TELKOM
NIKA
, Vol.11, No
.11, Novemb
er 201
3, pp. 6600
~6
610
e-ISSN: 2087
-278X
6600
Re
cei
v
ed Ma
rch 1
9
, 2013;
Re
vised June
21, 2013; Accepte
d
Jul
y
2
2
, 2013
Wideband Tuning of Impedance Matching for actual RF
Networks using AQPSO
Yinxin
1
, Tany
anghong*
1
, Liaoji
w
a
ng
2
1
Colle
ge of Ele
c
trical & Information En
gi
neer
i
ng, Hun
an U
n
i
v
ersit
y
, 4
100
82
, China
2
Huna
n Col
l
e
g
e
of
Information
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: tan
y
ho@
12
6.com
A
b
st
r
a
ct
An ad
aptiv
e w
i
deb
an
d i
m
p
e
d
ance
matchi
ng
techni
qu
e us
in
g a
passiv
e
PI-
netw
o
rk is pr
es
ented
in
the p
aper
o
n
the b
a
sis
of a
daptiv
e q
u
a
n
tu
m
particl
e sw
arm opti
m
i
z
a
t
i
on
alg
o
rith
ms
(AQPSO), w
h
i
c
h
avoi
din
g
th
e d
r
aw
backs of t
he sta
ndar
d p
a
rticle sw
ar
m
opti
m
i
z
at
ion
(
PSO) alg
o
rith
m. The
Wide
b
a
n
d
T
unin
g
tech
ni
que
is for act
ual RF
ch
ips.
So on-
c
h
ip
mo
de
ls for R
F
capacitors
and
ind
u
ctors
are
consi
dere
d
. A
nd th
e
effects of th
e p
a
ras
i
tic co
mpon
ent
l
o
ss i
n
th
e PI-
netw
o
rk are
a
naly
z
e
d
. T
h
en
th
e
circuit is si
mpl
i
f
ied accor
d
in
g to the sensitivit
ies of
the ele
m
ents. T
herefor
e, t
he computa
t
iona
l co
mpl
e
xi
ty
is dra
m
atic
ally
reduce
d
. F
i
na
lly,
the AQPSO algorith
m
is
adopte
d
to
maxi
mi
z
e
th
e po
w
e
r transmissi
o
n
efficiency. S
i
mulati
on r
e
sults
show
that t
h
e pr
opos
ed
tu
nin
g
tech
ni
que
can
ach
i
ev
e
goo
d acc
u
racy
of
impe
danc
e
ma
tching
an
d lo
a
d
pow
er. T
he r
e
flectio
n
co
efficient a
nd VSW
R obta
i
ne
d ar
e
also s
a
tisfacto
ry.
Moreov
er, the prop
osed
meth
od can b
e
usef
ul for
softw
are defin
ed ra
dio s
ystems usi
ng a
singl
e ante
n
n
a
for multi
p
l
e
mo
bile
and w
i
rel
e
ss bands.
Ke
y
w
ords
: i
m
ped
anc
e
matc
hin
g
, pass
i
ve
PI-netw
o
rk, parti
cle sw
arm o
p
timi
z
a
ti
on, p
a
r
asitic co
mpon
ent,
AQPSO
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Today’s m
obi
le and wi
rel
e
ss
comm
uni
cation devices are u
s
ed in
almost all im
agina
ble
environ
ment
s, su
ch
as in
cell ph
one, i
n
cars, i
n
tal
k
in
g po
sition
ne
ar the
he
ad.
The e
n
viro
nm
ent
of the antenn
a and the resulting field di
stributio
n
aro
und it has
un
fortunately an
eminent imp
a
ct
on its impe
d
ance [1]. And the mismat
ch bet
wee
n
antenn
a and
sou
r
ce/tran
smitter red
u
ce
s its
power efficie
n
cy, linearity and lowers th
e powe
r
of the input/output
signal. More
over, maximu
m
power i
s
expected
to tra
n
s
mit to the
a
n
tenna
to a
c
hieve maxim
u
m tra
n
smi
ssion efficie
n
cy
. So
the goal of obtaining fast
antenn
a tuning system
s, whi
c
h are
ca
pable of offering impe
dan
ce
matchin
g
(I
M) with
ch
a
nging l
oad
and e
n
viron
m
ental a
s
pe
cts, ha
s b
e
c
ome i
n
cre
a
singly
signifi
cant. T
he lowpa
ss
Pi and lowpa
ss T
circuit
s
, as the mo
st
popula
r
imp
edan
ce m
a
tching
config
uratio
n
s
a
r
e
cha
r
a
c
terize
d by th
eir
si
mple
st
ructu
r
e
s
, wi
d
e
ra
nge
of l
oad im
ped
an
ce
accomm
odati
ons a
nd hig
h
harm
oni
c reje
ction capa
bilities [2].
Variou
s co
m
m
unication st
anda
rd
s hav
e also be
en develop
ed to contain a variety of
appli
c
ation
s
at different frequ
en
cy ba
nds, such a
s
cell
ular
co
mmuni
cation
s at 900 a
nd
1800M
Hz, global po
sitioni
ng syste
m
(G
PS) at 1.
2 and 1.5GHz, an
d Bluetooth a
nd WiFi at 2.
4
and
5.2G
Hz.
Du
e to
high
ope
ration
freque
ncy,
a
s
well as
lo
w voltage
a
nd small
si
ze
trend,
impeda
nce
matchin
g
net
work are ve
ry difficult to
desi
gn. T
o
p
u
t them
on
to sili
co
n
chip
ha
s
proved
even
more difficult in view of no
n-ide
a
litie
s in
fabrication te
chn
o
logy an
d
para
s
itic effe
cts
at high frequency. Thus the imped
ance
matching net
works will
suf
f
er from the power loss. T
h
us
it is cruci
a
l to be a
b
le to
evaluate
su
ch p
o
wer lo
ss in the
de
si
gn an
d analy
s
is
of matchi
ng
netwo
rks for
budg
eting sy
stem po
we
r.
Mismat
ch
of the a
n
tenn
a impe
dan
ce
suffers
sig
n
i
fi
cant d
egen
eration
of th
e po
we
r
ef
fi
ci
en
cy of the radi
o link.
Automatic m
a
tchin
g
networks a
r
e therefore
devel
op
ed to match
any
cha
nge in
an
tenna imp
e
d
ance in man
y
RF appli
c
a
t
ions [3-18]. Single freq
ue
ncy impe
dan
ce
matchin
g
m
e
thod
s a
r
e
wi
d
e
ly appli
c
a
b
l
e
in
many
areas such a
s
power amplifi
e
rs a
nd
ante
nna
tuning sy
ste
m
s. Automati
c tuning
meth
ods h
a
ve be
e
n
investigate
d
usin
g ge
net
ic metho
d
[3].
A
particl
e swa
r
m optimizatio
n (PSO) ba
se
d algorith
m
s
have bee
n used in [4], [5-7]. [8] propo
se
d a
hiera
r
chi
c
al
geneti
c
algo
rithm. However, its
ability is limited to
improve a
u
tomatic mat
c
hing
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Wide
ban
d Tu
ning of Im
pedance Matchi
n
g
for ac
tu
al RF Networks u
s
ing AQPSO
(Yinxi
n)
6601
netwo
rk
nod
a
l
quality facto
r
s a
nd in
crea
se the ove
r
al
l matchin
g
ef
fi
ci
en
cy. It is
found that th
e
prop
osed tu
n
i
ng meth
od
s so
far are
all si
ngle
fre
quen
cy b
a
se
d mainly
at
HF
ran
ge. T
h
e
cha
nnel
s whi
c
h have ve
ry narro
w sig
nal
band
width,
thus ve
ry high
Q at high fre
quen
cie
s
can
be
tuned to, by swee
ping the f
r
equ
en
cy ra
n
ge of inte
rest.
They involve
swit
chin
g bet
wee
n
differe
nt
freque
nci
e
s,
cha
nnel
s a
n
d
band
s/sta
n
d
a
rd
s, a
s
may
be n
eed
ed.
More
over, th
ese
app
ro
aches
are all
ba
se
d on the i
d
e
a
l matchi
ng
netwo
rks. Th
at is, the pa
rasiti
c-aware
effects of t
h
e
cap
a
cito
rs a
n
d
indu
ctors in
the matching
netwo
rks are
not con
s
ide
r
ed. And the para
s
itic-a
ware
para
m
eters st
rongly affect t
he po
wer
effi
cien
cy of the radio [17].
Here we
con
s
ider a wi
deba
nd tuning issue and p
r
opo
se a metho
d
based on an
adaptive
quantum p
a
rticle swarm optimizatio
n algorith
m
(A
QPSO), whi
c
h is a
co
mbination of
an
improved Adaptive particl
e
swarm
opt
imizatio
n
algorithm (APSO) with a
Quantum
Parti
c
le
Swarm
Optim
i
zation
(QPS
O). By sha
r
in
g the two re
turne
d
extrem
e values
of the pa
rticle
s from
APSO and QPSO, the proposed method
enables to
adaptively search
thei
r opti
m
um solutions in
parall
e
l. Usin
g this metho
d
,
the impe
da
nce
matchi
ng
networks ca
n be
tuned
to
cove
r a
ba
n
d
of
freque
nci
e
s,
for all chan
n
e
ls of both
u
p
link
a
nd d
o
w
nlin
k, and i
ndee
d multi-band
s of several
stand
ard
s
.
In this pape
r,
we de
scrib
e
the impeda
n
c
e mat
c
hin
g
netwo
rk, a
nal
yze on
-chip i
ndu
ctor
model’
s
sen
s
itivity and si
mplify imped
ance matchi
ng structu
r
e
in Section 2.
The wid
eba
nd
impeda
nce tuning metho
d
usin
g AQPSO is pro
p
o
s
e
d
in Section 3. Simulations and results are
pre
s
ente
d
for GSM, UMTS and both in
Section
4. Co
nclu
sio
n
s a
r
e
drawn in Section 5.
2. Impedance Matc
hing Net
w
o
r
k
Ana
l
y
s
is
The maj
o
r
pu
rpo
s
e
of imp
edan
ce
matching n
e
two
r
k
is to maximi
ze po
we
r tran
smissio
n
efficien
cy if load im
ped
an
ce
s a
nd
cha
nnel
s
chan
g
e
. Figu
re 1
sho
w
s the tu
nable
matchi
ng
netwo
rk
stru
cture. The ant
enna lin
ks to
gether the si
g
nal sou
r
ce wit
h
powe
r
ampl
ifier throug
h the
tunable
matching n
e
two
r
k. Whe
n
a
n
ten
na imp
edan
ce or f
r
eq
uen
cy chan
nel
ch
ange, the
ce
ntral
pro
c
e
ssi
ng u
n
it gets information ab
out
antenna im
p
edan
ce a
nd
band of fre
q
u
ency throug
h
the
sen
s
o
r
. And
then p
r
o
c
e
s
sing u
n
it u
s
e
i
t
s inte
rnal
AQPSO alg
o
rit
h
m to
co
mpu
t
e the
eleme
n
t
para
m
eters i
n
imped
an
ce
matchin
g
ne
twork an
d th
rough the
exe
c
ution u
n
it ad
just the value
of
L, C1 and
C2
.
Figure 1. Tun
able Matchi
n
g
Net
w
or
k S
t
r
u
ct
ur
e
Let us co
nsi
d
er the parasiti
c
-a
wa
re effect of
the induct
o
r at first. (Kiyong Choi, David J.,
Allstot 20
06;
Gupta,
R., Ballwe
ber B.M. 2001
)
pre
s
e
n
ts a
n
in
du
ctor m
odel
for
use
in
pa
ra
sitic-
awa
r
e
synthe
sis. Pa
ra
sitic-awa
r
e m
odeli
ng be
gi
n
s
with the de
sig
n
and fa
bri
c
ati
on of
several
indu
ctors tha
t
span a ran
ge of indu
cta
n
ce
s wi
th a
d
equate q
ualit
y factor and
self-re
s
on
an
ce
freque
ncy val
ues for th
e a
n
ticipate
d
ap
plicatio
ns. Fi
gure
2
sh
ows a
n
a
c
curate pa
ram
e
tric
on-
chip i
ndu
ctor
model. Ea
ch
segm
ent i
s
m
odele
d
u
s
ing
a lump
ed e
q
u
ivalent ci
rcui
t comp
risi
ng
a
self-in
d
u
c
tan
c
e, a seri
es resi
stan
ce eq
ual to
the dc resi
stan
ce o
f
the metal segment, a sh
unt
cap
a
cita
nce repre
s
e
n
ting t
he capa
citive
cou
p
li
ng
bet
wee
n
the me
tal forming th
e se
gment a
nd
the sub
s
trate
,
an effective sub
s
trate lo
ss resi
st
an
ce
compute
d
si
mply as a lateral sp
rea
d
i
ng
resi
st
an
ce.
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 11, Novemb
er 201
3: 660
0 – 6610
6602
L
1
met
a
l
R
2
met
a
l
R
1
ox
C
2
ox
C
1
s
ub
R
2
s
ub
R
(a) Pa
ramet
r
i
c
indu
ctor m
o
del
L
L
r
'
1
L
C
'
2
L
C
1
C
g
2
C
g
(b) Equival
e
n
t
circuit of the inducto
r mod
e
l
Figure 2. Parametri
c
Mode
l and its Equi
valent Circuit of a Planar In
ducto
r
Figure 2
(
a) shows a
pa
ra
metr
ic plan
ar indu
ctor
mo
del, wh
ere Rmetal, Cox1,
Cox2,
Rsub a
r
e fun
c
tion
s of L ov
er a d
e
si
re
d range
of
indu
ctance
s
. Th
e floating in
du
ctor of the
ran
g
e
is
form 1nH to 18nH:
2
0.0278
1.741
2.3402
RL
L
me
t
a
l
(1)
2
0.0005
0.030
7
0
.0468
12
CC
L
L
ox
ox
(2)
3
.
5064
0.0894
32.151
RL
L
sub
(3)
Whe
r
e L is in
nH, Cox1 a
n
d
Cox2 are in
pF, and Rm
etal and Rsu
b
are in oh
ms.
The circuit m
a
y
be co
nverted
into Figu
re 2(b
)
th
rou
gh
Eq
uation
(4), (5
)
by
a
pplying
seri
es to pa
ra
llel conve
r
si
o
n
.
'
1
1
22
2
1
1
C
ox
C
L
CR
ox
su
b
(4)
22
'
1
1
22
2
1
1
CR
ox
s
u
b
g
C
CR
ox
s
u
b
(5)
So the Impedance matchi
n
g
netwo
rk is
sho
w
n in
Fig
u
re 3, if the para
s
itic pa
ra
meters of
the indu
ctor a
r
e con
s
ide
r
ed
. Let C1’=C1
+CL’, C2’
=
C2
+CL’, the circuit is tran
sformed into Figu
re
4. In ord
e
r to
con
s
id
er th
e
parasiti
c
-a
ware
effect
of the ind
u
cto
r
,
by cal
c
ulatin
g
the se
nsitivities
of Zin and PL
to s, where s={ rL , gc1, g
c
2}, we get:
Z
in
r
L
>>
2
Z
in
g
C
,
Z
in
r
L
>>
2
Z
in
g
C
P
L
r
L
>>
1
P
L
g
C
,
P
L
r
L
>>
2
P
L
g
C
(6)
Therefore, th
e effect of g
c
1 and
gc2 to
input
impe
da
nce
and
po
wer is far le
ss than that
of rL. It is approp
riate to ign
o
re g
c
1 an
d g
c
2. So the circuit ca
n be si
mplified as Fi
gure 5.
L
1
C
g
2
C
g
1
L
C
2
L
C
1
C
2
C
L
Z
s
R
L
r
Figure 3. Impedan
ce Mat
c
hing Ci
rcuit Consi
deri
ng Pa
rasiti
c Para
m
e
ters of the In
ducto
r
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Wide
ban
d Tu
ning of Im
pedance Matchi
n
g
for ac
tu
al RF Networks u
s
ing AQPSO
(Yinxi
n)
6603
L
L
r
1
C
2
C
1
C
g
2
C
g
s
R
L
Z
S
V
in
Z
L
L
r
1
C
2
C
s
R
L
Z
S
V
in
Z
Figure 4. Equivalent Circuit
of Figure 3
Fi
gure 5. The
Simplified Ci
rcuit of Figu
re
3
No
w co
nsid
er the para
s
itic-aware effect
of
the capa
citors. In Figure 3, the paral
lel arm
loss cond
uct
ance of the i
ndu
ctor i
s
tre
a
ted a
s
the p
a
ra
sitic
lo
ss condu
ct
an
ce
o
f
the ca
pacito
r
s,
say g
c
1 a
nd
gc2. An
d the
para
s
itic ca
p
a
citan
c
e,
say
CL1
and
CL
2 ca
n be
ab
sorbe
d
by
C1
and
C2 by letting
C1’
=
C1+CL’,
C2’
=
C2+CL’,
thus the e
ffe
cts of CL can
be eliminate
d
.
Loss
gCL i
s
in
parall
e
l with
C1 a
nd
C2
and the
r
efo
r
e ca
n be
tre
a
ted in the
same
way a
s
g
c
1 a
nd
g
c
2.
Therefore the
formula
s
pro
posed are no
w appli
c
a
b
le to both indu
ct
or and
cap
a
ci
tor model
s.
3. AQPSO M
e
thod
for Wi
deband Imp
e
danc
e Tuni
ng
PSO is a
pop
ulation-ba
sed
sto
c
ha
stic
o
p
ti
mization
te
chni
que
deve
l
oped i
n
(Ke
n
nedy J,
Eberh
a
rt R
1
995). It is
a
stocha
stic o
p
t
imiz
ation m
e
thod ba
se
d
on swa
r
m int
e
lligen
ce. The
fundame
n
tal idea is that th
e optimal can
be f
ound through
coo
peration and info
rmation
sha
r
i
n
g
among
indivi
dual
s in
the
swarm.
Thro
ugh
co
ope
ra
t
i
on a
nd
com
petition am
on
g the
pop
ulat
ion,
popul
ation-ba
sed
optimi
z
a
t
ion ap
proaches often
can find
go
o
d
solution
s
efficiently an
d
effec
t
ively. How
e
ver
,
it may eas
ily t
r
ap
i
n
to lo
cal
opti
m
al poi
nts
an
d may
difficul
t
ly obtain exa
c
t
solutio
n
s at the late of the iteration.
To overcome
the weakne
ss, some rese
ar
che
r
s have
employed m
e
thod
s with a
daptive
para
m
eters a
nd co
mbin
ed
quantum m
o
d
e
l. An adapt
ive mode a
d
ju
sts the pa
rame
ters a
c
cordin
g
to the feedba
ck inf
o
rmatio
n, su
ch
a
s
fu
zzy a
daptive i
nertia
weig
ht
(Y. Shi, R. Eberh
a
rt 20
01
).
A
combi
ned
qu
antum mod
e
l
is presented
in (Sun
J, Feng B, Xu WB 2004),
whe
r
e pa
rticle
s a
r
e
cha
nge
d accordin
g to qua
ntum movem
ent rule
s,
su
ch as pa
rticl
e
s having qua
ntum behavio
r.
In
this pap
er, a new p
a
rall
el adaptive qua
ntum parti
cle swarm o
p
timization alg
o
rit
h
m is pro
p
o
s
ed.
By sha
r
ing
t
he two extre
m
e of th
e p
a
r
ticle
s
,
the
p
r
opo
sed
meth
od a
daptively
se
arch
es th
eir
optimum
sol
u
tions in
pa
rallel. It co
mb
ines
the
opti
m
ization
s
of
an imp
r
ove
d
ada
ptive P
S
O
(APSO)
with
a quantum P
a
rticle Swarm
Optimi
zation
(QPSO). T
he
APSO thread and the QPS
O
thread o
p
e
r
at
e in parall
e
l.
3.1. APSO Thread.
Standard particle swarm o
p
timization m
i
ght unde
rgo
an und
esi
r
ed
process of d
i
versity
loss. Some p
a
rticle
s be
co
me ina
c
tively while lo
st
bo
th of the glob
al and lo
cal search
capa
bil
i
ty
in the next generation
s
. The lost of glo
bal and l
o
cal sea
r
ch ca
pa
bility means t
hat parti
cle
s
wil
l
be only movi
ng withi
n
a q
u
ite small
sp
ace,
whi
c
h
wi
ll be o
c
curs
whe
n
its lo
ca
tion and
pbe
st is
clo
s
e to
g
best (pb
e
st i
s
th
e optim
al p
o
s
ition
of
the
particl
e until now, gbe
st repre
s
e
n
t
the
past
optimal p
o
siti
on of the
swarm) an
d its
velocity is
clo
s
e to
ze
ro. T
o
overcom
e
t
h
is p
r
o
b
lem,
the
adaptive pa
rticle swa
r
m u
s
e
s
newly no
nlinea
r iner
ti
a weight
s an
d accele
ratio
n
coefficie
n
ts to
control the v
e
locity of pa
rticles
and
avoids
clu
s
teri
n
g
of parti
cle
s
and mai
n
tai
n
s dive
rsity
o
f
popul
ation in
the sea
r
ch sp
ace.
The ad
aptive
particl
e swa
r
m optimizatio
n co
ns
i
s
ts of,
at each tim
e
step, changi
ng the
velocity an
d
locatio
n
of e
a
ch
pa
rticle
towa
rd it
s p
b
e
st a
nd
gbe
st locatio
n
s a
c
cordi
ng to
the
Equation (7)
and (8
), re
sp
ectively:
*
*
(
)*
(
)
*
(
)*
(
)
11
2
V
w
V
c
rand
pbe
st
x
c
ra
nd
gbe
st
x
ii
i
i
(7)
1
x
xV
ii
i
(8)
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 11, Novemb
er 201
3: 660
0 – 6610
6604
Whe
r
e th
e lo
cation
of the i
t
h parti
cle i
s
repre
s
e
n
ted a
s
xi. The b
e
st
previo
us
po
sition of
the ith pa
rticl
e
is
re
co
rde
d
and
re
pre
s
e
n
ted a
s
p
b
e
s
t. The in
dex o
f
the be
st pb
est am
ong
all
the
particl
es i
s
re
pre
s
ente
d
by
the symb
ol
gbe
st. The
v
e
locity for th
e
ith parti
cle i
s
rep
r
e
s
ente
d
a
s
Vi. The ra
nd
() is
a rand
om
at interval [0
,l] with
uniformly distrib
u
tion. The fi
rst
part of Eq
uati
on
(7) represent
s the
p
r
eviou
s
velo
city, wh
ich
provi
des t
he n
e
cessa
r
y
mome
ntum f
o
r
parti
cle
s
to
fly
across the
search
spa
c
e.
The se
co
nd
part is kno
w
n a
s
the “cog
nitive” co
mpone
nt, wh
ich
rep
r
e
s
ent
s the person
a
l thinkin
g
of each
particle.
Thi
s
compo
nent e
n
co
ura
g
e
s
the particl
es to
fly
toward their
own b
e
st po
sition found so
far. The third part is kno
w
n a
s
the “social
”
co
mpo
nent,
whi
c
h
rep
r
e
s
ents th
e
coll
aborative effect of th
e
p
a
r
ticle
s
in
find
ing the
glob
a
l
optimum. T
h
is
comp
one
nt al
ways
pull
s
th
e parti
cle
s
to
ward the
glob
al be
st po
sition the
whol
e
swarm fo
und
so
far.
The ine
r
tia weights a
r
e up
dated a
s
:
20
/
1
0
max
/1
max
m
in
m
i
n
GG
ww
w
e
w
(9)
Whe
r
e wm
a
x
and wmin
are the initial and fi
nal values of the inertia weig
hts,
respe
c
tively, G is the
cu
rrent iteratio
n
numbe
r
and
Gmax i
s
the
maximum n
u
m
ber of all
o
wable
iteration
s
.
The accel
e
ration co
efficient
s are a
d
ju
ste
d
nonlin
early
as follo
ws:
3
*/
11
1
m
a
x
1
cc
c
G
G
c
f
ss
,
3
*/
22
2
m
a
x
2
cc
c
G
G
c
f
ss
(10)
Whe
r
e c1s a
nd c1f are th
e initial and final values of the accele
rati
on coeffici
ent
c1, c2s
and c2f are th
e initial and final value
s
of the accele
rati
on co
efficient
c2.
This
strategy
implies that
at the beginn
ing
of the se
arch the cog
n
itive compo
nent ha
s
more weight
than the so
ci
al comp
one
nt, so the particle
s
can
sea
r
ch ra
pidly an
d widely in the
whol
e sp
ace, while at the
latter
of se
arch, the soci
al
comp
one
nt plays mo
re i
m
porta
nt role
to
make p
a
rti
c
le
s co
nverg
e
to the global op
tima.
3.2. QPSO Thread
Quantum Pa
rticle Swarm
Optimizatio
n
algo
rithm (QPSO) i
s
p
r
opo
se
d in
2004. In
cla
ssi
cal P
S
O,
a part
i
cl
e
is st
at
ed b
y
it
s posit
ion
vector Xi a
nd velocity vector Vi, whi
c
h
determi
ne th
e traje
c
tory
of the parti
cl
e. The
move
ment of the particl
e alo
n
g
the dete
r
m
i
ned
trajecto
ry follows Ne
wtoni
an mechani
cs. Ho
weve
r,
in ca
se of q
uantum me
chani
cs the te
rm
trajecto
ry is
meanin
g
le
ss,
beca
u
se Xi and Vi of
a p
a
rticle
ca
nnot
be determin
ed sim
u
ltane
ously
according
to
un
certai
nty prin
ciple.
Th
erefo
r
e,
if in
dividual p
a
rti
c
le
s in
a PS
O sy
stem
h
a
ve
quantum
beh
avior, the pe
rforma
nce of
PSO will b
e
far fro
m
that of cla
s
si
cal PSO. In
the
quantum
mo
del of a PSO,
the state of
a par
ti
cle i
s
d
epicte
d
by wave functio
n
,
x
t
, instea
d
of
positio
n an
d
velocity. The
dynamic be
h
a
vior of th
e
p
a
rticle
is wid
e
ly diverg
ent
from that of t
he
particl
e in tra
d
itional PSO system
s in that the
exact values of Xi and Vi can
n
o
t
be determin
ed
simultan
eou
sl
y. In this
co
ntext, the probability of
t
he p
a
rticl
e
’s
appe
arin
g in
po
sition Xi f
r
om
probability density function
2
,
x
t
, the form of
whi
c
h dep
en
ds on the po
tential field in which
the parti
cle l
i
es. Employi
ng the M
ont
e Ca
rlo m
e
thod, the p
a
rticles
move
according to
the
followin
g
itera
t
ive Equation (11
)
:
1/
*
/
2
Xt
P
I
n
u
L
(11)
Whe
r
e
u i
s
ran
dom
nu
mber with
uniformly
distribution, L
is d
e
termi
n
e
d
from
12
Lt
m
b
e
s
t
X
t
, the param
eter
is calle
d Contractio
n-Expan
sio
n
(CE)
Coeffi
cient,
whi
c
h ca
n
be tuned to control the c
onverg
e
n
c
e
spee
d of the particle.
1
0
.
5
/
0.5
ma
x
m
ax
GG
G
. mbest calle
d Mean Best
Position, is defined a
s
the mean of
the pbe
st positions of all pa
rticle
s. That is :
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Wide
ban
d Tu
ning of Im
pedance Matchi
n
g
for ac
tu
al RF Networks u
s
ing AQPSO
(Yinxi
n)
6605
1/
*
(
,
,
,
)
12
11
1
MM
M
mbe
s
t
M
pbe
s
t
pbe
s
t
pb
e
s
t
i
i
iD
im
ii
i
(12)
Whe
r
e M i
s
the numb
e
r
of particl
es
o
f
t
he populat
ion, Dim i
s
the dime
nsi
o
n of the
particl
es, p
b
e
s
tiDim i
s
be
st position of
particl
es.
Co
nverge
nce of
the PSO algorithm m
a
y be
achi
eved if each p
a
rticl
e
converg
e
s to its lo
cal attract
o
r, P defined
at the coo
r
din
a
tes:
1
P
pbes
t
gbe
s
t
(13)
1l
n
1
/
X
tP
m
b
e
s
t
X
t
u
id
d
i
d
(14)
is ra
ndom
n
u
mbe
r
unifo
rmly distribute
d
on (0,l), pb
est an
d gbe
st present the
best
particl
e po
sition and the b
e
s
t positio
n of particl
es in th
e popul
ation.
Gene
rally, th
e value
of e
a
ch
co
mpo
n
ent in Xid
can be
cl
amp
ed to the
ra
nge that
distrib
u
te in feasi
b
le zone
aroun
d bo
rd
er to
co
ntrol
exce
ssive
ro
aming of pa
rticle out
side t
he
sea
r
ch sp
ace
.
The method
is given as fo
llow:
If Xid>
Xmax
Xid=
Xmax-c
*(Xid-Xmax)*rand( )
or
If Xid<
-Xmax
Xid=
-Xmax-c
*
(Xid+
X
max)
*rand( )
(15)
The valu
e of
c i
s
ap
plied
h
e
re to
adj
ust
the ra
nge
of t
he p
a
rticl
e
s,
and th
e Xma
x
is the
maximum moving distance.
As sh
own in
Figure 6, the
flow of the AQPSO
algo
rithm is
sho
w
n
in Figure 6. Explained
as
follows
:
id
X
Figure 6. AQPSO Flow Ch
art
Step 1: The initial para
m
et
ers a
r
e g
ene
rated at the be
ginnin
g
of pro
g
ram.
Step 2: APSO and QPS
O
threa
d
s
work
coll
abo
ra
tively. The position a
nd velocity of
partic
l
es
are randomly dis
t
ributed in APSO thread
. QPSO evaluat
e the fit
ness
f
o
r eac
h
partic
l
e
to get values
of pbest an
d gbe
st.
Step 3: In APSO thread,
the fitness
f
unc
tion
value of each partic
le is
c
a
lc
ulated to
determi
ne the gbest po
sit
i
on. The cu
rrent values
of
pbest and g
best are com
pare
d
with th
e
corre
s
p
ondin
g
returned va
lues fro
m
the
main thr
ead.
If the curre
n
t values are
better than the
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 11, Novemb
er 201
3: 660
0 – 6610
6606
store
d
p
b
e
s
t
and
gbe
st, th
en the
value
s
of pb
est
and
gbe
st in
mai
n
thread
are
repla
c
e
d
by t
he
c
u
rrent values
. In QPSO
threa
d
, the values of
and mbest i
s
eval
uate, and P i
s
found
usi
ng
Equation (13).
Step 4: APSO call
s pbest
and gbest in main
thread
to change the po
sition and velocity
of pa
rticle
s
a
c
cordi
n
g
to E
quation
(7
)
a
nd
(8).
Acco
rding to
Equ
a
t
ion (9) and
(10), th
e in
ertia
weig
hts a
nd
accele
ration
coeffici
ents a
r
e u
pdat
e
d
;
QPSO thre
a
d
chang
e the
po
sitions of
the
particl
es a
c
co
rding to Equ
a
t
ion (14
)
, whe
r
e Xid is limited by Equatio
n (15
)
.
Step 5: The end of each g
e
neratio
n, the term
in
al con
d
i
t
ion is examin
ed and the proce
s
s
is termi
nated
whe
n
the condition i
s
satisfied in A
PSO. When
the termin
al
con
d
ition is
not
satisfie
d the
pro
c
e
ss
prog
resse
s
into t
he next
step
. The fitness function of
each pa
rticle
is
determi
ned to obtain the gbe
st positio
n in QPSO.
Comp
ari
ng the current val
ues of pbe
st and
gbe
st with the return
ed val
ues from the
main
threa
d
, and the bette
r pbe
st and g
best are kept
fo
r
next iteration. Otherwise, t
he value
s
of pbe
st and gb
est from mai
n
thread a
r
e u
pdated.
Step 6: All thread
s a
r
e
exa
m
i
ned at
the end of
ea
ch
g
enerati
on. If the evol
utiona
ry cy
cle
or target valu
e is
compl
e
te
d, the AQPSO is te
rminat
ed an
d outp
u
t
s the pb
est
and g
b
e
s
t. If
not,
turn to Step 3.
4. Experiment
4.1. Fitness
Function
The
simplifie
d PI-network
is sho
w
ed
in
Figure 5.
Here, Vs i
s
the
si
gnal
sou
r
ce,
Rs i
s
th
e
transmissio
n impeda
nce (t
ypically 50 o
h
ms), rL is
Rmetal in Figu
re 2, and ZL
=R+jX re
pre
s
ents
the load imp
edan
ce
(e.g.
antenn
a inpu
t impedan
ce
), Zin rep
r
e
s
e
n
ts the inp
u
t impeda
nce a
nd
Zeq is the out
put impeda
nce.
We may ea
sil
y
obtain that the input po
wer wo
uld be:
2
2
4|
|
V
RZ
in
si
n
PP
in
a
v
a
RR
Z
R
in
in
in
s
(16)
Whe
r
e
2
/(
4
)
P
VR
av
a
s
s
is the maximum po
wers, and Vs is the amplitude of the so
urce
voltage. Let
s
be the reflectio
n
coeffici
ent, we have:
()
/
(
)
RZ
R
Z
s
si
n
s
i
n
(17)
Voltage stan
d
i
ng wave ratio (VSWR) ca
n be written a
s
:
1/
1
VS
W
R
s
s
(18)
The co
st fun
c
tion can b
e
chosen a
s
:
=
i0
i=
1
M
f
fV
S
V
R
V
S
W
R
i
(19)
Whe
r
e
0
f
VSVR
VSWR
ii
, M is the total n
u
m
ber of the
sam
p
le
poi
nts ove
r
the
con
s
id
ere
d
freque
ncy ran
ge. VSWR0 i
s
the ta
r
get
value of the
voltage sta
n
d
i
ng wave rati
o,
whi
c
h is i
deal
ly 1. Our aim
is to minimize the co
st fu
nction fitne
s
s. As
f(w)
i
s
a
highly nonli
n
ear
function
of the co
mpon
ent
s C1, C2
an
d L, a di
re
ct
minimization
of the co
st fu
nction i
s
a v
e
ry
diffic
u
lt tas
k
.
4.2. Simulati
on Results
Several exa
m
ples h
a
ve
been calcul
ated to dem
onstrate the
performan
ce of th
e
prop
osed
a
ppro
a
ch. Th
e exam
ple
s
sh
owed
wi
deba
nd im
p
edan
ce
mat
c
hin
g
te
chni
que
comp
ared to never u
s
ing a
n
y techniq
ue.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Wide
ban
d Tu
ning of Im
pedance Matchi
n
g
for ac
tu
al RF Networks u
s
ing AQPSO
(Yinxi
n)
6607
Example 1:
GSM-1800
uses 17
10
–178
5MHz f
o
r u
p
lin
k a
n
d
18
05–
188
0MHz fo
r
downlin
k. He
re we
con
s
id
er the freq
ue
ncy ran
ge/ba
nd from 1.71
GHz to 1.88
GHz. The so
urce
resi
st
an
ce
is
50
Ω
. The
loa
d
impe
dan
ce
is
cap
a
citive
, Zload
=50
-
j7
5
Ω
. 100
poi
nts a
r
e
sam
p
led
in the freque
n
c
y rang
e.
The o
b
taine
d
matchi
ng
co
mpone
nt val
ues
are C1=0.0201
51pF,
C2
=0.01
809
7
p
F, and
L=5.0
835
1n
H. The curve
s
of load power and
re
fle
c
tion coeffici
e
n
t with IM, load po
we
r a
n
d
reflectio
n
coe
fficient witho
u
t IM are sh
own in Fi
gu
re 7.The ave
r
age value
s
o
f
load po
wer
and
reflectio
n
coe
fficient with i
m
peda
nce m
a
tchin
g
(IM)
are 3.9
797m
w and
0.178
95. While
wit
hout
IM, their values a
r
e 3.2m
w and 0.59
999
, resp
ective
ly. Thus
without
the prop
ose
d
techni
que, t
he
power tra
n
sm
issi
on efficie
n
c
y has b
een i
n
crea
sed by
19.59%.
Figure 7. Loa
d power, Refl
ecti
on
Coefficient for GSM-1800M
Hz
Example
2: UMTS spe
c
if
ies
the ban
d
s
19
00-202
5
M
Hz and
21
10-2
200
MHz for 3G
transmissio
n. Here we
co
nsid
er the b
a
nd from
1.9 t
o
2.025G
Hz
(we
will con
s
ider even
wi
der
band to in
clu
de 2.11
-2.2G
H
z i
n
Exampl
e 3). The
so
u
r
ce
re
si
stan
ce is 50
Ω
. T
h
e
load imp
eda
nce
is re
si
stive, Zload
=20
0
Ω
. A set of 100
sampl
ed poi
n
t
s are
con
s
id
ered in th
e test. The obtai
ned
matchin
g
co
mpone
nt
valu
es
a
r
e C1
=0.
3356
0pF, C2
=0.62
876
pF, and
L
=
6.6
5
0
31n
H.
The cu
rves
of load po
wer and refle
c
tio
n
coeffici
ent with IM,
load power an
d re
flection coefficient with
out IM
are sho
w
n in
Figure 8.
Figure 8. Loa
d power, Refl
ection
Coefficient for UMT
S
1.
7
1.
7
5
1.
8
1.
85
1.
9
3.
9
3.
9
5
4
4.
0
5
x 1
0
-3
f
r
equ
en
c
y
(
G
H
z
)
lo
a
d
p
o
w
e
r w
it
h
IM
1.
7
1.
7
5
1.
8
1.
8
5
1.
9
0.
1
0.
1
5
0.
2
0.
2
5
f
r
eq
ue
nc
y
(
G
H
z
)
re
fl
e
ct
i
o
n
co
e
f
f
ic
ie
n
t w
it
h
IM
1.
7
5
1.
8
1.
8
5
3
3.
2
3.
4
3.
6
3.
8
x 1
0
-3
f
r
eq
ue
nc
y
(
G
H
z
)
lo
a
d
p
o
w
e
r w
it
h
o
u
t IM
1.
7
5
1.
8
1.
8
5
0.
5
5
0.
6
0.
6
5
0.
7
f
r
eq
ue
nc
y
(
G
H
z
)
re
fl
e
ct
i
o
n
co
e
f
f
ic
ie
n
t w
it
h
o
u
t I
M
(W
)
(W
)
1.
9
1.
95
2
3.
8
4
3.
8
6
3.
8
8
3.
9
x 1
0
-3
f
r
eq
ue
nc
y
(
G
H
z
)
lo
a
d
p
o
w
er
w
it
h
I
M
1.
9
1.
9
5
2
0.
2
2
0.
2
4
0.
2
6
0.
2
8
f
r
eq
ue
nc
y
(
G
H
z
)
r
e
fl
e
c
ti
o
n
c
o
e
ffi
c
i
e
n
t
w
it
h
I
M
1.
9
1.
95
2
3
3.
2
3.
4
3.
6
3.
8
x 1
0
-3
f
r
eq
ue
nc
y
(
G
H
z
)
l
o
ad
pow
er
w
i
t
ho
ut
I
M
1.
9
1.
9
5
2
0.
5
5
0.
6
0.
6
5
0.
7
f
r
eq
ue
nc
y
(
G
H
z
)
r
e
f
l
e
c
ti
o
n
c
o
e
ffi
c
i
e
n
t w
i
t
ho
ut
I
M
(W
)
(W
)
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 11, Novemb
er 201
3: 660
0 – 6610
6608
With the pro
p
o
se
d IM unit, the average
values of loa
d
powe
r
and
reflectio
n
co
e
fficien
t
are in
crea
sed
from 3.200
0
m
w an
d 0.60
01 to 3.
853
8
7
mw a
nd 0.2
4985. Th
e propo
sed te
chn
i
que
has e
nha
nce
d
power tra
n
smissi
on effici
ency by 16.9
6
%.
Example 3: We no
w con
s
ider du
al mod
e
s GSM
-
180
0 and UM
TS. The frequ
en
cy rang
es
from 1.7 to 2.
2GHz fo
r the
dual
stand
ard
s
. The
so
urce re
sist
ance i
s
50
Ω
. T
he l
oad imp
eda
n
c
e
is indu
ctive, Zload
=60
+
j1
2
0
Ω
. A set of 500 sample
d
points a
r
e consi
dered in
this expe
rime
nt.
The o
b
taine
d
matching
com
pon
ent
values are
C1
=0.38
0
4
67pF,
C2
=0.
9563
31pF,
and
L=3.1
874
0n
H. The curve
s
of load power and
re
fle
c
tion coeffici
e
n
t with IM, load po
we
r a
n
d
reflectio
n
co
e
fficient withou
t IM are sho
w
n in Figure 9.
The
p
e
rce
n
ta
ge
of po
wer transmissio
n efficien
cy
ha
s
e
nha
nced 38.02% with our
IM
techni
que form 2.2641m
w
to 3.6535m
w.
Figure 9. Loa
d power, Refl
ection
Coefficient for UMT
S
Figure 7 th
ro
ugh Fig
u
re 9
indicate that the value
of lo
ad po
we
r is close to th
e m
a
ximum
power
with
our IM u
n
it. Thro
ugh th
e pro
p
o
s
ed
techni
que,
we ca
n a
c
hie
v
e satisfa
c
to
ry
transmissio
n
efficien
cy and
reflectio
n
co
efficient. So the propo
se
d i
m
peda
nce m
a
tchin
g
net
work
is ca
pabl
e of offering imp
e
dan
ce matchi
ng with chan
ging loa
d
and
chan
nel
s.
In additio
n
, i
n
orde
r to
asse
ssm
ent the
pro
p
o
s
ed
IM
unit for a
c
tu
al RF
Netwo
r
ks,
we
comp
are the
returned
resu
lts by
applyin
g
the
sim
p
lified m
odel
an
d the
ide
a
l m
odel. T
he i
d
e
a
l
model prese
n
ts the mode
l without con
s
ide
r
ing a
n
y para
s
itic-a
wa
re effects of
the indu
ctor
and
the ca
pa
citors in the IM
un
it, while the
si
mplifi
ed mo
d
e
l is
sho
w
n i
n
Figure 5. Th
e re
sult
s can
be
see
n
in Table
1.
Table 1. The
Powe
r Lo
ss
Comp
ari
s
o
n
of Simplified Model an
d Ideal Model
Power
loss(w
)
(GHz
)
(
Ω
)
1.71-1.88
ZL=50-10j
1.9-2.025
ZL=20
1.7-2.2
ZL=60
Simplified model
Best
5.3953E-04
1.1261E-03
4.3527E-04
Worst
5.4973E-04
1.1543E-03
5.0555E-04
Mean
5.4463E-04
1.1401E-03
4.6930E-04
Ideal model
Best
1.6957E-03
1.1639E-03
5.2793E-04
Worst
1.7149E-03
1.2002E-03
5.4916E-04
Mean
1.7051E-03
1.1802E-03
5.3672E-04
Table
1 sho
w
s that the
po
wer lo
ss
of the sim
p
lified
model i
s
le
ss than that of t
he ide
a
l
model fo
r G
S
M, UMTS
and d
ual
sta
ndards from
1.7 to 2.2
G
Hz. T
he
po
wer lo
sse
s
of
the
obtaine
d sim
p
lified model
over the freq
uen
cy
rang
e are 5.44
63E-04, 1.1401E
-03 and 4.69
3
0
E-
1.
8
2
2.
2
3.
4
3.
6
3.
8
x 1
0
-3
f
r
equ
enc
y
(
G
H
z
)
l
oad p
o
w
er
w
it
h
I
M
1.
8
2
2.
2
0.
1
0.
2
0.
3
0.
4
0.
5
f
r
equenc
y
(
G
H
z
)
r
e
f
l
ec
t
i
on
c
o
ef
f
i
c
i
e
n
t
w
ith
I
M
1.
8
2
2.
2
2.
2
2.
3
x 1
0
-3
f
r
equ
enc
y
(
G
H
z
)
l
o
ad p
o
w
er
w
i
t
hou
t
I
M
1.
8
2
2.
2
0.
7
0.
75
0.
8
f
r
equenc
y
(
G
H
z
)
r
e
fl
e
c
ti
o
n
c
o
e
f
fi
c
i
e
n
t w
i
t
h
out
I
M
(W
)
(W
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Wide
ban
d Tu
ning of Im
pedance Matchi
n
g
for ac
tu
al RF Networks u
s
ing AQPSO
(Yinxi
n)
6609
04, re
spe
c
tively. Howeve
r,
the co
rre
sp
o
nding p
o
we
r l
o
sse
s
of the i
deal mo
del a
r
e 1.70
51E-0
3
,
1.1802E
-03 a
nd 5.367
2E-0
4. The power
transfe
rred to
load gro
w
s b
y
an average
of 28%.
Simulation re
sults
sho
w
that the prop
ose
d
wide
ba
nd imped
an
ce matchin
g
tech
niqu
e
can im
prove the efficie
n
cy
of load
p
o
we
r, control
refle
c
tion
coeffici
e
n
t in a satisfa
c
tory
sco
pe.
So
the pro
p
o
s
ed
adaptive tuni
ng techniq
ue
can
ac
hieve
good
accu
ra
cy of impedan
ce mat
c
hin
g
and
load
po
wer fo
r different a
n
tenna
impe
da
nce
s
. T
he
re
f
l
ection
coefficient al
so
sati
sfies the
nee
ds
of the project.
More
over, th
e pro
p
o
s
ed
method
c
an b
e
useful
for software defin
ed
ra
dio syst
ems
usin
g a sin
g
le
antenna for
multiple mobil
e
and wi
rele
ss ban
ds.
4. Conclusio
n
This pa
pe
r pre
s
ent
s an
adaptive wid
eban
d imped
ance tuning
techni
que for multi-
stand
ard
mo
bile commu
ni
cation
s. Th
e
on-chi
p
mo
dels fo
r
RF
cap
a
cito
rs a
nd ind
u
cto
r
s are
con
s
id
ere
d
, a
nd the
ci
rcuit
is
simplified
a
c
cordi
ng to
th
e sen
s
itivities of the
elem
e
n
ts. Th
erefo
r
e
,
the com
putational complex
i
ty is dramati
c
ally
red
u
ced
.
AQPSO is chosen to improve impeda
n
c
e
matchin
g
. Example
s
are gi
ven for GSM, UMTS and
d
ual stand
ards with three different anten
n
a
impeda
nces.
The frequ
en
cy rang
e fo
r t
he d
ual
stan
dard
s
i
s
f
r
o
m
1.7 to
2.2
G
Hz. Simulat
i
on
results show goo
d a
c
cu
ra
cy of in
put i
m
peda
nce
m
a
tchin
g
a
nd
power t
r
an
sf
er. Th
e
refle
c
tion
coeffici
ent a
n
d
VSWR a
r
e
also
satisfa
c
tory. The
prop
ose
d
tunin
g
method may also
be suita
b
le
for even wid
e
r
ran
ge of fre
quen
cie
s
to cover othe
r wireless sta
nda
rds.
Ackn
o
w
l
e
dg
ement
This
work
wa
s suppo
rted b
y
National
Na
tural Scie
nce
Found
ation o
f
China u
nde
r Grant
No.61
102
039
and the Fun
damental
Re
sea
r
ch Fun
d
s for the Central Universitie
s
.
Referen
ces
[1]
KR Bo
yl
e, Y
Yuan,
LP Li
gthart. Anal
ys
is
of m
obi
le pho
ne
a
n
ten
na im
ped
anc
e
vari
ations
w
i
th use
r
pro
x
imit
y, Ante
nnas a
nd Pro
p
agati
on.
IEEE Transactions on.
200; 55: 36
4
-
372.
[2]
Matthias Schmidt, Errikos Lourandakis.
A Co
mp
ariso
n
of T
unabl
e F
e
rroel
ectric
Π
- and T
-Matchin
g
Networks.
Proceed
ings
of the 37th Euro
pe
an
Micro
w
ave C
o
nferenc
e. 200
7
:
98-101.
[3]
M T
hompson, JK F
i
dler.
Ap
p
licatio
n of th
e
gen
etic a
l
gor
ithm
an
d si
mu
l
a
ted a
n
n
eal
in
g
to LC filt
er
tuning, Circuits
,
Devices and System
s.
IEE Procee
din
g
s. 2
001; 14
8(4): 17
7-18
2.
[4]
Yang
ya
n
g
Z
H
ANG, W
a
sim Q MALIK. Analo
gue
F
ilt
er
T
uning for A
n
tenn
a Match
i
ng
w
i
th M
u
ltip
l
e
Objective P
a
r
t
icle S
w
a
rm
Opti
mizatio
n
, Advanc
es i
n
W
i
red a
n
d
W
i
reless
C
o
mmunic
a
tio
n
,
IEEE/Sarnoff Sym
p
osium
.
20
05: 169-
17
4.
[5]
EW
C Neo, Y
L
i
n,
X L
i
u, L
C
N
de Vr
eed
e, et
c. Adaptiv
e mu
lti-ba
nd m
u
lti-
mode
po
w
e
r
a
m
pli
fi
er
usi
n
g
integr
ated var
a
ctor-base
d
tun
abl
e matchi
ng
net
w
o
rks.
IEEE J. Solid-Stat
e
Circuits
. 200
6; 41(9): 21
66-
217
6.
[6]
H Ferdin
and
o
a
, F Pasila, H Ku
s
w
a
n
to. Enha
nce
d
Ne
uro-F
u
zz
y
Architecture F
o
r
Electrical
Loa
d
Forecasting.
T
E
LKOMNIKA Indo
nesi
an Jo
u
r
nal of Electric
al
. 201
0; 8(2): 87-9
6
.
[7]
Sun J, Feng
B, X
u
WB.
Particle sw
ar
m
opti
m
i
z
at
ion
w
i
th partic
l
es
h
a
vin
g
q
uant
u
m
beh
avi
our.
Procee
din
g
s of
Congr
ess on
Evolut
i
on Com
putatio
n. 200
4: 325-3
31.
[8]
Yang
Hon
g
T
a
n
,
RuF
a
n
g
Yi,
a
nd Y
iCh
ua
ng
Sun. W
i
d
e
b
a
n
d
T
uning
of Im
ped
anc
e Matc
hin
g
N
e
t
w
ork
s
usin
g Hi
erarc
h
ical
Gen
e
tic
Algor
ithms f
o
r Multista
nd
ard Mo
bil
e
C
o
mmunic
a
tio
n
s
.
Journ
a
l of
Co
mp
uters.
20
12; 7(2): 35
6–3
61.
[9]
EL F
i
rrao, AJ
Ennem
a, B N
a
uta.
Anten
n
a
b
ehav
iour
in
the
prese
n
ce
of h
u
man
bo
dy.
Pr
oc. 15th
Ann
u
.
W
o
rkshop C
i
rcuits, S
y
st. Sign
al Process, (Pr
o
RISC). 200
4: 487
–4
90.
[10]
H Song, B Bakkalo
glu, JT
Ab
erle.
A CMOS
ada
ptative a
n
-tenn
a-i
m
p
e
d
a
n
c
e-tuni
ng IC o
perati
ng in t
h
e
850 MH
z
-
to-
2
GH
z
b
and.
Pro
c
. IEEE
Int. Solid-State Circ
u
it
s Conf. 200
9: 384–
38
6.
[11]
P Sjöbl
om, H Sjöl
and. An a
daptiv
e imp
e
d
anc
e tu
nin
g
C
M
OS cir-cuit for ISM 2.4 GHz ban
d.
IE
EE
T
r
ans. Circuits Syst. I, Reg. Papers
. 20
05; 5
2
(6): 111
5–
112
4.
[12]
A van
Bez
ooi
j
en, MA
de J
o
ngh,
C C
h
a
n
l
o
,
LC
H R
u
ijs,
F
van
Strate
n, R M
ahmo
u
di, AHM v
a
n
Roermu
nd. A
GSM/EDGE/WCDMA a
d
a
p
tiv
e
seri
es
LC m
a
tchin
g
n
e
t
w
or
k usi
ng
RF
-MEMS s
w
itch
es.
IEEE J. Solid-State Circuits
. 200
8; 43(1
0
): 2259
–2
268.
[13]
EL F
i
rra
o, AJ
Ann
e
ma, B
Nauta. A
n
au
tomatic a
n
ten
na tu
ni
ng s
y
s
t
em usi
n
g
on
l
y
RF
-Sig
na
l
amplit
udes.
IEEE Trans. Circuits Syst.II
, Ex
p. Briefs
. 2008;
55(9): 833
–8
3
7
.
[14] H Song, SH Oh, JT
Aberle
, B Bakkalo
glu, C
Chakrab
a
rti.
Automatic ante
n
na tuni
ng un
it for softw
are-
de
fi
n
ed a
nd co
gnitiv
e
radi
o
. in
Proc. IEEE Int. Sy
m
p
.Anten
n
Propag. 20
07
: 85–88.
Evaluation Warning : The document was created with Spire.PDF for Python.