white)LEDs used in PHY III. The standard also sup-ports the use of different clock rates with thesame device for transmitting and receiving datasince the transmitter (LED) and receiver (PD)are independent circuits. The infrastructurecould be transmitting at a lower clock rate usingslower but brighter LEDs while receiving at ahigher clock rate from a portable device that hasfaster but weaker LEDs.
OOK MODULATION AND DIMMING METHODOOK modulation is the simplest modulationscheme for VLC, where the LEDs are turned onor off dependent on the data bits being 1 or 0.While the modulation is logically OOK, OOK“off” does not necessarily mean the light is com-pletely turned off; rather, the intensity of thelight may simply be reduced as long as one can
Figure 1. Human eye shows nonlinear sensitivity to dimming [6], motivatingthe need for high-resolution dimming support.
Perceived light (%)
Need for low dimming levels
100
100
10-1
Mea
sure
d lig
ht (
%) 101
102
20 30 40 50 60 70 80 90 100
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 73
IEEE Communications Magazine • March 201274
distinguish clearly between the “on” and “off”levels. The OOK mode transmitter block dia-gram in Fig. 2 shows the bits from the upper lay-ers entering the FEC before being ManchesterRLL coded. The Manchester encoding embedsthe clock into the data by representing a logiczero as an OOK symbol “01” and a logic one asan OOK symbol “10,” providing a DC balancedcode.
OOK dimming can be achieved by eitherredefining the “on” or “off” levels of the OOKsymbol to have a lower intensity, or the levelscan remain the same and the average duty cycleof the waveform can be changed by the insertionof “compensation” time into the modulationwaveform. The compensation time is realized byfully turning on or off the light source for therequired duration to provide dimming. Thisallows a DC component, which determines thelight intensity, to be added to the waveform con-trolling the light source. For example, if thebrightness of data is A percent with period T1and the compensation symbols have an averagebrightness B percent with period T2, the result-ing average brightness N (percent) can be givenby
(2)
The two methods impact performance in dif-ferent ways. The first method, redefining the“on” and “off” levels of OOK, gives a constantbit rate as the light dims, which means the rangemust decrease, but it also risks color shift due tothe LEDs being underdriven. On the other hand,the insertion of compensation time results in alower bit rate as the light dims, which implies areduced bit rate with constant range. However,there has also been related work using compres-sion techniques to reduce the bit rate reduction[8].
The OOK dimming frame’s structure is shownin Fig. 3. The frame structure consists of apreamble for synchronization, a PHY headerthat provides details on the frame such as theframe length, modulation, and coding, and thedata payload frame. When compensation time isadded, it is possible for the receiver to lose syn-chronization for long compensation times, sincethe clock at the receiver is typically recoveredfrom the data. Hence, in the OOK dimmingframe structure, the data frame is broken into
NAT BT
T T=
++
1 2
1 2
Table 1. PHY I operating modes.
Modulation RLL code Opticalclock rate
FECData rate
Outer code (RS) Inner code (CC)
OOK Manchester 200 kHz
(15,7) 1/4 11.67 kb/s
(15,11) 1/3 24.44 kb/s
(15,11) 2/3 48.89 kb/s
(15,11) None 73.3 kb/s
None None 100 kb/s
VPPM 4B6B 400 kHz
(15,2) None 35.56 kb/s
(15,4) None 71.11 kb/s
(15,7) None 124.4 kb/s
None None 266.6 kb/s
Table 2. PHY II operating modes.
Modulation RLL code Opticalclock rate FEC Data rate
VPPM 4B6B
3.75 MHzRS(64,32) 1.25 Mb/s
RS(160,128) 2 Mb/s
7.5 MHz
RS(64,32) 2.5 Mb/s
RS(160,128) 4 Mb/s
None 5 Mb/s
OOK 8B10B
15 MHzRS(64,32) 6 Mb/s
RS(160,128) 9.6 Mb/s
30 MHzRS(64,32) 12 Mb/s
RS(160,128) 19.2 Mb/s
60 MHzRS(64,32) 24 Mb/s
RS(160,128) 38.4 Mb/s
120 MHz
RS(64,32) 48 Mb/s
RS(160,128) 76.8 Mb/s
None 96 Mb/s
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 74
IEEE Communications Magazine • March 2012 75
subframes, and each subframe can be precededby a resynchronization (resync) field using a1010… maximum transition sequence patternthat aids in readjusting the data clock after thecompensation time. The data frame is fragment-ed into subframes of the appropriate length afterthe FCS has been calculated and the FECapplied. An example of OOK dimming toincrease brightness by adding compensation sym-bols is shown in Fig. 4. The average brightness(AB) of N percent is achieved by adjusting thebrightness of the data and the compensationsymbols. When OOK modulation is used fordata communication, the data inherently has a50 percent duty cycle due to Manchester RLLcoding. In order to adjust the duty cycle withinthe frame, the compensation symbols of theappropriate duration and brightness (as definedin Eq. 2) need to be applied to maintain the ABof N percent. Outside the data communicationframe, idle patterns of N percent average bright-ness are sent to ensure the net average bright-ness from the illumination source remainsconsistent.
VPPM MODULATION AND DIMMING METHODThe use of modulation techniques such as pulseposition modulation (PPM) for dimming supporthas been proposed for VLC [7]. Variable pulse-position modulation (VPPM) changes the dutycycle of each optical symbol to encode bits. Thevariable term in VPPM represents the change inthe duty cycle (pulse width) in response to therequested dimming level. VPPM optical symbolsare distinguished by the pulse position. As shownin Fig. 5a, VPPM is similar to 2-PPM when theduty cycle is 50 percent. The logic 0 and logic 1symbols are pulse width modulated depending onthe dimming duty cycle requirement. As shown inFig. 5b, the pulse width ratio (b/a) of PPM canbe adjusted to produce the required duty cyclefor supporting dimming by pulse width modula-tion (PWM). Figure 6 shows an example wave-form of how VPPM can attain a 75 percentdimming duty cycle requirement, where bothlogic 0 and logic 1 have a 75 percent pulse width.
The VPPM mode transmitter block diagram isshown in Fig. 7. The input is sent through an RSFEC encoder for error protection, followed by a4B6B RLL code for DC balance and flicker miti-gation. The 4B6B coding takes a random 4-bitsymbol and changes it into a DC balanced 6-bitcode as shown in Table 4. The counts of 1 and 0in every VPPM encoded symbol are always equalto 3. Since the bit rate is constant regardless ofthe requested dimming level, as the light isdimmed, the range decreases with the dimminglevel. The features of the 4B6B RLL code are:• Always 50 percent duty cycle during one
encoded symbol• DC balanced run length limiting code• Error detection capability• Run length limited to four• Reasonable clock recovery
Figure 8 shows how the light intensity for thepayload can be adjusted by adapting the pulsewidth of VPPM symbols. The light intensity forthe preamble and header can be adjusted byinserting compensation symbols of the appropri-ate length and intensity before the frame. The
AB of N percent is achieved by adjusting thebrightness of the data and the compensationsymbols. When VPPM modulation is used fordata communication, the data inherently havethe required duty cycle and hence, the require-ment for compensation time within the dataframe is mitigated. However, the preamble andheader are always at 50 percent duty cycle, sincethey are not aware of the VPPM modulation.Therefore, some adjustment in compensationtime may be required to keep the average bright-ness of N percent consistent. However, it may bedifficult to achieve arbitrary duty cycles usingVPPM to support large dimming ranges. Hence,the VPPM symbols also time-multiplex with dif-ferent dimming levels within the frame in orderto attain high resolution up to 0.1 percent. Thisis discussed in detail later.
CSK MODULATION AND DIMMING METHODWhite LED lights are generated by using a mix-ture of different colors in typically two differentmethods. White LEDs can be generated using
Figure 2. OOK transmitter block diagram. The figure shows the bits sent fromthe upper layers being encoded for error protection and to provide DC balance.
OOKmodulated
light
ManchesterRLL
code
InnerCC
encoder
OuterRS
encoderbits
Figure 3. OOK dimming structure. The figure shows the frame structure forOOK dimming, where compensation symbols are added in time to maintainthe desired visibility level, and short sync fields are used to resync the receiverbefore the data subframes.
Frame length
Preamble Short sync fieldi.e. 1010 pattern Data subframe
DimmedOOK
extensionCompensation
symbolsPHY
header
Table 3. PHY III operating modes.
Modulation Optical clock rate FEC Data rate
4-CSK12 MHz
RS(64,32) 12 Mb/s
8-CSK RS(64,32) 18 Mb/s
4-CSK
24 MHz
RS(64,32) 24 Mb/s
8-CSK RS(64,32) 36 Mb/s
16-CSK RS(64,32) 48 Mb/s
8-CSK None 72 Mb/s
16-CSK None 96 Mb/s
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 75
IEEE Communications Magazine • March 201276
blue LEDs with yellow phosphor. However, yel-low phosphor slows down the switching responseof the white LEDs. Alternately, faster whiteLEDs that can be more useful for communica-tion can be generated by simultaneously excitingred, green and blue LEDs [9, 10]. The use ofsuch multi-color LEDs forms the principlebehind CSK modulation. Color shift keyingmodulation is similar to frequency shift keying inthat the bit patterns are encoded to color (wave-length) combinations. For example, for 4-CSK(two bits per symbol) the light source is wave-length keyed such that one of four possiblewavelengths (colors) is transmitted per bit paircombination. In order to define various colorsfor communication, the IEEE 802.15.7 standardbreaks the spectrum into 7 color bands in orderto provide support for multiple LED color choic-
es for communication. Figure 9 shows the centerof the seven color bands on xy color coordinatesas defined by CIE 1931 color coordinates [11].The 3-bit values in Fig. 9 represent each of theseven color bands. The CSK signal is generatedby using three color light sources out of theseven color bands. The three vertices of the CSKconstellation triangle are decided by the centerwavelength of the three color bands on xy colorcoordinates. CSK has the following advantages:
•The final output color (e.g., white) is guar-anteed by the color coordinates shown in Fig. 9:CSK channels are determined by mixed colorsthat are allocated in the color coordinates plane.
•The total power of all CSK light sources isconstant, although each light source may have adifferent instantaneous output power. CSK dim-ming ensures that the average optical power
Figure 4. Dimming support using OOK modulation. The figure shows the use of idle patterns (when there isno data) and the use of compensation symbols (in the presence of data using OOK) to maintain an aver-age brightness (AB) of N percent.
P: PreambleH: HeaderCS: Compensation symbolsSSF: Short sync fieldDS: Data subframe
Idlepatterns
IdlepatternsCS
50%
N%
P H CS SSF SSF SSFDS DS ... DS
TimeFrame length
CS CS
Brightness
PHY frame
AB=N%AB=N% AB=N% AB=N% AB=N% AB=N%
Figure 5. Basic concept of VPPM. VPPM is similar to 2-PPM, as shown in a) for 50 percent visibility, andthe duty cycle is adapted using PWM for other visibility levels as shown in b).
PWM2-PPM0 0 01 1
(b)(a)
a
b
Figure 6. Waveform of VPPM signal with 75 percent pulse width.
2T 3TT
“0” “0” “0” “0”“1” “1” “1” “1”
8T7T5T4T 6T
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 76
IEEE Communications Magazine • March 2012 77
from the light sources is kept constant and main-tains the requisite intensity of the center color ofthe color constellation. Thus, there is no flickerissue associated with CSK due to amplitude vari-ations. CSK dimming employs amplitude dim-ming and controls the brightness by changing thecurrent driving the light source. However, careneeds to be observed during CSK dimming toavoid unexpected color shift in the light source.
•CSK supports amplitude changes with digi-tal-to-analog (D/A) converters (higher complexi-ty), thus allowing higher order modulationsupport to provide higher data rates at a loweroptical clock frequency. PHY I and PHY IIallow only OOK modulation, thereby limitingtheir data rate to 1 b/clock.
The communication using CSK modulationwith 4-CSK symbol points, for example, aredefined by the design rule in Fig. 10. Points I, Jand K show the center of the three color bandson xy color coordinates. S0 to S3 are four symbolpoints of 4-CSK. S1, S2, and S3 are three verticesof the triangle IJK. S0 is the centroid of the trian-gle IJK. The absolute values for 4-CSK for multi-ple combinations of the optical sources, assumingthe spectral peak of the optical source is at thecenter of the band plan, can be obtained in [12].
Figure 11 shows the CSK system configura-tion for PHY III with three color (bands i, j andk) light sources. After scrambling and channelcoding, the logical data values (zeros and ones)are transformed into xy values, according to amapping rule on the xy color coordinates by thecolor coding block. The scrambler is necessary tocreate pseudo-random data and prevent data-pattern-dependent color shifts. The data parts ofthe frame are subject to the FEC block for errorprotection. These xy values are transformed intointensity Pi, Pj and Pk. The relation between thecoordinates and the intensity is shown in Eqs.3–5. On the receiver side, xy values are calculat-ed from the received light powers of 3 colors,and xy values are decoded into the received data.
xp = Pi xi + Pjxj + Pkxk (3)
yp = Pi yi + Pjyj + Pkyk (4)
Pi + Pj + Pk = 1 (5)
DIMMING MECHANISMS FORFLICKER-FREE COMMUNICATION
This section outlines the system design for dim-ming support that is enabled by the modulationmethods and physical layer support discussed inthe previous section.
IDLE PATTERN DIMMINGAn idle pattern is defined as a pattern whoseduty cycle variation results in a change of bright-ness to support dimming and may be transmittedduring idle or receive mode. An idle pattern canbe transmitted during medium access layer(MAC) idle (no data transmission) or receive(RX) operation states for infrastructure lightsources. This helps maintain visibility and flick-er-free operation. The data and the idle patternshould have the same duty cycle to minimize
flicker. This idle pattern, and its dependence onthe dimmer setting, is shown in Fig. 12. Thetransition between active communication andidle operation occurs on a large timescale (blockactive/idle/RX). However, within an active com-munication session, there can be small timescaletransition between active, idle, and RX modes.The final output from the light source shouldshow a constant duty cycle irrespective of theMAC state of the device. Dimmer setting (a)illustrates a higher switching frequency for high-er brightness, while dimmer setting (b) illustratesa lower switching frequency for lower brightness.
HYBRID IDLE PATTERN ANDCOMPENSATION TIME DIMMING
The idle pattern mechanism allows an idle pat-tern to be inserted between the data frames forlight dimming, as shown in Fig. 13. The dutycycle of the idle pattern can be adjusted to vary
Figure 7. VPPM transmitter block diagram. The figure shows the bits sent fromthe upper layers being encoded for error protection and to provide DC balance.
VPPMmodulated
light4B6BRLL
RS FECencoderbits
Table 4. Mapping input 4B to output 6B.
4B (input) 6B (output) Hex
0000 001110 0
0001 001101 1
0010 010011 2
0011 010110 3
0100 010101 4
0101 100011 5
0110 100110 6
0111 100101 7
1000 011001 8
1001 011010 9
1010 011100 A
1011 110001 B
1100 110010 C
1101 101001 D
1110 101010 E
1111 101100 F
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 77
IEEE Communications Magazine • March 201278
the brightness. An idle pattern can either be in-band or out-of-band as defined by the modula-tion domain spectrum shown in Fig. 14 (i.e., thespectrum observed at the output of the PD).Both types of idle patterns are supported. Figure14 shows the modulation spectrum divided intothree regions: flicker, out-of-band, and in-band.The flicker region is defined from DC to < 200Hz where eye safety may be of concern. Ambi-ent light interference (50-60 Hz) also falls in thisregion. Hence, this region should be avoided for
communication. The region between the flickerand in-band modulation is defined as out-of-band. An out-of-band region is needed for multi-ple PHYs to coexist. For example, PHY I can liein the out-of-band region of PHY II since it sup-ports a lower data rate. An in-band idle patterndoes not require any change in the clock and canbe seen by the receiver. An out-of-band idle pat-tern is typically sent at a much lower opticalclock rate (including the option of maintainingvisibility via a DC bias only) and is not seen bythe receiver (i.e., is not in the receiver modula-tion domain bandpass). The compensation timeis defined as the idle time inserted in the idlepattern or in the data frame where the light isturned on and off with the appropriate ratio tomeet dimming duty cycle requirements.
VISIBILITY PATTERN DIMMINGTo support continuous illumination from infras-tructures, the standard provides support forframes that do not contain any data but sendonly idle patterns to maintain continuous visibili-ty support. Visibility pattern dimming is similarto idle pattern dimming except that the patternsare used inside the payload of a visibility frame.However, these visibility patterns need to sup-port the high resolution required for dimming of0.1 percent discussed earlier. The visibility pat-terns support features such as flicker mitigation,continuous visibility, device discovery and colorstabilization. The visibility patterns are notencoded in the PHY layer and do not have aframe check sequence (FCS) associated withthem. To generate high resolution visibility pat-terns from 0 to 100 percent in steps of 0.1 per-cent, certain constraints must be considered indesigning the visibility patterns.• The number of transitions between 0s and
1s can be maximized to provide high-fre-quency switching to avoid flicker and helpthe CDR circuit at the receiver for synchro-nization purposes.
Figure 9. CIE 1931 xy color coordinates [11], where x and y are the chromatici-ty values. The outer curve is the spectral locus with wavelengths shown in nm.The three-digit values refer to the center wavelength of the seven bands definedin the IEEE 802.15.7 band plan.
x
y
700
0.1
0.1
0.0
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.0 0.2 0.3 0.4 0.5 0.6 0.7 0.8
001
000
011
100101110
010
620
600
580
560
540
500
490
480
470460 380
520
Figure 8. Frame structure to meet dimming requirements using VPPM. When VPPM is used for the data, thevisibility level for dimming is automatically satisfied. However, since the preamble and header is encodedusing OOK, compensation symbols can be used to accommodate the difference in brightness during thepreamble and header.
CS: Compensation symbolsP: PreambleH: PHY headerAB: Average brightness
Idlepatterns
Idlepatterns PayloadHPCS
Time
Fixed symbol width
Variable dimmingwidth
PHY frame
. . . .
AB=N% AB=N% AB=N% AB=N%Brightness
N%
50%
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 78
IEEE Communications Magazine • March 2012 79
• A simple approach should be used for visi-bility pattern generation. Designing 1000patterns to support low resolutions (as lowas 0.1 percent resolution) is not practical,and makes visibility pattern generation anduse very complex.
• Since visibility patterns are transmittedwithout changing the clock frequency (in-band), avoid patterns that conflict withexisting RLL code words.A set of 11 base low resolution visibility pat-
terns with 10 percent step size can be used fordimming. In fact, a set of 11 base low resolutionvisibility patterns of any length can be used aslong as no conflict exists between the visibilitypattern and a valid RLL code. Figure 15 pro-vides such a set as an example. The low resolu-tion patterns can be used to develop highresolution visibility patterns by averaging themacross time to generate the required high resolu-tion pattern. For example, if visibility patternsare available at 10 percent resolution, a 25 per-cent visibility pattern can be attained by alter-nately sending a 20 percent visibility patternfollowed by a 30 percent visibility pattern. Thismethod guarantees that all visibility patterns willretain the same properties as the base low reso-lution visibility patterns. The high resolution visi-bility pattern generation can be generalized byusing the low resolution patterns according tothe algorithm specified in the appendix. Thehigh resolution dimming algorithm provides twopatterns “sel1pat” and “sel2pat” out of the set of“K+1” available patterns with number of repeti-tions as “reppat1” and “reppat2” respectively.These two patterns are the closest available pat-terns on both sides of the precision requirement.For example, if visibility patterns are at 10 per-cent resolution, there are 11 patterns (K = 10),and any requirement between 20.01 and 29.99percent will time-multiplex the 20 and 30 per-cent visibility patterns as shown in Fig. 15, andthen adjust the repetition ratio of 20 and 30 per-cent patterns to get the required precision withthe minimum number of repetitions.
FLICKER MITIGATIONThe flicker in VLC is classified into two cate-gories according to its generation mechanism:intra-frame flicker and inter-frame flicker. Intra-frame flicker is defined as the perceivable bright-
ness fluctuation within a frame. Inter-frameflicker is defined as the perceivable brightnessfluctuation between adjacent frame transmis-sions. This section outlines the methods used forflicker mitigation in VLC.
INTRA-FRAME FLICKER MITIGATIONIntra-frame flicker mitigation refers to mitiga-tion flicker within the transmission of a dataframe. Intra-frame flicker in OOK is avoided byusing the dimmed OOK mode and RLL coding.VPPM uses RLL code and does not cause anyinherent inter-frame flicker. Intra-frame flickeris avoided in CSK modulation by ensuring con-stant average power across multiple light sourcesalong with scrambling and high optical clockrates (megahertz).
INTER-FRAME FLICKER MITIGATIONInter-frame flicker mitigation applies to bothdata transmission (RX mode) and idle periods.While idling, visibility patterns or idle patterns
Figure 10. Constellation design rule for 4-CSK. The x and y refer to chromatici-ty values. P1, P2 and P3 correspond to three points chosen from the sevenwavelengths in the band plan. P0 is chosen to be the centroid of the triangle.
x
CSK constellation
P2
P1
K I
J
P3
P0
0-0.2
0
-0.2
y
0.2
0.4
0.6
0.8
0.2 0.4 0.6 0.8 1
i bandj bandk bandCSK symbols
Figure 11. CSK system diagram for PHY III. The data is scrambled to ensure randomness before encodingand mapping to intensity values which are then sent to three distinct wavelength optical sources.
BandiD/APi
Optical source
BandjD/APj
x
BandkD/APk
xyto
Pi, Pj, PkChannelencoder y
ColorcodingScramblerData
Channelestimationsequence
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 79
IEEE Communications Magazine • March 201280
may be used to ensure light emission by theVLC transmitters have the same average bright-ness over adjacent MFTPs as during data trans-mission. These patterns can be modulatedin-band or out-of-band (Fig. 14). When the dim-mer setting is changed, the MAC and PHY lay-ers adjust the data transmission and idle timetransmission to fit the new dimmer settings. Asummary of the different mitigation techniquesfor inter-frame and intra-frame flicker is provid-ed in Table 5.
CONCLUSIONVLC provides the potential for multi-gigabit-per-second data rate communication at shortdistances with ~300 THz of available visiblelight spectrum at low power and cost, usingsimple LEDs and PDs. With the growing inte-
gration of LEDs in indoor and outdoor lightsources, and advances in the design of low-cost LEDs with fast subnanosecond switchingresponse times, the integration of lightingand communicat ion provides s ignif icantpotential for this technology. The two mainchallenges in communication in this spectrumare flicker mitigation and support for dim-ming. This article presents mechanisms tomit igate f l icker and support dimming asdefined in the IEEE 802.15.7 visible l ightcommunication standard.
Several technical challenges must be addressedto realize the full potential of VLC technology.First, channel models for VLC are not well under-stood, especially for outdoor non-line-of-sight(NLOS) environments, and there is an active areaof research for channel models and platforms forVLC [13, 14]. Also, the networking of the light
Figure 12. Adapting dimmer pattern and data duty cycle based on dimmer setting using the MAC knowledgeof active and idle states and keeping a constant duty cycle to mitigate flicker.
Active
MAC scheduled TX
ActiveIdle/RX
Active ActiveIdle/R X
(a)
(b)
Time
On
Off
BlockactiveBlock
idle/RX
Block active
I: idle / R XA: Active
AA IIAVLC activity I
Idle pattern
Idle pattern
Infrastructure TX output
Infrastructure TX output
MAC scheduled TX
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 80
IEEE Communications Magazine • March 2012 81
sources and upgrading current infrastructures tosupport communication is another challenge,which requires support from the lighting industry.With continued growth in LED based lightsources and the need for multi-Gb/s data distribu-tion, VLC, being developed as a global industrystandard in IEEE 802.15.7, promises to be a veryattractive candidate as a future high data rate andpower-efficient technology.
ACKNOWLEDGMENTThe authors are grateful to their colleagues atSamsung Electronics, Intel, ETRI and in IEEE802.15.7 for discussions and assistance with thiswork. ETRI’s work was supported by the ICTStandardization program of MKE (The Ministryof Knowledge Economy) and the IT R&D pro-gram of MKE/KEIT. The authors also thank theanonymous reviewers, the Series Editor, Dr.Mostafa Hashem Sherif, and Dr. Tracy Volzfrom Rice University for their comments andreview of this article.
REFERENCES[1] T. Komine and M. Nakagawa, “Fundamental Analysis
for Visible-Light Communication System using LEDLights,” IEEE Trans. Consumer Electronics, vol. 50, no.1, Feb. 2004, pp. 100–07.
[2] M. Kavehrad, “Sustainable Energy-Efficient WirelessApplications Using Light,” IEEE Commun. Mag., vol. 48,no. 12, Dec. 2010, pp. 66–73.
[3] D. O’Brien et al., “Home Access Networks Using OpticalWireless Transmission,” IEEE PIMRC, Sept. 2008, pp.1–5.
[4] L. Zeng et al., “High Data Rate Multiple Input MultipleOutput (MIMO) Optical Wireless Communications UsingWhite LED Lighting,” IEEE JSAC, vol. 27, no. 9, Dec.2009, pp. 1654–62.
[5] S. Berman et al., “Human Electroretinogram Responsesto Video Displays, Fluorescent Lighting and Other HighFrequency Sources,” Optometry and Vision Science, vol.68, 1991, pp. 645–62.
[6] M. Rea, Illumination Engineering Society of NorthAmerica (IESNA) Lighting Handbook, 9th ed., July 2000.
[7] B. Bai, Z. Xu, and Y. Fan, “Joint LED Dimming and HighCapacity Visible Light Communication by OverlappingPPM,” Proc. IEEE Annual Wireless and Opt. Commun.Conf., May 2010, pp. 71–75.
[8] J. K. Kwon, “Inverse Source Coding for Dimming in Visi-
Figure 13. Idle pattern and compensation time dimming where the compensation time and idle times areinserted during data transmission with appropriate duration to mitigate flicker.
Idle pattern
The compensation time, “ON” or“OFF” time, can be inserted for
dimming.
001001001001001i.e. duty cycle of 1/2 → 1/3
VLC data frameIdle pattern
Figure 14. Modulation domain spectrum for visible light communication. Theuse of flicker mitigation schemes ensure that there is no data transmitted in thelower frequencies, where flicker occurs.
Modulationfrequency
Ambient lightinterference
Flicker
Amplitude
Out-of-band
In-band
Figure 15. High resolution dimming with visibility patterns. High resolution visibility patterns are generatedby time-division multiplexing low resolution visibility patterns.
Visibility pattern Percentage visibility
11111 11111 100%
11110 11111 90%
11110 11110 80%
70%
11001 11100 60%
10001 11100 50%
00001 11100 40%
00001 11000 30%
00001 10000 20%
00001 00000 10%
00000 00000 0%
New high resolutionvisibility patterns
Mix and match(time division multiplex)
Fixed low resolutionvisibility patterns
11101 11100
RAJAGOPAL LAYOUT_Layout 1 2/22/12 4:27 PM Page 81
IEEE Communications Magazine • March 201282
ble Light Communications Using NRZ-OOK on ReliableLinks,” IEEE Trans. Photon. Technology Lett., vol. 22,no. 19, Oct. 2010, pp. 1455–57.
[9] H. L. Minh et al., “100-Mb/s NRZ Visible Light Commu-nications Using a Postequalized White LED,” IEEE Trans.Photon. Technology Lett., vol. 21, no. 15, Aug. 2009,pp. 1063–65.
[10] J. Vucvicv et al., “513 Mb/s Visible Light Communica-tions Link Based on DMT-Modulation of a White LED,”IEEE/OSA J. Lightwave Tech., vol. 28, no. 24, Dec.2010, pp. 3512–18.
[11] CIE, Commission Internationale de l’Eclairage Proc.,Cambridge Univ. Press, 1931.
[12] A. Yokoi et al., “More Description about CSK Constel-lation,” IEEE 802.15.7 Contrib. 15-10-0724-00-0007,Sept. 2010.
[13] J. Carruthers et al., “Propagation Modelling for IndoorOptical Wireless Communications Using Fast Multi-Receiver Channel Estimation,” IEE Proc. Optoelectronics,vol. 150, no. 5, Oct. 2003, pp. 473–81.
[14] K. Cui et al., “Line-of-Sight Visible Light Communica-tion System Design and Demonstration,” 7th IEEE Int’l.Symp. Commun. Sys. Networks and Digital Signal Proc.,July 2010, pp. 621–25.
APPENDIX
Algorithm for high resolution dimming:Let the following values be defined as
– ‘K+1’ Visibility patterns: V0, V1, …, VK– Desired visibility = dv (expressed as a per-
centage value) e.g., for a 25.3 percent visi-bility, dv = 25.3
– Desired precision = p, p ≤ 0, p is an integer(expressed as a logarithm value) e.g., for0.01 percent, precision, p = –2Then define the selected patterns as sel1pat
and sel2pat.
(6)
(7)
The number of repetitions of the selectedpatterns is given as reppat1 and reppat2. Thenumber of repetitions for these patterns can beselected as follows.
BIOGRAPHIESSRIDHAR RAJAGOPAL [SM] ([email protected]) is astaff engineer in the Standards Research Laboratory atSamsung Electronics R&D center in Dallas, Texas. Hereceived his M.S. and Ph.D. degrees from Rice University inelectrical and computer engineering in 2000 and 2004,respectively. His research interests are in algorithm andarchitecture designs for short-range and low-power wire-less communications, including RF, millimeter wave, tera-hertz, and optical communication. He is one of the keycontributors in developing the IEEE 802.15.7 standard onvisible light communication.
RICHARD D. ROBERTS [SM] ([email protected]) iswith Intel Labs Oregon where he is a wireless research sci-entist concentrating on the area of short-wavelength, high-ly directional, physical layer communications. This includes60 GHz, terahertz, and free space optics (FSO). In particu-lar, he is actively researching visible light communicationsand positioning (VLCP). He has been a member of the IEEEfor over 30 years and an active member of IEEE 802.15 forover a decade. His undergraduate degree is from the Uni-versity of Wisconsin (1980), and his graduate (1985) andpost-graduate degrees (1992) are from Florida Institute ofTechnology.
SANG-KYU LIM [M] ([email protected]) is a principal memberof engineering staff in the Electronics and Telecommunica-tions Research Institute (ETRI) in Daejeon, Korea. Hereceived his B.S degree in physics in 1995, and M.S andPh.D. degrees in electronics engineering in 1997 and 2001from Sogang University, Seoul, Korea. In 2001, he joinedETRI, where he has worked on high-speed optical transmis-sion systems and the microwave/millimeter-wave circuitdesign and fabrication for high-speed optical communica-tion. He is concentrating in the area of visible light com-munication. He is one of the major contributors indeveloping the IEEE 802.15.7 standard on VLC.
reppat2 10100 1
= −⎛⎝⎜
⎞⎠⎟
− p dvsel pat
K
*
sel pat210
=⎡
⎢⎢⎢
⎤
⎥⎥⎥
−dv
K
p*
sel pat110
=⎢
⎣⎢⎢
⎥
⎦⎥⎥
−dv
K
p*
Table 5. Flicker mitigation schemes.
Flicker mitigation Data transmission (Intra-frame flicker)
Idle or RX periods(Inter-frame flicker)
OOK modulation Dimmed OOK mode, RLL code
Idle/visibility patterns
VPPM modulation VPPM guarantees no intra-frame flicker, RLL code
CSK modulation
Constant average poweracross multiple light sources,scrambler, high optical clockrates (MHz)
