Excem Research Papers in Electronics and Electromagnetics

Number 1 June 2020 ISBN 978-2-909056-21-0

EXCEM

A Typology of Antenna Tuner Control

Schemes, for One or More Antennas
FRÉDÉRIC BROYDÉ1 , and
EVELYNE CLAVELIER2
1
Excem, 12 chemin des Hauts de Clairefontaine, 78580 Maule, France
2
Eurexcem, 12 chemin des Hauts de Clairefontaine, 78580 Maule, France
Corresponding author: Frédéric Broydé (e-mail: [email protected]).

ABSTRACT We identify five types of antenna tuner control scheme, which are suitable for a wireless
transmitter using a single antenna. Four of them use a sensing unit measuring electrical variables either
at the radio port or at the antenna port of the antenna tuner. We also define and discuss four types of
antenna tuner control scheme which are relevant to a transmitter utilizing several antennas for MIMO radio
transmission. The accuracy and other characteristics of the different schemes are discussed and compared.

INDEX TERMS Antenna, impedance matching, antenna tuner, radio transmitter, MIMO, control system.

I. INTRODUCTION in Fig. 1 is a single-antenna-port (SAP) AT, but a multiple-

The characteristics of an antenna may be modified by the antenna-port (MAP) AT may be needed for MIMO radio
effects of the electromagnetic characteristics of the surround- transmission.
ings (EECS). For a portable wireless device, a cause of the Several authors have proposed descriptions of some SAP
EECS is the electromagnetic interaction between the antenna AT control schemes applicable to a transmitter, and defined
and a person holding the portable wireless device, often categories [6]–[8]. Sections II to VIII provide a review of
referred to as “user interaction”. existing SAP AT control schemes for wireless transmitters,
In current flagship mobile phone designs, automatic an- and a new classification into 5 types. Appendices A and B
tenna tuning, which automatically adjusts a tunable passive present a new analysis of their accuracy. In Section IX, we
antenna (TPA) and/or an antenna tuner (AT), has become qualitatively compare the different types. Section X provides
increasingly prominent as a method to support the grow- simulations of some properties of the different schemes, in a
ing range of LTE or 5G frequencies, mitigate the EECS, particular case.
reduce the size of the antennas, increase overall power effi- Sections XI and XII cover the MAP AT control schemes
ciency and signal consistency, and obtain the highest possible for a radio transmitter using multiple antennas, including a
data transmission rates [1]–[3]. Automatic antenna tuning is new classification into 4 types. Appendix C explains some
also common in land mobile, marine and tactical HF radio aspects of the signal processing used in these schemes, and
transceivers, as well as in radio transceivers for the amateur Appendix D presents a new analysis of their accuracy.
service [4]-[5].
This article is about control schemes which can be used, II. DEFINITIONS AND ASSUMPTIONS
in a radio transceiver or radio transmitter, to automatically Several control schemes, which can be used to automatically
adjust an AT. Each AT shown in Fig. 1 has a port, referred to adjust a SAP AT of a wireless transmitter, are based on
as “antenna port”, which is directly or indirectly coupled to one of the two configurations shown in Fig. 1. In both
an antenna, and another port, referred to as “radio port” (or configurations, the transmitter comprises: an antenna; the AT;
as “user port”), for transmitting and/or receiving radio signals a sensing unit (SU); a control unit (CU); and a transmission
through the AT and the antenna. In the case of a transceiver and signal processing unit (TSPU) which consists of all parts
using time-division duplex (TDD), each port may be an of the transmitter not shown elsewhere in Fig. 1.
input port or an output port, depending on whether emission The TX port of the TSPU delivers an excitation which
or reception is taking place. In the case of a transceiver is a bandpass signal having a carrier frequency denoted by
using frequency-division duplex (FDD), both ports are input fC . The SU delivers, to the TSPU, one or more sensing unit
ports and output ports, simultaneously. The AT is linear with output signals determined by one or more electrical variables
respect to its antenna port and radio port. Each AT shown (such as voltage, current, incident or forward voltage, etc)

Copyright © 2020 by Excem 1

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

one TCS (e.g., an AID whose reactance is determined

by one or more variable capacitance diodes, or barium
strontium titanate (BST) varactors [2]);
• category 3 refers to an AID which neither belongs to
category 1 nor to category 2 (e.g., an AID such as a
motorized variable capacitor, a motorized variometer or
a motorized roller inductor, in which the one or more
TCSs applied to the motor cause a variation of the
nominal reactance value at fC [4]).
AIDs of categories 1 and 2 are commonly used in low-
power applications (e.g., mobile phones). AIDs of categories
1 and 3 are commonly used in medium and high-power ap-
plications. Though AIDs of categories 1 and 2 often include
non-linear components that may cause non-linear effects
during emission, we assume that the AT behaves, with respect
to its radio port and antenna port, substantially as a passive
linear 2-port device.
We use ZSant to denote the impedance seen by the antenna
port, and ZU to denote the impedance presented by the radio
port, which depends on ZSant and on the impedances of the
AIDs. A wanted value of ZU being denoted by ZW , the user
FIGURE 1. Two possible configurations of a transmitter comprising an antenna, port tuning range, denoted by DU T R (ZW ), is the set of all
a SAP antenna tuner (AT), a sensing unit (SU), a control unit (CU) and a ZSant for which there exist achievable values of the nominal
transmission and signal processing unit (TSPU).
reactances of the AIDs, such that ZU = ZW at fC [10].
In the literature on ATs, “open-loop” often erroneously
caused by the excitation, sensed at the radio port in the case refers only to a control scheme without SU, so that the
of Fig. 1(a) or at the antenna port in the case of Fig. 1(b). The AT is typically adjusted only as a function of the operating
CU is an interface that delivers at least one tuning control frequency, which is known to the TSPU [2, Sec. 4.5.1], [11].
signal (TCS) to the AT. In this article, using the correct terminology, “open-loop con-
The AT comprises one or more adjustable impedance trol” means: control which does not utilize a measurement of
devices (AIDs) each having a reactance which is adjustable the controlled variable [12].
by electrical means. Adjusting an AT means adjusting the In contrast, “closed-loop control” (which is also referred
reactance of one or more of its AIDs. At a given time τ , a to as “feedback control”) means control in which the control
nominal reactance of an AID at fC is determined by initial action is made to depend on a measurement of the controlled
conditions of the AID at an earlier time τ0 , and by the history variable [12]. This definition does not imply that the control
of at least one TCS in the time interval [τ0 , τ ]. The reactance action repetitively or continuously depends on a repetitive or
and the resistance of the AID are functions of the nominal continuous measurement of the controlled variable.
reactance of the AID and of other variables such as tempera- In what follows, “model-based” refers to a control scheme
ture, humidity, aging, uncertainties, etc. The resistance of the which uses a model that describes relevant properties of the
AID is unwanted because it entails an unwanted loss. AT and the CU, and which can use a single sample of each
We identify 3 AID categories, each requiring particular of the one or more sensing unit output signals to produce
TCSs. They may be defined as follows: nominal AID reactance values intended to provide the wanted
• category 1 refers to an AID in which controlled switches adjustment of the AT.
are used to obtain a finite number of nominal reac-
tance values at fC (e.g., an AID which is a network III. TYPE 0 CONTROL SCHEMES
comprising capacitors or coils or open-circuited stubs Type 0 designates the open-loop AT control schemes which
or short-circuited stubs, and one or more electrically do not use any SU. In subtype a of type 0, the nominal
controlled switches or change-over switches, such as reactance (or an equivalent variable) of each AID is de-
electro-mechanical relays, or MEMS switches, or PIN termined only as a function of an operating frequency fO ,
diodes, or insulated-gate FETs, used to cause different typically by utilizing a lookup table, the entries of which
capacitors or coils or stubs of the network to contribute have for instance been determined based on experiments. In
to the reactance [2], [9]); subtype b, the nominal reactance of at least one AID (or
• category 2 refers to an AID which does not belong to an equivalent variable) is determined as a function of an
category 1, and such that, after a delay larger than the operating frequency fO and of at least one auxiliary variable
response time of the AID, its nominal reactance at fC which is assumed to be correlated with some electromagnetic
is mainly determined by the present value of at least characteristics of the surroundings of the transmitter. In a

2 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

FIGURE 2. A SAP AT having a L-network structure. FIGURE 3. An AT having a π-network structure.

mobile phone, such an auxiliary variable may for instance ZAU each correspond to an AID [25]. If we ignore losses
be [13]: in the AT, we have, at the frequency fC ,
• a localization variable assumed to depend on a distance 1
between a part of a human body and a zone of the ZU = + jXAU , (1)
1
transmitter, determined by a sensor such as a capacitive + jBA
proximity sensor, an infrared proximity sensor, etc; ZSant
• a localization variable determined by a change of state where the real susceptance BA and the real reactance XAU
of a switch of a keypad, or by a touchscreen; are such that YA = jBA and ZAU = jXAU . Let us assume
• a communication type variable that indicates whether a that ZW is real. Since ∂ZU /∂XAU = j, an inner feedback
radio communication session is a voice session or a data loop using Im(ZU ) or arg(ZU ) as feedback signal may easily
session; or be designed to control ZAU and provide a zero Im(ZU ). At
• a speakerphone mode activation indicator, etc. the frequency fC and ignoring losses in the AT, we have
The idea of subtype b is that such auxiliary variables can GSant
be used to mitigate the EECS, statistically, if a correlation Re(ZU ) = 2 , (2)
G2Sant + (BSant + BA )
exists between fO and the one or more auxiliary variables
on the one part, and optimal nominal reactance values on where the real conductance GSant and the real susceptance
the other part. In practice, the nominal reactance value (or BSant are such that GSant + jBSant = 1/ZSant . It follows
the value of the equivalent variable) of each AID may be that ZU = ZW is possible only if Re(ZW ) ≤ 1/GSant , and
obtained from a lookup table, as a function of fO and of that
a typical use configuration determined based on the one or ∂Re(ZU ) − 2GSant (BSant + BA )
more auxiliary variables. =h i2 . (3)
∂ BA 2
G2Sant + (BSant + BA )
IV. TYPE 1 SAP AT CONTROL SCHEMES Thus, if the antenna and the AIDs are such that the sign
For a SAP AT, type 1 designates the control schemes which of BSant + BA is known by design, an outer feedback
use the configuration of Fig. 1 (a) and are such that the TCSs loop using kRe(ZU ), where k is real, as feedback signal
are determined by a feedback control system which seeks to may be designed to control YA and provide Re(ZU ) =
obtain a wanted value of ZU at fC , without implementing Re(ZW ) = ZW . Also, if the outer feedback loop is made
extremum-seeking control. The schemes using the configura- much slower than the inner feedback loop, k|ZU | may be
tion of Fig. 1 (a) and extremum-seeking control are the type used as a feedback signal of the outer feedback loop.
2 control schemes presented in Section V, in which a detailed If the sign of BSant + BA is not known, a stable subtype
definition of extremum-seeking control is provided. a or b control scheme is more difficult to design for the
In a type 1 control scheme, the TSPU estimates q real AT shown in Fig. 2. For instance, a possible route would
quantities depending on ZU , where q is typically equal to be to ensure that, when the control system is switched on,
two. We can identify 3 subtypes. Subtype a designates the BSant + BA has always the wanted sign, so that this sign
schemes using continuous-time control, such as the ones would be maintained thereafter by the feedback control sys-
described in [4], [7, Sec. IV-A], and [14]-[16]. Subtype b des- tem, provided BSant and fC never vary too rapidly, and
ignates the non-model-based schemes which utilize discrete- provided ZSant remains in or close to the user port tuning
time control, such as the schemes described in [7, Sec. IV- range DU T R (ZW ) of the AT at fC .
B] and [17]-[19]. Subtype c designates the model-based Let us now consider a subtype a or b control scheme using
discrete-time control schemes, such as the ones described in the AT shown in Fig. 3, which has a π-network structure
[20]-[24]. [25]-[26]. Here, the series impedance ZAU is fixed, and the
Let ZW be the wanted value of ZU . In practice, a type 1 variable parallel admittances YA and YB each correspond
control scheme is designed for a particular SAP AT structure. to an AID. If we ignore losses in the AT, we have, at the
For subtypes a and b, the reactance of each AID is typically frequency fC ,
determined by a separate feedback control loop which uses
one of the q real quantities depending on ZU as feedback 1 1
= + jBU , (4)
signal. For instance, let us assume that the AT has the L- ZU 1
+ jXAU
network structure shown in Fig. 2, in which the variable 1
parallel admittance YA and the variable series impedance + jBA
ZSant

Copyright © 2020 by Excem 3

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

where the real susceptance BU is such that YU = jBU . Since Extremum-seeking control is a family of nonlinear control
∂(1/ZU )/∂BU = j, an inner feedback loop using Im(1/ZU ) methods whose purpose is to autonomously find either a
or Im(ZU ) or arg(ZU ) as feedback signal may be designed maximum or a minimum of a performance variable which
to control YU and provide a zero Im(1/ZU ). However, the is a real function of one or more outputs of a controlled
reader can easily check that ∂(1/ZU )/∂BA and ∂ZU /∂BA system, by controlling one or more inputs of the controlled
are involved, so that a stable subtype a or b control scheme is system. In an extremum-seeking control algorithm, one or
difficult to design for the AT shown in Fig. 3. more signals varying over time are caused to appear at
The error of a subtype b control scheme is determined in these one or more inputs of the controlled system, in a way
subsection B of Appendix A. that allows the algorithm to probe the nonlinearity of the
For subtype c, the control scheme is based on a model performance variable with respect to the one or more inputs
of the SAP AT and of the CU. For instance, let us assume of the controlled system, and to get closer to an extremum.
that the AT is the one shown in Fig. 3, that the nominal Thus, extremum-seeking control algorithms are based on the
reactance of any one of the AIDs is determined by a tuning information that the extremum exists, but they do not need
unit adjustment instruction received by the CU, and that the an exact knowledge of the controlled system to find the
absolute value and the phase of ZU have been estimated, extremum. For this reason, extremum seeking control is said
at fC , for nominal AID reactances determined by an initial to be a non-model-based real-time optimization approach. A
tuning unit adjustment instruction [23]-[24]. In a first step, type of extremum-seeking control which uses one or more
the TSPU estimates YA , ZAU and YU using the model and periodical perturbations is usually referred to as perturba-
the initial tuning unit adjustment instruction, and then ZSant tion based extremum-seeking control [28]. There are many
using other types of extremum-seeking control, such as sliding
1 1 mode extremum-seeking control, neural network extremum-
= − YA . (5)
ZSant 1 seeking control, relay extremum seeking control, perturb and
− ZAU observe, numerical optimization based extremum-seeking
1
− YU control, stochastic extremum-seeking control, etc [29]-[31].
ZU
In an automatic AT control scheme, the nominal reactances
If the estimated ZSant lies in or close to the user port
of the one or more AIDs may be regarded as the “one or more
tuning range DU T R (ZW ) of the AT at fC , the TSPU can, in
inputs of the controlled system”. Thus, the extremum-seeking
a second step, compute a subsequent tuning unit adjustment
control algorithm controls and varies the AID reactances
instruction such that, according to the model, ZU at fC is
over time, to get closer to an extremum of the performance
close to ZW . This computation may for instance use the fast
variable.
algorithm proposed in [27, Appendix C], which takes losses
in the AT into account. The operation of the control system The performance variable may be substantially the abso-
may stop at this point, so that the nominal AID reactances lute value of the reflection coefficient at the radio port, or
directly jump from the ones determined by the initial tuning any monotone function of this quantity [31, ch. 7], [32]-
unit adjustment instruction, to the ones determined by the [35]. The absolute value of this reflection coefficient is a
subsequent tuning unit adjustment instruction, by utilizing performance variable which typically varies very little far
the model twice. The error of a subtype c control scheme, from the sought global minimum, and which may present
at the end of the adjustment sequence that we have just several local minima at a given frequency. Thus, a type 2
described, is determined in subsection E of Appendix A. control scheme must be designed to avoid that the extremum-
If the model of the AT and of the CU is not accurate, the seeking control algorithm fails to converge, or converges to
subsequent tuning unit adjustment instruction may produce a local extremum which is not the wanted global extremum.
a ZU at fC which is not close to ZW . In subsection F of For this reason, in a typical type 2 control scheme, suitable
Appendix A, it is shown that, in this case, repeating the steps initial values of the nominal AID reactances are generated
that we have just described typically reduces the error of before extremum seeking starts, as a function of fC , using
the subtype c control scheme, and makes the error almost one of the type 0 control schemes.
independent of the accuracy of the model. We define 2 subtypes. Subtype a designates the schemes
In practice, one or more lookup tables are needed to obtain using continuous-time extremum-seeking [28]. Subtype b
an accurate model. Additionally, some of the computations designates the schemes using discrete-time extremum-
can be replaced with interpolations, if a suitable lookup table, seeking, such as the ones described in [32]-[35]. For subtype
or a suitable set of lookup tables, is provided. b, the error is computed in subsection B of Appendix A.
For AIDs of category 1, subtype b control schemes using It is worth mentioning that subtype b includes a brute force
digital processing or subtype c control schemes are preferred. extremum seeking technique applicable to the case where
each AID can provide a finite (and small) number of nominal
V. TYPE 2 SAP AT CONTROL SCHEMES reactance values: all combinations of AID reactance values
For a SAP AT, type 2 designates the control schemes which are tested, and a combination providing either the larger or
use the configuration of Fig. 1 (a) and in which the TSPU the smaller value of the performance variable is selected [36]-
uses extremum-seeking control to obtain that ZU at fC [38]. This approach does not use initial values of the nominal
approximates a wanted value ZW . AID reactances determined as a function of fC .

4 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

VI. TYPE 3 SAP AT CONTROL SCHEMES fC and of one or more real quantities depending on ZSant ,
For a SAP AT, type 3 designates the model-based control using a type 3 control scheme.
schemes which use the configuration of Fig. 1(b) and are The extremum-seeking control algorithm seeks to max-
such that: the TSPU estimates q real quantities depending imize or to minimize a performance variable estimated as
on ZSant at fC ; and the nominal reactance (or an equivalent a function of one or more sensing unit output signals. To
variable) of at least one AID is determined as a function of discuss possible performance variables, let sE (t) be the
fC and of these real quantities, using a model of the AT complex envelope of the excitation delivered by the TX port,
and of the CU. Typically, q = 2 [39]-[40]. Since type 3 is sA (t) be the complex envelope of an electrical variable (e.g.,
an open-loop control scheme, an accurate knowledge of the a voltage, a current, an incident voltage, etc) sensed at the
characteristics of the AT is essential for good results. If these antenna port, and f be a function which is differentiable and
characteristics depend on temperature, it is advantageous to strictly monotone over the set of positive real numbers.
take into account one or more temperatures in the AT to If the excitation is not amplitude modulated, that is to
determine the TCSs [41]. The aim of a type 3 control scheme say if |sE (t)| is constant, it is easily seen that a possible
is unconstrained, that is to say: it may be arbitrarily defined. performance variable is f (|sA (t)|). For instance, if f is an in-
If the aim of the control scheme is to obtain a wanted value creasing function, maximizing f (|sA (t)|) clearly maximizes
ZW of ZU at fC , we observe that the type 3 control scheme PSant [45]-[47].
has much in common with the second step of the operation of If the excitation is amplitude modulated, this approach
a type 1 subtype c control scheme, presented above in Section does not work, because a variation in |sE (t)| creates a varia-
IV. For instance, in the case of an AT having the structure tion in f (|sA (t)|). In this case, we observe that, for given val-
of a π-network, suitable TCSs may be determined using the ues of ZSant and of the AID reactances, sE (t) is substantially
iterative computation technique of [27, Appendix C], or a proportional to a modulating signal sM (t), so that |sA (t)| is
numerical algorithm that minimizes a suitable performance substantially equal to λ|sM (t)|, where λ is a positive real
variable, for instance |ZU − ZW |2 computed using the model which we want to maximize. Here, a possible performance
of the AT and of the CU. A detailed algorithm which directly variable is f (|sA (t)|)/f (|sM (t)|), provided f is such that,
takes into account the set of the nominal reactance values of for any positive λ, the ratio f (λ|sM (t)|)/f (|sM (t)|) is inde-
the AIDs has been disclosed [42]. pendent of |sM (t)|. The function f must therefore be such
If the aim of the control scheme is to maximize the aver- that, for any positive λ and for any positive x, we have
age power delivered by the antenna port at fC , denoted by
PSant , suitable TCSs may be determined using a numerical f (λ x) f (λ)
= . (6)
algorithm that maximizes PSant computed using the model f (x) f (1)
of the AT and of the CU. Thus, we have
The error of a type 3 control scheme is determined in
subsection D of Appendix A, or in Appendix B, according f (x) f (λ)
f (λ x) = . (7)
to the aim. This error depends on the accuracy of the model. f (1)
For any aim of the control scheme, some or all of the com- Taking a partial derivative of (7) with respect to x, and a
putations can be replaced with interpolations, if a suitable partial derivative of (7) with respect to λ, we obtain
lookup table is provided.
f ′ (x) f (λ) f (x) f ′ (λ)
= , (8)
VII. TYPE 4 SAP AT CONTROL SCHEMES λ f (1) x f (1)
For a SAP AT, type 4 designates the control schemes using
in which f ′ is the derivative of f . For λ = 1, we obtain the
the configuration of Fig. 1 (b) and such that [43]-[44]:
differential equation
• an initial value of each nominal AID reactance is gener-
ated, using open-loop control; and f ′ (x) 1 f ′ (1)
• to increase as much as possible the average power deliv- = . (9)
f (x) x f (1)
ered by the antenna port at fC , denoted by PSant , one
or more subsequent values of one or more of the nom- This is not a linear differential equation. However, if we
inal AID reactances are generated, using an extremum- first consider that k = f ′ (1)/f (1) is an arbitrary variable,
seeking control algorithm. we can integrate the resulting first-order linear differential
Generating initial nominal AID reactance values which are equation of parameter k. This allows us to find that the
not too far from the one that would maximize PSant has solutions of (9) are the functions which satisfy
two advantages: it avoids that the extremum-seeking control f (x) = K xk , (10)
algorithm converges to a local extremum which is not the
wanted global extremum, and it speeds up the convergence, where k and K are real constants, K being nonzero. Con-
for a given accuracy. For subtype a, the initial nominal AID versely, all functions given by (10) satisfy (6), and are strictly
reactance values are obtained as a function of fC , using one monotone for k nonzero. Thus, for an amplitude modulated
of the type 0 control schemes. For subtype b, the initial excitation, the suitable functions f are given by (10) where k
nominal AID reactance values are obtained as a function of and K are nonzero real constants.

Copyright © 2020 by Excem 5

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

combination, shown in Fig. 4, can provide a broader tuning

range than the one provided by the AT alone, and reduced
feeder losses [53]-[57]. An adjustment of a TPA may provide
a change in its directivity pattern, and/or a change in its
impedance. If the TPA is such that adjusting it produces a
change in its impedance, an automatic control system may
use the adjustment of the TPA to obtain a coarse adjustment
of ZSant , and then use the AT to obtain a fine adjustment,
using any type of control scheme.
Here, it is worth noting that some authors strangely refer
to the adjustment of a TPA as “aperture tuning”, and to the
adjustment of an AT as “impedance tuning”.

IX. COMPARISON OF SAP AT CONTROL SCHEMES

A. CLOSED-LOOP VERSUS OPEN-LOOP
Since an extremum-seeking control algorithm is based on
closed-loop control, we can say that types 1, 2 and 4 utilize
closed-loop control, whereas type 0 and type 3 only utilize
open-loop control. A remarkable characteristic of a typical
AT control system is the severe non-linearity of the equations
that govern the AT, for instance visible in (1)-(4). The relation
between the reactance of an AID and the TCS(s) it receives
FIGURE 4. Two possible configurations of a transmitter comprising a tunable
passive antenna (TPA), a SAP antenna tuner (AT), a sensing unit (SU), a control is typically also involved. For these reasons, a continuous-
unit (CU) and a transmission and signal processing unit (TSPU). time (analog) closed-loop control system, if started far from
its goal, is typically unable to reach it, or to reach it in a
reasonable time. Consequently, we may assume that practical
Ideally, a type 4 control scheme maximizes PSant with type 1 and type 2 control schemes include a preliminary
respect to the reactances of the AIDs. This maximization open-loop step, of type 0. Type 4 always includes a prelimi-
does not entail conjugate matching at the antenna port (as nary open-loop step, of type 0 or type 3. Thus, most current
opposed to a maximization of PSant with respect to the closed-loop control designs use digital circuits and lookup
resistance and the reactance seen by the antenna port). The tables, because they are unavoidable for an open-loop control
error of the control scheme is determined in Appendix B. step; and closed-loop control systems without digital circuits
If the radio port sees a linear source of impedance ZS , an (which can only be of type 1 subtype a or type 2 subtype a)
ideal type 4 control scheme maximizes the transducer power are outdated for most applications.
gain of the AT at fC . If the AT is a part of a transceiver using
TDD, what was said above applies to emission. If we further B. MEASUREMENTS AND MITIGATION OF EECS
assume that the AT is reciprocal with respect to its radio port The requirements on the SU and the processing of sensing
and antenna port, and that the radio port sees a linear load unit output signals depend on the control scheme, as shown in
of impedance ZS during reception, then the AT adjustment Table 1. The easiest measurements are: the scalar reflection
obtained for emission maximizes the transducer power gain coefficient measurements at the radio port, used in type 2,
of the AT during reception. This follows from a well-known which need only be accurate in the vicinity of ZU = ZW if
reciprocal power theorem [48, Sec. IV], according to which an effective preliminary open-loop step has been used; and
the transducer power gain of the AT during reception is equal the scalar measurements at the antenna port, used in type 4
to the transducer power gain of the AT during emission. subtype a, which can be relative voltage or current measure-
ments, since they are only used to find a maximum power.
VIII. OTHER SAP AT CONTROL SCHEMES Vector impedance measurements at the radio port, used in
We have defined five types of control scheme which can be type 1, are more involved, but they need only be accurate in
used to automatically adjust a SAP AT of a radio transmitter. the vicinity of ZU = ZW if an effective preliminary open-
They cover most schemes described in the literature, but not loop step has been used. Vector impedance measurements at
all of them. the antenna port, used in type 3 and type 4 subtype b, are the
First, some AT control schemes use more than one SU, for most challenging, because accuracy is needed in the entire
instance: one SU at the radio port and one SU inside the AT set of possible values of ZSant .
[8], [49]; or one SU at the radio port and another one at the As shown in Table 1, mitigation of the EECS is obtained
antenna port [50]-[51]. Many variations are possible. for all schemes using a SU, that is to say, for all schemes
Second, an AT may be used in conjunction with a tunable except type 0. A scheme that can mitigate the EECS can also
passive antenna (TPA), also referred to as “reconfigurable compensate other causes of antenna performance degrada-
antenna”, which may be a frequency-agile antenna [52]. This tion: manufacturing tolerances, aging, temperature, etc.

6 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

TABLE 1. Some possible characteristics of the SAP AT control scheme types and subtypes defined in Section III to Section VII.

Type Figure Subtype Measurement Mitigation of EECS Aim of control Accuracy Speed

a none no any poor very fast

0 –
b auxiliary variable limited any poor very fast
a vector at radio port yes ZU = ZW very good slow / fast
1 1(a) b vector at radio port yes ZU = ZW very good fast
c vector at radio port yes ZU = ZW good / very good very fast / fast
a scalar at radio port yes ZU = ZW very good very slow / medium
2 1(a)
b scalar at radio port yes ZU = ZW very good medium
3 1(b) – vector at antenna port yes any good very fast
a scalar at antenna port yes maximizing PSant very good medium
4 1(b)
b vector at antenna port yes maximizing PSant good / very good very fast / fast

C. AIM OF THE CONTROL SCHEME AND DESIGN GOAL control systems, based on the following considerations:
The aims of the different control schemes are shown in Table • all schemes using only open-loop control are very fast,
1. How do these aims correspond to possible design goals? but cannot be very accurate, because their accuracy
Let us for a while assume that the design goal is a max- depends on a model of the AT, and models are imperfect;
imization of PSant , in a context where the TX port of the • all schemes using closed-loop control are very accurate,
TSPU need not be linear, and where the SU is transparent to but type 1 subtype c is special, because it is very
the signals intended for the antenna. In the case of a lossless accurate if used with a sufficient number of repetitions,
AT, the average power delivered by the TX port is equal to but has a reduced accuracy if used without repetition;
PSant , so that a maximum power delivered by the TX port • the schemes using closed-loop discrete-time control are
corresponds to a maximum PSant (if the TX port was linear, supposed to include a preliminary open-loop step, so
a maximum power delivered by the TX port would imply a that the characteristics of the open-loop and closed-loop
conjugate matching at the TX port). Thus, in the case of a steps interact to provide accuracy and speed;
lossless AT, a type 1, 2 or 3 control scheme, configured to • if they do not include a preliminary open-loop step,
provide a value of ZU which maximizes the power delivered the schemes using closed-loop continuous-time control
by the TX port at fC (we assume that this value is known), (type 1 subtype a and type 2 subtype a) are slow at best;
maximizes PSant , like a type 4 control scheme. If losses in in the opposite case, their speed is similar to the one of
the AT are not very small, a maximum power delivered by the discrete-time control scheme of same type;
the TX port need not correspond to a maximum PSant , so • all schemes using closed-loop control are significantly
that the types 1, 2 and 3 control schemes considered above slower than an open-loop scheme, but type 1 subtype c
are not optimal for the design goal, while type 4 is optimal. It and type 4 subtype b are special, the latter because it
is advisable to investigate if the degradation in PSant inherent includes an accurate and very fast preliminary type 3
to a type 1, 2 or 3 control scheme can be ignored. step, so that a value of PSant which is very close to the
The possible maximization of the transducer power gain aimed maximum value is obtained very quickly;
during TDD reception, explained in Section VII, is another • type 2 subtype a is slower than type 1 subtype a, and
advantage of type 4. type 2 subtype b is slower than type 1 subtype b, be-
Let us now assume that the design goal is ZU = ZW , cause in type 2 schemes, a non-model-based extremum-
for instance because it provides a wanted linearity or spectral seeking control algorithm must probe the non-linearity
purity, or a wanted efficiency of a power amplifier, or simply of the performance variable, so that it follows an indirect
because ZW is the nominal load of the TX port. Here, a type path toward its aim.
1, 2 or 3 control scheme can be optimal, if it is configured to The model-based control schemes are type 1 subtype c,
provide ZU = ZW , while type 4 is not optimal (except in the type 3 and type 4 subtype b. The subtypes a and b of type 1
case of a lossless AT). This may require an investigation. are not model-based, even though they use a model of the AT
to determine in which direction the nominal reactance of each
D. ACCURACY, SPEED AND DEPENDENCE ON A AID must vary, in order to move from the current value of
MODEL OF THE AT ZU toward ZW . Model-based control schemes are very fast
The performance of a control system depends on many but computationally demanding, in particular type 1 subtype
implementation details. However, as a guideline, Table 1 c because it uses the model twice, at least, in an adjustment
indicates the relative accuracy and speed of the different sequence.

Copyright © 2020 by Excem 7

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

FIGURE 5. Real part R and imaginary part X of the impedance ZSant seen FIGURE 8. Reflection coefficient at the radio port versus frequency, for d =
by the antenna port, for d = 0.1 m. 0.1 m. Curve A: effect of discretization on type 1, 2 or 3. Curve B: worst case
of type 3 with 1% uncertainty. Curve C: type 4.

FIGURE 6. An AT having a π-network structure.

FIGURE 9. Insertion gain versus frequency, for d = 0.1 m. Curve A: type 1,

type 2 or type 3 using an accurate AT model. Curve B: type 4.

FIGURE 7. Quality factor QL of the coil and quality factor QC of each

adjustable impedance device, for the AT of Section X.

X. EXAMPLE FOR A QUANTITATIVE COMPARISON

We now want to illustrate the qualitative comparison of
Section IX with simulation results suitable for a quantitative FIGURE 10. Insertion gain versus d, at 800 MHz. Curve A: type 1, type 2 or
comparison of SAP AT control schemes. type 3 using an accurate AT model. Curve B: type 4.

We are going to investigate several SAP AT control

schemes, implemented on a theoretical system comprising a
74.9-mm-long dipole antenna, used in the frequency range Sec. 5], with LN = 10 nH, RS ≃ 641 mΩ, RP ≃ 6.74 kΩ
600 MHz to 1 GHz, a lossy feeder, and the AT shown in and C ≃ 63.4 fF. Fig. 7 shows the quality factor QL of
Fig. 6, which consists of a coil of inductance LAU and two the coil. For CA and CU , we use the AID model of [27,
AIDs each presenting a negative reactance, of capacitances Sec. 5], with ωP = 37 × 106 rd/s and ωS = 650 × 109 rd/s.
CA and CU , respectively. The configuration also includes According to this model, the quality factor is independent
a large plate made of a perfect electrical conductor (PEC) of the capacitance value. Fig. 7 shows the quality factor QC
lying parallel to the antenna, at a distance d of the antenna, of the AIDs. We also assume accurate sensing unit output
used to create and vary the EECS. The impedance ZSant signals, and an accurate analog or digital signal processing.
presented by the antenna and the feeder varies as a function We say that an AID is continuously adjustable (CA) if its
of the frequency and of d. The computed values of ZSant for set of nominal reactance values is an interval. Types 1 and
d = 0.1 m are shown in Fig. 5. 2 can exactly provide ZU = ZW , if each AID is CA. The
We clearly have an electrically small antenna with a small same applies to a type 3 control scheme aiming ZU = ZW ,
radiation power factor R/X, for which an impedance match- if, in addition, it uses an exact model of the AT. In Fig. 8, we
ing network can provide a match to an arbitrary resistance, assume ZW = 50 Ω, and we show the reflection coefficient
but only over a narrow bandwidth [58]. at the radio port, defined with respect to ZW , after automatic
The antenna and PEC are physically realizable, but not adjustment of the AT, in some cases which do not exactly
meant to represent an actual use configuration. However, provide ZU = ZW . For types 1, 2 and 3 control schemes
they are suitable for our purpose, and allow the reader to using AIDs which are not CA, Fig. 8(A) shows the effect of
easily reproduce our results. In contrast, our model of the a discretization of the capacitance values (64 logarithmically
AT is representative of what can be achieved in a hand-held spaced nominal values for CA and CU ). For type 3 and CA
transceiver. For LAU , we use the coil model presented in [27, AIDs, Fig. 8 (B) shows the effect of an inaccurate AT model

8 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

(1% uncertainty of the nominal values for CA and CU ). The

results for type 4 and CA AIDs are shown in Fig. 8 (C).
In Fig. 9 and Fig. 10, we assume that the TX port is linear
and presents an impedance of 50 Ω, we assume continuous
AIDs, and we show the insertion gain of the AT, i.e., the
ratio of the power delivered by the antenna port of the au-
tomatically adjusted AT, to the power received by the feeder
if it was directly coupled to the TX port. For type 3, we also
assume an exact model of the AT. Type 4 maximizing the
power delivered by the antenna port, the plots show that type
1, 2 and 3 are not optimal, by an amount ranging from about
0.14 dB to 0.29 dB in this example.
For type 4, the maximum reflection coefficient shown in
curve C of Fig. 8 is about −12.9 dB, corresponding to
a VSWR of about 1.58. The power amplifier of a typical
transmitter operates without problem with a VSWR less than
2. If this applies to our transmitter, a type 4 control scheme
can be used, and provides a more efficient transmitter than
the other types, as shown in Fig. 9 and Fig. 10.

XI. CONTROL SCHEMES FOR

MULTIPLE-ANTENNA-PORT ANTENNA TUNERS
A. NEW DEFINITIONS AND ASSUMPTIONS
We will now look at the control schemes which can be
used to automatically adjust a multiple-antenna-port (MAP)
AT of a transmitter using any number n > 2 of antennas
simultaneously in the same frequency band, for instance for
MIMO radio transmission.
A MAP AT has n antenna ports, each of which is directly
or indirectly coupled to an antenna, and m radio ports (or
“user ports”), for transmitting and/or receiving radio signals
through the AT and the antennas. The AT is linear with FIGURE 11. Two possible configurations of a transmitter comprising 2 anten-
respect to its antenna ports and radio ports. A MAP AT may nas, a MAP antenna tuner (AT) having 2 antenna ports and 2 radio ports, 2
sensing units (SU), a control unit (CU) and a transmission and signal processing
be used to adjust the impedance matrix presented by the unit (TSPU).
radio ports, denoted by ZU , so that it approximates a wanted
impedance matrix ZW . Typically, ZW is a diagonal matrix
because, if ZU is diagonal, it follows that: configuration in which the SUs are coupled to the n antenna
• the radio ports are uncoupled, and therefore suitable ports, shown in Fig. 11(b) for n = m = 2. The m TX ports
for maximum power transfer from m uncoupled power of the TSPU deliver excitations which are bandpass signals
amplifier outputs [59]; having a common carrier frequency denoted by fC . The SUs
• there is no power amplifier performance degradation deliver, to the TSPU, sensing unit output signals determined
caused by active load modulation resulting from mutu- by electrical variables caused by the excitations, sensed at the
ally coupled signals [60]-[62]; and m radio ports or at the n antenna ports. The CU is an interface
• if losses are sufficiently small, the radio ports have that delivers tuning control signals (TCSs) to the AT.
orthogonal radiation patterns, suitable for maximum
capacity [63]-[66]. B. ABOUT THE MAP AT
Two possible configurations of a wireless transmitter com- The MAP AT comprises AIDs each having a reactance which
prising a MAP AT such that n = m = 2 are shown is adjustable by electrical means, and adjusting the AT means
in Fig. 11. A radio transmitter, which uses any number adjusting the reactance of one or more of its AIDs. Let
n > 2 of antennas simultaneously in the same frequency ZSant be the impedance matrix seen by the antenna ports.
band, and which can automatically adjust a MAP AT, may ZU depends on ZSant and on the impedances of the AIDs. If
be composed of: the antennas; the MAP AT; sensing units ZSant is symmetric (i.e., the antennas are reciprocal), an AT
(SUs); a control unit (CU); and a transmission and signal which is reciprocal with respect to its radio ports and antenna
processing unit (TSPU) which consists of all other parts of ports is said to have a full tuning capability if [10]:
the transmitter. Several possible control schemes are based • any small symmetric variation δZU in ZU can be ob-
either on a configuration in which the SUs are coupled to the tained, if suitable achievable values of the reactances of
m radio ports, shown in Fig. 11(a) for n = m = 2, or on a the AIDs exist; and

Copyright © 2020 by Excem 9

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

admittance matrix of the AIDs coupled in parallel with the

radio ports.
Though (11) is applicable to both of them, the two ATs
produce different results, because YA , ZAU , and YU are
diagonal in Fig. 12, whereas they need not be diagonal in
Fig. 13. A full tuning capability is possible in Fig. 13, since
p = m(m + 1) = 6.
For a MAP AT which is reciprocal with respect to its radio
ports and antenna ports, and for a duplex technique which
utilizes the same frequencies for emission and reception,
FIGURE 12. A MAP AT having n = 2 antenna ports (denoted by AP1 and if we further assume that the antenna array is reciprocal
AP2) and m = 2 radio ports (denoted by RP1 and RP2). This AT is made of 2 and that the radio ports see a linear multiport generator of
uncoupled π-networks.
internal impedance matrix ZS during emission, and a linear
and reciprocal multiport load of impedance matrix ZS during
reception, we can assert that [81]-[82]:
• the set of the values of the transducer power gain during
emission, obtained for all nonzero excitations, has a
least element referred to as “minimum value”, and a
greatest element referred to as “maximum value”;
• the set of the values of the transducer power gain during
reception, obtained for all nonzero excitations, has a
least element referred to as “minimum value”, and a
greatest element referred to as “maximum value”;
• the maximum value of the transducer power gain during
FIGURE 13. A MAP AT having n = 2 antenna ports (denoted by AP1 and AP2) emission and the maximum value of the transducer
and m = 2 radio ports (denoted by RP1 and RP2). This AT has the structure
of a multidimensional π-network. power gain during reception are equal; and
• if m = n, the minimum value of the transducer power
gain during emission and the minimum value of the
• any small symmetric variation δZSant in ZSant can be transducer power gain during reception are equal.
compensated to the first order in ||δZSant ||, if suitable It follows that, for a TDD transceiver using the same an-
achievable values of the reactances of the AIDs exist. tennas and the same frequencies for emission and reception,
Let p be the number of AIDs in the AT. To obtain a full an adjustment of the MAP AT which is good (or optimal) for
tuning capability, it is necessary that p > m(m + 1), but this emission, is also good (or optimal) for reception, if the stated
condition is not sufficient [10]. assumptions hold.
A MAP AT may consist of uncoupled SAP ATs, one for
each antenna [67]. For m = n = 2, an example of such C. SOME RELEVANT MAP AT CONTROL SCHEMES
a MAP AT, made of 2 uncoupled π-networks, is shown in The definition of type 0 control schemes in Section III can
Fig. 12. A MAP AT consisting of n uncoupled SAP ATs need be used without modification with any MAP AT. Subtype a
not be able to provide a ZU that approximates ZW because, does not require any comment. For subtype b, it is advisable
for n > 3, a MAP AT made of n uncoupled SAP ATs to use a plurality of localization variables, because antennas
does not have a full tuning capability [10], [68]. However, at different locations have different relative positions with
interesting results may be obtained with this kind of MAP nearby things or body parts [83].
AT [69]-[74]. We now look at the control schemes which use the config-
Recently, several MAP ATs which cannot be separated into urations comprising SUs coupled either to the radio ports or
uncoupled SAP ATs have been proposed and investigated to the antenna ports, to see if and how each type and subtype
[10], [27], [68], [75]-[80]. For m = n = 2, an example of defined for SAP ATs can be adapted to MAP ATs.
such a MAP AT, having the structure of a multidimensional If there is no significant coupling between the n antennas,
π-networks, is shown in Fig. 13. Such a MAP AT may be for all relevant EECS, it is possible (and highly advisable)
able to provide a ZU that approximates an arbitrary ZW . to use a MAP AT consisting of n uncoupled SAP ATs, each
The MAP ATs of Fig. 12 and Fig. 13 are both such that, at having a separate control system using a single SU according
the frequency fC , to one of the control schemes considered above in Section
ZU = (((Z−1 −1
+ ZAU )−1 + YU )−1 , IV to Section IX. Preferably, such separate control systems
Sant + YA ) (11)
could adjust the SAP ATs during tuning sequences which
where YA is the admittance matrix of the AIDs coupled in do not overlap in time, to avoid any interaction between the
parallel with the antenna ports, where ZAU is an impedance tuning sequences, through antenna coupling.
matrix of the device(s) connected to a terminal of an antenna If a significant coupling exists or may exist between at least
port and to a terminal of a radio port, and where YU is the 2 of the n > 2 antennas, we may say that:

10 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

TABLE 2. Some possible characteristics of the MAP AT control scheme types and subtypes defined in Section XI.

Type Figure Subtype Measurement Mitigation of EECS Aim of control Accuracy Speed

a none no any poor very fast

0 –
b auxiliary variable limited any poor very fast
b vector at radio port yes Z U = ZW very good fast
1 11(a)
c vector at radio port yes Z U = ZW good / very good very fast / fast
2 11(a) b vector at radio port yes Z U = ZW very good slow / very slow
3 11(b) – vector at antenna port yes any good very fast

• a MAP AT consisting of n uncoupled SAP ATs is the error of a type 1 subtype b control scheme is determined
usually not optimal, since it does not provide a full in subsection B of Appendix D.
tuning capability, but it can nevertheless be selected, for Type 1 subtype c designates the model-based discrete-time
instance based on cost or size considerations; control schemes [87]. More precisely, such a control scheme
• it is incorrect to control each SAP AT of this MAP AT is based on a model of the AT and of the CU. For instance, let
by utilizing a separate control system using a single SU, us assume that the AT is the one shown in Fig. 12 or the one
as if coupling through the antennas could be ignored. shown in Fig. 13, that the nominal reactance of any one of the
In what follows, we consider control schemes that are AIDs is determined by a tuning unit adjustment instruction
compatible with coupling between the antennas. Such control received by the CU, and that ZU has been estimated, at fC ,
schemes can be used with any type of MAP AT. for nominal AID reactances determined by an initial tuning
To obtain suitable information from the sensing unit output unit adjustment instruction. In a first step, the TSPU estimates
signals, the m TX ports of the TSPU deliver m excitations, YA , ZAU and YU using the model and the initial tuning unit
one and only one of the excitations being delivered by each adjustment instruction, and then ZSant using
TX port, so that one and only one of the excitations is applied ZSant = (((Z−1 −1
− ZAU )−1 − YA )−1 .
U − YU ) (12)
to each of the radio ports of the AT. It is possible that the m
excitations are delivered successively by the TX ports [84]. If the estimated ZSant lies in or close to the user port
A more general choice is to use excitations such that, with tuning range DU T R (ZW ) of the AT at fC , the TSPU can, in
respect to the carrier frequency fC , the m complex envelopes a second step, compute a subsequent tuning unit adjustment
of the m excitations are linearly independent in the set of instruction such that, according to the model, ZU at fC is
complex functions of one real variable, regarded as a vector close to ZW . This computation may for instance use the fast
space over the field of complex numbers [85]-[87]. algorithm proposed in [27, Section 4], which takes losses
Extracting suitable information from the sensing unit out- in the AT into account. The operation of the control system
put signals, for such excitations, requires some computation, may stop at this point, so that the nominal AID reactances
as explained in Appendix C. This precludes continuous-time directly jump from the ones determined by the initial tuning
control schemes. Thus, the type 1 subtype a and type 2 unit adjustment instruction, to the ones determined by the
subtype a defined for SAP AT cannot be adapted to a MAP subsequent tuning unit adjustment instruction, by utilizing
AT and will not be considered further. the model twice. The error of a type 1 subtype c control
In the following presentation of the possible types, we scheme, at the end of the adjustment sequence that we have
shall use the user port tuning range, denoted by DU T R (ZW ), just described, is determined in subsection E of Appendix D.
defined as the set of all ZSant for which there exist achievable If the model of the AT and of the CU is not accurate, the
values of the nominal reactances of the AIDs, such that subsequent tuning unit adjustment instruction may produce
ZU = ZW at fC [10]. a ZU at fC which is not close to ZW . In subsection F of
Type 1 designates the closed-loop control schemes which Appendix D, it is shown that, in this case, repeating the steps
use SUs coupled to the radio ports, and are such that the TCSs that we have just described typically reduces the error of the
are determined by a feedback control system which seeks to type 1 subtype c control scheme, and makes the error almost
obtain a wanted value ZW of ZU at fC , without implement- independent of the accuracy of the model.
ing extremum-seeking control. Typically, the TSPU estimates Type 2 designates the closed-loop control schemes which
q real quantities depending on ZU , where q = 2m2 . Since use SUs coupled to the radio ports, and in which the TSPU
continuous-time control is ruled out, we can only identify 2 uses extremum-seeking control to obtain that ZU at fC
subtypes. approximates a wanted value ZW [84], [85]. Continuous-
Type 1 subtype b designates the non-model-based schemes time control being ruled out, only subtype b, which utilizes
utilizing discrete-time control. We are not aware of any discrete-time extremum-seeking, exists. The performance
description of a type 1 subtype b control scheme. However, variable may be any monotone function of a norm of the

Copyright © 2020 by Excem 11

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

image of ZU under a suitable matrix function, the matrix

function being for instance one of the functions h mentioned
in subsection A of Appendix D. In contrast to the SAP AT
case, we are not aware of a simple manipulation providing
a suitable performance variable from scalar measurements.
Thus, unfortunately, we must assume that vector measure-
ments at the radio ports are required for type 2 in the MAP
AT case. Also, type 2 is slower for a MAP AT than for a
SAP AT. In fact, the greater p, the slower the control scheme,
because the extremum-seeking control algorithm must seek
an extremum in a space of dimension p.
The error of a type 2 control scheme is computed in
subsection B of Appendix D.
Type 3 designates the model-based control schemes which
use SUs coupled to the antenna ports, and in which: the
TSPU estimates q real quantities depending on ZSant at fC ;
and the nominal reactance (or an equivalent variable) of at
least one AID is determined as a function of fC and of these
real quantities, using a model of the AT and of the CU [86].
Typically, q = 2n2 . Type 3 is an open-loop control scheme
which relies on an accurate model. Thus, if temperature is
a relevant parameter, it should be measured and taken into FIGURE 14. A MAP AT having 4 antenna ports (denoted by AP1 to AP4)
account. The aim of a type 3 control scheme is unconstrained. and 4 radio ports (denoted by RP1 to RP4). This AT has the structure of a
multidimensional π-network.
If the aim of the control scheme is to obtain a wanted
value ZW of ZU at fC , the type 3 control scheme is similar
to the second step of the operation of a type 1 subtype c obtain a fine adjustment, using any type of control scheme.
control scheme. In the case of an AT having the structure It is also possible to combine antenna selection, that is to say
of a multidimensional π-network, suitable TCSs may be the selection of n > 2 antennas or TPAs among N > n
determined using the iterative computation technique of [27, antennas or TPAs, with a MAP AT using any one of the
Section 4], or a numerical algorithm that minimizes a suitable control schemes [93].
performance variable, for instance the square of a norm of An automatic MAP AT control system using a type 1,
ZU − ZW computed using the model of the AT and of the type 2 or type 3 control scheme may also, if this is relevant,
CU. seek and obtain an authorization from a wireless network,
The error of a type 3 control scheme is determined in prior to let the TSPU deliver the excitations needed to per-
subsection D of Appendix D. form the adjustment of the AT [94]-[96].
Last and least, type 4 cannot be adapted to MAP ATs,
because maximizing the average power delivered by the XII. EXAMPLE USING A MAP AT
antenna ports is not a legitimate aim for wireless transmission Let q be a nonzero integer, 1q be the identity matrix of size
using multiple antennas. q by q, and r0 be an arbitrary resistance. For an arbitrary
The applicable considerations of Section IX-A lead us to impedance matrix Z of size q and the reference resistance r0 ,
consider that any practical type 1 or type 2 control scheme for the return figure F (Z) is the nonnegative real number defined
a MAP AT includes a preliminary open-loop step, of type 0. by [10]:
Thus, closed-loop control designs use lookup tables, because F (Z) = |||S(Z)|||2 , (13)
they are unavoidable for an open-loop control step.
Based on the foregoing, we have established Table 2, by where S(Z) is a scattering matrix defined by
utilizing an approach similar to the one used in Section IX. S(Z) = (Z + r0 1q )−1 (Z − r0 1q )

D. OTHER MAP AT CONTROL SCHEMES

= (Z − r0 1q )(Z + r0 1q )−1 , (14)
We have defined four types of control scheme which can be and where the spectral norm |||A|||2 of a square matrix
used to automatically adjust a MAP AT of a radio transmitter. A is the largest singular value of A [97, Section 5.6.6].
Other control schemes are of course possible. F (Z) expressed in decibels is FdB (Z) = 20 log(F (Z)). It
A MAP AT may be used in conjunction with TPAs [88]- is the largest value of the ratio, expressed in decibels, of the
[92]. As said for a SAP AT in Section VIII, this combination reflected power to the incident power, for all possible nonzero
can provide a broader tuning range than the one provided by excitations. In what follows, “decoupling and matching”
the AT alone, and reduced feeder losses. An automatic AT means ZU = ZW , where ZW = r0 1m , and r0 = 50 Ω.
control system may use the adjustment of the TPAs to obtain Decoupling and matching is exactly obtained if and only if
a coarse adjustment of ZSant , and then use the MAP AT to F (ZU ) = 0 or FdB (ZU ) = −∞ dB.

12 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

The MAP AT shown in Fig. 14 has the structure of a mul-

tidimensional π-network. It comprises p = m(m + 1) = 20
AIDs presenting a negative reactance, depicted using the
variable capacitor symbols in Fig. 14. Examples showed that
it may have a full tuning capability, and that it can be adjusted
to provide decoupling and matching over broad frequency
ranges, and in the presence of EECS [10], [27], [68], [79].
It has also been shown, with a theoretical example, that
F (ZU ) is not very sensitive to the values of the 20 AIDs [27].
This point is important for a preliminary open-loop step of a
closed-loop control scheme (type 1 or type 2), and essential
for the open-loop control schemes (type 0 or type 3). It is
therefore interesting to summarize the results obtained about
the sensitivity of the MAP AT of Fig. 14 to AID values, which
are based on the following assumptions: FIGURE 15. Histogram of the relative frequency of F (ZU ) in dB, for σC =
1%, obtained with 10000 samples, and normal distribution having the same
• the antennas are n = 4 side-by-side parallel lossless mean and the same standard deviation (dashed curve).
dipole antennas, each having a total length of 224.8 mm,
the radius of the array being 56.2 mm, each antenna
having a 60 mm long lossy feeder; and
• the components of the MAP AT have loss characteristics
defined by a model similar to the one used in Section X,
with LN = 2.7 nH, RS ≃ 119 mΩ, RP ≃ 20.7 kΩ,
C ≃ 48.8 fF, ωP = 9×106 rd/s and ωS = 3×1012 rd/s.
In [27, Sec. 5], it is shown that an adjustment providing
decoupling and matching exists at any tuning frequency in
the frequency range 700 MHz to 900 MHz, and the cor-
responding capacitance values of the adjustable impedance
devices are computed.
To obtain an accurate picture of the effect of simulta-
neous deviations of the capacitance of the 20 AIDs, we
can assume independent normally distributed capacitance
deviations, with a zero mean deviation from the computed FIGURE 16. mF and σF versus σC , and linear regression lines based on the
values obtained for 0.01% 6 σF 6 10%.
values, and a specified relative standard deviation σC . At the
tuning frequency of 800 MHz, we can determine the statistic
of the return figure F (ZU ) in dB, the mean of F (ZU ) in dB,
denoted by mF , and the corrected sample standard deviation XIII. CONCLUSION
of F (ZU ) in dB, denoted by σF . The histogram of Fig. 15 An appropriate AT with a suitable control scheme can be used
shows the relative frequency of F (ZU ) in dB, obtained for to obtain the best possible performance from one or more
σC = 1 %, with 10000 samples (of the MAP AT). The antennas or TPAs, over the broadest frequency range, and
assumption σC = 1 % could for instance correspond to to adaptively compensate the EECS. It may also compensate
a ±3% tolerance defined by a 3-sigma deviation. For the some other causes of antenna performance degradation, such
statistics shown in Fig. 15, we have mF ≈ −27.22 dB and as manufacturing tolerances, aging, temperature, etc. Thus,
σF ≈ 2.79 dB. The maximum value of F (ZU ) is −19.00 dB, it may be used to reduce antenna size and to simplify an-
whereas, for a normal distribution the probability of having tenna or TPA design. In the case of MIMO transmission,
F (ZU ) greater than mF +3σF ≈ −18.86 dB would be about an appropriate MAP AT with a suitable control scheme can
1.35 × 10−3 , corresponding to an expectation of about 13.5 also adaptively compensate the antenna interaction, that is,
for 10000 samples. provide decoupling and matching. In this case, the acceptable
This data shows that a specified maximum return figure coupling between antennas is increased, so that their separa-
of mF + 3σF ≈ −18.86 dB, at the tuning frequency, can tion could be reduced.
reliably be obtained with σC = 1 %. The specified maximum Five types of control scheme applicable to a SAP AT have
return figures of mF + 3σF , at the tuning frequency, which been presented. Four of them are also applicable to a MAP
can reliably be obtained with a given value of σC , may be AT. They have been compared as regards: their use of closed-
derived from Fig. 16, which shows mF and σF , as a function loop control and/or open-loop control; the measurements
of σC , obtained with 1000 samples for each value of σC . used; their ability to mitigate the EECS; the aim of the control
Thus, in this example, accurate preliminary open-loop scheme; their accuracy and speed; and their dependence on a
steps and accurate open-loop control schemes are realisable, model of the AT and CU. Some aspects of the comparison
since the sensitivity of F (ZU ) to AID values is reasonable. are summarized in Table 1 and Table 2. This material should

Copyright © 2020 by Excem 13

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

help system designers to select the most appropriate control We say that Z is (very) close to ZW if and only if h(Z) is
schemes for their applications. (very) close to zero; we say that Z is as close as possible to
Type 3 is an accurate and very fast open-loop scheme for ZW if and only if h(Z) is as close as possible to zero; etc.
which it might be useful to take into account one or more We assume a digital control system in which the nominal
temperatures in the AT to determine the TCSs. reactances (or equivalent variables) of the AIDs are, at a
Type 4 control schemes are only applicable to SAP ATs, given point in time, determined by the CU as a function of
for which they provide the best transmitter efficiency, be- a tuning unit adjustment instruction delivered by the TSPU.
cause they are accurate and their aim is maximizing PSant . An exact numerical model of the AT and of the CU may be
They leave a residual VSWR, which should be small enough. put in the form of a mapping denoted by gEU and defined by
Type 4 subtype a is inexpensive (because it uses scalar mea-
surement), but not fast. Type 4 subtype b is more expensive, gEU (f, ZSant , tC , aT ) = ZU , (18)
but it is faster.
where f is the frequency, where tC is the applicable tuning
For SAP ATs, the fastest schemes providing a good accu-
unit adjustment instruction, and where aT is a real vector of
racy are type 1 subtype c, type 3, and type 4 subtype b, all of
temperatures, which is sufficient to characterize the effects of
which use vector measurements. Type 1 subtype c combines
temperature on ZU . As an example, if the impedance of each
high speed and accuracy. In contrast to type 3 and type 4
AID depends on its temperature and if the characteristics
subtype b, it uses vector impedance measurements at the
of the CU do not significantly depend on temperature, the
radio port. For this reason, it might be the best choice in most
elements of aT could be the temperatures of the AIDs.
fast and accurate SAP designs.
At the frequency f and for the temperatures specified in
If several antennas are used simultaneously and coupling aT , the user port impedance range of the AT is given by
between all antennas cannot be ignored, multiple indepen-
dent SAP AT control schemes are not appropriate, even if the DU R (ZSant ) = {gEU (f, ZSant , tC , aT )|tC ∈ TC } , (19)
MAP AT is made of uncoupled SAP ATs.
For MAP ATs, the fastest schemes providing a good where TC is the set of the possible tuning unit adjustment
accuracy are type 1 subtype c and type 3, which both use instructions [10].
vector measurements. Type 1 subtype c combines high speed
and accuracy. In contrast to type 3, it uses vector impedance B. NON-MODEL-BASED DIGITAL CLOSED-LOOP SAP
measurements at the radio ports. For this reason, it might be AT CONTROL SCHEMES
the best choice in most fast and accurate MAP designs. In a non-model-based digital closed-loop control scheme
Being not essential to define our typology of AT control (that is, a type 1 subtype b or type 2 subtype b scheme), a full
schemes, the following subject matters have not been treated automatic adjustment of the AT requires several iterations,
in this article: the design of SAP ATs, the design of MAP each iteration comprising the following steps: applying an
ATs, the design of SUs and the processing of sensing unit excitation to the radio port; sensing electrical variables at the
output signals. They are of course essential for the design of radio port; delivering a tuning unit adjustment instruction;
an optimal automatic antenna tuning system. and delivering TCSs. After a sufficient number of iterations,
a final tuning unit adjustment instruction tCF is reached. If
the control scheme is well-designed, the measured value of
APPENDIX A ZU at fC while tCF is applicable, denoted by ZU M , satisfies
A. PURPOSE OF THIS APPENDIX AND NOTATIONS
In this Appendix A, we want to further explain and investi- ZU M ≃ ZW − dQCL1 (fC , ZSant , tC , aT ) , (20)
gate different types of SAP AT control scheme which seek to
where the mapping dQCL1 represents a quantization error
obtain that ZU at fC is very close, or as close as possible, to
which is known to the control system, but which cannot be
a wanted impedance ZW .
avoided because there is no tC in TC such that ZU M is closer
We need to clarify the meaning of “very close, or as close to ZW . Thus, the error of the control system is given by
as possible, to a wanted impedance ZW ”. Let us choose a
complex function of a complex variable, denoted by h, the
ZU − ZW ≃ ZU − ZU M
function being continuous and smooth where it is defined,
and such that h(ZW ) = 0. For instance, the function may be − dQCL1 (fC , ZSant , tC , aT ) , (21)
defined by
where ZU − ZU M is the measurement error.
h(Z) = Z − ZW , (15)
C. ADDITIONAL ASSUMPTIONS FOR MODEL-BASED
or by SAP AT CONTROL SCHEMES
−1
h(Z) = Z −1 − ZW , (16) In a model-based digital control scheme (that is, a type 1
subtype c or type 3 scheme), we assume that the TSPU,
or by instead of knowing the exact numerical model of the AT
h(Z) = (Z − ZW )(Z + ZW )−1 . (17) and of the CU, corresponding to gEU , knows an approximate

14 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

numerical model of the AT and of the CU, which corresponds with respect to the unknown ZSantE , to obtain an estimated
to a mapping gAU such that value ZSantE of ZSant . Thus, we have

gAU (f, ZSant , tC , aT ) ZU I − ZU IM = gAU (fC , ZSant , tCI , aT )

+ dAU (f, ZSant , tC , aT ) = ZU , (22) − gAU (fC , ZSantE , tCI , aT M )
where the mapping dAU represents the error of the approx- + dAU (fC , ZSant , tCI , aT ) . (28)
imate numerical model, and is not known to the control
ZSantE and aT M are then used by a suitable algorithm to
system.
determine tCS such that gAU (fC , ZSantE , tCS , aT M ) is very
close, or as close as possible, to the wanted impedance ZW .
D. TYPE 3 SAP AT CONTROL SCHEME
We note that this step is similar to the one leading to (23). We
A type 3 control scheme uses a measurement ZSantM may write
of ZSant at fC , and possibly a measurement aT M of
aT . Here, a suitable algorithm is used to find a tun-
gAU (fC , ZSantE , tCS , aT M )
ing unit adjustment instruction, denoted by tCS , such that
gAU (fC , ZSantM , tCS , aT M ) is very close, or as close as + dQCL2 (fC , ZSantE , tCS , aT M ) = ZW , (29)
possible, to the wanted impedance ZW . We write
where the mapping dQCL2 represents a quantization error
which is known to the control system, but which can-
gAU (fC , ZSantM , tCS , aT M )
not be avoided because there is no tC in TC such that
+ dQOL (fC , ZSantM , tCS , aT M ) = ZW , (23) gAU (fC , ZSantE , tC , aT M ) is closer to ZW . The resulting
ZU at fC while tCS is applicable is given by
where the mapping dQOL represents a quantization error
which is known to the control system, but which can-
not be avoided because there is no tC in TC such that gAU (fC , ZSant , tCS , aT )
gAU (fC , ZSantM , tC , aT M ) is closer to ZW . The resulting + dAU (fC , ZSant , tCS , aT ) = ZU . (30)
ZU at fC is given by
Thus, the error of the control system while tCS is applica-
gAU (fC , ZSant , tCS , aT ) ble is given by
+ dAU (fC , ZSant , tCS , aT ) = ZU . (24)
ZU − ZW = gAU (fC , ZSant , tCS , aT )
Thus, the error of the control system is given by − gAU (fC , ZSantE , tCS , aT M )
+ dAU (fC , ZSant , tCS , aT )
ZU − ZW = gAU (fC , ZSant , tCS , aT )
− dQCL2 (fC , ZSantE , tCS , aT M ) , (31)
− gAU (fC , ZSantM , tCS , aT M )
+ dAU (fC , ZSant , tCS , aT ) in which the first 3 terms of the right-hand side vanish for
− dQOL (fC , ZSantM , tCS , aT M ) , (25) exact measurements and an exact numerical model. Let us
use ΛAU to denote the mapping such that
in which the first 2 terms of the right-hand side vanish for
exact measurements. ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) =
= gAU (fC , ZSant , tCS , aT )−gAU (fC , ZSantE , tCS , aT M )
E. TYPE 1 SUBTYPE C SAP AT CONTROL SCHEME
+ dAU (fC , ZSant , tCS , aT )
In a type 1 subtype c control scheme, an adjustment sequence
comprises the following steps: an initial tuning unit adjust- − [gAU (fC , ZSant , tCI , aT ) − gAU (fC , ZSantE , tCI , aT M )
ment instruction tCI is delivered by the TSPU; a measure- + dAU (fC , ZSant , tCI , aT )] . (32)
ment ZU IM of ZU I is obtained, where ZU I is the value of
ZU at fC while tCI is applicable; and a subsequent tuning For any values of fC , ZSant , ZSantE , tCI , aT and aT M ,
unit adjustment instruction tCS is computed as explained we have
below, and delivered by the TSPU [23]–[24]. While tCI is
applicable, we have ΛAU (fC , ZSant , ZSantE , tCI , tCI , aT , aT M ) = 0 . (33)

gAU (fC , ZSant , tCI , aT ) It follows from (28) and (32) that
+ dAU (fC , ZSant , tCI , aT ) = ZU I . (26)
ZU I − ZU IM
Let aT M be an estimate of aT , possibly based on a + ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) =
measurement. The TSPU solves the equation = gAU (fC , ZSant , tCS , aT )−gAU (fC , ZSantE , tCS , aT M )
gAU (fC , ZSantE , tCI , aT M ) = ZU IM (27) + dAU (fC , ZSant , tCS , aT ) . (34)

Copyright © 2020 by Excem 15

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

Substituting (34) in (31), we can write that the error of the first adjustment sequence is quickly followed by a second
control system while tCS is applicable is given by adjustment sequence, such that the subsequent tuning unit
adjustment instruction of the first adjustment sequence be-
ZU − ZW = ZU I − ZU IM comes the initial tuning unit adjustment instruction of a sec-
+ ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) ond adjustment sequence. If the first adjustment sequence is
− dQCL2 (fC , ZSantE , tCS , aT M ) . (35) sufficiently accurate, the second adjustment sequence uses an
initial tuning unit adjustment instruction such that ZU IM is
By (27), ZSantE may be regarded as a function of fC , sufficiently close to ZW to obtain that the error of the control
tCI , aT M and ZU IM . Thus, by (29), tCS may be regarded system at the end of the second adjustment sequence satisfies
as a function of fC , tCI , aT M , ZU IM and ZW . Thus, by (38). Thus, under this assumption, thanks to the iteration, a
(32), ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) may be cancellation of errors is obtained, such that the error of the
regarded as a function of fC , ZSant , tCI , aT , aT M , ZU IM control system at the end of the second adjustment sequence
and ZW . Thus, we can define the mapping EAU such that becomes to a large extent independent of the accuracy of the
approximate numerical model.
EAU (fC , ZSant , tCI , aT , aT M , ZU IM , ZW )
= ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) . (36) APPENDIX B
If ZU IM = ZW , the control system believes that it has In this Appendix B, we want to further explain and investi-
reached ZW , so that tCS = tCI . Thus, using (33) and (36), gate different types of SAP AT control scheme which seek to
we obtain that for any values of fC , ZSant , tCI , aT , aT M maximize PSant at fC .
and ZW , we have We assume a digital control system in which the nominal
reactances (or equivalent variables) of the AIDs are, at a
EAU (fC , ZSant , tCI , aT , aT M , ZW , ZW ) = 0 . (37) given point in time, determined by the CU as a function of
With respect to the variable ZU IM of (36), the mapping a tuning unit adjustment instruction delivered by the TSPU.
EAU is probably neither smooth nor continuous, because of An exact numerical model of the AT, of the CU, and of the
the quantization error and possibly other reasons. However, TX port of the TSPU while the TSPU delivers the excitation
let us assume that the control system and its numerical (this port need not be linear), may be put in the form of a
model are such that, with respect to the variable ZU IM , the mapping denoted by pE and defined by
mapping EAU may be approximately considered as contin-
uous. Thus, by (37), if ZU IM is sufficiently close to ZW , pE (f, ZSant , tC , aT ) = PSant , (39)
then EAU (fC , ZSant , tCI , aT , aT M , ZU IM , ZW ) is close to
0 and ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) is close where f is the frequency, where tC is the applicable tuning
to 0. Thus, if ZU IM is sufficiently close to ZW , the error of unit adjustment instruction, and where aT is a real vector of
the control system while tCS is applicable satisfies temperatures, which is sufficient to characterize the effects of
temperature on PSant .
ZU − ZW ≃ ZU I − ZU IM In a type 4 subtype a or type 4 subtype b control scheme,
− dQCL2 (fC , ZSantE , tCS , aT M ) . (38) a full automatic adjustment of the AT requires several non-
model-based iterations, each iteration comprising the follow-
According to (38), the error of the control system while ing steps: applying an excitation to the radio port; sensing one
tCS is applicable is almost equal to the measurement error or more electrical variables at the antenna port; estimating
ZU I − ZU IM less the quantization error. If we compare (38) the performance variable; delivering a tuning unit adjustment
to (35), we observe that a cancellation of errors has occurred. instruction; and delivering TCSs. After a sufficient number
Also, the error given by (38) is to a large extent independent of iterations, a final tuning unit adjustment instruction tCF is
of the accuracy of the approximate numerical model, and reached. If the control scheme is well-designed, tCF maxi-
(38) is similar to (21) established for a non-model-based mizes PSant at fC . Thus, tCF satisfies
closed-loop control system. This advantage is a consequence
of the fact that the control system has used the approximate
tCF = argmax pE (fC , ZSant , tC , aT ) , (40)
numerical model of the AT and of the CU twice: the first time tC ∈TC
to estimate ZSantE , and the second time to determine tCS .
where TC is the set of the possible tuning unit adjustment
F. TYPE 1 SUBTYPE C WITH ITERATION FOR A SAP AT instructions. Here, the control system reaches its aim, ex-
Let us now assume that the adjustment sequence considered actly, and there is no effect of measurement errors, under
so far has used an initial tuning unit adjustment instruction the assumption that the performance variable is a monotone
tCI such that ZU IM need not be sufficiently close to ZW function of PSant .
to obtain that the error of the control system at the end In a type 3 control scheme, which is model-based, we
of this first adjustment sequence satisfies (38). Thus, at the assume that the TSPU, instead of knowing the exact nu-
end of this first adjustment sequence, the error is given merical model of the AT, of the CU, and of the TX port
by (35). We can now introduce an iteration, in which the of the TSPU, corresponding to pE , knows an approximate

16 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

numerical model of the AT, of the CU, and of the TX port of B. PSEUDO-COORDINATES
the TSPU, which corresponds to a mapping pA such that Let W be a bandwidth of the excitations. Since, for any
a ∈ {1, . . . , m}, the real signal vOC a (t) is a bandpass signal
pA (f, ZSant , tC , aT ) of carrier frequency fC and of complex envelope vEOC a (t),
+ qA (f, ZSant , tC , aT ) = PSant , (41) it follows that vOC a (t) has a Fourier transform, denoted
by VOC a (f ), and that vEOC a (t) has a Fourier transform,
where the mapping qA represents the error of the approximate denoted by VEOC a (f ) and such that [98, Sec. 2.2.2]-[99,
numerical model, and is not known to the control system. Sec. 2.1-2]
The type 3 control scheme uses a measurement ZSantM
of ZSant at fC , and possibly a measurement aT M of / [−W/2, W/2] ⇒ VEOC a (f ) = 0 ,
f∈ (44)
aT . Here, a suitable algorithm is used to find a tun-
ing unit adjustment instruction, denoted by tCS , such that and
pA (fC , ZSantM , tCS , aT M ) is as large as possible. Thus, W W
f ∈ [− , ] ⇒ VEOC a (f ) = kVOC a (f + fC ) , (45)
tCS satisfies 2 2
where k √is a dimensionless real constant which is chosen
tCS = argmax pA (fC , ZSantM , tC , aT M ) , (42) equal to 2 by some authors.
tC ∈TC
Let x(t) be any voltage or current measured at anyone of
and the error of the control system is given by the antenna ports or radio ports. We assume that the effects
on the antennas, of electromagnetic fields caused by external
∆PSant = pE (fC , ZSant , tCS , aT ) sources, can be ignored. The AT being linear with respect to
− max pE (fC , ZSant , tC , aT ) . (43) its antenna ports and radio ports, x(t) has a Fourier transform,
tC ∈TC denoted by X(f ), which at any frequency f satisfies
This error depends on the accuracy of the model. It van- m
X
ishes for exact measurements and an exact model. X(f ) = λa (f )VOC a (f ) , (46)
a=1

APPENDIX C where, for any a ∈ {1, . . . , m}, λa (f ) is a dimensionless

A. PURPOSE AND ASSUMPTIONS complex function if x(t) is a voltage, or a complex function
In this Appendix C, we provide theoretical examples that having the dimensions of admittance if x(t) is a current.
clarify how, in a MAP AT control scheme, the sensing unit Thus, x(t) is a bandpass signal of carrier frequency fC , of
output signals may be processed to fully determine ZU at complex envelope xE (t), the Fourier transform of xE (t),
fC in the configuration where the SUs are coupled to the denoted by XE (f ) being such that
m radio ports, shown in Fig. 11(a) for m = n = 2; or to f∈
/ [−W/2, W/2] ⇒ XE (f ) = 0 , (47)
fully determine ZSant at fC in the configuration where the
SUs are coupled to n antenna ports, shown in Fig. 11(b) for and
m = n = 2.
Let us number the radio ports from 1 to m, and let us f ∈ [−W/2, W/2]
number the excitations from 1 to m, in such a way that, if we m
X
use t to denote time, for any a ∈ {1, . . . , m}, the excitation ⇒ XE (f ) = k λa (f + fC )VOC a (f + fC ) . (48)
number a consists of an open-circuit voltage vOC a (t) applied a=1

to the radio port number a. We assume that vOC a (t) is Using (44) and (45) in (47) and (48), respectively, we obtain
a bandpass signal of carrier frequency fC and of complex m
envelope vEOC a (t). Let E be any suitable subspace of the
X
XE (f ) = λa (f + fC )VEOC a (f ) . (49)
set of complex functions of one real variable, regarded as a a=1
vector space over the field of complex numbers. We assume
that the complex envelopes vEOC 1 (t), . . . , vEOC m (t) are If W is sufficiently small compared to fC , then λa (f +fC )
linearly independent in E. To obtain this result, we may for is almost constant over [−W/2, W/2], so that
instance use m excitations that are delivered successively m
X
by the TX ports, or orthogonal excitations, or excitations XE (f ) ≃ λa (fC )VEOC a (f ) , (50)
obtained as explained below at the end of subsection E. We a=1
assume that, while the excitations are delivered by the TSPU, and
m
the radio ports see a linear multiport source of impedance X
matrix ZL , which has a positive definite hermitian part, so xE (t) ≃ λa (fC )vEOC a (t) . (51)
a=1
that it is the impedance matrix of a strictly passive network.
Note that, if ZL is created by the outputs of uncoupled Let S be the span of vEOC 1 (t), . . . , vEOC m (t) in E.
power amplifiers, ZL is diagonal, but we do not need this Since vEOC 1 (t), . . . , vEOC m (t) are assumed to be linearly
assumption. independent, they form a basis B of S. In what follows, based

Copyright © 2020 by Excem 17

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

on (51), λ1 (fC ), . . . , λm (fC ) are referred to as the “pseudo- invertible [81, Lemma 1]. Thus, JER is correctly determined
coordinates” of xE (t) in B. These pseudo-coordinates are by (54). By (53), we have
complex and time-independent.
UER ≃ ZU JER . (55)
By (46), x(t) is exactly a sum of m components, each
being the part of x(t) caused by one of the excitations, and It follows from (54) that JER is invertible, so that
each being a bandpass signal of carrier frequency fC . By
(49), xE (t) is exactly a sum of m components, each being ZU ≃ UER J−1
ER . (56)
the part of xE (t) caused by one of the excitations, and each The entries of UER are pseudo-coordinates of the com-
being the complex envelope of the part of x(t) caused by this plex envelopes vER 1 (t), . . . , vER m (t), and the entries of
excitation. Thus, if W is sufficiently narrow, it follows from JER are pseudo-coordinates of the complex envelopes
(51) that, for any a ∈ {1, . . . , m}, the complex envelope of iER 1 (t), . . . , iER m (t). In subsection “E. Implementation”
the part of x(t) caused by excitation a, which is equal to below, we explain how sensing unit output signals each
the part of xE (t) caused by excitation a, is approximately proportional to a voltage or current at one of the radio ports
λa (fC )vEOC a (t). We have consequently shown that, if W can be processed to obtain such pseudo-coordinates, hence
is sufficiently narrow, then for any a ∈ {1, . . . , m}, the part the entries of UER and JER . Once this is done, it is possible
caused by excitation a, of a voltage or current x(t) measured to compute J−1 ER and use (56) to obtain the wanted result: ZU
at anyone of the antenna ports or radio ports, is a bandpass at fC .
signal of carrier frequency fC , the complex envelope of
this bandpass signal being nearly vEOC a (t) times the a-th D. SENSING UNITS AT THE ANTENNA PORTS
pseudo-coordinate of the complex envelope of x(t) in B. We now outline an example of signal processing for the
configuration in which the sensing unit output signals are de-
C. SENSING UNITS AT THE RADIO PORTS termined by electrical variables sensed at the n antenna ports.
We now outline an example of signal processing for the For simplicity, we may for instance assume that the electrical
configuration in which the sensing unit output signals are variables are the voltages and currents at the antenna ports.
determined by electrical variables sensed at the m radio ports. For any b ∈ {1, . . . , n}, we can use iA b (t) to denote the
For simplicity, we may for instance assume that the electrical current flowing out of antenna port b, and vA b (t) to denote
variables are the voltages and currents at the radio ports. the voltage across antenna port b. As explained above, iA b (t)
For any b ∈ {1, . . . , m}, we can use iR b (t) to denote and vA b (t) are bandpass signals of carrier frequency fC . Let
the current flowing into radio port b, and vR b (t) to denote iEA b (t) and vEA b (t) be the complex envelopes of iA b (t) and
the voltage across radio port b. As explained above, iR b (t) vA b (t), respectively, for this carrier frequency.
and vR b (t) are bandpass signals of carrier frequency fC . Let For any a ∈ {1, . . . , m}, we can use uEA a to de-
iER b (t) and vER b (t) be the complex envelopes of iR b (t) and note the column vector of the a-th pseudo-coordinates
vR b (t), respectively, for this carrier frequency. of the complex envelopes vEA 1 (t), . . . , vEA n (t) in B,
For any a ∈ {1, . . . , m}, we can use uER a to denote the and we can use jEA a to denote the column vector of
column vector of the a-th pseudo-coordinates of the complex the a-th pseudo-coordinates of the complex envelopes
envelopes vER 1 (t), . . . , vER m (t) in B, and jER a to denote iEA 1 (t), . . . , iER n (t) in B. If the bandwidth of the complex
the column vector of the a-th pseudo-coordinates of the envelopes vEOC 1 (t), . . . , vEOC m (t) is sufficiently narrow,
complex envelopes iER 1 (t), . . . , iER m (t) in B. If the band- by (46) and (51), for any a ∈ {1, . . . , m}, we have
width of the complex envelopes vEOC 1 (t), . . . , vEOC m (t) is
sufficiently narrow, by (46) and (51), for any a ∈ {1, . . . , m}, jEA a ≃ K(ZU + ZL )−1 Ea , (57)
we have where Ea has the same meaning as above and where K is a
jER a ≃ (ZU + ZL )−1 Ea , (52) dimensionless complex matrix of size n by m, which depends
where Ea is a complex vector of size m, the entries of which on ZSant , and
are zero except the entry of row a which is equal to 1, and uEA a ≃ ZSant jEA a . (58)
uER a ≃ ZU jER a . (53) In (57) and (58), K, ZU and ZL are considered at fC .
Let UEA be the complex matrix of size n by m whose
In (52) and (53), ZU and ZL are considered at fC .
column vectors are uEA 1 , . . . , uEA m , and JEA be the
Let UER be the complex matrix of size m by m whose
complex matrix of size n by m whose column vectors are
column vectors are uER 1 , . . . , uER m , and JER be the com-
jEA 1 , . . . , jEA m . UEA is dimensionless, and JEA has the
plex matrix of size m by m whose column vectors are
dimensions of admittance. By (57) we have
jER 1 , . . . , jER m . UER is dimensionless, and JER has the
dimensions of admittance. By (52) we have JEA ≃ K(ZU + ZL )−1 . (59)
JER ≃ (ZU + ZL )−1 . (54) Since ZU + ZL is invertible, as explained above, JEA is
correctly determined by (59). By (58), we have
ZU + ZL being the impedance matrix of a strictly passive
network, its hermitian part is positive definite, so that it is UEA ≃ ZSant JEA . (60)

18 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

If m = n, we may assume that K(ZU + ZL )−1 is and digital quadrature demodulation, to obtain digital signals
invertible. In this case, it follows from (60) that representing the samples x̌E [1], . . . , x̌E [µ] of xE (t).
Let X̌E be the column vector of size µ whose entries
ZSant ≃ UEA J−1
EA . (61)
are x̌E [1], . . . , x̌E [µ]. By (51), for a well-designed signal
The entries of UEA are pseudo-coordinates of the com- processing, we have
plex envelopes vEA 1 (t), . . . , vEA n (t), and the entries of m
X
JEA are pseudo-coordinates of the complex envelopes X̌E ≃ λa (fC )V̌EOC a . (62)
iEA 1 (t), . . . , iEA n (t). In subsection “E. Implementation” a=1
below, we explain how sensing unit output signals each in which we want to determine the pseudo-coordinates
proportional to a voltage or current at one of the antenna ports λ1 (fC ), . . . , λm (fC ). There are five reasons for the ≃ sym-
can be processed to obtain such pseudo-coordinates, hence bol in (62): the ≃ symbol in (51); inaccuracies in the pro-
the entries of UEA and JEA . Once this is done, if m = n, it cess used to generate the excitations vOC 1 (t), . . . , vOC m (t);
is possible to compute J−1 EA and use (61) to obtain the wanted inaccuracies in the process used to obtain the samples
result: ZSant at fC . x̌E [1], . . . , x̌E [µ]; noise in these processes; and signals or
noise received by the antennas while x(t) is sensed.
E. IMPLEMENTATION
Let Λ(fC ) be the column vector whose entries are
To obtain ZU in subsection “C. Sensing units at the radio λ1 (fC ), . . . , λm (fC ), and let V̌EOC be the matrix of size µ
ports”, we have used (56) which involves a matrix inversion. by m whose column vectors are V̌EOC 1 , . . . , V̌EOC m . By
Likewise, to obtain ZSant in subsection “D. Sensing units at (62), we have
the antenna ports”, we have used (61) which also involves
V̌EOC Λ(fC ) ≃ X̌E . (63)
a matrix inversion. Thus, continuous-time control is not
possible, and we consider a discrete-time control scheme in We now want to obtain some solution of (63), regarded as
which we use a number µ > m of samples of each relevant an equation of unknown Λ(fC ). A least square solution of
variable to determine ZU or ZSant . the linear system (63) is a column vector Λ(fC ) such that
For any a ∈ {1, . . . , m}, we may for instance assume ||Λ(fC )||2 is minimal among all vectors Λ(fC ) for which
that the µ samples v̌EOC a [1], . . . , v̌EOC a [µ] of vEOC a (t) ||V̌EOC Λ(fC ) − X̌E ||2 is minimal, where the euclidian
are known as digital signals. The TSPU uses these samples norm of a column vector V is denoted by ||V||2 . The unique
to generate the excitation vOC a (t). The TSPU could for least square solution of (63) is
instance subject v̌EOC a [1], . . . , v̌EOC a [µ] to a digital to
+
analog conversion producing in-phase and quadrature analog Λ(fC ) = V̌EOC X̌E , (64)
signals used for in-phase/quadrature (I/Q) modulation, either
where M+ is used to denote the Moore-Penrose generalized
at the frequency fC to directly obtain the excitation vOC a (t),
inverse of a matrix M [97, Sec. 7.3.P7 and Sec. 7.3.P9], [100,
or at an intermediate frequency which is up-converted to
Sec. 4.3], [101, Sec. 5.7 and 5.8]. We required above that
obtain the excitation vOC a (t). Alternatively, the TSPU could
V̌EOC 1 , . . . , V̌EOC m are linearly independent. It entails
for instance subject v̌EOC a [1], . . . , v̌EOC a [µ] to a bandpass ∗
that rank V̌EOC = m, so that V̌EOC V̌EOC is of rank m
digital to analog conversion followed by an up-conversion, to
by [97, Sec. 0.4.6], where we have used M∗ to denote the
obtain the excitation vOC a (t). ∗
hermitian adjoint of a matrix M. Since V̌EOC V̌EOC is of
For any a ∈ {1, . . . , m}, let V̌EOC a be the column
size m by m, it is invertible so that all singular values of
vector of size µ whose entries are v̌EOC a [1], . . . , v̌EOC a [µ].
V̌EOC are positive and
Since, as explained above in the subsection “A. Purpose
and assumptions”, we want vEOC 1 (t), . . . , vEOC m (t) to be +
V̌EOC ∗
= (V̌EOC V̌EOC )−1 V̌EOC
∗
, (65)
linearly independent, we require that V̌EOC 1 , . . . , V̌EOC m
are linearly independent. This is possible because µ > m. where we have used [100, Sec. 4.3.7]. Thus, there is no need
+
Let x(t) be any voltage or current sensed by a SU. We to perform a singular value decomposition to obtain V̌EOC .
+
can assume that this SU delivers a sensing unit output signal It follows from (65) that rank V̌EOC = m. This ensures that
proportional to x(t). Several techniques may be used to get the solution given by (64) can reach any arbitrary value of
the samples of xE (t), denoted by x̌E [1], . . . , x̌E [µ]. The Λ(fC ). In the special case µ = m, rank V̌EOC = m entails
+ −1
TSPU could for instance subject x(t) to an I/Q demodulation that V̌EOC is invertible, so that V̌EOC = V̌EOC .
+
at the frequency fC (homodyne I/Q demodulation), or to We note that V̌EOC can be used to process the samples of
a down-conversion followed by an I/Q demodulation at an the complex envelop of all voltages and currents sensed by
intermediate frequency (heterodyne I/Q demodulation), to the SUs. Moreover, if V̌EOC 1 , . . . , V̌EOC m are determined
+
obtain two analog signals: the real part of xE (t) and the in advance, they can be stored in memory, and V̌EOC can be
imaginary part of xE (t). These analog signals may then stored in memory so that it does not need to be computed in
be converted into digital signals representing the samples real time. Thus, extracting Λ(fC ) from X̌E requires only a
x̌E [1], . . . , x̌E [µ] of xE (t). Alternatively, the TSPU could matrix multiplication, so that the pseudo-coordinates of the
for instance perform a down-conversion of x(t), followed complex envelope of each voltage or current sensed by the
by a conversion into digital signals using bandpass sampling SUs can be easily obtained.

Copyright © 2020 by Excem 19

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

An important point is that there is a lot of flexibility in the given point in time, determined by the CU as a function of
choice of V̌EOC 1 , . . . , V̌EOC m , so that it is possible to use a tuning unit adjustment instruction delivered by the TSPU.
a choice which is compatible with the requirements of the An exact numerical model of the AT and of the CU may be
standards applicable to a MIMO wireless network [85]-[87]. put in the form of a mapping denoted by gEU and defined by
For a ∈ {1, . . . , m}, we may for instance consider the
column vectors V̌CX a , V̌DX a and V̌EX a , of size µ and gEU (f, ZSant , tC , aT ) = ZU , (72)
such that where f is the frequency, where tC is the applicable tuning
V̌EX a = V̌CX a + V̌DX a (66) unit adjustment instruction, and where aT is a real vector of
where V̌CX a 6= 0, and where, for any b ∈ {1, . . . , m}, temperatures, which is sufficient to characterize the effects of
∗ ∗ temperature on ZU .
V̌CX a V̌DX b = 0 and (a 6= b) ⇒ V̌CX a V̌CX b = 0.
This assumption is for instance easily satisfied in a MIMO At the frequency f and for the temperatures specified in
OFDM signal, if V̌CX 1 , . . . , V̌CX m correspond to orthog- aT , the user port impedance range of the AT is given by
onal reference signals existing on one or more subcarriers,
DU R (ZSant ) = {gEU (f, ZSant , tC , aT )|tC ∈ TC } , (73)
and V̌DX 1 , . . . , V̌DX m correspond to any signals existing
on the other subcarriers. where TC is the set of the possible tuning unit adjustment
For any complex numbers µ1 , . . . , µm such that instructions [10].
m
X
µa V̌EX a = 0 , (67) B. NON-MODEL-BASED DIGITAL MAP AT
a=1 CLOSED-LOOP CONTROL SCHEMES
we have, for any b ∈ {1, . . . , m}, In a non-model-based digital closed-loop control scheme
m
X (that is, a type 1 subtype b or type 2 subtype b scheme), a full
∗
V̌CX b
∗
µa V̌EX a = µb V̌CX b V̌CX b = 0 . (68) automatic adjustment of the AT requires several iterations,
a=1 each iteration comprising the following steps: applying exci-
∗ tations to the radio ports; sensing electrical variables at the
Since V̌CX b 6= 0, we have V̌CX b V̌CX b 6= 0. It follows radio ports; delivering a tuning unit adjustment instruction;
that µb = 0.
and delivering TCSs. After a sufficient number of iterations,
Thus, V̌EX 1 , . . . , V̌EX m are linearly independent. It
a final tuning unit adjustment instruction tCF is reached. If
follows that a possible choice of V̌EOC 1 , . . . , V̌EOC m is
the control scheme is well-designed, the measured value of
V̌EOC 1 = V̌EX 1 , . . . , V̌EOC m = V̌EX m .
ZU at fC while tCF is applicable, denoted by ZU M , satisfies
APPENDIX D ZU M ≃ ZW − dQCL1 (fC , ZSant , tC , aT ) , (74)
A. PURPOSE OF THIS APPENDIX AND NOTATIONS
In this Appendix D, we want to further explain and investi- where the mapping dQCL1 represents a quantization error
gate different types of MAP AT control scheme which seek which is known to the control system, but which cannot be
to obtain that ZU at fC is very close, or as close as possible, avoided because there is no tC in TC such that ZU M is closer
to a wanted impedance matrix ZW . to ZW . Thus, the error of the control system is given by
We need to clarify the meaning of “very close, or as close
as possible, to a wanted impedance matrix ZW ”. Let us ZU − ZW ≃ ZU − ZU M
choose a matrix function, denoted by h, the matrix function − dQCL1 (fC , ZSant , tC , aT ) , (75)
being a function from a set of square complex matrices into
the same set of square complex matrices, the matrix function where ZU − ZU M is the measurement error.
being continuous and smooth where it is defined, and such
that h(ZW ) is a null matrix. For instance, the function may C. ADDITIONAL ASSUMPTIONS FOR MODEL-BASED
be defined by MAP AT CONTROL SCHEMES
h(Z) = Z − ZW , (69) In a model-based digital control scheme (that is, a type 1
subtype c or type 3 scheme), we assume that the TSPU,
or by instead of knowing the exact numerical model of the AT
h(Z) = Z−1 − Z−1
W , (70) and of the CU, corresponding to gEU , knows an approximate
or by numerical model of the AT and of the CU, which corresponds
h(Z) = (Z − ZW )(Z + ZW )−1 . (71) to a mapping gAU such that

We say that Z is (very) close to ZW if and only if a norm gAU (f, ZSant , tC , aT )
of h(Z) is (very) close to zero; we say that Z is as close as + dAU (f, ZSant , tC , aT ) = ZU , (76)
possible to ZW if and only if a norm of h(Z) is as close as
possible to zero; etc. where the mapping dAU represents the error of the approx-
We assume a digital control system in which the nominal imate numerical model, and is not known to the control
reactances (or equivalent variables) of the AIDs are, at a system.

20 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

D. TYPE 3 MAP AT CONTROL SCHEME matrix ZW . We note that this step is similar to the one leading
A type 3 control scheme uses a measurement ZSantM to (77). We may write
of ZSant at fC , and possibly a measurement aT M of
aT . Here, a suitable algorithm is used to find a tun- gAU (fC , ZSantE , tCS , aT M )
ing unit adjustment instruction, denoted by tCS , such that + dQCL2 (fC , ZSantE , tCS , aT M ) = ZW , (83)
gAU (fC , ZSantM , tCS , aT M ) is very close, or as close as
possible, to the wanted impedance matrix ZW . We write where the mapping dQCL2 represents a quantization error
which is known to the control system, but which can-
gAU (fC , ZSantM , tCS , aT M ) not be avoided because there is no tC in TC such that
+ dQOL (fC , ZSantM , tCS , aT M ) = ZW , (77) gAU (fC , ZSantE , tC , aT M ) is closer to ZW . The resulting
ZU at fC while tCS is applicable is given by
where the mapping dQOL represents a quantization error
which is known to the control system, but which can- gAU (fC , ZSant , tCS , aT )
not be avoided because there is no tC in TC such that + dAU (fC , ZSant , tCS , aT ) = ZU . (84)
gAU (fC , ZSantM , tC , aT M ) is closer to ZW . The resulting
ZU at fC is given by Thus, the error of the control system while tCS is applica-
ble is given by
gAU (fC , ZSant , tCS , aT )
+ dAU (fC , ZSant , tCS , aT ) = ZU . (78) ZU − ZW = gAU (fC , ZSant , tCS , aT )
− gAU (fC , ZSantE , tCS , aT M )
Thus, the error of the control system is given by
+ dAU (fC , ZSant , tCS , aT )
ZU − ZW = gAU (fC , ZSant , tCS , aT ) − dQCL2 (fC , ZSantE , tCS , aT M ) , (85)
− gAU (fC , ZSantM , tCS , aT M )
in which the first 3 terms of the right-hand side vanish for
+ dAU (fC , ZSant , tCS , aT ) exact measurements and an exact numerical model. Let us
− dQOL (fC , ZSantM , tCS , aT M ) , (79) use ΛAU to denote the mapping such that
in which the first 2 terms of the right-hand side vanish for ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) =
exact measurements.
= gAU (fC , ZSant , tCS , aT )−gAU (fC , ZSantE , tCS , aT M )
E. TYPE 1 SUBTYPE C MAP AT CONTROL SCHEME + dAU (fC , ZSant , tCS , aT )
In a type 1 subtype c control scheme, an adjustment sequence − [gAU (fC , ZSant , tCI , aT ) − gAU (fC , ZSantE , tCI , aT M )
comprises the following steps: an initial tuning unit adjust- + dAU (fC , ZSant , tCI , aT )] . (86)
ment instruction tCI is delivered by the TSPU; a measure-
ment ZU IM of ZU I is obtained, where ZU I is the value of For any values of fC , ZSant , ZSantE , tCI , aT and aT M ,
ZU at fC while tCI is applicable; and a subsequent tuning we have
unit adjustment instruction tCS is computed as explained be-
low, and delivered by the TSPU [87]. While tCI is applicable, ΛAU (fC , ZSant , ZSantE , tCI , tCI , aT , aT M ) = 0 . (87)
we have
It follows from (82) and (86) that
gAU (fC , ZSant , tCI , aT )
ZU I − ZU IM
+ dAU (fC , ZSant , tCI , aT ) = ZU I . (80)
+ ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) =
Let aT M be an estimate of aT , possibly based on a = gAU (fC , ZSant , tCS , aT )−gAU (fC , ZSantE , tCS , aT M )
measurement. The TSPU solves the equation + dAU (fC , ZSant , tCS , aT ) . (88)
gAU (fC , ZSantE , tCI , aT M ) = ZU IM (81)
Substituting (88) in (85), we can write that the error of the
with respect to the unknown ZSantE , to obtain an estimated control system while tCS is applicable is given by
value ZSantE of ZSant . Thus, we have
ZU − ZW = ZU I − ZU IM
ZU I − ZU IM = gAU (fC , ZSant , tCI , aT ) + ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M )
− gAU (fC , ZSantE , tCI , aT M ) − dQCL2 (fC , ZSantE , tCS , aT M ) . (89)
+ dAU (fC , ZSant , tCI , aT ) . (82)
By (81), ZSantE may be regarded as a function of fC ,
ZSantE and aT M are then used by a suitable algorithm tCI , aT M and ZU IM . Thus, by (83), tCS may be regarded
to determine tCS such that gAU (fC , ZSantE , tCS , aT M ) is as a function of fC , tCI , aT M , ZU IM and ZW . Thus, by
very close, or as close as possible, to the wanted impedance (86), ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) may be

Copyright © 2020 by Excem 21

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

regarded as a function of fC , ZSant , tCI , aT , aT M , ZU IM control system at the end of the second adjustment sequence
and ZW . Thus, we can define the mapping EAU such that becomes to a large extent independent of the accuracy of the
approximate numerical model.
EAU (fC , ZSant , tCI , aT , aT M , ZU IM , ZW )
= ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) . (90) REFERENCES
[1] Y. Chen, R. Martens, R. Valkonen, and D. Manteuffel, “Evaluation
If ZU IM = ZW , the control system believes that it has of adaptive impedance tuning for reducing the form factor of handset
reached ZW , so that tCS = tCI . Thus, using (87) and (90), antennas,” IEEE Trans. Antennas Propag., vol. 63, no. 2, pp. 703-710,
we obtain that for any values of fC , ZSant , tCI , aT , aT M Feb. 2015.
[2] J.L. Hilbert, Tunable RF Components and Circuits − Applications in
and ZW , we have Mobile Handsets, Boca Raton, FL, USA: CRC Press, 2016.
[3] W. Lam and B. Shaffer, The RF front-end: Unsung Hero of the Premium
EAU (fC , ZSant , tCI , aT , aT M , ZW , ZW ) = 0 . (91) Smartphone, Jul. 2017 [Online]. Available: https://technology.ihs.com.
[4] HF transceiver / All-mode multi bander – TS-450S/690S – Service
With respect to the variable ZU IM of (90), the mapping manual, Tokyo, Japan: Kenwood, 1991.
EAU is probably neither smooth nor continuous, because of [5] HF Antenna Tuning Unit R&S FK855U, data sheet, Rohde & Schwarz &
the quantization error and possibly other reasons. However, Co. KG, May 2006.
let us assume that the control system and its numerical [6] Y. Sun and J.K. Fidler, “High-speed automatic antenna tuning units,”
Proc. IEE Ninth Int. Conf. on Antennas and Propagation, (IEE Conf.
model are such that, with respect to the variable ZU IM , the Publ. No. 407), pp. 218-222, Apr. 1995.
mapping EAU may be approximately considered as contin- [7] E.L. Firrao, A.-J. Annema, and B. Nauta, “An automatic antenna tuning
uous. Thus, by (91), if ZU IM is sufficiently close to ZW , system using only RF signal amplitudes,” IEEE Trans. Circuits Syst. II,
Exp. Briefs, vol. 55, no. 9, pp. 833-837, Sep. 2008.
then EAU (fC , ZSant , tCI , aT , aT M , ZU IM , ZW ) is close to [8] A. van Bezooijen, M.A. de Jongh, F. van Straten, R. Mahmoudi, and
0 and ΛAU (fC , ZSant , ZSantE , tCS , tCI , aT , aT M ) is close A.H.M. van Roermund, “Adaptive impedance-matching techniques for
to 0. Thus, if ZU IM is sufficiently close to ZW , the error of controlling L networks,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol.
57, no.2, pp. 495-505, Feb. 2010.
the control system while tCS is applicable satisfies [9] E.L. Firrao, A.-J. Annema, F.E. van Vliet, and B. Nauta, “On the min-
imum number of states of switchable matching networks,” IEEE Trans.
ZU − ZW ≃ ZU I − ZU IM Circuits Syst. I, Reg. Papers, vol. 62, no. 2, pp. 433-440, Feb. 2015.
[10] F. Broydé and E. Clavelier, “Some properties of multiple-antenna-port
− dQCL2 (fC , ZSantE , tCS , aT M ) . (92) and multiple-user-port antenna tuners,” IEEE Trans. Circuits Syst. I, Reg.
Papers, vol. 62, no. 2, pp. 423-432, Feb. 2015.
According to (92), the error of the control system while [11] G. Weale, P. McIntosh, and S. Haddow, “Closed-loop antenna tuning:
tCS is applicable is almost equal to the measurement error why, what, where,” Proc. 2016 IEEE 5th Asia-Pacific Conference on
ZU I − ZU IM less the quantization error. If we compare (92) Antennas and Propagation, APCAP 2016, pp. 257-258, Jul. 2016.
[12] Dictionnaire CEI multilingue de l’électricité – IEC multilingual dictio-
to (89), we observe that a cancellation of errors has occurred. nary of electricity, vol. 1, Genève, Suisse: Bureau Central de la Commis-
Also, the error given by (92) is to a large extent independent sion Electrotechnique Internationale, 1983.
of the accuracy of the approximate numerical model, and [13] R. Scheer, E. Krenz, and I. Szini, “Adaptive antenna tuning systems and
methods,” U.S. Patent 8 204 446, filed Oct. 29, 2009.
(92) is similar to (75) established for a non-model-based
[14] V. True and B. Fisk, “Automatic Impedance Matching Apparatus,” U.S.
closed-loop control system. This advantage is a consequence Patent 2 745 067, filed Jun. 28, 1951.
of the fact that the control system has used the approximate [15] The A. R. R. L. Antenna Book, 8th edition, West Hartford, CT, USA: The
numerical model of the AT and of the CU twice: the first time American Radio Relay League, 1956.
[16] W.B. Bruene, ”Antenna Tuner Discriminator,” U.S. Patent 4 493 112,
to estimate ZSantE , and the second time to determine tCS . filed Nov. 19, 1981.
[17] M. Mileusnić, P. Petrović, and J. Todorović, “Design and implementation
F. TYPE 1 SUBTYPE C WITH ITERATION FOR A MAP AT of fast antenna tuners for HF radio systems,” Proc. 1997 IEEE Int. Conf.
on Information, Communications and Signal Processing (ICICS ’97), pp.
Let us now assume that the adjustment sequence considered 1722-1726, Sep. 1997.
so far has used an initial tuning unit adjustment instruction [18] A. van Bezooijen, M. de Jongh, C. Chanlo, L. Ruijs, H. J. ten Dolle, P.
tCI such that ZU IM need not be sufficiently close to ZW Lok, F. van Straaten, J. Sneep, R. Mahmoudi, and A.H.M. van Roermund,
“RF-MEMS based adaptive antenna matching module,” Proc. 2007 IEEE
to obtain that the error of the control system at the end Radio Frequency Integrated Circuits Symp., pp. 573-576, Jun. 2007.
of this first adjustment sequence satisfies (92). Thus, at the [19] A. van Bezooijen, M. A. de Jongh, C. Chanlo, L.C.H. Ruijs, F. van
end of this first adjustment sequence, the error is given Straten, R. Mahmoudi, and A.H.M. van Roermund, “A GSM/EDGE-
/WCDMA adaptive series LC matching network using RF-MEMS
by (89). We can now introduce an iteration, in which the switches,” IEEE J. Solid-State Circuits, vol. 43, no. 10, pp. 2259-2268,
first adjustment sequence is quickly followed by a second Oct. 2008.
adjustment sequence, such that the subsequent tuning unit [20] F. Chan Wai Po, E. de Foucauld, and P. Vincent, “Method for automatic
impedance matching for a radiofrequency circuit and transmission or
adjustment instruction of the first adjustment sequence be- reception system with automatic matching,” U.S. patent 8 140 033, filed
comes the initial tuning unit adjustment instruction of a sec- Sep. 10, 2008.
ond adjustment sequence. If the first adjustment sequence is [21] S.M. Ali, M.H. Bakr, and J. Warden, “Dynamic real-time calibration for
antenna matching in the transmission mode”, Proc. 2010 IEEE Antenna
sufficiently accurate, the second adjustment sequence uses an and Propagation Society Int. Symp., APS/URSI 2010, pp. 1-4, July 2010.
initial tuning unit adjustment instruction such that ZU IM is [22] F. Chan Wai Po, E. de Foucauld, D. Morche, P. Vincent, and E. Kerhervé,
sufficiently close to ZW to obtain that the error of the control “A Novel Method for Synthesizing an Automatic Matching Network and
system at the end of the second adjustment sequence satisfies Its Control Unit,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 58, no.
9, pp. 2225-2236, Sep. 2011.
(92). Thus, under this assumption, thanks to the iteration, a [23] F. Broydé and E. Clavelier, “Method for automatic adjustment of
cancellation of errors is obtained, such that the error of the a tuning unit, and apparatus for radio communication using this

22 Copyright © 2020 by Excem

Excem Research Papers in Electronics and Electromagnetics, no. 1, June 2020

EXCEM

method,” Patent Cooperation Treaty application PCT/IB2019/056447 [49] J.E. Roza, “Impedance matching system,” U.S. Patent 3 443 231, filed
(WO 2020/035756), filed 29 Jul. 2019. Apr. 27, 1966.
[24] F. Broydé and E. Clavelier, “Method for automatically adjusting [50] N.J. Smith, C.-C. Chen, and J.L. Volakis, “An improved topology for
a tuning unit, and apparatus for radio communication using this adaptive agile impedance tuners,” IEEE Antennas and Wireless Propagat.
method,” Patent Cooperation Treaty application PCT/IB2019/056484 Letters, vol. 12, pp. 92-95, Mar. 2013.
(WO 2020/039284), filed 30 Jul. 2019. [51] Q. Gu and A.S. Morris III, “A new method for matching network adaptive
[25] M. Thompson and J.K. Fidler, “Determination of the impedance match- control,” IEEE Trans. Microwave Theory Techn., vol. 61, no. 1, pp. 587-
ing domain of impedance matching networks,” IEEE Trans. Circuits Syst. 595, Jan. 2013.
I, Reg. Papers, Vol. 51, No. 10, pp. 2098-2106, Oct. 2004. [52] A. Petosa, Frequency-Agile Antennas for Wireless Communications, Nor-
[26] Y. Sun and J.K. Fidler, “Design of Π impedance matching networks,” wood, MA, USA: Artech House, 2014.
Proc. 1994 IEEE Int. Symp. on Circuits and Systems, ISCAS ’94, vol.5, [53] F. Broydé and E. Clavelier, “Method for automatic adjustment of a
pp. 5-8, Jun. 1994. tunable passive antenna and a tuning unit, and apparatus for radio
[27] F. Broydé and E. Clavelier, “A tuning computation technique for a communication using this method,” U.S. Patent 9 960 491, filed Oct. 4,
multiple-antenna-port and multiple-user-port antenna tuner,” Interna- 2017.
tional Journal of Antennas and Propagation, vol. 2016, Article ID [54] F. Broydé and E. Clavelier, “Method for automatically adjusting a tunable
4758486, Nov. 2016. passive antenna and a tuning unit, and apparatus for radio communication
[28] K.B. Ariyur and M. Krstic, Real-Time Optimization by Extremum- using this method,” U.S. Patent 10 008 777, filed Oct. 12, 2017.
Seeking Control, Hoboken, NJ, USA: Wiley-Interscience, 2003. [55] F. Broydé and E. Clavelier, “Method for automatic adjustment of a
[29] C. Olalla, M.I. Arteaga, R. Leyva, and A.E. Aroudi, “Analysis and tunable passive antenna and a tuning unit, and apparatus for radio
comparison of extremum seeking control techniques,” Proc. 2007 IEEE communication using this method,” U.S. Patent 10 044 380, filed Nov.
Int. Symp. on Industrial Electronics, pp. 72-76, 4-7 Jun. 2007. 2, 2017.
[30] B. Calli, W. Caarls, P. Jonker, and M. Wisse, “Comparison of ex- [56] F. Broydé and E. Clavelier, “Method for automatically adjusting a tunable
tremum seeking control algorithms for robotic applications,” Proc. 2012 passive antenna and a tuning unit, and apparatus for radio communication
IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, pp. 3195-3202, using this method,” U.S. Patent 9 991 911, filed Nov. 16, 2017.
Oct. 2012. [57] F. Broydé and E. Clavelier, “Method of automatic adjustment of a
[31] C. Zhang and R. Ordóñez, Extremum-Seeking Control and Application, tunable passive antenna and a tuning unit, and apparatus for radio com-
London, U.K.: Springer-Verlag, 2012. munication using this method,” Patent Cooperation Treaty application
[32] D. Armitage, “Automatic impedance matching apparatus,” U.S. Patent PCT/IB2020/055006, filed 27 May 2020.
4 356 458, filed Aug. 31, 1981. [58] H.A. Wheeler, “Small Antennas,” in Antenna Engineering Handbook,
[33] D.A. Roberts and B.T. DeWitt, “Automatic antenna tuning system,” U.S. R.C. Johnson, Ed., 3rd ed, New York, NY, USA: McGraw-Hill, 1993.
Patent 5 225 847, filed Feb. 7, 1991. [59] C.A. Desoer, “The maximum power transfer theorem for n-ports,” IEEE
Trans. Circuit Theory, vol. 20, no. 3, pp. 328-330, May 1973.
[34] Y. Sun and W.K. Lau, “Evolutionary tuning method for automatic
[60] S.A. Bassam, M. Helaoui, and F.M. Ghannouchi, “Crossover Digital Pre-
impedance matching in communication systems,” Proc. 1998 IEEE Int.
distorter for the Compensation of Crosstalk and Nonlinearity in MIMO
Conf. on Electronics, Circuits and Systems (ICECS 1998), pp. 73-77, Sep.
Transmitters”, IEEE Trans. Microw. Theory Tech., vol. 57, no. 5, pp.
1998.
1119-1128, May 2009.
[35] J. de Mingo, A. Valdovinos, A. Crespo, D. Navarro, and P. Garcia,
[61] K. Hausmair, P.N. Landin, U. Gustavsson, C. Fager, and T. Eriksson,
“An RF electronically controlled impedance tuning network design and
“Digital Predistortion for Multi-Antenna Transmitters Affected by An-
its application to an antenna input impedance matching system,” IEEE
tenna Crosstalk”, IEEE Trans. Microw. Theory Tech., vol. 66, no. 3, pp.
Trans. Microw. Theory Techn., vol. 52, no. 2, pp. 489-497, Feb. 2004.
1524-1535, Mar. 2018.
[36] P. Sjöblom and H. Sjöland, “An adaptive impedance tuning CMOS circuit
[62] S.K. Dhar, A. Abdelhafiz, M. Aziz, M. Helaoui, and F.M. Ghannouchi, “A
for ISM 2.4 GHz band,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol.
reflection-aware unified modeling and linearization approach for power
52, no. 6, pp. 1115-1124, Jun. 2005.
amplifier under mismatch and mutual coupling,” IEEE Trans. Microw.
[37] H. Song, S.-H. Oh, J.T. Aberle, B. Bakkaloglu, and C. Chakrabarti,
Theory Tech., vol. 66, no. 9, pp. 4147-4157, Sep. 2018.
“Automatic antenna tuning unit for software-defined and cognitive radio,”
[63] S. Stein, “On cross coupling in multiple-beam antennas,” IRE Trans.
Proc. 2007 IEEE Antennas and Propagation Society Int. Symp., pp. 85-
Antennas Propagat., vol. 10, no. 5, pp. 548-557, Sep. 1962.
88, Jun. 2007.
[64] J.W. Wallace and M.A. Jensen, “Termination-dependent diversity per-
[38] H. Song, B. Bakkaloglu, and J.T. Aberle, “A CMOS adaptative antenna- formance of coupled antennas: network theory analysis,” IEEE Trans.
impedance-tuning IC operating in the 850 MHz-to-2 GHz band,” Proc. Antennas Propagat., vol. 52, no. 1, pp. 98-105, Jan. 2004.
2009 IEEE Int. Solid-State Circuits Conf., pp. 384-386, Feb. 2009. [65] M.L. Morris and M.A. Jensen, “Network model for MIMO systems with
[39] L.F. Cygan, P.H. Gailus, W.J. Turney, and F.R. Yester, Jr., “Method and coupled antennas and noisy amplifiers,” IEEE Trans. Antennas Propagat.,
apparatus for enhancing an operating characteristic of a radio transmit- vol. 53, no. 1, pp. 545-552, Jan. 2005.
ter,” U.S. Patent 5 564 086, filed Nov. 29, 1993. [66] F. Broydé and E. Clavelier, “About the beam cosines and the radiation
[40] G.J.L. Bouisse and J.E. Morgan, “Circuit and method for impedance efficiency of a multiport antenna array,” Proc. 12th European Conference
matching,” U.S. Patent 6 414 562, filed May 27, 1997. on Antenna and Propagation, EuCAP 2018, pp. 1-5, Apr. 2018.
[41] F. Broydé and E. Clavelier, “Method for automatically adjusting a tunable [67] R. Mohammadkhani and J.S. Thompson, “Adaptive uncoupled termi-
matching circuit, and automatic tuning system using this method,” U.S. nation for coupled arrays in MIMO systems,” IEEE Trans. Antennas
Patent 9 628 135, filed Nov. 23, 2016. Propagat., vol. 61, no. 8, pp. 4284-4295, Aug. 2013.
[42] Q. Gu, J.R. De Luis, A.S. Morris III, and J. Hilbert, “An analytical [68] F. Broydé and E. Clavelier, “Two multiple-antenna-port and multiple-
algorithm for pi-network impedance tuners,” IEEE Trans. Circuits Syst. user-port antenna tuners,” Proc. 9th European Conference on Antenna
I, Reg. Papers, vol. 58, no. 12, pp. 2894-2905, Dec. 2011. and Propagation, EuCAP 2015, pp. 1-5, Apr. 2015.
[43] F. Broydé and E. Clavelier, “Method of automatic adjustment of a tunable [69] M.A. Jensen and B. Booth, “Optimal uncoupled impedance matching for
matching circuit, and automatic tuning system using this method,” U.S. coupled MIMO arrays,” Proc. First European Conf. on Antennas and
Patent 9 935 607, filed Oct. 20, 2017. Propagation, EuCAP 2006, pp. 1-4, Nov. 2006.
[44] F. Broydé and E. Clavelier, “Method of automatic adjustment of a tunable [70] Y. Fei, Y. Fan, B.K. Lau, and J.S. Thompson, “Optimal single-port match-
impedance matching circuit, and automatic tuning system using this ing impedance for capacity maximization in compact MIMO arrays,”
method,” U.S. Patent 9 966 924, filed Oct. 27, 2017. IEEE Trans. Antennas Propagat., vol. 56, no. 11, pp. 3566-3575, Nov.
[45] W.E. McKinzie III, “Adaptive impedance matching apparatus, system 2008.
and method,” U.S. Patent 7 714 676, filed Nov. 8, 2006. [71] B.K. Lau and J.B. Andersen, “On closely coupled dipoles with load
[46] W.E. McKinzie III, “Adaptive impedance matching apparatus, system matching in a random field,” Proc. 2006 IEEE 17th Int. Symp. on
and method with improved dynamic range,” U.S. Patent 7 535 312, filed Personal, Indoor and Mobile Radio Communications (PIMRC’06), pp.
Nov. 8, 2006. 1-5, Sep. 2006.
[47] W.E. McKinkie, III, “Adaptive impedance matching module,” U.S. Patent [72] B.K. Lau, J.B. Andersen, A.F. Molisch, and G. Kritensson, “Antenna
8 299 867, filed: Nov. 8, 2006. matching for capacity maximization in compact MIMO systems,” Proc.
[48] K. Kurokawa, “Power waves and the scattering matrix”, IEEE Trans. 3rd Int. Symp. on Wireless Communication Systems (ISWCS’06), pp. 253-
Microwaves Theory Techn., vol. 13, no. 2, pp. 194-202, Mar. 1965. 257, Sep. 2006.

Copyright © 2020 by Excem 23

 F. Broyde and E. Clavelier: A Typology of Antenna Tuner Control Schemes, for One or More Antennas

EXCEM

[73] J.B. Andersen and B.K. Lau, “On closely coupled dipoles in a random [94] F. Broydé and E. Clavelier, “Method for automatic adjustment of a tuning
field,” IEEE Antennas and Wireless Propagat. Letters, vol. 5, pp. 73-75, unit, and radio transceiver using this method,” Patent Cooperation Treaty
2006. application application PCT/IB2019/052360 (WO 2019/197923) , filed
[74] M.A. Jensen and B.K. Lau, “Uncoupled matching for active and passive 22 Mar. 2019.
impedances of coupled arrays in MIMO systems,” IEEE Trans. Antennas [95] F. Broydé and E. Clavelier, “Method for automatically adjusting a tuning
Propagat., vol. 58, no. 10, pp. 3336-3343, Oct. 2010. unit, and radio transceiver using this method,” Patent Cooperation Treaty
[75] N. Murtaza, M. Hein, and E. Zameshaeva, “Reconfigurable decoupling application PCT/IB2019/052374 (WO 2019/197925), filed 23 Mar. 2019.
and matching network for a cognitive Antenna,” Proc. 41st European [96] F. Broydé and E. Clavelier, “Method for automatic adjustment of tunable
Microwave Conference (EuMC), pp. 874-877, Oct. 2011. passive antennas, and radio transceiver using this method,” Patent Coop-
[76] X. Tang, K. Mouthaan, and J.C. Coetzee, “Tunable decoupling and eration Treaty application PCT/IB2019/052454 (WO 2019/197930), filed
matching network for diversity enhancement of closely spaced antennas,” 26 Mar. 2019.
IEEE Antennas and Wireless Propagat. Letters, vol. 11, pp. 268-271, [97] R.A. Horn and C.R. Johnson, Matrix analysis, Second Edition, New
2012. York, NY, USA: Cambridge University Press, 2013.
[77] A. Krewski and W.L. Schroeder, “Electrically tunable mode decomposi- [98] D. Tse and P. Viswanath, Fundamentals of Wireless Communication,
tion network for 2-port MIMO antennas,” Proc. of the 2013 Loughbor- New York, NY, USA: Cambridge University Press, 2005.
ough Antennas and Propagation Conference (LAPC), pp. 553-558, Nov. [99] J.G. Proakis and M. Salehi, Digital Communications, Fifth Edition, New
2013. York, NY, USA: McGraw-Hill, 2008.
[78] F. Broydé and E. Clavelier, “Antenna tuning apparatus for a multiport [100] J.M. Ortega, Matrix Theory, New York, NY, USA: Plenum Press, 1987.
antenna array,” U.S. Patent 9 621 132, filed Jun. 25, 2014. [101] L. Hogben, Handbook of Linear Algebra, Boca Raton, FL, USA: Chap-
[79] F. Broydé and E. Clavelier, “A new multiple-antenna-port and multiple- man & Hall/CRC, 2007.
user-port antenna tuner,” Proc. 2015 IEEE Radio & Wireless Week, RWW
2015, pp. 41-43, Jan. 2015.
[80] F. Broydé and E. Clavelier, “Antenna tuning apparatus for a multiport FRÉDÉRIC BROYDÉ was born in France in
antenna array,” U.S. Patent 10 187 033, filed Apr. 10, 2017. 1960. He received the M.S. degree in physics
[81] F. Broydé and E. Clavelier, “Two reciprocal power theorems for passive engineering from the Ecole Nationale Supérieure
linear time-invariant multiports,” IEEE Trans. Circuits Syst. I, Reg. Pa-
d’Ingénieurs Electriciens de Grenoble (ENSIEG)
pers, vol. 67, no. 1, pp. 86-97, Jan. 2020.
and the Ph.D. in microwaves and microtechnolo-
[82] F. Broydé and E. Clavelier, “Corrections to ‘Two reciprocal power
theorems for passive linear time-invariant multiports,” ’ in IEEE Trans. gies from the Université des Sciences et Technolo-
Circuits Syst. I, Reg. Papers, DOI: 10.1109/TCSI.2020.2975533. gies de Lille (USTL).
[83] F. Broydé and E. Clavelier, “Radio communication using multiple an- He co-founded the Excem corporation in May
tennas and localization variables,” U.S. Patent 9 654 162, filed Oct. 15, 1988, a company providing engineering and re-
2015. search and development services. He is president
[84] F. Broydé and E. Clavelier, “Method and apparatus for automatically and CTO of Excem. Most of his activity is allocated to engineering and re-
tuning an impedance matrix, and radio transmitter using this apparatus,” search in electronics, radio, antennas, electromagnetic compatibility (EMC)
U.S. Patent 9 077 317, filed Nov. 24, 2014. and signal integrity. Currently, his most active research areas are automatic
[85] F. Broydé and E. Clavelier, “Method and apparatus for automatic tuning antenna tuners, measuring antennas and active antennas.
of an impedance matrix, and radio transmitter using this apparatus”, U.S. Dr. Broydé is author or co-author of about 100 technical papers, and
Patent 10 116 057, filed Oct. 18, 2016. inventor or co-inventor of about 80 patent families, for which 48 US patents
[86] F. Broydé and E. Clavelier, “Method for automatically adjusting a tun- have been granted. He is a Senior Member of the IEEE since 2001. He is a
ing unit, and automatic tuning system using this method”, U.S. Patent licensed radio amateur (F5OYE).
9 966 930, filed Oct. 21, 2016.
[87] F. Broydé and E. Clavelier, “Method of automatic adjustment of
a tuning unit, and apparatus for radio communication using this EVELYNE CLAVELIER was born in France in
method”, Patent Cooperation Treaty application PCT/IB2019/051501 1961. She received the M.S. degree in physics
(WO 2019/180520), filed Feb. 25, 2019. engineering from the Ecole Nationale Supérieure
[88] F. Broydé and E. Clavelier, “Radio communication using tunable anten- d’Ingénieurs Electriciens de Grenoble (ENSIEG).
nas and an antenna tuning apparatus,” U.S. Patent 9 680 510, filed Oct. She is co-founder of the Excem corporation,
20, 2015. based in Maule, France. She is CEO of Excem.
[89] F. Broydé and E. Clavelier, “Method for automatically adjusting tunable She is also President of Eurexcem (a subsidiary
passive antennas, and automatically tunable antenna array using this of Excem) and President of Tekcem, a company
method”, U.S. Patent 9 698 484, filed Nov. 3, 2016. selling or licensing intellectual property rights to
[90] F. Broydé and E. Clavelier, “Method for automatic adjustment of tunable foster research. She is also an active engineer and
passive antennas and a tuning unit, and apparatus for radio communica-
researcher. Her current research area is radio communications. She has also
tion using this method”, U.S. Patent 9 929 460, filed Sep. 1, 2017.
done research work in the areas of electromagnetic compatibility (EMC) and
[91] F. Broydé and E. Clavelier, “Method for automatically adjusting tunable
passive antennas and a tuning unit, and apparatus for radio communica-
signal integrity. She has taken part in many electronic design and software
tion using this method”, U.S. Patent 9 912 075, filed Sep. 6, 2017. design projects of Excem.
[92] F. Broydé and E. Clavelier, “Method of automatic adjustment of tun- Prior to starting Excem in 1988, she worked for Schneider Electrics (in
able passive antennas and a tuning unit, and apparatus for radio com- Grenoble, France), STMicroelectronics (in Grenoble, France), and Signetics
munication using this method,” Patent Cooperation Treaty application (in Mountain View, USA).
PCT/IB2020/054953, filed 26 May 2020. Ms. Clavelier is the author or a co-author of about 80 technical papers.
[93] F. Broydé and E. Clavelier, “Radio communication using a plurality of She is co-inventor of about 80 patent families. She is a Senior Member of
selected antennas,” U.S. Patent 10 224 901, filed Oct. 28, 2015. the IEEE since 2002. She is a licensed radio amateur (F1PHQ).

Open access license: you may reproduce, store and distribute this document, provided the document remains complete and unmodified.
All other rights reserved. Published by Excem. ISBN 978-2-909056-21-0
Excem — 12, chemin des Hauts de Clairefontaine — 78580 Maule — France

24 Copyright © 2020 by Excem