Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
12
,
No.
3
,
Decem
ber
201
8
, p
p.
1
0
71
~
1
0
80
IS
S
N: 25
02
-
4752, DO
I:
10
.11
591/ijeecs
.v1
2
.i
3
.pp
1
0
71
-
1
0
80
1071
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
PAPR R
eduction
at Lar
ge Multi
-
User
-
M
I
MO
-
OF
DM usin
g
Adaptiv
e Data D
etection
Algorith
m
N.
Pra
ba
1
,
K.
M.
R
av
ik
um
ar
2
1
Depa
rtment of
ECE
,
GCE,
Ramanaga
r
am
,
Indi
a
2
SJ
CIT,
chi
kk
ab
al
apur
,
Ind
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r 8
, 2
018
Re
v
ise
d A
pr 27
, 2
018
Accepte
d
Ma
y
11
, 201
8
W
ire
le
ss
comm
unic
a
ti
on
in
pre
sent
era
contain
s
la
rge
-
sc
al
e
MIM
O
net
work
arc
hi
te
c
ture
that
nee
d
to
deliv
er
an
opt
imize
-
QoS
to
m
ult
i
-
user
(MU
).
The
opti
m
ize
da
ta
rate
tra
nsm
ission
in
m
assive
MU
-
MIMO
wir
el
ess
s
y
stems
is
one
of
the
m
ost
diffi
cult
ta
sk
due
to
the
ex
tr
emel
y
h
igh
implementation
complexi
t
y
.
The
pra
ctical
wire
l
ess
sy
st
em
cha
n
nel
s
gene
r
al
l
y
e
xhibi
ts
th
e
PA
PR
and
fre
quency
se
le
c
ti
ve
fa
ding,
i
t
is
al
so n
e
ce
ss
ar
y
to
have a pre
cod
ing
soluti
on
in
PA
PR
for
the
sel
e
cted
desira
b
le
ch
anne
ls.
A
soluti
on
for
the
designe
d
probl
e
m
of
a
noble
err
or
-
cor
recti
ng
co
de
for
OF
DM
proc
ess
with
a
low
PA
PR
,
in
the
ca
se
of
impuls
e
noise
should
b
e
conside
r
ed.
In
thi
s
pape
r
,
Adapti
ve
-
Da
ta
-
d
et
e
ct
ion
(AD
D)
al
gorit
hm
is
proposed
to
obta
in
lower
-
complexi
t
y
da
ta
-
det
e
ct
ion
th
at
cor
responds
to
high
throughput
design
and
impuls
e
noise
r
e
m
oval
for
l
arg
e
MU
I
-
MIMO
wire
le
ss
s
y
st
ems
b
y
the
OF
DM
m
odula
ti
on
tec
hnique
.
Th
at
c
onta
ins
som
e
s
te
ps
such
as;
ini
ti
a
li
z
ation,
pre
-
proc
essing
a
nd
equa
l
iz
a
ti
on
s
te
ps
in
orde
r
to
get
no
per
form
a
nce
loss
and
to
m
ini
m
al
i
ze
t
he
re
cur
ren
t
am
ount
a
t
e
ac
h
ite
rat
ions
dur
ing
o
per
ation.
In
orde
r
to
use
sim
pli
f
y
m
odel
,
he
re
we
assum
e
suit
ably
per
f
ec
t
s
y
nchr
oni
zation,
la
rg
e
c
y
c
lic
pre
fix
and
per
fect
-
CS
I
(c
hanne
l
-
st
at
e
-
informa
ti
on)
which
has
bee
n
de
vel
oped
through
the
pil
ot
depe
n
ded
tra
in
ing
.
Sim
ula
ti
on
res
ult
s
ana
l
y
sis
show
the
proposed
m
et
hod
subs
ta
nti
al
improvem
ent
over
the
exi
stin
g
al
gorit
hm
in
te
rm
s
of
both
‘Er
ror
-
rat
e
’
m
ini
m
iz
at
ion
an
d
PA
PR
red
uct
io
n.
Ke
yw
or
ds:
Ad
a
ptive
data
detect
ion (
ADD)
Mult
iple
-
in
pu
t
m
ul
ti
ple
-
ou
t
put
(MIMO
)
Or
t
hogonal
fr
e
qu
e
ncy
-
div
isi
on
m
ul
ti
plexing (OF
DM)
Peak
-
to
-
a
ve
rage p
ow
e
r rat
io
(P
A
PR)
Copyright
©
201
8
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
N.
Pr
a
ba,
Dep
a
rtm
ent o
f EC
E,
GCE,
Ra
m
anag
aram
, I
ndia
.
Em
a
il
:
np
ra
ba2010
@r
e
diff
m
ai
l.com
1.
INTROD
U
CTION
Pr
ese
nt
wirele
ss
com
m
un
ic
at
ion
syst
em
con
ta
ins
la
r
ge
-
sc
al
e
MIM
O
net
work
arc
hitec
ture
t
hat
nee
d
to
deli
ver
a
pr
om
isi
ng
res
ult
in
or
der
t
o
m
eet
gro
win
g
de
m
and
s
for
m
ulti
-
us
er
qual
it
y
-
of
-
ser
vice
(
)
a
nd
higher
thr
oughput
in
wirele
ss
syst
em
s
[1
]
.
Th
e
la
r
ge
-
sc
al
e
MIM
O
ne
twork
al
s
o
s
houl
d
hav
e
abi
li
ty
to
decr
ease
the
c
on
s
um
ption
of
operati
onal
powe
r
(
)
at
the
transm
itter
sid
e,
al
so
s
upport
the
us
a
ge
of
l
ow
com
plexity
structu
res
f
or
c
oncu
rr
i
ng
m
ulti
-
us
e
r
inter
face
(
).
T
hese
ki
nd
of
pro
per
ti
e
s
of
la
r
ge
-
scal
e
MIM
O
m
akes
a
pr
om
isi
ng
-
te
chnolo
gy
f
or
the
upcom
ing
ge
ner
at
io
n
of
wi
reless
syst
e
m
s.
OF
DM
te
ch
niq
ue
is
well
est
ablishe
d
an
d
the
at
tract
ive
process
t
o
deal
with
the
‘freque
ncy
sel
ect
ive’
cha
nnel
s.
Mor
eo
ver,
in
ord
e
r
to
sim
plify
th
e
receive
r
side
equ
al
iz
at
io
n,
OFDM
proce
s
s
al
lows
sc
he
duli
ng,
bit
al
lo
cat
ion
,
a
nd
pe
r
-
to
ne
powe
r
in
t
he
sp
ect
r
um
sh
ap
ing
a
nd
f
requ
ency
dom
ai
n.
The
the
oret
ic
al
view
of
la
rge
-
scal
e
MIM
O
ha
ve
acqu
i
red
lot
of
sign
ific
a
nt
at
te
ntion
f
ro
m
the
researc
hers,
bu
t
the
kn
ow
le
dge
of
pract
ic
al
transm
issi
on
pro
cess
is
ver
y
le
ss.
A
s
per
the
pa
per
[2
]
,
the
pr
act
ic
al
op
erati
on
s
of
la
r
ge
-
M
IM
O
net
work
will
need
the
le
ss
rad
i
o
fr
e
qu
e
ncy
(RF
)
powe
r
an
d
low
c
os
t
com
po
ne
nts.
I
n
ad
di
ti
on
in
[2
]
,
t
he
pro
posed
m
od
el
of
m
ult
i
us
er
pr
ec
odin
g
for
flat
-
fr
e
quen
cy
channels
de
pe
nd
on
pr
e
-
ante
nn
a
‘c
onsta
nt
-
env
el
op
e
’
tran
sm
issi
on
to
al
l
ow
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
0
71
–
1
0
80
1072
eff
ic
ie
nt
opera
ti
on
thr
ough
non
-
li
near
RF
-
c
om
po
ne
nts.
T
he
pract
ic
al
wireless
syst
e
m
channels
ge
ne
rall
y
exh
i
bits
the
PA
PR
an
d
fr
e
qu
ency
sel
ect
ive
fad
i
ng,
it
al
so
necessa
ry
to
ha
ve
a
pr
ec
odin
g
so
luti
on
of
P
A
PR
fo
r
sel
ect
ed
desira
ble
channels.
The
so
l
ution
f
or
s
uch
kind
of
com
plexity
r
equ
i
res
an
in
div
id
ual
m
ob
il
e
te
rm
inal
to
ha
ve
le
ss
P
AP
R
beca
us
e
of
power
c
on
st
r
ai
nts
an
d
strin
ge
nt
area;
al
so
,
i
t
can
a
ffo
rd
the
hea
vy
process
ing
at
the b
a
se stat
io
n (BS).
Howe
ver,
the
MIM
O
-
OFDM
is
popula
r
to
unde
rgo
with
a
le
sser
P
AP
R
that
dem
and
s
the
us
a
ge
of
‘linear
-
RF’
c
om
po
nen
ts
(e
xa
m
ple,
po
wer
a
m
pl
ifie
rs)
in
orde
r
to
overc
om
e
the
s
ign
al
distor
ti
on
an
d
ou
t
-
of
-
band
rad
ia
ti
on.
Unf
or
t
un
at
el
y,
the
li
nea
r
-
RF
Com
po
ne
nt
ge
ner
al
ly
com
es
with
le
ss
e
ff
ic
ie
nt
powe
r
a
nd
m
or
e
cost
c
om
par
ed
to
the
no
n
-
li
ne
ar
c
om
po
ne
nts,
w
hich
cau
ses
t
he
e
xcessi
ve
c
os
t
f
or
la
r
ge
-
sc
al
e
i
m
ple
m
ent
at
ions
of
BS
t
hat
m
a
y
con
ta
ins
hundre
ds
of
a
nten
nas.
In
t
his
pa
per
[3
]
,
T
hey
pro
po
se
d
a
res
idu
e
nu
m
ber
-
s
yst
e
m
(‘
RN
S’)
in
ord
er
to
re
duce
P
AP
R
in
MIM
O
-
OFDM
syst
e
m
s
and
the
pro
po
s
ed
m
od
el
a
ppr
oach
m
akes
us
e
of
RNS
pro
pe
rtie
s
to
gr
eat
ly
de
crease
the
‘c
om
pu
ta
ti
on
al
com
plexit
y’
as
well
as
the
PA
PR.
This
m
e
thodo
l
og
y
com
par
ed
with
the
PT
S
-
sc
he
m
e
(p
arti
al
tra
ns
m
it
sequ
enc
e)
al
so,
w
he
re
the
RN
S
-
base
d
sc
hem
e
of
P
AP
R
reducti
on
perform
ance can
be a
naly
ze f
or
co
m
pu
ta
ti
on
al
com
plexit
y wit
h m
ini
m
al
side i
nfor
m
at
ion
.
PA
PR
reducti
on
ap
proac
hes
su
c
h
as
cl
ip
-
sig
nal
has
pr
opos
e
d
in
[
4]
,
[
5],
sel
ect
ed
m
app
ing
(S
LM)
[6
]
,
[
7]
and
pa
rtia
l
transm
it
sequ
e
nc
e
(P
TS
)
[
8]
,
[9
]
ha
ve
bee
n
pro
po
se
d
in
order
to
ov
e
rc
om
e
the
eff
ect
s
of
P
APR
in
MIM
O
-
O
FD
M
sig
nals.
Ther
e
f
or
e,
us
i
ng
t
he
te
ch
niques
the
‘PAPR
’
can
be
re
duc
e,
but
these
te
chn
i
qu
es
al
so
com
e
up
with
thei
r
re
sp
ect
ive
dra
wbacks.
Cl
ip
ping
the
peak
am
plit
ud
es
un
der
a
tim
e
do
m
ai
n
reason
s
beh
i
nd
disto
rtion
of
si
gn
al
,
wh
ic
h
re
duc
e
the
perform
ance
of
BER
in
the
receive
r
side.
Partit
ion
s
of
PTS
data
i
nto
the
sub
-
b
loc
ks
,
wh
e
re
in
di
vidual
sub
-
bloc
ks
are
c
om
bin
ed
base
d
on
ph
as
e
seq
uen
ce
an
d
t
he
ph
a
se
se
quences
c
om
bin
at
ion
pro
vid
es
t
he
lo
we
r
P
APR
at
OFDM
si
gn
al
,
w
hich
w
il
l
be
us
e
d
at
data
tr
ansm
issi
on
.
T
he
SLM
c
om
bin
es
phase
se
quence
s
with
t
he
ir
duplica
te
s
cop
y
of
OFD
M
data,
and
the
n
perform
s
sever
al
it
erati
on
s
of
IF
F
T
(in
ve
rse
fast
-
Fou
rier
-
t
ran
s
f
or
m
)
to
c
hoose
the
blo
c
k
of
data
a
t
lowest
PAPR
.
Both
of
t
he
m
et
ho
d
SLM
and
PT
S
re
qu
ire
the
side
i
nfor
m
at
ion
tra
ns
m
issi
on
(in
ph
a
se
seq
uen
ce
)
to
r
ecov
e
r
the
ori
gin
al
transm
it
t
e
d
data
that
m
ay
req
uire
s
om
e
extra
ban
dwidth
.
The
refo
re,
it
is
necessa
ry
to
re
du
ce
the
am
ount
of
P
AP
R
of
OFDM
de
pe
ndent
m
ulti
-
us
er
MIM
O
netw
ork
syst
em
to
prov
i
de
corres
pondin
g
low
-
po
wer
a
nd
lo
w
-
c
os
t
B
S
operati
on
s
.
More
ov
e
r,
t
he
MUI
cancel
a
ti
on
ca
n
be
giv
en
as
unde
rd
et
erm
ined
li
nea
r
r
eg
ul
arizat
ion
pro
bl
e
m
that require
s opti
m
iz
e so
lu
ti
on
.
In
this
pap
e
r,
we
pro
po
se
d
a
lowe
r
-
c
om
plexity
data
-
detect
ion
al
go
rithm
that
co
rr
es
pondin
g
to
hi
gh
thr
oughput
de
sign
f
or
la
r
ge
MUI
-
MIM
O
wireless
syst
e
m
s
by
the
O
FD
M
m
od
ulat
io
n
te
c
hn
i
qu
e
.
DDA
(D
at
a
detect
io
n
al
gorithm
)
c
an
i
m
pr
ove
the
com
pu
ta
ti
on
al
eff
ic
ie
ncy
in
a
har
dwa
re
prot
otype
that
avo
i
ds
the
excess
com
pu
t
at
ion
.
A
DDA
is
a
well
-
know
n
it
erati
ve
fr
am
ewo
r
k
to
so
l
ve
a
huge
nu
m
ber
of
conve
x
diff
ic
ulty
(e
xa
ct
ly
or
a
ppr
ox
i
m
at
el
y)
thr
ough
c
oor
din
at
e
-
w
ise
updates
,
but
her
e
we
pro
posin
g
A
dap
ti
ve
-
Data
-
detect
ion
(
A
D
D)
al
gorithm
t
hat
con
ta
i
ns
th
e
init
ia
li
za
ti
on
,
pr
e
-
proces
sin
g
an
d
eq
ualiz
at
ion
ste
ps
in
orde
r
to
get
low
pe
rform
ance
loss,
m
ini
m
iz
at
ion
of
i
m
pu
lse
no
is
e
and
m
ini
m
a
li
ze
the
recu
rrent
a
m
ou
nt
at
each
it
erati
on
s
durin
g
op
e
rati
on.
In
orde
r
to
us
e
si
m
pl
ify
m
od
el
,
we
ass
um
e
su
it
ably
pe
rf
ect
sy
nchr
on
iz
at
io
n,
la
rg
e
cy
cl
ic
pr
efix
a
nd
pe
rf
ect
-
CSI
is
bee
n
dev
el
oped
thr
ough
t
he
pilot
dep
e
nd
ed
trai
ning.
T
hi
s
al
gorithm
per
f
or
m
s
thr
ough
co
ns
id
ered
regulariz
a
ti
on
pa
ram
et
er
s,
w
hich
e
nab
l
es
the
neig
hbori
ng
m
axi
m
u
m
li
kelihood
value
s
that
co
rr
es
pond
to
perform
a
nce
of
outp
ut
data
detect
ion
in
la
r
ge
M
U
-
MIM
O
syst
em
s,
m
or
eo
ver
th
e
la
r
ge
‘BS
-
to
-
us
er
-
a
nt
enn
a
’
rati
o.
Si
m
ula
ti
on
res
ults
analy
sis
sho
w
th
e
pro
pose
d
m
et
ho
d
s
ubs
ta
ntial
i
m
pr
ove
m
ent
ov
e
r
t
he
e
xisti
ng alg
or
it
hm
in
te
rm
s o
f
both
‘
E
rror rate’
m
i
nim
iz
at
ion
and
PAPR re
duct
ion.
The
pa
pe
r
is
orga
nised
as
f
ollows;
S
ect
ion
2
re
presents
li
te
ratur
e
su
r
vey,
S
ect
io
n
3
sh
ows
t
he
pro
po
se
d
syst
e
m
m
od
el
,
S
ect
i
on
4
re
presen
t
the
ex
per
im
e
ntal
res
ult
and
analy
sis,
an
d
S
ect
ion
5
c
oncl
ud
e
our wor
k.
2.
LIT
ERATUR
E SU
RV
E
Y
In
s
pite
of
m
any
adv
a
ntages
of
MIM
O
-
O
FDM
sign
al
s,
the
high
P
AP
R
va
lue
is
m
a
in
dr
a
wb
ac
k
tha
t
com
ing
in
sig
na
ls.
I
n
or
der
t
o
decr
ea
se
the
hi
gh
PA
PR
,
the
r
e
are
m
any
PAPR
re
du
ct
io
ns
appr
oach
ha
ve
been
pro
po
se
d
by
t
he
researc
he
rs
an
d
i
n
t
his
s
ect
ion
we
a
re
goin
g
el
a
bor
at
e
the
dif
fer
e
nt
P
AP
R
re
duct
ion
te
chn
iq
ue a
nd t
he pr
oble
m
ass
ociat
ed wit
h
it
.
The
sel
ect
ed
m
app
in
g
(SL
M)
ap
proach
i
s
an
ef
fici
ent
s
chem
e
beca
us
e
of
it
s
high
P
AP
R
re
duct
ion
capab
il
it
y
with
ou
t
t
he
disto
rtion
of
sig
nal,
wh
il
e
t
he
c
onve
ntion
al
SLM
(‘
CSLM
’)
ap
proac
h
c
om
e
up
wit
h
m
ajo
r
com
plexity
in
com
puta
ti
on
.
T
he
re
are
se
ver
al
m
et
hodo
l
og
ie
s
ha
s
bee
n
pro
posed
t
o
ov
e
rcom
e
the
diff
ic
ulty
,
ABC
(A
rtifi
ci
al
Be
e
Colon
y)
te
c
hn
i
qu
e
is
a
ne
wly
propose
d
swar
m
based
optim
iz
at
ion
app
r
oa
c
h
wh
ic
h
pro
vi
de
the
high
perfor
m
ance
in
re
gardin
g
of
P
AP
R
m
ini
m
iz
at
ion
.
The
pe
r
form
ance
of
ABC
ap
proac
h
gen
e
rall
y
base
d
upon
the
bee
s
search
strat
e
gies
an
d
there
are
al
so
so
m
e
m
od
ifie
d
searc
h
strat
egies
are
there.
In
pa
per
[
10]
,
a
pa
rall
el
AB
C
(P
-
ABC)
ha
s
pro
pose
d
to
m
od
ify
the
sea
rch
strat
egy,
a
nd
im
pr
ov
e
d
(
I
-
ABC)
appr
oach
is
be
en
intr
oduce
d
to
dec
rease
the
com
plexity
in
com
pu
ta
ti
on
at
SLM
appro
ac
h.
A
PA
BC
f
ollo
w
s
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
PAPR Re
duct
ion at
Large
M
ulti
-
User
-
MI
M
O
-
OF
DM usin
g
A
daptive
Da
t
a
…
(
N. Pr
aba1
)
1073
strat
egies
to
obta
in
best
ex
pl
oitat
ion
an
d
e
xp
l
or
at
io
n.
I
n
I
-
ABC,
an
in
di
vid
ual
be
e
search
f
or
the
optim
a
l
neig
hborh
ood
t
o
obta
in
best
s
olu
ti
on
f
or
m
utati
on
process
.
Af
te
r
wa
rd
s
,
th
e
ob
ta
ine
d
so
l
utions
is
sh
are
d
with
diff
e
re
nt sw
a
r
m
an
d
novel
‘s
earch
d
ire
ct
ions’ is
ob
ta
i
ned to
fin
d op
ti
m
um
so
ur
ces.
In
(single
-
i
nput
sing
le
-
outp
ut)
SISO
-
OF
D
M
wireless
sy
stem
s,
the
bes
t
proj
ect
in
g
te
chn
i
qu
e
s
ar
e
cl
ipp
in
g
[
11
]
,
act
ive
co
ns
te
ll
at
ion
e
xten
sio
n
(‘
’)
[
12]
,
to
ne
rese
rv
at
i
on
(
‘
’)
[13],
pa
rtia
l
tran
sm
issi
on
seq
uen
ce
(‘
’)
[14],
sel
ect
ed
m
app
in
g
(
‘
’)
[15],
an
d
oth
ers
.
I
n
[
16]
;
they
pro
vid
e
d
a
detai
le
d
ov
e
r
view,
in
w
hich
t
he
PA
PR
-
re
duct
io
n
a
ppr
oach
c
an
be
stret
c
he
d
easi
ly
f
rom
SI
SO
t
o
MIM
O
syst
e
m
s
[1
7]
,
[
16
]
,
t
he
e
xtens
ion
of
m
ulti
us
er
(M
U)
MIM
O
syst
em
is
not
that
m
uch
ea
sy,
beca
us
e
t
he
joint
process
of
rec
ei
ver
-
side
si
gnal
transm
issi
on
is
al
m
os
t
ver
y
dif
ficult
in
pr
act
ic
al
at
presence
of
distr
ibu
te
d
us
ers
.
M
IMO
-
OFDM
syst
e
m
has
been
e
xtensi
vely
acce
pted
to
se
r
ve
f
or
the
‘w
i
re
le
ss
com
m
un
ic
at
ion
’
syst
e
m
s.
Thou
gh
it
is
sti
ll
unde
r
go
es
from
the
m
axi
m
a
l
PA
PR
t
hat
is
the
m
ai
n
disadv
a
ntage
of
MIM
O
-
OFDM
-
based s
yst
e
m
s.
In
pa
per
[18],
they
co
ns
ide
r
the
ap
proac
h
t
o
reduce
the
P
AP
R
in
c
oded
OFDM
syst
em
s.
A
novel
PA
PR
m
ini
m
i
zat
ion
a
ppro
a
c
h
usi
ng
the
‘la
bel
-
inse
rted
’
e
ncode
r
of
s
of
t
a
m
plit
ud
e
lim
i
te
r
(‘
’)
an
d
us
ing
this
ap
proach
they
go
t
5.5dB
PA
PR
m
ini
m
iz
at
ion
(in
OFDM
syst
e
m
),
w
hich
c
onta
in
3dB
cl
ipp
i
ng,
128
su
bc
ar
riers,
an
d
4
-
bit
sel
ect
ion.
E
xclu
ding
the
sig
nifica
nc
e
of
P
AP
R
m
inim
iz
at
ion
this
ap
proac
h
al
s
o
ha
ve
adv
a
ntage
of
low
c
om
ple
xity
,
sm
all
over
hea
d,
sm
all
perf
or
m
ance
loss,
a
nd
no
e
xtra
inf
or
m
at
ion
transm
issi
on
.
Am
on
gs
t
se
veral
rando
m
pro
cess,
an
irre
gu
la
r
re
peat
accu
m
ula
te
(‘
’)
c
od
e
is
finest
op
ti
o
n
for
sim
pler
enc
od
e
r
process
.
The
prototype
can
be
ap
plied
directl
y
to
MI
MO
-
OFDM
syst
e
m
and
the
c
apacit
y
of
cl
ip
MIM
O
-
OFDM
can
be
analy
ze
through
Ga
us
sia
n
appr
ox
im
at
ion
.
Her
e,
they
c
on
si
der
s
of
t
MAP
detect
or
with
it
erati
ve
recei
ve
r
in
M
IMO
-
O
F
DM
syst
em
.
In
dif
fer
e
nt
num
ber
of
a
nten
nas
syst
e
m
s,
the
hy
br
id
appr
oach
is
be
com
e
ind
epe
ndent
pa
rt
of
e
nc
od
e
r,
there
f
ore
there
is
no
e
xtra
c
onstrai
nt
has
a
ppli
ed
at
IRA
-
cod
e
optim
iz
ation
.
T
he
desig
ning
of
IRA
c
od
e
s
is
com
pl
et
e
accord
i
ng
to
the
er
godi
c
MIM
O
syst
em
by
consi
der
se
veral
PA
PR
decrem
ent
set
ti
ng
,
m
or
eover,
th
e
diff
e
ren
t
cl
ipp
i
ng
rati
os
is
based
on
ex
trinsi
c
inf
or
m
at
ion
transf
e
r
(
‘EI
T
’)
plans.
T
he
hi
gh
value
of
P
AP
R
le
ads
to
the
in
-
ba
nd
distor
ti
on
al
on
g
with
fr
e
qu
e
ncy
spre
ad
sp
ect
ru
m
,
due
to
the
pr
ese
nce
of
no
n
-
li
ne
arit
y
in
the
‘h
i
gh
-
power’
am
plifie
rs
a
nd
the
la
rg
e
r
PA
PR
value
es
sentia
l
to
dec
r
ease
al
though
i
ts
prob
a
bili
ty
sh
ould
not
m
uch
hi
gh.
Wh
il
e
el
i
m
inati
ng
the
high
peaks
the
signa
l
per
f
or
m
ance
is
al
so
aff
ec
te
d,
there
f
or
e,
a
sp
eci
al
m
a
i
nt
ena
nce
nee
ds
to
be
co
ns
id
er
t
o
decr
ease
t
he
P
AP
R
in
M
IMO
-
OFDM.
In
t
hi
s
pa
per
[19],
t
hey
pro
posed
a
novel
al
gorith
m
that
is
based
upon
the
gro
upin
g
of
PTS
an
d
SL
M
m
et
ho
ds
f
or
the
P
AP
R
reducti
on.
I
n
pr
e
vi
ou
s
e
xisti
ng
te
chn
i
qu
e
s
of
P
AP
R
-
reducti
on
only
us
es
ei
ther
PTS
or
SL
M
as
a
separ
at
e
te
chn
iq
ue
on
MIM
O
-
OFDM.
He
re,
they
consi
der
QPS
K
m
od
ulate
d
te
chn
iq
ue
i
n
MIM
O
-
O
F
DM
syst
e
m
s
throu
gh
c
on
siderin
g
f
our
nu
m
ber
of ante
nn
as
.
In
[
20]
and
[21]
propose
d
a
Pu
lse
sh
a
ping
m
et
ho
d
that
is
eff
ect
ive
in
dec
reasin
g
the
PAP
R
of
MIM
O
-
OFDM
sign
al
,
al
so
pro
vid
e
the
lowe
r
com
pu
ta
ti
on
al
com
ple
xity
.
Sele
ct
ing
the
pr
oper
pu
ls
e
,
the
P
AP
R
ca
n
be
reduce
d,
m
or
eov
er
the
-
pu
lse
s
as
‘squ
are
-
root
raise
d
cosin
e’
(S
RR
C)
an
d
raised
cosine
(RC)
ar
e
usual
ly
us
e
d
pulse
s.
Co
nv
e
rsel
y,
the
ideal
-
pulse
s
a
re
nor
m
al
l
y
non
-
ca
usa
l,
that
i
s
wh
y
it
is
no
t
pract
ic
al
ly
reachab
le
a
nd
the
ideal
-
pulse
s
ca
n
be
s
horten
at
tim
e
do
m
ai
n
t
o
pr
ese
rve
the
ca
us
al
it
y,
but
this
m
od
el
intr
oduces
the
undesire
d
side
lob
es
of
non
-
z
ero
in
f
reque
nc
y
do
m
ai
n.
T
o
desig
n
a
fun
dam
ental
pu
lse
us
i
ng
an
ef
ficie
nt
co
m
pu
ta
ti
on
al
optim
iz
at
ion
m
et
hod
[
22]
,
w
hich
offer
e
d
a
m
uc
h
eff
ect
ive
m
et
ho
d
for
exe
cutin
g
pulse
sh
a
ping
m
et
ho
d
in
M
IMO
-
OFDM
s
yst
e
m
s
in
or
de
r
to
dec
rease
t
he
hi
gh
PA
PR
values
.
Howe
ve
r
,
thi
s
pulse
-
s
hap
i
ng
ap
proac
h
m
ay
pr
ese
nt
int
er
-
ca
rr
ie
r
i
nter
fer
e
nce
(
‘
’)
a
nd
el
i
m
inate
the
or
th
ogonal
it
y
fu
nction
i
n
the
MUI
-
M
IMO
-
OFDM
syst
e
m
.
Theref
or
e
,
it
causes
the
degrad
at
i
on
in the pe
rfor
m
ance
of d
em
odulati
on
i
n OF
D
M sy
stem
[
23
]
.
Ther
e
are
al
wa
ys
so
m
e
residual
that
le
adin
g
to
sig
nal
-
to
-
in
te
rf
ere
nce
a
nd
no
ise
rati
o.
To
overc
om
e
this
pro
blem
,
in
[
24
]
the
y
propose
d
a
zero
-
f
orci
ng
equal
iz
at
ion
te
chn
i
qu
e
.
The
a
dv
a
nce
m
ent
in
com
m
un
ic
at
ion
fiel
d
outc
ome
require
th
e
hi
gh
e
r
data
r
at
es
with
lo
wer
bi
t
-
er
r
or
-
rate
(B
ER)
an
d
inc
re
ase
in
the
po
wer
ef
fici
ency.
MIM
O
-
OFDM
is
c
om
pr
ehe
ns
ivel
y
us
e
i
n
t
he
pr
ese
nt
a
nd
s
ub
s
eq
ue
nt
ge
ne
rati
on
broa
db
a
nd
or
wireless
c
omm
un
ic
at
ion
s
[
25]
to
ge
ner
at
e
a
higher
data
-
rate
tra
ns
m
issi
on
,
al
so
the
s
pectral
eff
ic
ie
ncy
at
s
om
e
m
ult
i
path
channels
fad
i
ng.
T
he
M
IMO
syst
e
m
s
hav
in
g
the
m
ajo
r
drawb
ac
k,
w
hich
suffe
r
s
from
the
hig
h
PA
PR
that
re
quires
im
ple
m
e
ntati
on
of
high
ly
eff
ic
ie
nt
power
am
plifie
r,
in
orde
r
to
co
nt
inu
e
a
broa
der
li
near
re
gion
for
a
voidin
g
the
sig
nal
cl
ip
ping,
t
her
e
fore
it
i
nc
reases
t
he
po
wer
co
nsum
pti
on
a
nd
hard
war
e
c
omplexit
y.
T
he
se
ver
al
reducti
on
te
chn
i
qu
e
of
P
AP
R
has
bee
n
pro
po
se
d
t
hat
pro
vid
e
t
he
m
a
xim
a
l
reducti
on
of
P
AP
R,
w
hich
r
ang
i
ng
f
ro
m
0.5
dB
to
6dB.
In
pa
per
[
26]
,
disc
us
se
d
the
PAPR
perfor
m
ance
analy
sis,
al
so
th
e
al
go
rith
m
s
fo
r
beam
-
form
ing
(BF
)
te
chn
i
ques
t
hat
incl
ud
e
MR
T
(Maxim
um
Ra
ti
o
Transm
issi
on
),
EGT
(E
qu
al
Gain
Tra
ns
m
issi
on)
an
d
c
om
bin
ing
of
r
ecei
ver
te
c
hn
i
qu
e
s
s
uch
a
s
MR
C
(Maxim
u
m
R
at
io
Com
bin
ing
)
,
EGC
(Equal
Gain
C
om
bin
ing
)
f
or
MIM
O
-
OFDM
syst
e
m
.
Moreo
ve
r
,
the MIMO
-
OFDM sy
ste
m
m
e
asur
em
ent is c
om
plete
in
te
rm
s o
f
PAPR
r
e
du
ct
io
n, S
NR a
nd BER
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
0
71
–
1
0
80
1074
3.
PROP
OSE
D SYSTE
M MO
DEL
In
this
sect
io
n,
we
co
ns
ide
r
a
la
rg
e
MU
I
-
M
I
MO
-
OFDM
syst
e
m
to
add
res
s
the
MUI
ca
nc
el
la
ti
on
an
d
PA
PR
re
duct
ion.
Pr
eci
sel
y,
the
MUI
canc
el
at
ion
can
be
giv
en
as
un
de
rd
et
erm
ined
li
near
re
gu
la
rizat
io
n
pro
blem
that
req
ui
res
op
ti
m
i
ze
so
l
ution.
I
n
MIM
O
-
OFDM
m
od
el
,
is
an
a
nten
na
us
er
te
r
m
inals
that
tra
ns
fe
r
the d
at
a sim
ult
aneously
BS
-
a
nten
na
≫
ov
e
r
the
subcar
riers a
nd the
us
e
r
enc
odes it
s p
e
rs
on
a
l
bit stream
(
=
1
,
…
,
)
throu
gh
for
ward
er
ror
c
orrecti
on
a
ppr
oa
ch.
Af
te
r
wards
,
the
co
de
d
bit
s
sign
al
on
t
o
gathe
rin
g
po
i
nt
at
a
finit
e
set
i.e
.,
16
-
Q
AM,
with
a
verage
t
ran
sm
it
powe
r
in
unit
y
[
|
|
2
]
=
1
at
=
log
2
|
|
(i.e.
,
bits/
con
ste
ll
at
ion
point
)
a
nd
∈
.
Figure
1
.
Flo
w
char
t
of P
rop
ose
d
M
od
el
The
n
the
res
ulti
ng
f
re
qu
e
ncy
do
m
ai
n
(
)
sym
bo
ls
[
1
,
.
.
.
,
]
are
trans
f
or
m
ed
in
a
tim
e
do
m
ain
sy
m
bo
l
us
in
g
I
DF
T
[27]
(in
ve
rse
disc
rete
‘
Four
ie
r
Tra
ns
f
or
m
’)
.
Af
te
r
pre
-
pe
ndin
g
cy
cl
ic
sta
rt,
al
l
the
us
er
s
trans
fer
t
heir
tim
e
do
m
ain
sig
nals
at
the
sel
ect
iv
e
‘w
i
reless
f
reque
ncy
cha
nn
el
’
at
sam
e
per
i
od.
Af
te
r
,
discar
di
ng
cy
cl
ic
pr
e
fi
x,
t
he
ti
m
e
dom
ai
n
sign
al
s
ha
s
recei
ved
in
each
of
the
BS
-
a
nten
nas
a
nd
t
he
n
trans
form
ed
into
fr
e
quency
do
m
ai
n
us
in
g
discrete
‘
Four
i
er
Tra
ns
f
orm
’.
Figure
1
s
ho
ws
the
flo
wchart
of
pro
po
se
d
m
od
e
l.
In
orde
r
to
us
e
si
m
plify
m
odel
,
we
ass
um
e
su
it
ably
pe
rf
ec
t
synch
ronizat
ion,
la
r
ge
cy
cl
ic
pr
e
fix
an
d
perfect
-
CS
I
(c
hannel
-
sta
te
in
form
a
t
ion
)
is
be
en
de
velo
pe
d
throu
gh
t
he
pi
lot
dep
e
nded
t
rainin
g.
C
on
si
der
i
ng
this ass
um
ption
, t
he fre
qu
e
nc
y do
m
ai
n
relat
ion o
f
i
nput a
nd outp
ut w
it
h
ℎ
s
ubcar
rier ca
n be
m
od
el
ed
as;
=
+
(1)
Wh
e
re,
t
he
∈
can
be
a
ss
ociat
ed
with
f
re
quency
dom
ai
n
receive
d
vec
tor,
∈
co
ntains
the
transm
itted sym
bo
l t
hr
ou
gh a
ll
u
sers
(
)
a
nd
channel m
at
rix
is re
pr
ese
nted
as;
∈
∈
×
(2)
The
sym
bo
ls
is
tra
ns
m
itted
thr
u
us
e
r
at
-
subcar
rie
r
an
d
t
her
m
al
no
ise
m
od
el
ed
as
∈
at
sy
m
m
e
tric
com
plex
Gau
ssia
n vecto
r wit
h n
oise
var
ia
nc
e
and im
pu
lse
noise.
In
pract
ic
al
,
th
e
unde
rd
et
e
rm
i
ned
li
near
re
gula
rizat
ion
probl
e
m
can
be
s
ol
ve
by
re
gu
la
rize
the
le
ast
-
sq
ua
re
dif
ficult
y;
=
∈
‖
−
‖
2
2
+
‖
‖
2
2
(3)
Since
the
E
qua
ti
on
(
3)
is
qua
dr
at
ic
f
or
m
in
r,
so
t
he
regula
rizat
ion
pa
ram
et
er
pro
blem
has
in
cl
os
e
d
form
so
luti
on
.
W
hile
the
pro
blem
occu
r
in
ab
ov
e
E
quat
ion
(
3)
ca
n
be
cal
c
ul
at
ed
im
plicitl
y
t
hro
ug
h
data
-
detect
ion
al
gorithm
(D
DA
).
DDA
can
i
m
pr
ove
the
com
pu
ta
ti
on
al
eff
ic
ie
ncy
in
a
hard
war
e
pr
oto
ty
pe
that
avo
i
ds
the
excess
com
puta
ti
on
.
A
DDA
is
a
well
-
kn
own
it
erati
ve
f
ram
ewo
rk
to
s
olv
e
a
hu
ge
num
ber
of
c
onve
x
dif
f
ic
ulty
(ex
act
l
y
or
ap
pro
xim
a
te
ly
)
throu
gh
coor
din
at
e
-
wis
e
updates
a
nd
we
ca
n
def
i
ne
the
functi
on as;
(
1
,
.
.
.
,
)
=
(
)
=
‖
−
‖
2
2
+
(
)
(4)
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
PAPR Re
duct
ion at
Large
M
ulti
-
User
-
MI
M
O
-
OF
DM usin
g
A
daptive
Da
t
a
…
(
N. Pr
aba1
)
1075
Wh
e
re,
the
(
)
is
‘
conve
x
regular
iz
er’
a
nd
it
is
neces
sary
t
o
reali
ze
the
E
quat
ion
(
3),
w
hi
le
m
ini
m
iz
ing
E
quat
ion
(4)
thr
ough
sel
ect
ing
(
)
=
‖
‖
2
2
.
He
re,
m
in
i
m
izing
is
si
m
il
ar
to
s
olv
e
the
s
quare
er
r
or
regulariz
at
io
n pro
blem
thr
ough,
(
)
=
(
∈
)
(5)
Wh
e
re,
(
∈
)
def
ine
the fu
nctional
char
act
e
risti
cs that i
s zer
o
if
the
∈
and
oth
e
r
wise it
is inf
init
y.
Mi
ni
m
iz
ing
E
qu
at
io
n
(
5)
base
d
e
q
ualiz
at
ion
functi
on
ev
olv
es
(
1
,
.
.
.
,
)
se
quentia
ll
y
to
coor
din
at
e eac
h varia
ble
an
d
=
1
,
.
.
.
,
.
Her
e
,
we
ass
um
e
DD
A
ba
sed
s
quare
e
r
ror
e
qu
al
iz
at
io
n
to
fin
d
ℎ
op
ti
m
u
m
rate
fo
r
error
equ
al
iz
at
io
n prob
le
m
in
(
3)
.
̂
=
∈
‖
−
‖
2
2
+
‖
‖
2
2
(6)
Wh
ic
h
ho
l
ds
t
he
oth
e
r
values
of
,
∀
≠
fixe
d.
This
involve
d
a
quadr
at
ic
pro
blem
and
it
can
s
olve
thr
ough sett
ing t
he fu
nctio
n of g
rad
ie
nt in (
6) w
.
r.
t t
he
ℎ
com
po
ne
nt
to
zer
o.
0
=
∇
(
)
=
(
−
)
+
(7)
Thro
ugh dec
om
po
sing
E
quat
ion
(
8)
,
the
E
quat
ion (
7) for
in
ord
e
r
to
g
et
cl
os
e
d
f
r
om
ex
pressi
on
:
=
∑
l
≠
+
l
(8)
̂
=
[
−
∑
l
≠
]
(
‖
‖
2
2
+
)
⁄
(9)
The
e
xpressio
n
is
accuratel
y
update DDA
r
ul
e
for
ℎ
entry
of
,
a
nd
f
or
eac
h
it
er
at
ion
,
E
quat
ion
(9)
can
cal
culat
e
seq
uen
ti
al
ly
for
us
e
r
=
1
,
.
.
.
,
,
wh
e
re
fr
e
qu
e
ntly
reuse
al
l
new
no
vel
outc
om
e
̂
for
ℎ
us
er
. T
her
e
f
or
e
, w
e
r
e
peat the
process
for
tot
al
n
um
ber
of it
erati
on to
est
i
m
at
e;
̃
=
(
)
(10)
Wh
e
re,
(
)
is t
he
f
inal res
ult o
f D
DA b
ase
d desc
ribe
d
it
erati
ve pr
ocess
.
DDA
ena
bles
a
non
-
li
nea
r
da
ta
-
detect
ion
a
ppr
oac
h,
w
hich
op
e
rate
openly
in
‘f
re
quency
do
m
ai
n’
on
each
subca
rr
i
er
basis.
T
hi
s
update
de
r
ive
the
box
-
const
raine
d
e
qu
al
iz
at
io
n
pr
ob
le
m
.
More
ov
e
r
,
the
c
har
act
erist
ic
s
functi
on
ca
n
be
pro
vid
e
thr
ough
(5)
is
not
di
ff
e
ren
ti
abl
e,
but
it
is
sim
il
ar
m
e
th
od
th
at
ca
n
be use s
ub
-
gr
a
dients t
o
e
nab
l
es close
d form
of ex
pr
e
ssio
n,
̂
=
(
×
−
∑
l
≠
‖
ℎ
‖
2
2
)
(11)
Her
e
,
(
∙
)
sh
ows t
he
orth
ogonal
pr
oj
ect
io
n at
the
conve
x po
ly
to
pe
and
giv
e
n by;
(
)
=
{
∈
a
rg
∈
|
−
|
∉
(12)
The
a
r
gu
m
ent
∈
is
un
der
the
set
of
,
an
d
t
hen
i
s
pro
j
ect
io
n
ou
tpu
ts;
the
is
e
xt
ern
al
set
of
in
the
E
uclidea
n
dista
nce.
T
he
re
are
se
ver
al
r
el
evan
t
co
ns
te
l
la
ti
on
set
s,
an
d
(12)
pro
j
ect
io
n
can
be
ob
ta
i
n
eff
ect
ively
for
Q
AM
c
onste
ll
at
ion
.
Where
,
im
aginar
y
an
d
real
par
t
ca
n
be
cl
ip
i
nde
pende
ntly
of
on
t
o
the
[
−
,
+
]
inter
val a
nd
is a rad
i
us
of
cl
os
e
-
fitt
ing
box,
w
hich
c
over
s the
‘QAM’
c
on
ste
ll
at
ion.
In
t
his
pap
e
r, we p
r
opos
i
ng A
da
ptive d
at
a
-
detect
ion
(
ADD)
al
gorithm
t
hat
co
ntains
th
e
init
ia
li
zat
ion
,
pr
e
processi
ng
and
eq
ualiz
at
ion
ste
ps
i
n
order
t
o
get
no
perform
ance
loss
a
nd
to
m
i
nim
a
li
ze
the
r
ecurrent
a
m
ou
nt
at
each
it
erati
on
s
duri
ng
operati
on
(
=
1
,
.
.
.
,
)
.
I
ns
te
ad
,
to
com
pu
te
blind
ly
up
dates
in
(
9)
a
nd
(
11)
for
re
gu
la
rizat
ion
pa
ram
et
er,
the
pr
e
proces
sing
a
nd
restr
uctu
rin
g
ste
p
are
pe
rfor
m
ed.
In
pre
process
ing
is
perform
ed
to
de
crease
t
he
op
e
rati
on
al
c
om
plexity
,
A
DD
pr
e
-
cal
culat
e
the
sever
al
m
ajor
qu
a
n
ti
ti
es,
w
hi
ch
ca
n
be
re
proce
ssed
at
ind
i
vidual
it
erati
on
of
=
1
,
.
.
.
,
.
This
ty
pe
of
prep
ro
ce
ssin
g
is
not
only
pro
vi
de
the
sign
ific
a
nt
sav
ing
du
rin
g
co
m
plexit
y
at
it
e
rati
ve
proces
s
(su
c
h
as
D
D
A
),
but
al
so
sim
plifie
s
the
ha
r
dw
a
re
i
m
ple
m
entat
io
ns
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
0
71
–
1
0
80
1076
I
n pa
rtic
ular w
ay
, w
e
pre
-
cal
c
ulate
r
e
gu
la
rized
(in
ver
se
squ
ared col
um
n)
nor
m
s o
f
, th
at
is
;
−
1
=
(
‖
‖
2
2
+
)
−
1
(13)
Wh
e
re,
=
1
,
.
.
.
,
, wit
h
≥
0
and
the
r
e
gula
riz
ed param
et
er gai
n
is;
=
−
1
‖
‖
2
2
(14)
In
the
e
rror
m
od
e
of
re
gula
rizat
ion
pa
ram
et
er
=
an
d
in
box
-
c
on
st
rained
re
gula
riza
ti
on
m
od
e
=
0
, th
at
yi
el
ds
=
1
,
=
1
,
.
.
.
,
.
To
av
oid
t
he
r
ecurrent
proce
ss
at
equ
al
iz
at
ion
op
e
rati
on,
ADD
al
gorith
m
pr
ovide
the
increm
ental
updates
that
can
re
us
e
i
nterm
ediat
e
quantit
ie
s
at
each
phase
of
it
erati
on
=
1
,
.
.
.
,
.
Ther
e
fore,
we per
form
ing
seque
ntial
u
pg
rad
at
io
n, al
so c
an be cal
le
d as
appr
ox
im
at
e residu
al
vecto
rs and de
fine
d
as;
=
−
∑
l
(
)
=
1
(
15)
Howe
ver,
we
are
no
t
goin
g
t
o
re
-
cal
c
ulate
the
a
ppr
ox
im
ate
resi
du
al
vect
or
s
f
or
in
div
i
dual
it
erati
on.
In co
nf
li
ct
in
g,
we
init
ia
ll
y co
m
pu
te
the sym
bo
l e
stim
at
ion
(
)
by the a
ppr
oxim
at
e residu
al
ve
ct
or
s.
(
)
=
(
−
1
+
(
−
1
)
×
)
(16)
The del
ta
v
al
ue
of e
qu
at
io
n (
16)
ca
n be c
ompu
te
a
s;
∆
(
)
=
(
)
−
(
−
1
)
(17)
Eq
uation (
17) e
nab
le
s
the
(
)
r
esi
du
al
update
in (1
8)
with
ou
t
co
m
pu
ti
ng the
r
esi
dual
e
xp
li
c
it
ly
.
←
−
∆
(
)
(18)
The recei
ve
d o
utput vect
or ca
n be c
om
pu
te
d as;
̃
=
[
1
(
)
,
.
.
.
,
(
)
]
(19)
As
we
de
scri
be
d
a
bove,
A
D
D
al
gorithm
prov
i
de
the
sig
nificantl
y
le
ss
co
m
pu
ta
ti
on
al
co
m
plexit
y.
In
DDA
a
ppr
oac
h,
it
re
quires
a
com
plex
inn
e
r
pr
oduct
value
a
nd
(
−
1
)
scal
ar
-
by
-
ve
ct
or
c
om
plex
m
ul
ti
plica
ti
on
s
at
pe
r
it
erati
on
of
,
w
he
reas
t
he
pro
po
se
d
A
DD
ap
proac
h
r
equ
i
res
on
ly
one
scal
ar
-
by
-
ve
ct
or
com
plex
m
ultip
li
cat
ion
a
nd
one in
ne
r produ
ct
v
al
ue.
4.
E
X
PERI
MEN
TAL RES
UL
TS A
ND AN
A
LYSIS
The
sim
ulatio
n
is
par
t
is
do
ne
us
ing
Ma
tl
ab
2016b,
syst
e
m
config
ur
at
io
n;
8G
B
RAM,
2GB
gr
a
phic
s
card,
1TB
RO
M
and
intel
i5
process
or
.
In
order
t
o
evalu
at
e
the
per
f
or
m
ance
of
er
r
or
-
rate
f
or
our
pro
pose
d
ADD
m
od
el
with
res
pect
to
Zer
o
Forci
ng
[24]
al
gorithm
,
we
per
for
m
ing
‘Mo
nte
-
Ca
rlo’
sim
ulatio
ns
in
MIM
O
-
OFDM
syst
e
m
,
wh
er
e
LTE
A
dvan
ced
(L
TE
-
A)
t
echn
i
qu
e
is
use
d
f
or
data
tr
ansm
issi
on
wi
th
the
su
bc
ar
riers
[
28
]
.
In
this
stu
dy
,
we
us
e
16
-
Q
AM
with
gray
m
app
in
g
an
d
to
acco
unt
the
f
reque
ncy
and
s
pati
al
correla
ti
on
we
create
c
ha
nn
e
l
est
i
m
at
ion
m
at
rices
us
i
ng
W
i
nner
-
P
hase
-
2
prototype
[
29]
with
c
on
si
de
rin
g
7.8cm
antenn
a
distance
(s
pac
ing).
W
e
re
por
t
the
sy
m
bo
l
e
rror
rate
(
SER)
,
Bi
t
err
or
rate
(BER)
an
d
pe
ak
-
to
-
aver
a
ge
powe
r
rati
o
(
‘PAPR
’
)
is
co
ns
i
der
a
s
pea
k
am
plit
ud
e
s
qu
a
re
d
val
ue
div
ide
d
by
‘roo
t
m
ean
square
’
in
orde
r
to
get
aver
a
ge
powe
r.
T
he
BER
a
nd
SER
perfor
m
ance
of
16
-
QA
M
a
rr
a
nge
m
ent
with
O
F
DM
in
pr
ese
nce
of
G
aussian
noise
and
im
pu
lsi
ve
no
ise
will
del
iver
the
syst
em
per
fo
rm
ance
us
in
g
our
propos
e
d
ADD
m
od
el
w
it
h
res
pect to
Z
ero F
or
ci
ng al
gorithm
.
Her
e
,
we
c
ons
iderin
g
tw
o
sc
enar
i
os
w
he
re
we
fi
x
the
num
ber
of
recei
ve
anten
nas
a
nd
var
yi
ng
th
e
transm
it
antenn
as
us
in
g
16Q
AM
m
od
ulati
on
te
ch
nique.
Mon
te
-
Ca
rlo
t
rial
s
is
co
ns
id
er
as
1000
0
a
nd
the
sign
al
-
to
-
n
oise
-
rati
o
(
SN
R
)
is var
yi
ng f
r
om
-
10
to 10dB (
with the in
te
rv
al
of
2d
B
).
In sce
nar
i
o
-
a,
we
co
ns
ide
r
32
num
ber
of
receive
a
nte
nnas
with
16
tr
ansm
it
antennas
an
d,
32
num
ber
of
receive
a
nten
nas
w
it
h
32
transm
it
antenn
as
in
or
der
t
o
eval
uate
the
err
or
rate
an
d
PAPR
.
Sim
i
la
rly
in
scenar
io
-
b,
we
c
onsider
64
nu
m
ber
of
r
ec
ei
ve
anten
nas
with
32
tra
ns
m
it
antenn
as
a
nd
,
64
num
ber
of
receive
a
nten
nas
wit
h
64
tra
ns
m
i
t
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
PAPR Re
duct
ion at
Large
M
ulti
-
User
-
MI
M
O
-
OF
DM usin
g
A
daptive
Da
t
a
…
(
N. Pr
aba1
)
1077
anten
nas
in
or
der
t
o
e
valuate
PA
PR
a
nd
the
error
rate.
T
hr
ough
var
yi
ng
t
he
a
nten
nas
num
ber
at
SN
R,
we
ca
n
analy
ze the
be
hav
i
or of
our p
rop
os
ed
m
od
el
.
Figure
2
sho
w
s
f
or
the
SN
R
vs
SER
at
32
nu
m
ber
of
rec
ei
vin
g
a
nten
na
(RA
s)
a
nd
16
transm
it
ti
ng
anten
nas
(TAs
),
wh
e
re
our
pro
po
se
d
m
odel
perform
con
side
rab
ly
well
with
inc
reasing
num
ber
of
SN
R,
in
10dB
S
NR
the
A
DD
go
t
6.17%
le
ss
SE
R
than
the
ZF
[
24
]
.
Sim
il
arly
,
in
F
ig
ure
3
s
hows
f
or
the
S
NR
vs
SER
at
32
RA
s
and
32
TAs
,
wh
e
re
the
SE
R
is
li
t
tl
e
m
or
e
com
par
e
to
F
ig
ure
2,
w
he
reas
our
pro
pose
d
m
od
el
go
t
41%
le
ss
S
ER
at
10dB
S
NR.
Fi
g
ure
4
s
hows
for
the
S
NR
vs
BER
at
32
RAs
a
nd
16
T
As,
at
10dB
SN
R
our
pro
pose
d
m
od
el
AD
D
pe
rfor
m
s
6.
6%
l
ess
com
par
ed
to
the
ZF
ap
pr
oach.
Sim
i
la
rl
y,
F
ig
ure
5
shows
f
or
the
S
NR
vs
BE
R
at
32
RAs
a
nd
32
T
As,
w
he
re
in
-
10dB
Z
F
got
11%
m
or
e
BER
com
pared
to
A
DD.
Fi
gure
6
and
7
s
hows
for
the
S
NR
vs
P
A
PR
at
32
RA
s
with
16
a
nd
3
2
T
As;
c
on
si
der
i
ng
10dB
S
NR
in
F
ig
ure
6
A
D
D
go
t
1.2
5%
le
ss
PA
PR t
hen
ZF
and in
F
ig
ure
7
A
D
D go
t
24% less PA
PR
then
ZF
a
ppr
oa
ch.
Bel
ow
F
ig
ure 8
t
o
13
s
how
s
f
or
t
he
sce
nar
io b
; wh
e
re
F
ig
ure
8
and
9
s
hows
f
or
t
he
SN
R vs
SER
at
64
R
As
with
32
a
nd
64
TA
s,
F
ig
ure
10
a
nd
11
s
how
s
f
or
t
he
S
NR
vs
BE
R
at
64
RA
s
with
32
a
nd
64
TA
.
Mo
reov
er,
F
ig
ure
12
a
nd
13
sh
ows
f
or
the
SN
R
vs
PAPR
at
64
R
A
s
with
32
an
d
64
T
As,
w
her
e
in
F
ig
ure
12,
propose
d
A
D
D
s
ho
ws
th
e
1.5%
le
ss
PAP
R
at
10dB
SNR
and
in
F
ig
ure
13,
pro
posed
A
DD
sho
ws
t
he
35
%
le
ss
P
AP
R
at
10dB
SN
R
.
Table
1
sho
w
f
or
P
AP
R
val
ue
at
diff
ere
nt
SN
R
per
recei
ve
d
anten
nas
[
dB
]
,
wh
ere
c
on
si
der
e
d
total
num
ber
of
receivin
g
a
nte
nn
a
s
is
12
8
a
nd
c
on
si
der
e
d
t
ran
sm
it
antenna
is
64.
In
-
2dB
SN
R,
A
DD
pro
po
se
d
m
odel
go
t
23%
le
ss
PAP
R
value
com
pa
red
to
ZF
a
pp
ro
ac
h
a
nd
in
4dB
S
NR,
A
D
D
m
od
el
go
t
7.7%
le
ss
P
APR
value
com
par
ed
t
o
ZF
ap
proac
h.
Ther
e
f
or
e,
f
rom
the
analy
sis
of
res
ult
we
can
say
t
hat
increasi
ng
in
S
NR
pe
r
receive
d
a
nten
nas
t
he
a
ver
a
ge
v
al
ue
of
PAP
R i
s d
ec
reasin
g.
Scenari
o
-
a
Figure
2
.
S
NR
vs
S
ER (R
As=
32 and
TAs=
16)
Figure
3
.
S
NR
vs
S
ER (R
As=
32 and
TAs=
32)
Figure
4
.
S
NR
vs
BER
(RAs=
32 and
TAs=
16)
Figure
5
.
S
NR
vs
BER
(RAs=
32 and
TAs=
32)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
0
71
–
1
0
80
1078
Figure
6
.
S
NR
vs
P
AP
R
(
R
As
=32 an
d TAs=
16)
Figure
7
.
S
NR
vs
P
AP
R
(
R
As
=32 an
d TAs=
32)
Scenar
i
o
-
b
Figure
8
.
S
NR
vs
S
ER
(R
As=
64 and
TAs=
32)
Figure
9
.
S
NR
vs
S
ER (R
As=
64 and
TAs=
64)
Figure
10
.
SNR
v
s BER
(
R
A
s=64 a
nd TAs
=32)
Figure
11
.
SNR
v
s BER
(
R
A
s=64 a
nd TAs
=64)
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
PAPR Re
duct
ion at
Large
M
ulti
-
User
-
MI
M
O
-
OF
DM usin
g
A
daptive
Da
t
a
…
(
N. Pr
aba1
)
1079
Figure
12
.
SNR
v
s
PAPR
(
R
As=64
a
nd T
A
s=32)
Figure
13
.
SNR
v
s
PAPR
(
R
As=64
a
nd T
A
s=64)
T
able
1
.
PAPR
v
al
ue
at di
ff
e
r
ent S
NR (
R
As
=128
a
nd T
As
=64)
SNR[dB
]
-
10
-
8
-
6
-
4
-
2
0
2
4
6
8
10
Z
F
[24
]
3
.47
3
.36
3
.24
3
.11
2
.98
2
.88
2
.81
2
.77
2
.76
2
.76
2
.76
ADD
2
.11
2
.19
2
.28
2
.35
2
.41
2
.47
2
.52
2
.57
2
.63
2
.67
2
.71
5.
CONCLUS
I
ON
We
ha
ve
pr
opos
e
d
A
dap
ti
ve
data
detect
ion
(ADD)
al
gorit
hm
based
dete
ct
ion
te
ch
niqu
e
for
the
la
r
ge
MUI
-
M
IMO
s
yst
e
m
by
the
use
of
ort
h
ogon
al
-
fr
e
quency
-
di
vision
-
m
ulti
pl
exin
g
(‘
’)
with
the
pr
ese
nc
e
of
Im
pu
lse
and
Ga
us
sia
n
no
ise
.
This
pro
posed
A
DD
al
go
rit
hm
h
avin
g
both
li
near
a
nd
non
-
li
nea
r
regulariz
at
io
n
const
raint
to
pr
ov
i
de
sim
ple
detect
ion
operat
ion
.
Her
e
,
we
consi
der
e
d
tw
o
scenari
os
;
scenari
o
-
a
and
sce
nar
i
o
-
b
to
valid
at
e
the
pe
rfor
m
ance
of
our
pro
posed
m
od
el
.
Wh
ere
we
fix
th
e
nu
m
ber
of
r
ecei
ve
anten
nas
a
nd
var
yi
ng
the
t
r
ansm
it
antenn
as
us
in
g
16Q
AM
m
od
ulati
on
te
ch
nique
a
nd
1000
0
Mo
nte
-
Ca
rl
o
tria
ls
has
been
ta
ken
at
diff
e
ren
t
sig
nal
-
to
-
no
ise
-
rati
o
(SNR).
Our
res
ul
ts
sh
ow
s
that
ADD
is
su
it
able
f
or
pr
act
ic
al
OFD
M
based
la
rg
e
MUI
-
MIM
O
syst
e
m
to
su
pp
or
t
se
ver
al
us
e
rs
com
m
un
ic
at
ing
with
hundr
eds
of
base
sta
ti
ons
by
achievi
ng
be
tt
er
data
t
ran
s
act
ion
rate
a
nd
P
AP
R
re
duc
ti
on
with
pres
ence
of
Im
pu
l
se
an
d
Gau
s
sia
n
noise
.
I
n
c
onside
rin
g
128
R
As
12
8
a
nd
64
TAs
,
we
got
1.8
%
reducti
on
in
P
AP
R
at
10dB
SN
R.
In
f
ut
ur
e
work,
the
A
DD
m
od
el
ca
n
be
i
m
ple
m
ent
in
VLSI
f
or
f
ast
er
co
nv
e
r
gen
ce
and
ca
n
ena
bl
e
high
transm
issi
on
at sam
e PA
PR.
REFERE
NCE
S
[1]
F.
Rusek,
D.
Per
ss
on,
B.
K.
La
u
,
E.
G.
La
rss
on,
O.
Edfor
s,
F.
Tu
fve
ss
on,
and
T
.
L.
Marz
etta
,
“
Scal
ing
up
MIM
O:
opportuni
ties a
n
d
challe
ng
es
wit
h
ver
y
la
rg
e array
s,
”
a
rXiv:1201.3210v1,
Jan
.
20
12.
[2]
S.
K.
Moham
me
d
and
E.
G
.
L
arsson,
“
Per
-
antenna
consta
nt
e
nvel
ope
pr
ec
od
i
ng
for
la
rg
e
m
ult
i
-
user
MIM
O
s
y
st
ems
,
” arXiv:1111.3752v1, Ja
n.
2012
.
[3]
Hala
M.
Abd
Elkade
r,
Gam
al
M
.
Abdel
-
Ham
id
,
Adl
y
Ta
g
El
-
Di
en,
As
m
aa
A.
Nass
if,
“
Com
bine
d
Bea
m
form
ing
with
Orthogona
l
Space
T
ime
Bloc
k
Code
for
MIM
O
-
O
FDM
with
Sim
ple
F
ee
dba
ck”,
Indon
esia
n
Journal
o
f
El
e
ct
ri
ca
l
Enginee
ring
and
Com
pute
r
Scie
nc
e
Vol.
4,
No.
3,
Desem
ber
2016,
pp.
58
0
~
585
DO
I
:
10.
11591/ijeecs.
v4.
i3.
pp580
-
585
.
[4]
X.L
ia
nd
L.
Cim
in
i,
“
Eff
e
ct
s
of
cl
ip
ping
and
filte
rin
g
on
the
per
form
anc
e
of
OF
DM
,
”
Comm
unic
at
ion
s
Le
tt
ers,
I
EEE
,
vol.
2
,
no
.
5
,
pp
.
131
–
133,
Ma
y
1
998.
[5]
X.
Zhu,
W
.
Pan,
H.
Li,
and
Y.
T
ang,
“
Sim
pli
fied
appr
oa
ch
to
opt
imize
d
i
te
r
at
iv
e
cl
ippi
ng
and
fil
t
eri
ng
for
PA
P
R
red
uction
of
OF
DM
signal
s,”
IE
EE
Tra
nsa
ct
ions
on
Com
m
unic
ations,
vol
.
61
,
no
.
5,
pp.
1891
–
190
1,
Ma
y
2013.
[6]
S.
Um
eda
,
S.
Suy
ama
,
H.
Suzuki
,
and
K.
Fuk
awa
,
“
PA
PR
red
uct
i
on
m
et
hod
for
bl
ock
dia
gon
al
i
zat
ion
in
m
ult
iuser
MIM
O
-
O
FD
M
s
y
stems
,
”
in
Veh
ic
ul
ar
Te
chno
lo
g
y
Confer
enc
e
(
VTC
2010
-
Spring),
2010
IEE
E
71st,
Ma
y
2010,
pp.
1
–
5.
[7]
Kalve
in
R
ant
e
lo
bo,
Hendro
La
m
i,
W
ira
wan
,
“
Video
Tra
nsm
iss
ion
using
Com
bine
d
Scalabilit
y
Vi
deo
Coding
over
MIM
O
-
O
FD
M
S
y
stems
”,
Indo
nesia
n
Journa
l
of
El
e
ct
r
ic
a
l
En
gine
er
ing
and
Com
pute
r
Sc
ie
n
ce
Vol
.
4,
No.
2,
Novem
ber
2016,
pp.
390
~ 396
D
OI:
10.
11591/
ij
e
ec
s.v4.i2.
pp390
-
396.
[8]
S.
H.
Han
and
J.
H.
Le
e
,
“
An
over
vie
w
of
pea
k
-
to
-
ave
r
age
power
rat
io
red
uct
ion
t
ec
hniqu
es
for
m
ult
ic
arr
ier
tra
nsm
ission,”
W
ire
le
ss
Com
muni
cations,
IEEE
,
vol
.
12
,
no
.
2
,
p
p.
56
–
65
,
April
2005.
[9]
J.
C.
Ch
en,
“
Partial
tra
nsm
it
s
eque
nc
es
for
P
AP
R
red
uct
ion
of
OF
DM
sign
al
s
with
sto
chas
ti
c
opt
imizatio
n
te
chn
ique
s,”
IE
E
E
Tr
ansa
c
ti
ons o
n
Consum
er
Ele
ct
roni
cs,
vo
l. 56, no. 3, pp. 1229
–
1234,
Aug 2010
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
0
71
–
1
0
80
1080
[10]
N.
Ta
şpınar
an
d
M.
Yıldı
rım,
"A
Novel
Para
llel
Artif
ic
i
al
Bee
Colon
y
Al
gorit
hm
and
Its
PA
PR
Reduc
ti
on
Perform
anc
e
Us
ing
SLM
Scheme
in
OF
DM
and
MIM
O
-
O
FD
M
S
y
stems
,
"
in
IE
EE
Com
m
unic
ations
Le
tters,
vol
.
19,
no
.
10
,
pp
.
1
830
-
1833,
Oc
t. 2
015.
[11]
R.
ON
ei
ll
and
L
.
B.
Lope
s,
“
Enve
lope
v
ari
a
ti
ons
a
nd
spec
tr
al
spl
a
tt
er
i
n
clipped
m
ult
icarri
er
signals
,
”
in
Proc
.
IEEE
PIM
RC’95,
Tor
onto,
C
ana
d
a, Se
p.
1995
,
pp
.
71
–
75.
[12]
B.
S.
Krongold
and
D.
L.
Jones,
“
PA
R
red
uct
ion
in
OF
DM
via
activ
e
const
el
l
at
ion
ex
te
nsio
n,
”
IE
EE
Tr
ans.
Broadc
ast
ing, vo
l.
49
,
no
.
3
,
pp
.
2
58
–
268,
Sep
.
20
03.
[13]
J.
Tell
ado, “P
ea
k
to ave
r
age po
wer
red
u
ct
ion
fo
r
m
ult
icarri
er
m
odula
ti
on
,
”
PhD
the
sis (Sta
nford
Univer
sit
y
,
2000
).
[14]
S.
H.
M
¨
u
l
le
r
a
nd
J.
B.
Huber
,
“
OF
D
M
with
red
uce
d
p
ea
k
-
to
-
ave
rag
e
power
r
at
io
b
y
op
ti
m
um
combinat
ion
of
par
tial tra
nsm
it
s
eque
nc
es,
”
IE
E El
e
c. L
e
tt
ers
,
vo
l.
33
,
no
.
5
,
pp
.
3
68
–
369,
Feb
.
19
97.
[15]
R.
W
.
B
¨
a
um
l
,
R
.
F.
Fis
ch
er,
and
J.
B.
Huber
,
“
Re
duci
ng
th
e
p
ea
k
-
toa
ver
age
pow
er
rat
io
of
m
ult
i
ca
r
rie
r
m
odulation
b
y
select
ed
m
ap
ping,
”
IE
E
E
lec.
Le
tters,
vol. 32,
no.
22
,
pp
.
2056
–
2057
,
Oct
.
199
6.
[16]
S.
H.
Han
and
J.
H.
Le
e
,
“
An
over
vie
w
of
pea
k
-
to
-
ave
r
age
power
rat
io
red
uct
ion
t
ec
hniqu
es
for
m
ult
ic
arr
ier
tra
nsm
ission,”
I
EE
E
W
ire
l
ess Com
m
un.
,
vol. 12
,
no
.
2
,
pp
.
56
–
6
5,
Apr.
2005.
[17]
T.
Jiang
and
Y.
W
u,
“
An
over
vie
w:
Peak
-
to
-
av
e
rag
e
power
rat
i
o
red
uct
ion
t
ec
hn
ique
s
for
OF
D
M
signal
s,”
IEEE
Tra
ns.
Bro
adcast
ing,
vo
l. 54, no.
2,
pp
.
257
–
268
,
Jun.
2008.
[18]
Dinesh
N.
Bhan
ge,
Chand
rashe
k
har
G.
Deth
e,
“
Perform
anc
e
of
Pilot
-
Aided
3D
-
OF
DM
Channe
l
Esti
m
at
ion
usin
g
Diffe
ren
t
Ant
en
na
Configura
t
ion
s”,
Indone
sia
n
J
ourna
l
of
Elec
trica
l
Eng
ineeri
ng
and
Com
pute
r
Scie
n
ce
Vol.
8
,
No.
1,
Octob
er
2017
,
pp.
77
~ 84
DO
I:
10.
11591
/i
j
eec
s.v8.
i1.
pp77
-
84.
[19]
H.
Ti
wari
,
R.
Roshan
and
R.
K.
Singh,
"P
APR
red
uction
in
MIM
O
-
O
FD
M
using
combined
m
et
hodolog
y
o
f
sele
c
te
d
m
appi
n
g
(SLM)
an
d
par
ti
al
tra
nsm
it
seque
nce
(PTS),"
2
014
9th
Inte
rna
t
i
onal
Confer
en
ce
on
Industria
l
an
d
Inform
at
ion
S
y
s
t
ems
(ICIIS),
Gw
al
ior
,
2014
,
pp
.
1
-
5.
[20]
S.
Slim
ane
,
“
Pea
k
-
to
-
ave
r
age
po
wer
rat
io
r
educ
t
i
on
of
OF
DM
sig
nal
s using
broa
d
band
pulse
shapi
ng,
”
in
Veh
ic
ul
a
r
Tec
hno
log
y
Con
fer
ence, 2002. Proce
ed
ings.
VT
C
2002
-
Fall.
200
2
IEEE
56
th, vol.
2
,
2002
,
pp
.
88
9
–
893
vol.
2
.
[21]
R.
Rei
ne
and
Z.
Za
ng,
“
Anal
y
s
i
s
and
compari
son
of
a
set
of
ISI
fre
e
wave
form
s
for
PA
P
R
red
uct
ion
in
OF
DM
s
y
stems
,
” in TE
NCO
N 2011
-
2
011
IEEE
R
egi
o
n
10
Confer
ence
,
Nov 2011
,
pp
.
246
–
250.
[22]
“
A
quadr
at
ic
pr
ogra
m
m
ing
appr
oac
h
in
pulse
shaping
filte
r
d
esign
to
r
educ
in
g
PA
PR
in
OFDM
sy
st
ems
,
”
i
n
Com
m
unic
at
ions (
AP
CC),
2013
19th
As
ia
-
Pac
ific
Confer
ence
on
,
Aug 2013, pp. 5
72
–
576.
[23]
M.
-
J.
Hao
and
C.
-
H.
Lai,
“
Puls
e
sh
api
ng
b
ase
d
PA
PR
red
uct
ion
for
OF
DM
signal
s
wi
th
m
ini
m
um
er
ror
proba
bil
i
t
y
,
”
in
Inte
lligen
t
Signa
l
Proce
ss
ing
an
d
Com
m
unic
at
i
ons
S
y
stems
,
2
008.
ISP
ACS
2008.
Int
ern
a
ti
on
al
S
y
m
posium
on,
Feb
2009,
pp.
1
–
4.
A.
Am
inj
ava
her
i
,
A
.
Farha
ng
,
A.
Re
za
z
ade
hr
e
y
hani,
L.
E
.
Do
y
le
and
B.
Farha
ng
-
Bor
ouje
n
y
,
"O
FD
M
W
it
hout
CP
in
Mass
ive
MIM
O,"
in
I
EE
E
Tr
ansa
ct
ions o
n
W
ire
l
e
ss
Comm
unic
at
i
ons,
vol. 16, no.
11,
pp
.
7619
-
76
33,
Nov.
2017.
[24]
G.
Maha
la
kshm
i
and
V.
M.
Bhaskar
an,
"M
an
agi
ng
m
obil
ity
in
wire
le
ss
ce
l
l
ula
r
net
works
:
A
profil
e
base
d
appr
oac
h
,
"
2014
Inte
rn
at
ion
al
C
onfe
ren
c
e
on
E
le
c
troni
cs
and
Com
m
unic
at
ion
S
y
stems
(ICE
CS
),
Coim
bat
or
e
,
2014,
pp
.
1
-
5.
[25]
Z.
A.
Sim
,
R
.
R
ei
ne
,
Z
.
Z
ang
an
d
L.
Gopa
l,
"P
AP
R
and
BER
red
uct
ion
in
MU
-
MIM
O
-
OFDM
sy
stems
via
a
se
t
of
wave
form
s,"
2017
IEE
E
In
te
rn
at
ion
al
Confer
e
nce
on
Signa
l
and
Im
age
Proce
ss
ing
Applica
ti
ons
(ICSIP
A),
Kuching,
2017
,
pp.
55
-
60
.
[26]
S.
Isam
and
I.
D
arwa
ze
h
,
"S
imple
DS
P
-
IDF
T
te
c
hnique
s
for
gen
e
rat
ing
spe
ct
r
al
l
y
eff
icient
FD
M
signal
s,"
2010
7th
Inte
rna
ti
ona
l
S
ym
posium
on
Com
m
unic
at
ion
S
ystems
,
Networks
&
Digi
ta
l
Sig
nal
Proc
essing
(
CS
NDSP
2010),
Newca
stle upon T
y
n
e, 2010, pp.
20
-
24.
[27]
3rd
Gene
r
at
ion
Partne
rship
Proj
ec
t
;
T
ec
hn
ic
a
l
Speci
fi
ca
t
ion
Gr
oup
Radi
o
Acc
ess
Network;
E
volve
d
Univ
ersa
l
Te
rre
str
ia
l
Radio
Acc
ess
(E
-
UT
RA);
Ph
y
sic
al
L
a
y
er
Proce
du
res
(Rel
e
ase
10),
3
GP
P
Organi
za
tional
Partn
ers
T
S
36.
213
ver
sion
1
0.
10.
0
,
Jul.
2013
.
[28]
L.
Hen
ti
l
¨
a
,
P.
K
y
¨
os
ti,
M.
K
¨
a
ske,
M.
Na
ran
d
zi
c
,
and
M.
Al
atos
sava
,
“
Matlab
imp
le
m
ent
a
ti
on
of
the
W
INN
ER
phase
II
cha
nn
el
m
odel
ver
1.
1
,
”
Dec
.
2007.
Evaluation Warning : The document was created with Spire.PDF for Python.