electronics

Article
Adapting Techniques to Improve Efficiency in Radio
Frequency Power Amplifiers for Visible
Light Communications
Daniel G. Aller 1, * , Diego G. Lamar 1 , Juan Rodriguez 2 , Pablo F. Miaja 1 ,
Valentin Francisco Romero 3 , Jose Mendiolagoitia 3 and Javier Sebastian 1
1 Power Supplies Group—Electrical Engineering Department of the University of Oviedo, 33204 Gijon, Spain;
gonzalezdiego@uniovi.es (D.G.L.); fernandezmiapablo@uniovi.es (P.F.M.); sebas@uniovi.es (J.S.)
2 Center for Industrial Electronics—Polytechnic University of Madrid, 28006 Madrid, Spain;
juan.rodriguezm@upm.es
3 Thyssenkrupp Elevator Innovation Center S A U—Thyssenkrupp AG, 33203 Gijon, Spain;
valentin-francisco.romero@thyssenkrupp.com (V.F.R.); jose.mendi@thyssenkrupp.com (J.M.)
* Correspondence: garciaadaniel@uniovi.es; Tel.: +34-985-182-578




Received: 30 November 2019; Accepted: 3 January 2020; Published: 10 January 2020


Abstract: It is well known that modern wireless communications systems need linear, wide bandwidth,
efficient Radio Frequency Power Amplifiers (RFPAs). However, conventional configurations of RFPAs
based on Class A, Class B, and Class AB exhibit extremely low efficiencies when they manage signals
with a high Peak-to-Average Power Ratio (PAPR). Traditionally, a number of techniques have been
proposed either to achieve linearity in the case of efficient Switching-Mode RFPAs or to improve the
efficiency of linear RFPAs. There are two categories in the application of aforementioned techniques.
First, techniques based on the use of Switching-Mode DC–DC converters with a very-fast-output
response (faster than 1 µs). Second, techniques based on the interaction of several RFPAs. The current
expansion of these techniques is mainly due to their application in cellphone networks, but they
can also be applied in other promising wireless communications systems such as Visible Light
Communication (VLC). The main contribution of this paper is to show how Envelope Tracking (ET),
Envelope and Elimination (EER), Outphasing, and Doherty techniques can be helpful in developing
more efficient VLC transmitters capable of reaching high bit-rates (higher than 1 Mbps) by using
advance modulation schemes. Finally, two examples based on the application of the Outphasing
technique and the use of a Linear-Assisted Envelope Amplifier (EA) to VLC are presented and
experimentally verified.

Keywords: visible light communication (VLC); light emitting diodes (LED); switching-mode; DC–DC
converters and radio frequency power amplifiers (RFPA)

1. Introduction
Currently, the use of wireless communication systems is essential for the present and the future of
our society [1]. Furthermore, the speed demanded by each communication service is continuously
growing due to the high bit-rate required (higher than 1 Mbps) by mainstream services. As a result,
the Radio Frequency (RF) spectrum is already close to congestion and hence enabling the data traffic
predicted for upcoming years requires further research into new wireless communication technologies.
Visible Light Communication (VLC) is one of the most promising solutions for alleviating the
saturation of the RF spectrum [2–4]. VLC uses the wide, unlicensed visible light spectrum (430–750 THz
range) to transmit information. The strength and potential of this approach become apparent when the
communication task is merged with the lighting functionality of Light-Emitting Diode (LED) lamps.

Electronics 2020, 9, 131; doi:10.3390/electronics9010131 www.mdpi.com/journal/electronics

Electronics 2020, 9, 131 2 of 19

Thus, the existing lighting infrastructure can be partially adapted to incorporate communication
features. Since its introduction in 2004 [1], VLC has gained attention due to the widespread use of
LEDs in lighting installations. Besides their longer lifespan, higher power efficiency, and the fact that
they are more environmentally friendly than other lighting technologies, LEDs are able to change
their light intensity very quickly, making this technology the most promising one in terms of enabling
communication functionalities.
The design of the LED driver with communication functionalities (VLC-LED driver) is one of the
main challenges of the VLC-transmitter. The VLC-LED driver is responsible for two tasks: guaranteeing
the desired lighting level and reproducing, with a high degree of accuracy, the communication signal
originating from the other kind circuitry of the VLC-transmitter, which processes the data.
Traditionally, the literature focuses on the most important VLC contributions in data processing,
advanced modulation schemes, LED modeling and experimental verification using low power prototypes.
In other words, the main efforts have been carried out to increase the spectral efficiency of VLC-LED
drivers in order to achieve very high bit-rates (higher than 10 Mbps). However, a major issue with
regard to VLC-LED drivers is not taken into account: the high efficiency of LED lighting is not only
due to the high efficiency of the LEDs in converting electrical power into lighting power, but also
to the high efficiency achieved by the VLC-LED driver that delivers the electrical power. Therefore,
the VLC-LED driver must also be power efficient [5]. The VLC-LED drivers for VLC-transmitters
proposed to date, which can be classified into two subsets, do not achieve both high power efficiency
(higher than 90%) and high bit-rates (higher than 1 Mbps):

• VLC-LED drivers based on switching the LEDs on and off. Figure 1a shows a conventional structure
based on a high efficiency, limited-output-response Switching-Mode DC–DC converter. In this
example, a power MOSFET is placed in series to turn the LED on and off. Other structures with
the MOSFET placed in parallel are also valid. This VLC-LED driver is simple and high-power
efficient (higher than 90%). Only Pulse-Based Modulation (PBM) schemes can be reproduced
using this solution [6–9]; however, PBM schemes are not the best option for providing a high
bit-rate and for avoiding the multipath issue (an issue arising from the different paths that the
light rays can follow from the light focus to the receiver). Although some PBM proposed solutions
achieve a very high bit-rate (10 Mbps) at higher power levels associated with the light emitted for
communications (up to 10 W) [8,9], they are far from the theoretically achievable bit-rate using
advanced modulation schemes.
• VLC-LED drivers for reproducing advanced modulation schemes. Figure 1b shows the conventional
structure, which involves the use of a high efficiency, limited-output-response Switching-Mode
converter (DC–DC or AC–DC) to bias the LED, in addition to the use of a low power efficiency,
very-fast-output response (i.e., wide bandwidth) RFPA to superpose the communication signal.
These VLC-LED drivers focus on reproducing Single-Carrier Modulation (SCM) and Multi-Carrier
Modulation (MCM) schemes that allow for providing a higher bit-rate (up to 1 Gbps). These solutions
are based on classical configurations of linear Radio Frequency Power Amplifiers (RFPAs) (Class
A, Class B, and Class AB) in order to obtain linearity and wide bandwidth. However, they exhibit
extremely low efficiency when they manage signals with a wide ratio between average power and
peak power [10–15] (i.e., a high Peak-to-Average Power Ratio, PAPR). Although the very low power
efficiency only refers to power processing for communication tasks (e.g., efficiencies between
10% and 40%), this fact affects the overall efficiency of the VLC-LED driver (e.g., efficiencies
below 70%). Moreover, all the proposed solutions reported to date using linear RFPAs limit the
application to very short distances (up to 1 m) with very low output power for both lighting and
communications tasks (up to 1 W).

As in any electronics device, a Switching-Mode Converter (either AC–DC or DC–DC [16–19])

will be in charge of generating the DC voltage level needed to supply power to any VLC system.
The voltage source represented as VCC in the circuits shown in Figure 1 will be the output port of that
Electronics 2020, 9, 131 3 of 19

converter.
ElectronicsNo
2020,specific design
9, x FOR PEER constrains will distinguish this first converter to any other converter.
REVIEW 3 of 19
The switching-mode DC–DC converters shown in Figure 1 are designed identically as traditional
post-regulators, which are being traditionally used in LED drivers.

V CC

Switching- i LED i LED

Mode dc-dc
Bias control
converter
LED
Communication control vPBM
vPBM

(a)

Switching-
Mode dc-dc
V CC converter
i bias_LED

Bias control

v 8QAM i ac_LED i LED

Linear i LED
v 8QAM - +
RFPA i ac_LED
LED

(b)

Figure 1. Conventional VLC-LED drivers presented to date: (a) VLC-LED driver based on switching
the LEDs on and off, reproducing a PBM scheme; (b) VLC-LED driver for reproducing advanced
modulation schemes, in this example, 8-Quadrature Amplitude Modulation (8-QAM).
Figure 1. Conventional VLC-LED drivers presented to date: (a) VLC-LED driver based on switching
Focusing
the our
LEDs onattention on the second
and off, reproducing a PBM subset of (b)
scheme; solutions
VLC-LED(i.e., VLC-LED
driver driversadvanced
for reproducing for reproducing
modulation schemes, in this example, 8-Quadrature Amplitude Modulation (8-QAM).
advanced modulation schemes), the low power efficiency of VLC-LED drivers is an important limitation
that may be overcome so as to obtain high power efficiency and a high bit-rate. The possible solution
As in any electronics device, a Switching-Mode Converter (either AC–DC or DC–DC [16–19])
arises from the application of traditional techniques used in the past either to improve the efficiency of
will be in charge of generating the DC voltage level needed to supply power to any VLC system. The
linear RFPAs or to achieve linear behavior in the case of efficient Switching-Mode RFPAs. In fact, this
voltage source represented as VCC in the circuits shown in Figure 1 will be the output port of that
is the main objective
converter. No specificof the present
design paper:will
constrains to describe
distinguishthethis
adaptation of traditional
first converter to any other techniques
converter.used
either
Thetoswitching-mode
improve the efficiency
DC–DC of or linearize
converters RFPAs
shown to VLC.
in Figure In some
1 are cases,
designed this adaptation
identically implies a
as traditional
significant change in the circuitry that defines these techniques,
post-regulators, which are being traditionally used in LED drivers. with major associated benefits. For all
these reasons, theour
Focusing aim of this paper
attention on the is the revision
second subset ofofsolutions
these techniques,
(i.e., VLC-LEDtheirdrivers
adaptation for VLC, and
for reproducing
the experimental validation
advanced modulation of two examples:
schemes), the low power Outphasing
efficiencytechnique
of VLC-LED and Linear-Assisted technique.
drivers is an important
limitation
The paperthatismay be overcome
organized so as toFirst,
as follows. obtainahigh power efficiency
summary and a high
of the traditional bit-rate. The
techniques possible
used in RF to
solution arises from the application of traditional techniques used in the past either
either improve the efficiency of RFPAs or to linearize the RFPAs are given in Section 2. The adaptation to improve the
efficiency of linear RFPAs or to achieve linear behavior in the case of efficient
of these techniques to VLC are given in Section 3. The adaptation of the Outphasing technique to VLC is Switching-Mode
RFPAs. In fact, this is the main objective of the present paper: to describe the adaptation of traditional
given in Section 4. The use of the envelope amplifiers in VLC is explained in Section 5. The adaptation
techniques used either to improve the efficiency of or linearize RFPAs to VLC. In some cases, this
of the Linear-Assisted technique is given in Section 6. Section 7 presents the circuitry design and the
adaptation implies a significant change in the circuitry that defines these techniques, with major
experimental results. Finally, Section 8 provides the conclusions of the paper.
associated benefits. For all these reasons, the aim of this paper is the revision of these techniques,
their adaptation for VLC, and the experimental validation of two examples: Outphasing technique
2. Review of Traditional Solutions Either to Improve the Efficiency of or to Linearize RFPAs
and Linear-Assisted technique.
Table
The1 summarizes the traditional
paper is organized techniques
as follows. used either
First, a summary to improve
of the traditional the efficiencyused
techniques of orintoRF
linearize
to
either
RFPAs. Onimprove
the onethe efficiency
hand, Tableof1 RFPAs
shows or to linearize
solutions for the RFPAs are
increasing thegiven in Section
efficiency 2. The adaptation
of Linear RFPAs, which
of these techniques
are naturally towhen
inefficient VLC are givenwith
signals in Section
very 3. ThePAPR
high adaptation of the Outphasing
are processed. technique
On the other hand, to there
VLC are
Electronics 2020, 9, 131 4 of 19

other kinds of techniques for providing linearity to Switching-Mode RFPAs (Class D, Class E, and Class
F). Switching-Mode RFPAs are very efficient (theoretical efficiency is 100%) because the active devices
(transistors) function as electronic switches instead of linear gain devices. However, they exhibit a
lack of linearity because of the nonlinear relationship between the pulses that are used to drive the
transistors and the output. In both cases (Linear or Switching-Mode RFPAs), these techniques are based
on the use of either very-fast-output response (faster than 1 µs) Switching-Mode DC–DC converters or
an arrangement of RFPAs.

Table 1. Traditional techniques either to improve the efficiency of or to linearize RFPAs.

Linear RFPAs (Class A & B) Switching-Mode RFPAs (Class D, E & F)

Using Switching-Mode DC–DC converters Envelope Tracking (ET) Envelope Elimination and Restoration (EER)
Arrangement of RFPAs Doherty Outphasing

In the past, very-fast-output response Switching-Mode DC–DC converters, capable of changing

their output voltage at a MHz rate, have been proposed to implement techniques used to increase
the efficiency of RFPAs (first row in Table 1). The most popular of these techniques are Envelope
Tracking (ET) for Linear RFPAs (Figure 2a) and Envelope Elimination and Restoration (EER) for
Switching-Mode RFPAs (Figure 2b) [19]. In both cases, a very-fast-output response Switching-Mode
DC–DC converter has to generate the time-varying output voltage that is used as the power supply for
the RFPA. The variations of this voltage are imposed by the envelope of the communication signal
processed by the
Electronics 2020, 9, x RFPA. This
FOR PEER fast-output response Switching-Mode DC–DC converter is often
REVIEW 5 ofknown
19
as an Envelope Modulator or Envelope Amplifier (EA).

v env
V CC
Envelope Amplifier
v env
(very-fast-output v env
Envelope
response
Detector Envelope v oRFPA
Switching-Mode
v env
control dc-dc converter)
vi
vi v oRFPA
Linear
Delay
RFPA

(a)

v env
V CC
Envelope Amplifier
venv (very-fast-output venv
Envelope
response
Detector Envelope v oRFPA
Switching-Mode
v env
control dc-dc converter)
vi
vi vPWM voRFPA
Switching
Limiter Delay
RFPA
vPWM

(b)

Figure 2. Conventional block diagram of techniques based on the use of Envelope Amplifiers (EA):
(a) Envelope Tracking (ET) technique; (b) Envelope Elimination and Restauration (EER) technique.
Figure 2. Conventional block diagram of techniques based on the use of Envelope Amplifiers (EA):
(a) Envelope Tracking (ET) technique; (b) Envelope Elimination and Restauration (EER) technique.

Moreover, techniques based on an arrangement of several RFPAs can achieve the desired
improvements, in this case without using very-fast-output response Switching-Mode DC–DC
converters: the Doherty technique for linear RFPAs and the Outphasing technique for Linear RFPAs
 Electronics 2020, 9, 131 5 of 19

Moreover, techniques based on an arrangement of several RFPAs can achieve the desired
improvements, in this case without using very-fast-output response Switching-Mode DC–DC converters:
the Doherty technique for linear RFPAs and the Outphasing technique for Linear RFPAs and
Switching-Mode RFPAs. W.H. Doherty originally proposed the Doherty technique in 1936 [20]
(Figure 3a). In the case at hand, at least two linear RFPAs must be used. One of the RFPAs operates with
the low part of the signal and is called the “carrier” RFPA. The other RFPA only works when the signal
level exceeds a specific value and is called the “peak” RFPA. The realization of this technique involves
the splitter of the input signal that is to be amplified and the combination of the output signal of each
RFPA. The realization of the circuitry that performs these functions (i.e., the splitter and the combiner)
is complex and sometimes constitutes the bottleneck of this technique. The outphasing technique
was first proposed in the 1930s [21,22] and is based on the following idea: two sinusoidal signals of
constant amplitude can be combined to obtain one sinusoidal signal with variable amplitude and
phase. Therefore, from this statement, two Switching-Mode RFPAs working at maximum efficiency
can be 2020,
Electronics combined
9, x FORto amplify
PEER any signal (Figure 3b) by means of appropriate phase shifting between
REVIEW 6 of 19
them. In this case, the design and construction of the splitter, the output network of each RFPA, and
the combiner are also key points.

i RFPA1

vRFPA1 vo
V CC
“Carrier” RFPA i RFPA1
Linear
RFPA1 vRFPA1 Impedance inverter
Splitter and combiner
vi vo
i RFPA2
vi V CC
i RFPA2
Linear
RFPA2
“Peak” RFPA

(a)

vRFPA1
V CC vo
vi
Switching Output
+α network 1 v
RFPA1 RFPA1
Splitter Combiner
vRFPA2 vo
vi

V CC

Phase control Output

−α Switching
RFPA2 network 1 vRFPA2

(b)

Figure 3. Conventional block diagram of techniques based on an arrangement of RFPAs. (a) Doherty
technique;
Figure (b) Outphasing
3. Conventional blocktechnique.
diagram of techniques based on an arrangement of RFPAs. (a) Doherty
technique; (b) Outphasing technique.

3. Traditional Solutions Either to Improve the Efficiency of or to Linearize RFPAs and Their
Adaptation to VLC
The main idea of this section is very simple: to export and adapt any of the techniques indicated
 Electronics 2020, 9, 131 6 of 19

3. Traditional Solutions Either to Improve the Efficiency of or to Linearize RFPAs and Their
Adaptation to VLC
The main idea of this section is very simple: to export and adapt any of the techniques indicated
in Table 1 to the RFPA of the VLC-LED driver shown in Figure 2b. Figure 4 shows an example of the
adaptation of the ET technique to VLC-LED drivers. As already stated, the ET technique requires a
very-fast-output response Switching-Mode converter. This converter must include the capability of
following
Electronics 2020,the
9, xenvelope
FOR PEERofREVIEW
the signal that is to be amplified (in the example in Figure 4, an 8-QAM signal).
7 of 19

Envelope Amplifier
Very-fast-output
response Switching- V CC
Mode dc-dc
converter)
Envelope control
venv Switching-
Mode dc-dc
Envelope
venv vdcRFPA converter
Detector i bias_LED
v8QAM
Bias control

vQAM vac_LED i LED

Linear
Delay RFPA - +
i LED LEDs
vac_LED vdcRFPA
i bias_LED

Figure
Figure 4. Appling thethe
4. Appling ETETtechnique
techniquetotoan
anRFPA
RFPA of
of aa VLC-LED driverreproducing
VLC-LED driver reproducingan
anMCM
MCMscheme
scheme
(i.e., 8QAM).
(i.e., 8QAM).

However, prior
However, totothe
prior theadaptation
adaptationofofall
allthe
thetechniques
techniques in Table
Table11totoVLC,
VLC,ititisisnecessary
necessarytoto point
point
out some aspects, which will eventually lead to the introduction of certain constraints in
out some aspects, which will eventually lead to the introduction of certain constraints in the real the real
implementation:
implementation:
(a) (a)OnOn thetheone
one hand,
hand, the themaximum
maximum bandwidth of theof
bandwidth LEDtheforLED
the electrical-to-optical power conversion
for the electrical-to-optical power
is a fundamental
conversion parameter that
is a fundamental must be taken
parameter that into
mustaccount for VLC.
be taken intoAaccount
blue GaN forLED
VLC.in combination
A blue GaN
LEDwith a yellow phosphor
in combination with LED is thephosphor
a yellow preferred approach
LED is the topreferred
obtain white light in Solid-State
approach to obtain whiteLighting
light
(SSL). However, this phosphor layer is optimized for color conversion, but not
in Solid-State Lighting (SSL). However, this phosphor layer is optimized for color conversion, for achieving a very
butrapid
not forresponse
achieving to changes
a very in lightresponse
rapid intensity. to
In changes
fact, it limits the LED
in light bandwidth
intensity. of it
In fact, white light
limits thetoLED
a
few MHz (i.e., 3–5 MHz). Moreover, the bandwidth for blue light emission
bandwidth of white light to a few MHz (i.e., 3–5 MHz). Moreover, the bandwidth for blue lightis limited to 10–20 MHz
due to is
emission LED technology.
limited to 10–20OnMHz
the other
due tohand,
LEDthe light intensity
technology. modulation
On the other hand,must belight
the fast enough
intensity
to be unappreciable to the human eye for safety reasons (i.e., higher than approximately 1 kHz).
modulation must be fast enough to be unappreciable to the human eye for safety reasons (i.e.,
The aforementioned limits define the usable frequency spectrum to modulate the amplitude and
higher than approximately 1 kHz). The aforementioned limits define the usable frequency
phase of the visible light intensity (VLC frequency spectrum).
spectrum to modulate the amplitude and phase of the visible light intensity (VLC frequency
(b) The greater part of the previously defined frequency spectrum for the use of visible light (i.e.,
spectrum).
1 kHz–20 MHz) could be used for many of the possible scenarios (in some cases, the entire VLC
(b) The greater part of the previously defined frequency spectrum for the use of visible light (i.e., 1
frequency spectrum). This fact facilitates the use of MCM schemes to achieve high bit-rates
kHz–20 MHz) could be used for many of the possible scenarios (in some cases, the entire VLC
because there are no spectrum restrictions. This means it is necessary to use wideband RFPAs for
frequency spectrum). This fact facilitates the use of MCM schemes to achieve high bit-rates
VLC-drivers. Another possible solution is to use several narrowband RFPAs, whose combination
because there are no spectrum restrictions. This means it is necessary to use wideband RFPAs
for VLC-drivers. Another possible solution is to use several narrowband RFPAs, whose
combination could cover the bandwidth required by the MCM scheme. However, the cost of this
solution makes it unfeasible.
(c) If very wideband RFPAs (when the BW is higher than 10% of the center frequency) operating at
relatively low frequencies (due to the bandwidth of the VLC frequency spectrum) are required
 Electronics 2020, 9, 131 7 of 19

could cover the bandwidth required by the MCM scheme. However, the cost of this solution
makes it unfeasible.
(c) If very wideband RFPAs (when the BW is higher than 10% of the center frequency) operating at
relatively low frequencies (due to the bandwidth of the VLC frequency spectrum) are required
for VLC-LED drivers, then the adaptation to VLC of some of the techniques mentioned in Table 1
is not possible. For example, Switching-Mode RFPAs require either band-pass resonant circuits
(e.g., Class D RFPAs) or transistor-integrated resonant circuits (Class E and F RFPAs) to operate
properly because the square waveform generated by transistors must be filtered in order to obtain
a sinusoidal carrier. This fact limits the exploitation of the VLC frequency spectrum because it is
very difficult to adapt EER and Outphasing techniques using nonlinear RFPAs operating within
Electronics 2020, 9, x FOR PEER REVIEW 8 of 19
this frequency range (e.g., from 1 kHz to 20 MHz). Moreover, in the case of the Outphasing
technique, the use of selective output networks for the combiner (Figure 3b) is mandatory. The real
mandatory. The real implementation of these output networks for the VLC frequency spectrum
implementation of these output networks for the VLC frequency spectrum is very complex and
is very complex and makes the adaptation of this technique unachievable.
makes the adaptation of this technique unachievable.
(d) In the case of the application of the Doherty technique, impedance inverters, made of discrete
(d) In the case of the application of the Doherty technique, impedance inverters, made of discrete
components, are needed. Impedance inverters are narrowband, which limit the bandwidth of
components, are needed. Impedance inverters are narrowband, which limit the bandwidth of
the VLC-driver.
the VLC-driver.
(e) From all the above, if MCM schemes are to be reproduced for VLC, then only the ET technique
(e) From all the above, if MCM schemes are to be reproduced for VLC, then only the ET technique
can be applied to single wideband linear RFPAs. In this case, the entire free VLC frequency
can be applied to single wideband linear RFPAs. In this case, the entire free VLC frequency
spectrum
spectrum can bebe
can exploited.
exploited.
Another option is either to give up the entire free frequency spectrum of visible light or split it
Another option is either to give up the entire free frequency spectrum of visible light or split it up
up using several narrowband RFPAs. In this case, all the techniques in Table 1 can be adapted to
using several narrowband RFPAs. In this case, all the techniques in Table 1 can be adapted to VLC.
VLC. Figure
Figure 5 shows
5 shows the adaptation
the adaptation of thetechnique
of the Doherty Doherty totechnique
VLC. As canto be
VLC.
seen,As canthebe“carrier”
both seen, both
RFPAthe
“carrier” RFPA and the “peak” RFPA are highlighted. In this case, the impedance inverter
and the “peak” RFPA are highlighted. In this case, the impedance inverter is designed with discreteis designed
with discrete components
components (twoC,
(two capacitors, capacitors, C, and one
and one inductor, inductor,
L) instead of L)
theinstead of the
traditional traditional quarter-
quarter-wavelength
wavelength transmission line, due to the frequencies
transmission line, due to the frequencies used in VLC. used in VLC.

Switching-
vRFPA1 i RFPA1 Mode dc-dc
converter
i bias_LED

Bias control
V CC
“Carrier” RFPA i RFPA1 L vac_LED i LED
Linear
RFPA1 v RFPA1 - +
Splitter C C
vi i RFPA2 LEDs

vi V CC
i RFPA2
Linear
RFPA2 v ac_LED i LED
“Peak” RFPA
i bias_LED

Figure 5. Appling the Doherty technique to a VLC-LED driver. This scheme is only valid for narrowband
Figure 5. Appling the Doherty technique to a VLC-LED driver. This scheme is only valid for
approaches.
narrowband approaches.

4. Adaptation of the Outphasing Technique to VLC

The Outphasing technique can be applied using both linear and Switching-Mode RFPAs. The
efficiency with linear RFPAs can be increased up to the maximum of each class (50% for Class A and
78% for B), whereas, in the case of Switching-Mode RFPAs, the maximum is theoretically 100%,
making this the most suitable adaptation to VLC.
Electronics 2020, 9, 131 8 of 19

4. Adaptation of the Outphasing Technique to VLC

The Outphasing technique can be applied using both linear and Switching-Mode RFPAs. The efficiency
with linear RFPAs can be increased up to the maximum of each class (50% for Class A and 78% for B),
whereas, in the case of Switching-Mode RFPAs, the maximum is theoretically 100%, making this the
most suitable adaptation to VLC.
The major difficulty in applying the Outphasing technique (Figure 3b) is the design of the output
combiner and the output networks, which are in charge of connecting the two RFPAs together and
summing the sinusoidal signals. On the one hand, the connection of the outputs of the RFPAs is not
straightforward, as the output impedance of each RFPA changes over time depending on the output
voltage due to the phase shift operation, thereby leading to an undesirable influence between them.
On the other hand, the combiner must be designed to prevent the differential mode connection of
the load (i.e., the load must be connected to the ground of the circuitry). Traditionally, the combiner
is performed by means of two quarter-wavelength transmission lines. Furthermore, the cross effect
between the RFPAs due to the output combiner is especially critical when the RFPA is based on a
resonant topology (i.e., Class E or Class F), in which the efficiency depends on the correct tuning of the
resonant circuit. In this case, the analysis and design of the combiner is complex in order to obtain a
trade-off between efficiency and bandwidth.
Another key part of the application of the Outphasing technique is the input splitter and phase
control. This circuitry is in charge of generating the input signals for each RFPA, whose sum is the
desired output communication signal. In the case of Switching-Mode RFPAs, this splitter and phase
control can be implemented on a digital platform, as the input signals for these amplifiers are square
waveforms. In the case of linear amplifiers, the splitter and phase control have to be implemented over
the analog signals, thus increasing the difficulty of the implementation.
At this point, we can exploit summing the light output in order to adapt the Outphasing technique
to VLC. The idea of light Outphasing is that of splitting the LED load between the two RFPAs and
summing the two sinusoidal signals in their light form, as can be seen in Figure 6. In the case of Figure 6,
both LED strings will generate the same DC value of light intensity due to the fact that ibias_LED1
equals ibias_LED2 (generated by a conventional Switching-Mode DC–DC converter). Furthermore, two
sinusoidal signals of constant amplitude and variable phase are superposed to each string of LEDs by
the actions of both Switching-Mode RFPAs. As the added quantities are two light intensities instead of
two electrical signals, the need to use a combiner and connect both Switching-Mode RFPAs electrically
is circumvented, thus leading to a huge simplification of the design and avoiding the influence between
the two signals (with the resulting increase in power efficiency). In other words, as the outputs of each
Switching-Mode RFPA are not connected together and the load is an LED string, the impedance from
the point of view of the RFPAs is known and constant. Hence, the possible condition of Zero Voltage
Switching (ZVS) of each RFPA does not depend on the other, thereby simplifying the overall design.
In a conventional VLC-LED driver, the LED string is connected via a bias T (Figure 1b). The bias
T is used to connect the amplifier and the Switching-Mode DC–DC converter together, blocking the
DC component from passing to the RFPA (using a series capacitor) and the signal components from
passing to the Switching-Mode DC–DC (using an inductor). As shown in Figure 6, there is no path
between the strings at high frequency, meaning that the assumption that there is no influence between
the RFPAs is still correct. In the case of Class E RFPAs, due to their intrinsic resonant output filter, DC
blocking is already implemented in each amplifier, so there is no need for an additional series capacitor
between each RFPA and the corresponding LED string.
avoiding the influence between the two signals (with the resulting increase in power efficiency). In
other words, as the outputs of each Switching-Mode RFPA are not connected together and the load
is an LED string, the impedance from the point of view of the RFPAs is known and constant. Hence,
the possible condition of Zero Voltage Switching (ZVS) of each RFPA does not depend on the other,
Electronics
thereby 2020, 9, 131 the overall design.
simplifying 9 of 19

Conventional
Switching-Mode
dc-dc converter i LED1
i bias_LED1
Bias control vRFPA1 i bias_LED2 i bias_LED1
V CC
vi i LED2
+α Switching i LED1
- + i bias_LED1
RFPA1 vRFPA1 i LED2
Splitter
vi vRFPA2 vRX
V CC

Phase control
−α Switching
RFPA2 vRFPA2 - +

LEDs

Electronics 2020, 9, x FOR PEER REVIEW 10 of 19

Figure 6. Adapting the Outphasing technique to a VLC-LED driver exploiting summing the light. This
Figure 6. Adapting the Outphasing technique to a VLC-LED driver exploiting summing the light.
scheme is only valid for narrowband approaches.
5. Exploiting Envelope
This scheme Amplifiers
is only valid (EA) for
for narrowband Implementing VLC-LED Drivers
approaches.
5. Exploiting Envelope Amplifiers (EA) for Implementing VLC-LED Drivers
As previously stated, ET and EER techniques (Table 1) require EAs (i.e., very-fast-output
In As
a conventional VLC-LED
previously stated, ET anddriver,
EER the LED string
techniques (Table2)is connected
1)capable
require EAs via a bias
(i.e., T (Figure 1b). The bias
very-fast-output
response Switching-Mode DC–DC converters, Figure of changing their outputresponse
voltage at
T is used to connect
Switching-Mode the
DC–DC amplifier and the Switching-Mode DC–DC converter together,
at ablocking the
a very high frequency rateconverters,
(i.e., MHz)Figure 2) capable
following of changing
the envelope their
of the output voltage
modulation scheme. very
The high
idea is
DCfrequency
component ratefrom
(i.e., passing
MHz) to the RFPA (using a ofseries modulation
capacitor) and the signal components from
now very simple: to use EAsfollowing the envelope
synthesizing the overall the
current passingscheme. throughThe the idea
LEDisstring.
now very
In this
passing
simple: to the Switching-Mode DC–DC (using an inductor). As shown in Figure 6, there is no path
case, onlytooneuseSwitched-Mode
EAs synthesizingDC–DC the overall currentispassing
converter also usedthrough
in thethe LEDimplementation
final string. In this case,
of aonly
VLC-
between
onedriver.the strings at
Switched-Mode high frequency, is meaning that
thethefinalassumption that there is no influence
LED However,DC–DC converter
this converter also
is very used in
challenging becauseimplementation
it is in chargeof ofagenerating
VLC-LED driver.
both the
between
However, this converter is very challenging because it is in charge of generating both the DC output
the RFPAs is still correct. In the case of Class E RFPAs, due to their intrinsic resonant bias
DC bias current for the lighting task and the AC current level corresponding to the information to be
filter, DC for
current blocking is already
the lighting task and implemented
the AC current inlevel
eachcorresponding
amplifier, sotothere is no needtofor
the information be an additional
transmitted
transmitted using advanced modulation schemes (Figure 7). Therefore, the cutoff frequency
series
usingcapacitor
advanced between
modulationeachschemes
RFPA and the corresponding
(Figure 7). Therefore, the LED string.
cutoff frequency corresponding to the
corresponding to the transfer function between the control variable (the converter duty cycle) and
transfer function between the control variable (the converter duty cycle) and the output voltage must
the output voltage must be as high as the highest frequency component of the ac current that is to be
be as high as the highest frequency component of the ac current that is to be used for communication
used for communication tasks. In practice, this means that very-fast-output response Switching-
tasks. In practice, this means that very-fast-output response Switching-Mode DC–DC converters must
Mode DC–DC
be used converters
for this purpose. must be used for this purpose.

Very-fast-output i LED
response Switching-
V CC Mode dc-dc
converter
vd_ac vd i LED
+
+
+ vd Switching-Mode
vd_dc dc-dc control
vd_dc
LEDs

Figure 7. Very-fast-output Switching-Mode DC–DC converter for synthesizing the overall current
Figure 7. Very-fast-output Switching-Mode DC–DC converter for synthesizing the overall current
passing through the LED string (VLC-LED driver). This scheme is valid for reproducing PBM, SCM
passing through the LED string (VLC-LED driver). This scheme is valid for reproducing PBM, SCM
and MCM schemes.
and MCM schemes.
Over the years, several designs of very-fast-output response Switching-Mode DC–DC converters
Over the
conceived years,
to work several
as EAs designs
have been of [23].
proposed very-fast-output response
Most of these designs Switching-Mode
derive DC–DC
from the buck DC–DC
converters
converter,conceived to workexhibits
as this converter as EAs have beenbetween
linearity proposedthe[23]. Most
control of output
and these designs
voltagederive frominthe
operating
buck DC–DC converter, as this converter exhibits linearity between the control and output voltage
operating in Continuous Conduction Mode (CCM). The basic structure of the buck DC–DC converter
(Figure 8a) presents a second-order filter at its output. Using Pulse-Width Modulation (PWM), the
buck DC–DC converter can reproduce signals with frequencies up to 10–20 times its switching
frequency. The challenge is now to propose circuit derivations aimed at reducing the gap between
Electronics 2020, 9, 131 10 of 19

Continuous Conduction Mode (CCM). The basic structure of the buck DC–DC converter (Figure 8a)
presents a second-order filter at its output. Using Pulse-Width Modulation (PWM), the buck
DC–DC converter can reproduce signals with frequencies up to 10–20 times its switching frequency.
The challenge is now to propose circuit derivations aimed at reducing the gap between the switching
frequency and the frequency of the signal to be reproduced without distortion:
(a) Increase the order of the output filter [24] (Figure 8b). For a certain switching frequency value,
the higher the filter order, the higher the rejection of the switching-related harmonics of the PWM
signal at the input of the buck DC–DC converter filter. In the case of VLC-LED drivers, as the
dynamic resistance of LEDs is quite constant, the conventional filter theory can be used to design
a 4th- or 6th-order output filter, thus reducing the gap between the switching frequency and the
frequency of the signal to be reproduced up to 4–5 times.
(b) Use multi-input structures [25] (Figure 8c). The idea is to reduce the power of the switching-related
harmonics of the PWM signal. Given that the voltage at the input of the buck DC–DC converter
filter is a PWM voltage waveform, the switching-related harmonics are determined by the
amplitude of the pulses. The lower the amplitude of the pulses, the lower the power of the
switching-related harmonics at the input of the buck DC–DC converter filter and hence the higher
the cutoff frequency of the filter for a specified switching frequency. Moreover, multi-input
structures of the buck DC–DC converter mean lower switching losses due to the fact that the
Electronics 2020, 9, x FOR PEER REVIEW 11 of 19
transistor switches between lower voltages (in the circuit in Figure 7, S3 switches when S2 is on
(c) Use and therefore
multiphase withstands the (Figure
structures voltage V8d).
CC3 -V CC2 ).multiphase operation allows the complete
The
(c) elimination
Use multiphase of the switching frequency componentoperation
structures (Figure 8d). The multiphase of PWM allows the complete
modulation. elimination
Although the
of the switching
sidebands frequency
components, component
which are also of PWM modulation.
generated Although
when working with athe sidebands components,
non-constant duty cycle,
which
can arebealso
only generatedMoreover,
attenuated. when working with a non-constant
this technique duty
and the use of cycle, can order
a higher only be attenuated.
output filter
Moreover,
become this useful
a very technique and the
solution foruse
EAsof[26].
a higher order output filter become a very useful solution
(d) The for EAs [26].
combination of multiphase and multi-input structures is also possible [27].
(d) The combination of multiphase and multi-input structures is also possible [27].

i LED
i LED

L L1 L2
S S
V CC D C LEDs
V CC D C1 C2
LEDs

(a) (b)

i LED i LED
L
S3 L A1 L2
V CC3 D C SA C2
LEDs V CC C1
D2 DA
L B1 LEDs
S2
V CC2
D1
S1 SB
V CC1 DB

(c) (d)
Figure
Figure 8. Derivations
Derivations of
of the
the buck
buck DC–DC
DC–DC converter.
converter. (a)
(a) conventional;
conventional;(b)
(b)higher
higherorder
order output
outputfilter;
filter;
(c) multi-input; (d) multiphase.

To increase
To increase the
thetransient
transientresponse
responseofofthe
theSwitching-Mode
Switching-Mode DC–DC
DC–DC converter,
converter, thethe combination
combination of
of switching
switching andand linear
linear stages
stages cancan provide
provide certain
certain benefits.
benefits. ThisThis solution
solution has has
alsoalso
beenbeen explored
explored for
for ET
applications [28,29]. The proposal is based on the following idea: the greater part of the power is
processed by a very-fast-output Switching-Mode DC–DC converter, while only extreme variations at
the output are tracked by the linear RFPA. A good trade-off between extremely-fast-output response
and high power efficiency is achieved using this solution. The two most common linear switching
combinations are depicted in Figure 9. Series combination (Figure 9b) is often regarded as less
 Electronics 2020, 9, 131 11 of 19

ET applications [28,29]. The proposal is based on the following idea: the greater part of the power is
processed by a very-fast-output Switching-Mode DC–DC converter, while only extreme variations at
the output are tracked by the linear RFPA. A good trade-off between extremely-fast-output response
Electronics
and high 2020,power
9, x FOR PEER REVIEW
efficiency is achieved using this solution. The two most common linear switching 12 of 19

combinations are depicted in Figure 9. Series combination (Figure 9b) is often regarded as less efficient,
as all the current through the LEDs has to pass through a dissipative linear stage. However, if the input
voltage to this i LEDoutperform the parallel combination [29]. One ofi the
Electronics 2020, 9, stage is carefully
x FOR PEER REVIEWselected, it can 12 of LED19

main difficulties is to drive the transistor, S, whose source terminal is not referred to ground.
V CC V CC
Very-fast-output Very-fast-output
response Switching- i LED
i LED
LED response Switching- i LED
Splitter Mode dc-dc Splitter Mode dc-dc
converter converter S
vi V CC
+ Very-fast-output
response Switching-
vi V CC
i LED
Very-fast-output + i LED
response Switching-
Splitter Mode dc-dc Splitter Mode dc-dc
vvi Linear
converter Linear
converter S
i + RFPA vi
vi
+ RFPA
LEDs LEDs
vi Linear vi Linear
(a)RFPA RFPA
(b)
LEDs LEDs
Figure 9. Linear-assisted technique applied to VLC-LED drivers. (a) parallel connection scheme; (b)
series connection scheme.
(a) (b)
Figure 9. Linear-assisted technique applied to VLC-LED drivers. (a) parallel connection scheme; (b)
The idea9. ofLinear-assisted
Figure adapting the technique applied totechnique
linear-assisted VLC-LED drivers.
for VLC (a)resides
parallel connection scheme;
in exploiting the light,
series connection
(b) series scheme.
connection scheme.
allowing summing both signals in terms of light, instead of electrically. This modification greatly
simplifies
Thethe
6. Adaptation
ideadesign.
of
of the On thethe
onelinear-assisted
Linear-Assisted
adapting hand, the design
Technique to of
VLCthe for
technique circuitry
VLC that combines
resides both signals
in exploiting is no
the light,
longer necessary,
allowing summing while, on the other, eliminating the electrical connection avoids the influence
Taking the blockboth signals
diagram in terms
in Figure 9a of
as light, instead
a reference, of electrically.
Figure 10 shows the This modification
proposed greatly
adaptation of
between the
simplifies different
the design.circuitry
On the involved.
one hand, This
the influence
design of is
the especially
circuitry critical
that in
combines resonant
both circuits
signals is such
no
the linear-assisted technique for VLC. The proposal follows the same principle: the high efficiency
as longer
Class Enecessary,
RFPAs, inwhile,whichon efficiency
the and
other, correct behavior
eliminating greatlyconnection
the electrical depend onavoids
the correct
the tuning of
influence
power stage comprises a Class E RFPA instead of a Switching-Mode DC–DC converter, which will be
thebetween
resonant thecircuit.
different circuitry involved. This influence is especially critical in resonant circuits such
linear-assisted by a linear RFPA.
as Class E RFPAs, in which efficiency and correct behavior greatly depend on the correct tuning of
the resonant circuit.
Conventional
Switching-Mode
dc-dc converter i LED1
Conventional i bias_LED1
Switching-Mode
Bias control
dc-dc vconverter
linear i bias_LED2 i bias_LED1 i LED1
V CC i bias_LED1
vPSK Bias control vlinear i bias_LED2 i bias_LED1 i LED2
Linear V CC i
vPSK - + LED1 i bias_LED1
RFPA1 vlinear i LED2 i LED2
Splitter Linear i
- + LED1 i bias_LED1
vi RFPA1 vlinear
vClassE i LED2
Splitter
vRX
vi V CC v
ClassE
vRX
V CC
Class E
RFPA2 vClassE - +
Class E
v - + LEDs
RFPA2 ClassE

LEDs
Figure 10. Adapting the Linear-Assisted technique to a Class E RFPA as a VLC-LED driver, exploiting
Figure 10. Adapting
summing the light.the Linear-Assisted
This scheme is only technique to amodulation
valid for PSK Class E RFPA as a VLC-LED driver, exploiting
schemes.
summing theAdapting
Figure 10. light. This
thescheme is only valid
Linear-Assisted for PSK
technique to amodulation
Class E RFPAschemes.
as a VLC-LED driver, exploiting
summing the light. This scheme is only valid for PSK modulation schemes.
In this case (Figure 10), instead of having one string of LEDs, the load is split into two strings,
each one In being connected
this case to the
(Figure 10), Classof
instead E having
RFPA or thestring
one linearofRFPA,
LEDs, respectively. Both
the load is split strings
into of LEDs
two strings,
areeach
biasedone being connected to the
by a conventional Class E RFPA orDC–DC
Switching-Mode the linear RFPA, respectively.
converter, Both strings
which controls of LEDs
the bias current
are biased
(ibias_LED1 by a).conventional
= ibias_LED2 By controllingSwitching-Mode DC–DC
the bias current, converter,
both LED which
strings controlsworking
are always the biaswithin
currentthe
(ibias_LED1 = ibias_LED2). By controlling the bias current, both LED strings are always working within the
Electronics 2020, 9, 131 12 of 19

The idea of adapting the linear-assisted technique for VLC resides in exploiting the light, allowing
summing both signals in terms of light, instead of electrically. This modification greatly simplifies the
design. On the one hand, the design of the circuitry that combines both signals is no longer necessary,
while, on the other, eliminating the electrical connection avoids the influence between the different
circuitry involved. This influence is especially critical in resonant circuits such as Class E RFPAs, in which
efficiency and correct behavior greatly depend on the correct tuning of the resonant circuit.
In this case (Figure 10), instead of having one string of LEDs, the load is split into two strings, each
one being connected to the Class E RFPA or the linear RFPA, respectively. Both strings of LEDs are
biased by a conventional Switching-Mode DC–DC converter, which controls the bias current (ibias_LED1
= ibias_LED2 ). By controlling the bias current, both LED strings are always working within the linear
region, regardless of the temperature effect. As can be seen in Figure 10, each string of LEDs has
the same bias current, thus generating the same DC light intensity in order to perform the task of
illumination. Moreover, due to the connection of each string of LEDs to either the Class E RFPA or
linear RFPA, its combination generates the AC light intensity related to the communication task.
With this proposed adaptation, only SCM schemes can be reproduced—in this particular case,
only a Phase Shift-Keying (PSK) modulation scheme. The Class E RFPA is used to provide the great
amount of power associated with the communication signal at high efficiency. The linear RFPA only
supplies high slew-rate parts (phase changes) and corrects the distortion at the output.

7. Circuitry Design and Experimental Section

This section reports the design, building, and experimental validation of two prototypes as a
proof of concept of the adaptation of the two aforementioned techniques for VLC.

7.1. Outphasing Technique Using Two Class E RFPAs (Exploiting Summing the Light)
The first approach is performed to adapt the Outphasing technique to VLC by exploiting summing
the light. Figure 11 shows the complete circuit of the transmitter made up of: Two Class E RFPAs, two
LED strings, a DC–DC converter, and an FPGA.

7.1.1. Circuitry Design

The Class E RFPAs were built and each one was connected to an LED string of 8 XLamp MX-3
HB-LEDs from Cree (Durham, North Carolina, USA). An external DC–DC converter is necessary to
bias and control the average current through each LED string. In order to make the LED string work in
the middle of its linear region, the average current is kept at 250 mA for each string. The entire process
is controlled by an Artix-7 FPGA from Digilent (Pullman, WA, USA). Control of the average current is
also necessary in order to avoid the temperature drift suffered by LEDs. This means that the DC–DC
converter ensures that the LED always works in its linear region, regardless of the temperature. It is
worth mentioning that the DC–DC converter does not required any adaptation and can be implemented
using the same topologies used in normal LED drivers. The only difference is that the DC–DC converter
is connected through a Lbias = 5.6 µH biasing inductor acting as a low-pass filter. The value of the
inductor is obtained as follows:
RL
Lbias = , (1)
2π fsw/10

in order to have a cutoff frequency 10 times lower than the communication signal frequency. RL = 17.6 Ω
is the resistive equivalent of the LED string and fsw is the communication frequency. This enables
the DC biasing current to pass to the LEDs and also provides a high impedance path for the
communication signal.
 𝐿 = , (1)
in order to have a cutoff frequency 10 times lower than the communication signal frequency. 𝑅 =
17.6 Ω is the resistive equivalent of the LED string and 𝑓 is the communication frequency. This
enables the
Electronics DC
2020, biasing current to pass to the LEDs and also provides a high impedance path13for
9, 131 the
of 19
communication signal.

Figure 11. Experimental setup of the adaptation of the Outphasing technique using two Class E RFPAs.
Figure 11. Experimental setup of the adaptation of the Outphasing technique using two Class E
RFPAs.
7.1.2. Communication Scheme
The first part of Scheme
7.1.2. Communication the communication design is the selection of the communication scheme to
be used. Amplitude and phase modulation are chosen to test the capability of the adaptation of
the The first parttechnique.
Outphasing of the communication design scheme
The communication is the selection of the communication
used is 16-QAM digital modulationscheme
withto abe
used. Amplitude
carrier frequencyand of 5 phase modulation
MHz. The are chosen
symbol period is fivetosignal
test the capability
periods, of the
reaching adaptation
a bit-rate of the
of 4 Mbps.
Outphasing technique. The communication scheme used is 16-QAM digital modulation
The communication scheme defines the minimum bandwidth necessary for the amplifiers. According with a carrier
frequency
to [30], an of 5 MHz. The of
approximation symbol period is
the minimum five signalBW
bandwidth, periods, reaching
min , of the a bit-rate
modulation is of 4 Mbps. The
communication scheme defines the minimum bandwidth necessary for the amplifiers. According to
2
[30], an approximation of the minimumBW bandwidth,
min = =BW min, of the modulation is
2 MHz, (2)
Ts

where Ts = 1 µs is the symbol period.

7.1.3. Class E RFPA Design

The circuit of the Class E RFPAs that comprises the Outphasing circuit is shown in Figure 11.
Each Class E RFPA comprises a PD84010S-E RF MOSFET, a high speed EL7155 driver, a bias inductor,
and an output resonant circuit. The high switching frequency needed for the communication signal
requires the use of a high frequency RF MOSFET and a high speed driver. The bias inductor works
as a low-pass filter to bias the MOSFET and creates a high impedance path for the communication
signal. This means that no communication components flow through the power supply. The switching
frequency of fsw = 5 MHz is chosen in accordance with the communication signal. The bandwidth
of the Class E RFPA depends on the quality factor, Q, of the output resonant filter. The higher the Q,
the smaller the bandwidth. According to the definition of Q

fsw 5 MHz
Q= = = 2.5, (3)
BW 2 MHz
the value of Q is 2.5 for the selected bandwidth and switching frequency. Following the design
guidelines [31], the values of L, C and C1 that make up the output filter of the Class E RFPA are
obtained as follows:
2.85RL 0.7124 0.219
L= , C= , C1 = , (4)
2π fsw 2πRL fsw 2πRL fsw
and the values are given in Figure 11.

7.1.4. Experimental Results

The main waveforms of the light-outphasing process are shown in Figure 12a. Iph1 and Iph2 are
the currents through each LED string. The light emitted by each string is proportional to the current
flowing through it. As the amplitude of the sine waves is kept constant, the amplitude of the light
emitted by them is also constant. Vrx is the light received by an optical receiver, PDA10A-EC, which
receives the contribution of both strings. It can be seen in the figure that the received light is able to
 7.1.4. Experimental Results
The main waveforms of the light-outphasing process are shown in Figure 12a. Iph1 and Iph2 are
the currents through each LED string. The light emitted by each string is proportional to the current
flowing2020,
Electronics through
9, 131 it. As the amplitude of the sine waves is kept constant, the amplitude of the14 light
of 19
emitted by them is also constant. Vrx is the light received by an optical receiver, PDA10A-EC, which
receives the contribution of both strings. It can be seen in the figure that the received light is able to
reproduce
reproduceamplitude
amplitude changes
changeseven
eventhough
though thethe
individual lights
individual of both
lights strings
of both are constant.
strings This effect
are constant. This
iseffect
achieved due to the phase shift between the two waveforms.
is achieved due to the phase shift between the two waveforms.
The
Thewaveforms
waveformsofofthe theClass
ClassEERFPARFPAwhenwhena aphase
phasestep
stepoccurs
occursare
areshown
shownininFigure
Figure12b.
12b.AAClass
ClassE
RFPA is able to reach high efficiency due to the zero voltage switching (ZVS) achieved
E RFPA is able to reach high efficiency due to the zero voltage switching (ZVS) achieved in the main in the main
switch. The condition of ZVS is met before and after a phase shift in the Class E RFPA,
switch. The condition of ZVS is met before and after a phase shift in the Class E RFPA, as can be seen as can be seen
ininthe
thedrain
draintotosource
sourcevoltage,
voltage,VVdsds,, but
but then
then aa phase
phase shift
shift occurs, ZVS is
occurs, ZVS is lost,
lost, and
and hence
hence the
the power
power
efficiency decreases. Moreover, the higher the bit-rate,
efficiency decreases. Moreover, the higher the bit-rate, the more the more the phase shifts per second and hence
per second and hence
the
thelower
lowerthetheefficiency.
efficiency.Accordingly,
Accordingly,aatrade-off
trade-offbetween
between efficiency
efficiency and bit-rate
bit-rate has
has to
to be
be defined
definedin in
order
ordertotokeep
keepthetheefficiency
efficiencyand
andthethebit-rate
bit-ratewithin
withinadmissible
admissible values.
values.

(a) (b)

Figure 12. (a) communication waveforms of the light-Outphasing system; (b) main waveforms of the
Class E RFPA during a phase shift.

The proposed transmitter reproduces 16-QAM digital modulation, achieving a bit-rate of up

to 4 Mbps at a distance of up to 1 m. The Class E reaches an efficiency of 78% (which is the signal
generation efficiency). This efficiency is measured by comparing the input DC power delivered by
the 9 V biasing voltages, shown in Figure 11, and the RMS value of the communication signal in
the output. The communication signal is being recorded by an oscilloscope and post processed by
computer. The efficiency of the Class E is higher than the maximum efficiency reached by the Class A
and B alternatives for the 16-QAM modulation. The overall efficiency reached by the whole prototype
is 92%, when the signal and the lighting tasks are considered. The efficiency of the lighting task is
calculated by comparing the input power of the DC–DC converter and its output power, which is the
biasing power on the LED strings.

7.2. Linear-Assisted Technique for a Class E RFPA (Exploiting Summing the Light)
The second experimental approach is made to adapt the linear-assisted technique for a Class E
RFPA. A linear-assisted Class E amplifier was suitably designed and built to provide experimental
results (Figure 13). The circuitry is divided into four parts: Class E amplifier, error amplifier, linear
amplifier, and external DC–DC converter. The design resides in the idea of summing the signals in
terms of light, instead of electrically.

7.2.1. Circuitry Design

The Class E amplifier and the linear amplifier are each connected to an individual LED string
comprising 8 XLamp MX-3 HB-LEDs. Both strings are biased by an external DC–DC converter which
controls the average current through them. The DC–DC converter can be implemented by using a
normal LED driving stage. The only modification required are the two biasing inductors Lbias = 28 µH
 Electronics 2020, 9, 131 15 of 19

that are used to create a biasing path for the biasing current and to avoid any communication current
passing through the converter. The inductor is obtained as follows:

RL
Lbias = , (5)
2π fsw/10

in order to have a cutoff frequency 10 times lower than the communication signal frequency. RL = 17.6 Ω
is the resistive equivalent of the LED string and fsw is the communication frequency.
The Class E RFPA reproduces the communication signal, while the linear amplifier (LT1206)
reproduces the error signal between the ideal communication signal and the one delivered by the Class
Electronics
E RFPA. 2020,
The 9, xentire
FOR PEER REVIEW
process is controlled by an Artix-7 FPGA. 16 of 19

Figure
Figure 13.13. Experimental
Experimental setup
setup of of
thethe adaptationofofthe
adaptation thelinear-assisted
linear-assistedtechnique
techniquefor
foraaClass
ClassEERFPA.
RFPA.

7.2.2. Communication Scheme

7.2.2. Communication Scheme
The first part of the design consists in the selection of the communication scheme to be used. In this
The
case, first part
digital phase ofmodulation
the design (16-PSK)
consists in thechosen.
was selection
TheofClass
the communication scheme
E RFPA reaches its maximumto beefficiency
used. In
this
whencase,the
digital phaseofmodulation
amplitude (16-PSK) signal
the communication was chosen. The
does not Classover
change E RFPA
time.reaches
In phase itsmodulation,
maximum
efficiency
the amplitude is kept constant while the phase varies over time. The chosen carrier frequency In
when the amplitude of the communication signal does not change over time. is 1phase
MHz,
modulation, the amplitude is kept constant while the phase varies over time. The
the modulation uses 16 different symbols, which encode four bits each, giving a maximum bit-rate chosen carrier
frequency is 1 MHz,
of 0.5 Mbps. In linethe modulation
with what wasuses 16 different
previously donesymbols, whichto
and according encode four
[30], an bits each, giving
approximation a
of the
maximum bit-rate of 0.5 Mbps. In line with
minimum bandwidth, BWmin , of the modulation is what was previously done and according to [30], an
approximation of the minimum bandwidth, BWmin, of the modulation is
𝐵𝑊
BWmin = = 2 = = 250
250KHz,𝐾𝐻𝑧, (6)
(6)
Ts
where Ts = 8 µs is the symbol period.
where Ts = 8 µs is the symbol period.

7.2.3. Class
7.2.3. E RFPA
Class Design
E RFPA Design
Figure
Figure1313shows
shows the
theClass
ClassE ERFPA
RFPAcircuit.
circuit.ItItcomprises
comprisesthe thesame
sameRFRFPD84010S-E
PD84010S-E MOSFET MOSFET and
and
EL7155
EL7155 high
high speed
speeddriver
driverused
usedpreviously,
previously,aabias biasinductor,
inductor, and
and aa resonant
resonant filter.
filter. The
The switching
switching
frequency
frequency is is
chosen
chosen according
accordingtotothe
themodulation
modulationscheme;
scheme;ininthis thiscase,
case,11MHz
MHz is is used.
used. The
The resonant
resonant
filter
filter is designed according to the previously calculated modulation BW. According to thedefinition
is designed according to the previously calculated modulation BW. According to the definition
of of
Q,Q,
fsw 1 MHz
Q𝑄== == = 4,= 4, (7)
(7)
BW 250 kHz
thethe
necessary
necessary QQvalue toto
value provide
provideenough
enoughBW BWforforthe
thecommunications
communicationssignalsignalisis4.4.Following
Followingthethedesign
design
guidelines in [31], the values of 𝐿, 𝐶, 𝑎𝑛𝑑 𝐶 that make up the output filter of the Class
guidelines in [31], the values of L, C, and C1 that make up the output filter of the Class E are obtained E are obtained
as as
follows:
follows:
3.75RL 0.4166 0.2150
L =. , C =. , C .= , (8)
𝐿= 2π,f𝐶sw= 2πR , 𝐶L fsw
= 1 , L fsw
2πR (8)

and given in Figure 13.

In order to be able to measure the error on the communication signal delivered by the Class E
RFPA, a shunt resistance, Rsense, is placed in series with the LED string of the Class E RFPA. The
voltage across the shunt is used in the error amplifier, generating verr.
Electronics 2020, 9, 131 16 of 19

and given in Figure 13.

In order to be able to measure the error on the communication signal delivered by the Class E
RFPA, a shunt resistance, Rsense , is placed in series with the LED string of the Class E RFPA. The voltage
across the shunt is used in the error amplifier, generating verr .

7.2.4. Linear-Assisted Circuit Design

The linear RFPA and the error amplifier are shown in Figure 13. The signal, vclass-E , passes through
a 2nd-order Butterworth high-pass filter (HPF) in order to obtain only the communication signal,
without the biasing contribution. The cutoff frequency is 10 kHz. Moreover, the FPGA generates the
input signal modulated in PWM. In order to obtain the ideal communication signal, vin , from the
PWM, a 2nd-order Butterworth HPF and low-pass filter (LPF) are respectively needed to eliminate
Electronics 2020, 9, x FOR PEER REVIEW
the
17 of 19
DC component and the PWM harmonics. The cutout frequencies are 10 kHz and 3 MHz, respectively.
Then, vclass-E Then,
respectively. is subtracted
vclass-E isfrom vin , thereby
subtracted from vobtaining
in, therebythe error signal,
obtaining verr. The
the error error
signal, verr.signal, verr ,
The error
is amplified by the linear RFPA, LT1206, whose output signal is delivered to a string of
signal, verr, is amplified by the linear RFPA, LT1206, whose output signal is delivered to a string of eight XLamp
MX-3 XLamp
eight HB-LEDs. MX-3 HB-LEDs.

7.2.5. Experimental
7.2.5. Experimental Results
Results
Figure 14
Figure 14 shows
shows the
the waveforms
waveforms of of the
the communication
communication systemsystem reproducing
reproducing 16-PSK 16-PSK digital
digital
modulation. The
modulation. The carrier
carrier frequency
frequency is is 11 MHz
MHz and and the
the achieved
achieved bit-rate
bit-rate is is 0.5
0.5 Mbps.
Mbps. vvin shows the
in shows the
desired and
desired andideal
idealcommunication
communicationsignal signal that is used
that as aasreference.
is used a reference.vclass-E is the
vclass-E signal
is the signal delivered
deliveredby the
by
Class E RFPA and measured across the shunt resistance, R sense. By subtracting v class-E
the Class E RFPA and measured across the shunt resistance, Rsense. By subtracting vclass-E from vin, the from vin , the error
v err is vobtained,
error amplified by the linear RFPA and delivered to the LED strings. An optical receiver,
err is obtained, amplified by the linear RFPA and delivered to the LED strings. An optical

receiver, PDA10A-EC,the
PDA10A-EC, receives vout signal,
receives the vthe result of the sum in light of v in light
out signal, the result of the sum class-E
andofvv err . The addition of
class-E and verr. The
v err makes
addition ofvout
verrcloser
makes to vthe ideal signal, v , by reducing the distortion and error introduced by the
out closer to the inideal signal, vin, by reducing the distortion and error

introduced by the Class E RFPA,phase

Class E RFPA, especially during shifts.during phase shifts.
especially

Figure
Figure 14.
14. Communication
Communication waveforms
waveforms of
of the
the linear-assisted
linear-assisted transmitter.
transmitter.

The proposal is is able

able to
to reproduce
reproduce16-PSK
16-PSKphase
phasedigital
digitalmodulation
modulationwith witha acarrier
carrierfrequency
frequencyofof
1
1MHz,
MHz,achieving
achievinga abit-rate
bit-rateofof0.5
0.5Mbps.
Mbps.Since
SincethetheClass
ClassEEamplifier
amplifierdelivers
deliversmostmost ofof the
the power
power at a
high efficiency and the linear amplifier stage only delivers the error signal, the electrical efficiency of
the signal generation
generationisis75%.
75%.ThisThisefficiency
efficiencyisisbeing
beingcalculated
calculated bybycomparing
comparing thethe
bias input
bias power
input of
power
both amplifiers (Class E and linear amplifier) and comparing it with
of both amplifiers (Class E and linear amplifier) and comparing it with the RMS value of the RMS value of the output
communication
communication signal. The The communication
communication signal signal isis being
being recorded
recorded by by an oscilloscope and post
processed by
processed by computer.
computer.TheThesignal
signalgeneration
generationefficiency
efficiency is is higher
higher than
than thethe alternatives
alternatives basedbased solely
solely on
on a linear amplifier. When the signal and biasing power are taken
a linear amplifier. When the signal and biasing power are taken into account, the into account, the overall efficiency
reached is up to 85%.

8. Conclusions
VLC technology is a promising alternative for alleviating the congested RF spectrum, although
one of the main drawbacks of this technology is the low power efficiency of the driver when the
communication capability is added. Most of the research work related to VLC focuses on increasing
the bit-rate, more complex modulations, or communication schemes, among other issues, whereas
Electronics 2020, 9, 131 17 of 19

8. Conclusions
VLC technology is a promising alternative for alleviating the congested RF spectrum, although
one of the main drawbacks of this technology is the low power efficiency of the driver when the
communication capability is added. Most of the research work related to VLC focuses on increasing
the bit-rate, more complex modulations, or communication schemes, among other issues, whereas
little has been published on the improvement of power efficiency. This paper presents the adaptation
of some RF techniques aimed at improving overall efficiency.
Two different techniques are developed and adapted to VLC by exploiting summing the light:
Outphasing and linear assistance. Both techniques are based on summing two signals, which is done
electrically. By using the sum of the light in both techniques, the design is greatly simplified.
On the one hand, an adaptation of the Outphasing technique is presented based on two Class E
RFPAs. The proposed transmitter reproduces 16-QAM digital modulation, achieving a bit-rate of up to
4 Mbps at a distance of up to 1 m. The prototype reaches an electrical efficiency of 78% in terms of
signal generation (higher than the Class A and B maximum efficiency for 16-QAM modulation) and an
overall efficiency of 92% when the communication and lighting tasks are considered.
On the other hand, a linear-assisted VLC transmitter comprising a Class E amplifier and a linear
amplifier is presented. The proposal is able to reproduce 16-PSK phase digital modulation with a
carrier frequency of 1 MHz, achieving a bit-rate of 0.5 Mbps. Due to the fact that the Class E amplifier
delivers most of the power and the linear amplifier stage only delivers the error signal, the proposal
achieves an electrical efficiency of 75% in terms of signal generation (higher than the alternatives based
solely on a linear amplifier) and an overall efficiency of 85% considering the signal and the biasing
contribution of the DC–DC converter.

Author Contributions: J.R. and P.F.M. conceived and performed the experiments. D.G.A. and J.S. developed the
adaptation of solutions to VLC and the design procedure. D.G.A. and D.G.L. analyzed the data and wrote the
paper. V.F.R. and J.M. contributed with the definition, the design, and the evaluation of the experiments and the
revision of the paper. All authors have read and agreed to the published version of the manuscript.
Funding: This work was funded by the Spanish Government under Project MINECO-17-DPI2016-75760-R, by the
Principado de Asturias under Projects SV-PA-17-RIS3-4 and FC-GRUPIN-IDI/2018/000179, and the scholarship
BP17-91, as well as by European Regional Development Fund (ERDF) grants.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Cisco Systems. Cisco Visual Networking Index: Global Mobile DataTraffic Forecast Update, 2015–2020. 2016.
Available online: https://www.cisco.com/c/dam/m/en_in/innovation/enterprise/assets/mobile-white-paper-
c11-520862.pdf (accessed on 6 January 2020).
2. Ghassemlooy, Z.; Alves, L.N.; Zvanovec, S.; Khalighi, A. Visible Light Communications: Theory and Applications;
CRC Press: Boca Raton, FL, USA, 2017.
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; IEEE: Piscataway, NJ, USA, 2011; pp. 1–309.
4. Haas, H.; Chen, C.; O’Brien, D. A guide to wireless networking by light. Prog. Quantum Electron. 2017, 55, 88–111.
[CrossRef]
5. Sebastián, J.; Lamar, D.G.; Aller, D.G.; Rodríguez, J.; Miaja, P.F. On the Role of Power Electronics in Visible
Light Communication. IEEE J. Emerg. Sel. Top. Power Electron. 2018, 6, 1210–1223. [CrossRef]
6. Zhao, S.; Xu, J.; Trescases, O. Burst-mode resonant LLC converter for an LED luminaire with integrated
visible light communication for smart buildings. IEEE Trans. Power Electron. 2014, 29, 4392–4402. [CrossRef]
7. Modepalli, K.; Parsa, L. Dual-purpose offline LED driver for illumination and visible light communication.
IEEE Trans. Ind. Appl. 2015, 51, 406–419. [CrossRef]
8. Deng, X.; Arulandu, K.; Wu, Y.; Mardanikorani, S.; Zhou, G.; Linnartz, J.M.G. Modeling and Analysis of
Transmitter Performance in Visible Light Communications. IEEE Trans. Veh. Technol. 2019, 68, 2316–2331.
[CrossRef]
Electronics 2020, 9, 131 18 of 19

9. Arulandu, K.K.; Linnartz, J.; Deng, X. Efficient Amplitude Modulator and Ripple Canceller for Visible Light
Communication. In Proceedings of the 2018 IEEE Globecom Workshops, Abu Dhabi, UAE, 9–13 December
2018; pp. 1–6.
10. Wang, Y.; Shao, Y.; Shang, H.; Lu, X.; Wang, Y.; Yu, J.; Chi, N. 875-Mb/s asynchronous bi-directional
64QAM-OFDM SCM-WDM transmission over RGB-LED-based visible light communication system.
In Proceedings of the 2013 Optical Fiber Communication Conference and Exposition and the National Fiber
Optic Engineers Conference, Anaheim, CA, USA, 17–21 March 2013; pp. 1–3.
11. Khalid, A.M.; Cossu, G.; Corsini, R.; Choudhury, P.; Ciaramella, E. 1-Gb/s Transmission Over a Phosphorescent
White LED by Using Rate-Adaptive Discrete Multitone Modulation. IEEE Photonics J. 2012, 4, 1465–1473.
[CrossRef]
12. Vučić, J.; Kottke, C.; Habel, K.; Langer, K.D. 803 Mbit/s visible light WDM link based on DMT modulation
of a single RGB LED luminary. In Proceedings of the 2011 Optical Fiber Communication Conference and
Exposition (OFC/NFOEC) and the National Fiber Optic Engineers Conference, Los Angeles, CA, USA,
6–10 March 2011; pp. 1–3.
13. Wang, Y.; Huang, X.; Zhang, J.; Wang, Y.; Chi, N. Enhanced performance of visible light communication
employing 512QAM N-SC-FDE and DD-LMS. Opt. Express 2014, 22, 15328–15334. [CrossRef]
14. Tsonev, D.; Chun, H.; Rajbhandari, S.; McKendry, J.J.; Videv, S.; Gu, E.; Haji, M.; Watson, S.; Kelly, A.E.;
Faulkner, G.; et al. A 3-Gb/s Single-LED OFDM-Based Wireless VLC Link Using a Gallium Nitride LEDs.
IEEE Photonics Technol. Lett. 2014, 26, 637–640. [CrossRef]
15. Li, S.; Tan, S.; Lee, C.K.; Waffenschmidt, E.; Hui, S.Y.; Tse, C.K. A survey, classification, and critical review of
light-emitting diode drivers. IEEE Trans. Power Electron. 2016, 31, 1503–1516. [CrossRef]
16. Wang, Y.; Alonso, J.M.; Ruan, X. A Review of LED Drivers and Related Technologies. IEEE Trans. Ind. Electron.
2017, 64, 5754–5765. [CrossRef]
17. Castro, I.; Vazquez, A.; Arias, M.; Lamar, D.G.; Hernando, M.M.; Sebastian, J. A Review on Flicker-Free
AC–DC LED Drivers for Single-Phase and Three-Phase AC Power Grids. IEEE Trans. Power Electron.
2019, 34, 10035–10057. [CrossRef]
18. Arias, M.; Vázquez, A.; Sebastián, J. An overview of the AC-DC and DC-DC converters for LED lighting
applications. Autom. J. Control Meas. Electron. Comput. Commun. 2012, 53, 156–172. [CrossRef]
19. Raab, F.H.; Asbeck, P.; Cripps, S.; Kenington, P.B.; Popovic, Z.B.; Pothecary, N.; Sevic, J.F.; Sokal, N.O. Power
amplifiers and transmitters for RF and microwave. IEEE Trans. Microw. Theory Tech. 2002, 50, 814–826.
[CrossRef]
20. Doherty, W.H. A new high-efficiency power amplifier for modulated waves. Bell Syst. Tech. J. 1936, 15, 469–475.
[CrossRef]
21. Chireix, H. High power outphasing modulation. Proc. Inst. Radio Eng. 1935, 23, 1370–1392. [CrossRef]
22. Gaudernack, L.F. A Phase-Opposition system of amplitude modulation. Proc. Inst. Radio Eng. 1938, 26, 983–1008.
[CrossRef]
23. García, O.; Vasić, M.; Alou, P.; Oliver, J.Á.; Cobos, J.A. An Overview of Fast DC–DC Converters for Envelope
Amplifier in RF Transmitters. IEEE Trans. Power Electron. 2013, 28, 4712–4722. [CrossRef]
24. Sebastián, J.; Miaja, P.F.; Rodríguez, A.; Rodríguez, M. Analysis and Design of the Output Filter for Buck
Envelope Amplifiers. IEEE Trans. Power Electron. 2014, 29, 213–233. [CrossRef]
25. Rodríguez, M.; Miaja, P.F.; Rodríguez, A.; Sebastián, J. A multiple input digitally controlled buck converter
for envelope tracking applications in radiofrequency power amplifiers. IEEE Trans. Power Electron. 2010, 25,
369–381. [CrossRef]
26. Sebastián, J.; Miaja, P.F.; Ortega-González, F.J.; Patiño, M.; Rodríguez, M. Design of a two-phase buck
converter with fourth-order output filter for envelope amplifiers of limited bandwidth. IEEE Trans. Power
Electron. 2014, 29, 5933–5948. [CrossRef]
27. Rodríguez, J.; Lamar, D.G.; Miaja, P.F.; Aller, D.G.; Sebastián, J. Power Efficient VLC Transmitter Based
on Pulse-Width Modulated DCDC Converters and the Split of the Power. IEEE Trans. Power Electron.
2019, 34, 1726–1746. [CrossRef]
28. Miaja, P.F.; Sebastián, J.; Marante, R.; García, J.A. A Linear Assisted Switching Envelope Amplifier for a UHF
Polar Transmitter. IEEE Trans. Power Electron. 2014, 29, 1850–1861. [CrossRef]
Electronics 2020, 9, 131 19 of 19

29. Vasić, M.; García, O.; Oliver, J.A.; Alou, P.; Cobos, J.A. Theoretical Efficiency Limits of a Serial and Parallel
Linear-Assisted Switching Converter as an Envelope Amplifier. IEEE Trans. Power Electron. 2014, 29, 719–728.
[CrossRef]
30. Proakis, J. Digital Communications; Electrical Engineering Series; McGraw-Hill: New York, NY, USA, 2008.
31. Kazimierczuk, M.; Puczko, K. Exact analysis of Class E tuned power amplifier at any Q and switch duty
cycle. IEEE Trans. Circuits Syst. 1987, 34, 149–159. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).