Indonesian J
ournal of Ele
c
trical Engin
eering and
Computer Sci
e
nce
Vol. 2, No. 1,
April 201
6, pp. 49 ~ 60
DOI: 10.115
9
1
/ijeecs.v2.i1.pp49
-60
49
Re
cei
v
ed
Jan
uary 15, 201
6
;
Revi
sed Ma
rch 1
6
, 2016;
Acce
pted Ma
rch 2
6
, 2016
Single Phase Variable Sampling Phase Locked Loop
using Composite Observer
K Aru
n
, K Selv
aj
y
o
thi*
Dep
a
rtment of Electron
ics En
gin
eeri
ng, Indi
an Institute of Informatio
n
T
e
chno
log
y
D
e
si
g
n
& Manufactur
i
ng,
Kanch
e
e
puram
, Chenn
ai – 1
2
7
, India
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: ksj
y
o
th
i@i
iitd
m
.ac.in
A
b
st
r
a
ct
An obs
erver
b
a
sed v
a
ri
abl
e
sampli
ng
per
io
d ph
ase
lock
e
d
lo
op
is i
n
trod
uced f
o
r gri
d
c
onn
ected
system
s. The compos
ite obs
erver acts as an efficien
t estimator of the fundam
ent
al com
p
onents from a
peri
odic i
n
p
u
t sign
al rich i
n
DC an
d har
mo
nics. T
he obs
e
r
ver gai
ns are
desi
gne
d usi
n
g pol
e pl
ace
m
ent
techni
qu
e, w
h
i
c
h i
nher
ently
e
n
sures
the
sta
b
ility
of this
es
timator.
Even
und
er dr
ift freq
uency,
a c
onst
ant
nu
mb
er of sa
mp
les (
512)
p
e
r cycle
are
mainta
ine
d
w
i
th
the he
lp
of the
nu
mer
i
cal
l
y c
ontrol
l
ed
oscil
l
a
tor.
T
h
is
makes
th
e osci
llat
o
r g
a
i
n e
l
e
m
ents i
n
the o
b
serv
er
a co
nstant
a
nd e
l
i
m
i
nates
the
trig
ono
metric
computati
on. T
h
is ph
ase lock
ed lo
op is fou
n
d
to
be w
o
rkin
g in a w
i
de ra
nge of freq
uen
cy 40- 70H
z
.
T
h
e
perfor
m
a
n
ce of
the pro
pos
ed
sche
m
e
is stud
ied w
i
th a
sy
nt
hetic h
a
r
m
on
ic
rich si
gna
l as
w
e
ll as va
lid
ate
d
by impl
ementi
n
g the PLL i
n
C
yclon
e IV F
P
GA w
i
th a real time gr
id vo
ltag
e.
Ke
y
w
ords
: co
mp
osite
obs
er
ver, har
mo
nics
, nu
meric
a
lly c
ontrol
l
ed
osci
ll
ator, phas
e l
o
c
k
ed l
o
o
p
, varia
b
l
e
sampling
Copy
right
©
2016 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Accu
rate e
s
ti
mation of magnitude, pha
se and fr
eq
ue
ncy play a key-rol
e
in interfaci
ng
grid
conne
cte
d
sy
stem
s
su
ch
asre
ne
wa
ble e
n
e
r
gy
int
egratio
n, a
c
ti
ve po
we
r filte
r
, uni
nterrupt
ed
power
sup
p
ly, dynamic vo
ltage re
sto
r
er, powe
r
qu
ali
t
y studiesetc [1-3]. Phase
Locke
d Loo
p
(PLL) i
s
wid
e
ly used to estimate the
phase
information. Several PLL sche
mes have b
e
e
n
discu
s
sed in
literature for
si
ngl
e pha
se
system
s. Th
ree mai
n
co
mpone
nts of
PLL are p
h
a
s
e
detecto
r, loop
filter and volt
age
controlle
d oscillato
r [4
]. The cla
s
sification
of PLL
is mainly b
a
sed
on the type o
f
phase
dete
c
tor u
s
e
d
, su
ch a
s
st
atio
n
a
ry (p
rod
u
ct) type and sy
nch
r
on
ou
s (P
ark
transfo
rmatio
n) type. In stationary type
phase
dete
c
tor the p
h
a
s
e e
rro
r is
calcul
ated a
s
the
prod
uct of in
put and the estimated q
u
adratu
r
e
co
mpone
nt fro
m
the PLL. Here, thedou
ble
freque
ncy
co
mpone
nts
will
be retai
ned i
n
the ph
as
e e
rro
r even
after ste
ady
state
is reached.T
h
e
low p
a
ss filte
r
intro
d
u
c
ed
to rem
o
ve the
doubl
e fre
q
u
ency
com
pon
ent, increa
se
s the
estimati
on
time of the
PLL. A modi
fied pha
se
d
e
tector [5
] d
e
sig
ned
usin
g state va
ria
b
le feed
ba
ck is
discu
s
sed to remove the d
ouble fre
que
n
c
y c
onte
n
t without usi
ng th
e low pa
ss filter.
Synchrono
us refe
ren
c
e
fra
m
e PLL
s
usi
ng Pa
rk
tra
n
sformation
type ph
ase d
e
te
ctor are
existingin lite
r
ature for
sin
g
le
ph
ase sy
stem
s. The
Park t
r
an
sformation h
e
lp
s in re
moving
the
doubl
e freq
u
ency
conten
t. Several schem
es
ca
n
be used t
o
gene
rate
the ortho
gon
al
comp
one
nts.
Delay by q
uarte
r cy
cle,
Hilbe
r
t tran
sform
a
tion,
all pa
ss filte
r
, se
con
d
o
r
der
gene
rali
sed i
n
tegrato
r
(S
OGI), inverse Park, s
lidi
ng discrete fourie
r tran
sf
orm (SDFT) and
Kalman ba
se
d PLLs a
r
e some metho
d
s
discu
s
sed
i
n
literature.Delay by quart
e
r cy
cle[6] fails
whe
n
the
r
e i
s
a d
r
ift in inp
u
t
signal
freq
u
ency. Mo
difie
d
delay
ba
se
d metho
d
s re
duces th
e e
r
rors
in the estimat
ed paramete
r
s of the PLL [7] even in
pre
s
en
ce of ha
rmonics an
d d
r
ift in frequen
c
y
con
d
ition.Hilb
ert tran
sform
a
tion ba
sed
method
s
sh
o
w
poo
r perfo
rmance wh
en
there is a d
r
ift in
freque
ncy [6]
.
All pass filter [5] is ad
ap
tive to
freque
ncy drifts, b
u
t
cann
ot be u
s
ed in
disto
r
t
e
d
con
d
ition
of the in
put
sign
al. Second
order gen
er
alized inte
grato
r
(SOGI) [8, 9]
PLL an
d inve
rse
Park PL
L me
thods
are
also ada
ptive to freque
nc
y
chang
es [1
0, 11].The pe
rfo
r
man
c
e
of these
two method
s are equivale
nt to each other [10]. In
[12] SDFT act as a pre-filt
er for DC an
d
harm
oni
cs of
the input sig
nal. Kalman
based PLL [1
3]
is also use
d
to eliminate
the harm
oni
cs
ofthe input si
gnal. Apart from the abov
e mentione
d
scheme
s
e
n
h
anced PLL (
EPLL) [14] is also
available in li
terature whi
c
h is eq
uivale
nt to
SRF-PL
L [15] and th
is metho
d
do
es not
req
u
ire
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 1, April 2016 : 49 – 60
50
addition
al ort
hogo
nal com
pone
nts to b
e
gene
rate
d. Whe
n
the DC offset is p
r
ese
n
t in the input
sign
al, a
DC
block i
s
i
n
tro
duced i
n
p
a
rallel
with
the
fundame
n
tal
block i
n
[16].
For a
ha
rmo
n
ic
rich
input
sig
nal, the in
-loo
p filter [17] a
nd pa
rallel
arrang
eme
n
t of multiple EPL
L [18] blo
c
ks
are
use
d
. All the
PLLs me
ntion
ed above
work on con
s
tant
samplin
g pe
riod.
Variable s
a
mpling period bas
ed
s
c
hem
es
like
moving average filter, SDFT and
ca
scade
d del
ayed sig
nal
can
c
ell
a
tion
use
d
as f
ilter are p
r
e
s
ent
ed [19-21] fo
r disto
r
ted g
r
id
con
d
itionin
si
ngle pha
se
systems. The
main advant
a
ge of
the va
ri
able
sa
mplin
g PLL
is that
it
gives
co
nsta
nt num
ber of
sampl
e
s pe
r
cycle
even
un
d
e
r dr
ift fr
e
q
u
e
nc
y. T
h
us
,
le
n
d
s
it
s
u
itab
le
for power q
u
a
lity applicatio
ns. In variabl
e samp
li
ng scheme
s
, the sampling p
u
lses are gen
erated
by dividing t
he ha
rd
wa
re
clo
c
k [19-20
], which
lim
its the
numb
e
r
of
sampl
e
s per
cycl
e.In this
pape
r, to ma
ke the
PLL in
depe
ndent
of hardware,
th
e sa
mpling
p
u
lse
s
a
r
e
gen
erated
thro
ug
h a
stand
ard
co
upled o
scill
a
t
or with automatic gai
n
control [21-2
2
]. Multireso
nant filters [23],
adaptive
notch filters [24]
a
nd Kalm
an fil
t
er b
a
sed
re
cursive
alg
o
rit
h
m [25]
are
u
s
ed
to e
s
tima
te
the fundame
n
tal and ha
rmonics. Co
m
posite o
b
se
rv
er
ba
sed h
a
rmonics extra
c
tion di
scussed in
[26] utilizes
simple observer
structure described
using pole placem
ent technique. These
scheme
s
u
s
e
parallel
stru
ctures fo
r esti
mation of indi
vidual com
p
o
n
ents.
In this pa
pe
r, a co
mpo
s
ite
observe
r is
use
d
to e
s
tim
a
te DC an
d h
a
rmo
n
ic
com
pone
nts
along
with t
he fund
ame
n
tal com
pon
ents to d
e
si
gn a PLL f
o
r ha
rmo
n
ically distorte
d
grid
environ
ment.
Study of PLL usin
g
si
mpl
e
ob
serve
r
h
a
s
sho
w
n tha
t
it is suitable
only whe
n
the
input si
gnal i
s
free
from
harm
oni
cs.
Compo
s
ite
ob
serve
r
ai
ds i
n
introd
uci
n
g
any numb
e
r of
blocks dep
en
ding on the
h
a
rmo
n
ic
cont
ent in the inp
u
t signal. Also there i
s
an
other a
d
vant
age
that the spe
e
d
of estimatio
n
ca
n be a
d
j
u
sted
by
cho
o
sin
g
the cl
o
s
ed lo
op p
o
le
location. T
h
us,
there i
s
a trade off between the e
s
timation time
and the ba
n
d
width fo
r the cho
s
e
n
po
le.
Simulation st
udy ha
s bee
n done fo
r variou
s
con
d
itions
su
ch a
s
sud
den
cha
n
g
e in am
plitude,
freque
ncy a
n
d
pha
se
angl
e in presen
ce of ha
rmoni
cs. T
h
is
pro
p
o
se
d sch
e
me
has
2
N
sa
m
p
les
per
cycl
e wh
ich e
nha
nces its suitability in any
sign
al processing
appli
c
ation
s
.
This
pap
e
r i
s
orga
nized a
s
follows:
se
ction 2, pro
p
o
s
es a va
ri
able
sampli
ng PL
L for a ha
rm
onically distorted
grid conditio
n
.
In section 3, transfe
r function
mo
del
of the sche
m
e is ded
uced and valid
ated
throug
h pha
se disturban
ce. Section 4,
discusse
s the re
sults
of the prop
ose
d
PLL throu
g
h
simulatio
n
an
d experim
ent.
2. Observ
e
r
Bas
e
d Varia
b
le Sampling SRF-PLL
The blo
c
k di
agra
m
of the
prop
osed va
riable
sa
mpli
ng PLL i
s
sh
own i
n
Figu
re 1. The
main
comp
o
nents
of this pro
p
o
s
ed P
LL a
r
e the
variabl
e
sampl
i
ng pe
rio
d
co
mposite
ob
se
rver
use
d
a
s
an
estimato
r an
d the sampli
ng ge
nerator
.In this
sc
heme, disc
rete time c
o
mpos
ite
observe
r
e
s
timates DC,
i
n
-pha
se and q
uadrature
axi
s
compo
nent
s of fund
ame
n
tal as
well
as
harm
oni
cs fro
m
a h
a
rm
oni
c rich p
e
ri
odi
c
input
sign
al.
The fun
d
am
e
n
tal in-pha
se
and
qua
d
ratu
re
sign
als a
r
e
transfo
rme
d
to dire
ct and qua
drature axi
s
comp
one
nts using the
Park
transfo
rmatio
n. The q
uad
rature axis compon
ent
i
s
pro
c
e
s
sed t
h
rou
gh
a PI cont
rolle
r. T
h
is
control sig
nal
(pha
se e
r
ror) is fed after
corre
c
tion in
to the sampli
ng gen
erato
r
to produ
c
e t
h
e
enabli
ng p
u
l
s
e
s
fo
r the
observe
r a
n
d
a
cou
n
ter.
The
sam
p
lin
g ge
nerator
is a
num
eri
c
ally
controlled
o
s
cillator (NCO) wih
aut
om
atic g
a
in
cont
rol
.
The
NC
O g
enerates a
consta
nt num
b
e
r
of sampl
e
s p
e
r cy
cle of the input sig
nal
indepe
nde
nt of hard
w
are
clo
ck frequ
en
cy. The co
unt
e
r
output is u
s
e
d
to ge
nerate the
u
n
it si
ne an
d
cosi
n
e
sig
nal
s for the Park tra
n
sformation.
The
comp
osit
r ob
serve
r
an
d the NCO for thi
s
PLL are
discu
s
sed bri
e
fly in the followi
ng su
bsectio
n
s.
Figure 1. Structure of p
r
op
ose
d
Varia
b
le
samplin
g SRF-PLL
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Single Pha
s
e
Variable Sa
m
p
ling Phase
Locked L
oop
using
Com
posite Ob
se
rver
(K Selvaj
yothi)
51
2.1. Compos
ite Obse
rv
er
A variabl
e
sa
mpling
pe
riod
co
mpo
s
ite
o
b
se
rver e
s
tim
a
tes th
e in
-p
hase a
nd
qu
adratu
r
e
comp
one
nts
of the inp
u
t si
gnal fo
r thi
s
Park type
PL
L. The
de
sign
of ob
se
rver f
eedb
ack g
a
in
s i
s
the sam
e
usi
ng pole
pla
c
ement techni
que in b
o
th consta
nt sam
p
ling co
mpo
s
it
e observe
r [2
6
]
and the va
ri
able sampli
n
g
com
p
o
s
ite
observer. A
n
y perio
dic
sign
al ca
n b
e
rep
r
e
s
ente
d
in
Fouri
e
r
se
rie
s
a
s
the
su
m
of intege
r m
u
ltiples
of the
fundam
ental
frequ
en
cy of the inp
u
t sig
nal.
Hen
c
e, the bl
ocks in the
compo
s
ite ob
server a
r
e
a
r
range
d in pa
rallel to estim
a
te the individual
freque
ncy co
mpone
nts.Th
e state sp
ace
model of an observe
r is gi
ven by:
(kT)
x
ˆ
C
(kT)
y
ˆ
(kT))
y
ˆ
L(y(kT)
(kT)
x
ˆ
A
1)T)
((k
x
ˆ
(1)
whe
r
e, A-syst
em matrix, L
-
observe
r fe
e
dba
ck gain
vector,
C-outp
u
t vector,
(kT)
x
ˆ
–
state
vector
and
(kT)
y
ˆ
– o
u
tput
of the o
b
serv
er.The
representati
on
of th
e state
spa
c
e
model
an
d t
he d
e
si
gn
requi
rem
ents
for simpl
e
an
d comp
osite
observe
r is ta
bulated in Ta
ble I.
Table 1. Re
prese
n
tation of state sp
ace m
odel an
d de
sign requi
rem
ents for
simpl
e
and
comp
osite o
b
s
erve
r
Simple observer
Composite obser
ver
S
y
stem Matrix (A
)
512
0.02
T
)
ω
T
cos(
α
where
α
1)
(
α
1)
(
α
α
A
1
1
1
1
1
m
m
m
m
1
1
1
1
α
1)
(
α
..
..
0
0
0
1)
(
α
α
..
..
0
0
0
:
:
:
:
:
:
:
:
:
:
:
:
:
:
0
0
..
..
α
1)
(
α
0
0
0
..
..
1)
-
(
α
α
0
0
0
..
..
0
0
1
A
n=0,1,3….m
Observer f
eedba
ck gain
vector
(
L
)
[L
11
L
12
][
L
0
L
11
L
12
..
.. L
m1
L
m2
]
Output vector
(C)
[1 0]
[1 1 0 .. .. 1 0]
State vector (
x
)
[x
11
x
12
][
x
0
x
11
x
12
.. ..
x
m1
x
m2
]
Closed loop char
acteristic
equation of the o
b
server
|zI – A + LC
| = 0
|
zI – A + LC
| = 0
Desired pole location
-w
=a
±
b
b=
w
w-
aT
11
1
ze
(
j
)
w
her
e
s
i
n
(
T
)
f
undam
ent
al
f
r
equenc
y
aT
aT
01
1
aT
mm
ze
,
e
(
j
)
,
.
...
e
(
j
)
-w
-w
-w
=a
a
±
b
a±
b
Desired characte
ristic
equation
0
))
j
β
(
α
e
))(z
j
β
(
α
e
(z
1
1
T
a
ω
1
1
T
a
ω
0
))
j
β
(
α
e
))(z
j
β
(
α
e
......(z
))...
j
β
(
α
e
))(z
j
β
(
α
e
).(z
α
e
-
(z
m
m
T
a
ω
m
m
T
a
ω
1
1
T
a
ω
1
1
T
a
ω
0
T
a
ω
Comp
ari
ng t
he coefficie
n
t
s of the o
b
s
erve
r
cha
r
a
c
teri
stic
equ
ation with
th
at of the
desi
r
ed
ch
aracteri
stic
equ
ation, the fee
dba
ck g
a
in
v
e
ctor
L is o
b
t
ained. He
re
‘a’ deci
d
e
s
the
observe
r pol
e location in
the z-plan
e. The st
ru
ct
ure of the co
mposite
ob
server i
s
sho
w
n in
Figure 2.
Th
e n
th
bl
ock
of this
ob
se
rver i
s
p
r
e
s
e
n
ted in
Fig
u
r
e
3. The
e
qual m
agnitu
de
quad
ratu
re compon
ent is
obtaine
d by multiplying the quad
ratu
re
sign
al with a
gain
n
n
n
β
1)
(
α
g
.
Let G
0
, G
1
, G
2
,………G
m
be the tran
sf
er fun
c
tion of
individual bl
ocks in th
e compo
s
ite
observe
r. He
nce, th
e transfer fun
c
tion
of
the fun
dame
n
tal blo
c
k in
case
of
simple
and
compo
s
i
t
e
observe
r are
(z)
G
1
(z)
G
1
1
and
m
0,1,...
n
n
1
)
(
G
1
(z)
G
z
respe
c
tively. The mappin
g
of po
les in
z-plan
e for the
comp
osite
ob
serve
r
and
the ma
gnitud
e
re
sp
on
se
o
f
the fund
am
ental bl
ock i
n
the
co
mpo
s
ite
observe
r de
si
gned fo
r 50
Hz with th
e pol
e location
correspon
ding to
‘a’ = 0.2
and
1are
shown i
n
Figure 4 a
n
d
Figure 5 respectively. Ma
gnitude
re
sp
onse sho
w
s t
hat the fund
a
m
ental blo
c
k
in
the pa
rallel
st
ructu
r
e
acce
p
t
s funda
ment
al freq
uen
cy and reje
cts al
l
other ha
rmo
n
ic
frequ
en
cies.
Also a
s
the
pole
s
move far interi
or to
the unit circle the ban
d width de
crea
se
s, re
sulting
in
sha
r
pe
r tunin
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 1, April 2016 : 49 – 60
52
Figure 2. Structure of
comp
osite
ob
se
rve
r
Figure 3. Structure of n
th
bl
ock in the co
mposite
observe
r
Figure 4. Poles of co
mpo
s
ite obse
r
ver i
n
z-
plane
corre
s
p
ondin
g
to ‘a’ = 0.2 and 1
Figure 5. Magnitude pl
ot of the fundame
n
tal block in
the comp
osit
e observe
r for the pole lo
cations
corre
s
p
ondin
g
to ‘a’ = 0.2 and 1
2.2. Samplin
g Gener
a
tor
The NCO i
s
u
s
ed a
s
the
sa
mpling g
ene
rator to create
enablin
g pul
s
e
s
for the
co
mposite
observe
r and
to generate the add
re
ss bi
ts for the lo
ok up table to prod
uce unit sine an
d co
si
ne
sign
als a
s
sh
own
in
Figu
re 1. T
he f
r
eq
uen
cy of
the
output sign
al gene
rated
by
the NCO
i
s
an
integer m
u
ltiple of frequ
e
n
cy of the in
put sign
al
. T
h
is p
r
op
erty make
s it suitable for va
ria
b
le
sampli
ng sch
e
mes. Th
e st
ructu
r
e of NCO gene
rating
sampl
e
pul
se
s is sho
w
n in Figure 6. Here,
NCO is
de
rived from
a sta
ndard
coupl
e
d
oscillato
r [2
1-22]. Th
e
state sp
ace m
odel of the
NCO
with automati
c
gain
control
is:
)
(kT
x
)
(kT
x
2
β
1
β
β
2
β
1
G
)
1)T
((k
x
)
1)T
((k
x
0
2
0
1
2
o
o
o
2
o
NCO
0
2
0
1
(2)
w
h
er
e
))
(kT
x
)
(kT
(x
2
3
G
o
2
2
o
2
1
NC
O
(3)
)
T
sin(
ω
β
o
o
o
(4)
oo
o
NCO
1
wher
e,
ω
2
π
fa
n
d
T
f
==
f
NCO
–
the enabling fre
que
ncy of the NCO (=5MHz),
f
o
, oscillator o
u
tput freque
n
c
y = N.f
f, input signal
frequen
cy,
x
1
and x
2
are the state vari
able
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Single Pha
s
e
Variable Sa
m
p
ling Phase
Locked L
oop
using
Com
posite Ob
se
rver
(K Selvaj
yothi)
53
When the in
put sign
al fre
quen
cy is
∆ω
ω
, th
e input to the oscillator o
r
the co
ntrolle
r
ouput
cha
nge
s t
o
)
)T
∆ω
si
n((
ω
o
o
o
so as to gen
erate N (51
2
) ti
mes the inp
u
t signal fre
q
ue
ncy.
Hen
c
e,
)
)T
ω
sin((
ω
β
β
o
o
o
o
o
(5)
The vari
ation
of
o
∆β
with res
p
ec
t
to
∆
f
is lin
ear. F
r
om
(5
), it is
clea
r
that unde
r d
r
ift in
freque
ncy, a
corre
c
tion n
e
ed to be pro
v
ided at t
he output of the
controller. T
h
is corre
c
tio
n
fac
t
or (K
cor
)
[21] is cal
c
ul
ated on the
assumption
that as
∆ω
o
0, cos(
∆ω
o
T
o
)
0,
sin(
∆ω
o
T
o
)
∆ω
o
T
o
.
ω
T)
(
∆
K
∆β
))
T
)(cos(
ω
)(T
T
N
ω
T)(
(
∆
)
T
.cos(
ω
T
∆ω
∆β
becomes,
(5)
Hence
co
r
o
o
o
o
o
o
o
o
o
(6)
From (6), correction fa
ctor i
s
obtain
ed a
s
))
T
)(cos(
ω
T
NT
(
K
o
o
o
cor
(7)
whe
r
e, T is th
e sampli
ng p
e
riod.
The PI controller
output
con
s
i
s
ts of
b
o
th pha
s
e
a
nd fre
que
ncy
informatio
n.
(f
+
∆
f)
is
dedu
ce
d fro
m
the inte
gral pa
rt of th
e controll
er
as
sh
own in
Figu
re
6, which
contain
s
the
freque
ncy inf
o
rmatio
n. As the frequen
cy informatio
n
is a sinusoi
dal quantity, the freque
ncy is
cal
c
ulate
d
usi
ng inverse si
ne functio
n
.
)
)T
ω
si
n
(
(
ω
β
β
y
Here,
o
o
o
o
o
(8)
)
6
y
(y
NT
2
π
1
∆
f
f
Hence,
3
o
(9)
The first two
terms of inve
rse
sine
seri
e
s
is con
s
ide
r
ed in (9) a
n
d
this cal
c
ulat
es the freq
ue
ncy
with an ac
curac
y
of ±0.02mHz
for a range of 40Hz
to 70Hz
.
Figure 6. NCO based sam
p
le pul
s
e ge
n
e
ration
3. Transfer F
unction Mo
d
e
l
This se
ction deal
s
with
th
e
analy
s
is of
the propo
se
d PLL by d
e
duci
ng the transfe
r
function m
o
d
e
l in continuo
us time
with two diffe
rent
estimato
rs-si
m
ple ob
se
rve
r
and
co
mpo
s
ite
observe
r. A step ch
ang
e in
pha
se i
s
give
n to the
s
e m
o
dels
and
the t
i
me re
sp
on
se
of these PLL
s
is studi
ed. Symmetrical opt
imum pro
c
e
d
u
re [10]
is u
s
ed to desi
gn the co
ntrolle
r
of PLL.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 1, April 2016 : 49 – 60
54
3.1. Simple
Observ
e
r as an Estimato
r
The ob
serve
r
is modelled
as a
fi
rs
t order s
y
s
t
em
with time c
o
ns
tant
a
ω
1
obs
. The
transfe
r fu
ncti
on m
odel
of t
he o
b
serve
r
based va
riabl
e samplin
g P
LL i
s
sho
w
n
i
n
Fig
u
re
7. T
he
simulatio
n
st
udy of this
model i
s
don
e for a time
con
s
tant
50
2
π
1
obs
. Figure
8 shows the
pha
se erro
r vs time for the PLL and
the transfe
r functio
n
mod
e
l for three d
i
ff
ere
n
t values of
pha
se ma
rgin
s su
ch a
s
30
, 45
and 60
res
p
ec
tively.
The tra
n
sfe
r
function m
o
d
e
l is de
rived
based on th
e assum
p
tion
that
ω
is co
nstant.
Whe
n
the ste
p
cha
ngei
n p
hase angl
e is given to
bot
h the tran
sfer function m
o
d
e
l and the PL
L,
initial transi
e
nt in freque
ncy deviates from theno
minal value and it is proportio
n
al to
th
e
controlle
r
g
a
i
n
s.
T
he co
ntroller gain
s
are
hig
her
with
lowe
r
p
h
a
s
e margi
n
(PM
)
and hen
ce
m
o
re
deviation is
seen in the re
spon
se
of the PLL com
pare
d
to the
mode
l as sh
own in Figure 8.
Figure 7. Tra
n
sfer fu
nction
model of
an ob
serve
r
b
a
se
d SRF-P
L
L
Figure 8. Simulation re
sult
s wh
en a ste
p
cha
nge in
pha
se of 40
i
s
appli
ed to transfe
r
functio
n
model an
d
the PLL (a) P
M
=30
(b
) PM=45
(c
)
PM=
6
0
3.2. Compos
ite Obse
rv
er
as an Estima
tor
Whe
n
DC an
d ha
rmo
n
ic compon
ents a
r
e p
r
e
s
e
n
t al
ong
with th
e
fundam
ental
blo
ck,
comp
osite o
b
s
erve
r is a
pproximated as
a first
ord
e
r
system having
time con
s
tan
t
given by:
N
m
3
m
dc
dc
ob
s
0.5
H
;
m
1
H
H
wh
e
r
e
;
a
ω
aH
1
(10
)
The respon
se of the v
a
riabl
e sam
p
ling
PLL
a
n
d the tran
sfer fu
nctio
n
model
areillustrated in Figure
9
for a phase change of 40
. The
controll
er i
s
de
sign
ed usin
g
symmetri
c
al
optimump
r
o
c
edure. The
p
r
opotio
nal int
egral
(PI)
co
e
ffi
cient
s h
a
ve bee
n evalu
a
ted
for di
ff
e
r
ent
pha
se ma
rgi
n
s (30
, 45
and 60
). He
re, the mod
e
l
closely follo
ws the
syste
m
.
Comp
osite
o
b
s
erve
r ba
se
d
PLL gives a
sluggi
sh
re
spon
se comp
ared to the
simple
ob
server
becau
se of the pre
s
en
ce of
DCa
nd ha
rm
onics blo
c
ks
along
with the fundame
n
ta
l.
Figure 9. Simulation re
sult
s wh
en a ste
p
cha
nge in p
h
ase of 40
is applied to trans
f
er func
tion
model an
d variable
sam
p
ling PLL. (a) P
M
=30
(b
) PM=45
(c
)
PM=
6
0
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Single Pha
s
e
Variable Sa
m
p
ling Phase
Locked L
oop
using
Com
posite Ob
se
rver
(K Selvaj
yothi)
55
4. Results a
nd Discu
ssi
on
The compo
s
i
t
e obse
r
ver
based varia
b
l
e sam
p
ling
PLL is si
mul
a
ted usi
ng
MATLAB
Simulink a
n
d
the re
sults
are ve
rified
experim
ent
all
y
by implementing the
same u
s
ing
DSP
Builder to
ol in
Altera
Cyclo
ne IV FPGA.
A simple
ob
server
as well
as a
compo
s
i
t
e observe
r
with
clo
s
ed lo
op p
o
les lo
cate
d at z = e
−
a
ω
T
in z-domai
n with ‘a’=1 a
r
e
use
d
as e
s
ti
mators in the
PLL
for a digitally
synthe
sized
harm
oni
c ri
ch
input sig
nal
as
well a
s
fo
r a real time
grid voltag
e
o
f
slo
w
ly drifting
frequen
cy.
4.1. Perform
a
nce o
f
Simple Observ
e
r Based PLL
Initially a sinu
soid
al si
gnal
of magnitud
e
unity
at 50Hz
is given to thi
s
PLL. At 0.2
s
, a DC
offset of 10
% is introdu
ced
to the
si
nusoidal i
npu
t sign
al. The
n
at 0.3
s
, th
e input
sig
n
a
l
is
swit
che
d
sud
denly to
a
h
a
rmo
n
ic ri
ch
sign
al of
THD 4
5
% h
a
vin
g
od
d h
a
rm
o
n
ics u
p
to 1
5
th
at
50Hz. Figu
re
10 re
present
s si
mulation
result of
e
s
tim
a
ted Vd, drift
in frequ
en
cy and p
h
a
s
e e
r
ror
of
thisPL
L.Wi
th
a sin
u
soid
al
inp
u
t
si
gn
al,
sim
p
le ob
serve
r
estim
a
tes pa
ramet
e
rs without a
n
y
error. But fo
r a ha
rmo
n
ic
rich in
put
sign
al, rippl
e is seen in
the e
s
timated value
s
, inferring
th
at
noise creep
s
into the syste
m
. The increa
se of filt
ering
cap
ability by moving pole
s
closer to o
r
ig
in
in z-pl
ane
wil
l
make the resp
on
se mo
re slug
gish.
Another fa
ct is that the simple ob
serve
r
is
estimating o
n
l
y fundament
al orthog
onal
comp
one
nts
and hen
ce, fo
r a harm
oni
c rich in
put sig
nal
this PLL
fail
to e
s
timate
the p
a
ram
e
ters
efficientl
y
. This n
e
ce
ssitate
s th
e
requi
rem
ent
o
f
estimating
DC a
nd in
dividual
harm
oni
c
comp
one
nts, which is p
o
ssible
thro
u
gh a
co
mpo
s
ite
observe
r. Fo
r the
sam
e
pol
e lo
cation,
he
re a
s
the
filtering cap
ability is
in
crea
sed
-
becau
se of
th
e
parall
e
ling of
simple o
b
se
rver blo
c
ks,
modelle
d for all the harm
onics in the
input sig
nal-t
he
estimated
pa
rameters
are
conve
r
gin
g
to
the d
e
si
re
d
value a
s
sh
o
w
n i
n
the
tra
n
sie
n
t resp
on
se
s
unde
r subse
c
tion 4.2. Wit
h
the sinu
soi
dal i
nput the
PLL estimat
e
s ma
gnitud
e
, freque
ncy
and
pha
se exactl
y where
a
s wi
th a 10% DC offset, the estimation g
oes o
s
cillatin
g
about a mean
value. W
hen
a h
a
rm
oni
c ri
ch
sign
al
is fed, th
e
PLL give
s
more
o
scill
atory respon
se
for
magnitud
e
a
nd frequ
en
cy. The DC off
s
et ha
s more
deterio
rating
effect on est
i
mation of ph
ase
than
with h
a
rmonics for thi
s
simpl
e
o
b
server ba
se
d
PLL. Here
th
e controller g
a
ins a
r
e
ch
osen
as K
p
= 13
0 a
nd K
i
=70
14.
Figure 10. Simulation resp
onse of estim
a
ted V
d
, drift i
n
freque
ncy a
nd pha
se e
r
ror wh
en
sinu
soi
dal an
d a harm
oni
c rich
sign
al are applie
d to the simpl
e
ob
serve
r
ba
se
d PLL
4.2. Perform
a
nce o
f
Com
posite Ob
ser
v
e
r
Based P
LL
In the follo
wi
ng
sub
s
e
c
tio
n
s th
e p
e
rfo
r
mance of
the
PLL
usin
g th
e compo
s
ite
observe
r
as estimato
r
i
s
studie
d
by giving
a step
cha
nge
in a
m
plitude, fre
q
u
ency
and
pha
se
of the in
p
u
t
sign
al. The controlle
r gain
s
are
cho
s
en
as K
p
= 100 and K
i
=3
500.
The experim
ental re
sults
are
coin
cid
ent wit
h
the simulati
on re
sults.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 1, April 2016 : 49 – 60
56
4.2.1. Step Change in Am
plitude
A 40% sag i
n
the harm
o
n
i
c rich input
signal (o
dd ha
rmoni
cs upto
15
th
) is prod
uce
d
at
0.2s and
restored
afte
r
0.4s. Fi
gure 11 (a)
and
(b) illustrates t
he
res
ponse of the variable
sampli
ng PL
L
throu
gh
sim
u
lation a
nd e
x
perime
n
t re
spectively. Wh
en the a
m
plit
ude of
sign
al
is
cha
nge
d the
PLL e
s
timate
s the
mag
n
itude of V
d
with
in a
cycle, freque
ncy ove
r
sho
o
t is
0.25
Hz
and pea
k ph
ase e
rro
r is a
bout 3.5
. Thi
s
estimatio
n
spe
ed is a
c
hi
eved with the
inclusi
on of DC
as
well a
s
ha
rmoni
cs by the estimato
r.Under
stea
dy
state all the pa
ramete
rs
are
confin
ed to th
e
desi
r
ed val
u
e
,
which is
cle
a
r fro
m
both t
he sim
u
la
tion
and exp
e
rim
ental re
sult
s
sho
w
n i
n
Fig
u
re
11.
4.2.2. Step Change in Fre
quency
For the
same
harmo
nic
rich input sig
nal
wi
th frequ
en
cy 47.5Hz is
fed to the PLL and
sud
denly
ch
a
nged
to 5
2
.5
Hz at 0.
2s a
nd
re
store
d
t
o
47.5
H
z aft
e
r
0.4s.
Figu
re 12
(a) and
(b
)
sho
w
s the i
n
put sig
nal, di
rect
axis volt
age, d
r
i
ft freq
uen
c
y and
p
h
ase e
rro
r th
roug
h
simulat
i
on
and exp
e
rim
e
nt. These results reveal th
at the PLL
i
s
estimating
th
e freq
uen
cy i
n
2.5
cycle
s
and
pea
k ph
ase d
e
viation of 19
.5
for +5
Hz step
cha
nge. For
a step
ch
ange
of -5
Hz
the freq
uen
cy is
estimated
in
3.3 cy
cle
s
an
d the m
a
ximu
m pha
se
erro
r is ab
out 20.
5
. Also
a
sm
oother tran
si
ent
can be
see
n
in the result
s with step d
i
sturb
a
n
c
e fo
r simulatio
n
and expe
rim
ent. Steady state
values a
r
e co
nvergin
g
to the desi
r
ed val
ues.
Figure 11(a).
Simulation re
spo
n
se of PLL on
estimation of
V
d
, drift frequency an
d pha
se
error with 4
0
%
sag at 0.2s and rein
state
d
at
0.4s
Figure 11(b).
Simulation re
spo
n
se of PLL on
estimation of
V
d
, drift frequency an
d pha
s
e erro
r
with 40% sa
g
at 0.2s and reinstate
d
at 0.4s
Figure 12(a).
Simulation re
spo
n
se of PLL on
estimation of
V
d
, drift frequency an
d pha
se erro
r
for step
chan
ge in frequ
en
cy from 47.5
H
z to
52.5Hz at 0.2
s
and rein
stated after 0.4
s
Figure 12(b).
Simulation re
spo
n
se of PLL on
estimation of
V
d
, drift frequency an
d pha
s
e
error for
step
cha
nge in fre
q
uen
cy from
47.5Hz to 52.
5
Hz at 0.2s a
nd rein
stated
after 0.4s
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Single Pha
s
e
Variable Sa
m
p
ling Phase
Locked L
oop
using
Com
posite Ob
se
rver
(K Selvaj
yothi)
57
The fre
que
ncy of this ha
rm
onic
rich si
gn
al is vari
ed in
step
s of 10
Hz from
40
Hz t
o
70
Hz
and the
estim
a
ted pa
ram
e
ters are ob
se
rved. Figur
e 1
3
illustrates the PLL
de
sig
ned fo
r 50
Hz is
workin
g effici
ently over a wide rang
e o
f
frequen
cy. Hen
c
e, this schem
e co
uld
be used for a
n
y
stand
ard g
r
id
freque
ncy.
4.2.3. Step Change in Ph
ase An
gle
A step cha
n
g
e
in phase of 40
is given to
this harmo
ni
c rich input si
gnal at 50Hz
and the
PLL is evalu
a
ted throu
g
h
simulation a
nd experim
e
n
t. The resul
t
s for estima
ted V
d
, drift
in
freque
ncy an
d pha
se erro
r are sh
own in Figure 14 (a) and
(b). T
he estimatio
n
of phase an
gle
occurs i
n
2.8
3
cy
cle
s
. The
pea
k p
h
a
s
e
overshoot i
s
about 1
8
.95
and p
e
a
k
fre
quen
cy devia
tion
is 4.25
Hz.
4.2.4. DC O
f
fset
A DC offset o
f
0.5Vis given to this harmo
ni
c ri
ch inp
u
t sign
al at 50Hz and the si
m
u
lation
and
experi
m
e
n
tal re
sp
on
se
s of th
e PL
L f
o
r e
s
timate
d
V
d
, drift in fre
quen
cy a
nd
p
hase e
r
ror
are
sho
w
n i
n
Fig
u
re
15
(a) an
d (b
) respe
c
tively. There i
s
no e
r
ror i
n
p
hase an
d fre
quen
cy e
s
tim
a
ted
by this PLL under
steady state.
4.3. Effec
t
of Unmodelled
Harmonics
A similar ha
rmonic
rich sig
nal co
ntainin
g
odd ha
rmo
n
ics upto 25
th
with THD
= 4
5
.22%
is
fed to a
com
posite
ob
se
rver mo
delle
d
with DC
and
odd h
a
rm
oni
cs upto
15
th
.
The u
n
mod
e
ll
ed
harm
oni
cs (T
HD = 4.47%)
pre
s
e
n
t
in
th
e
sig
nal are
17
th
, 19
th
, 21
st
, 23
rd
and
25
th
of 2% ea
ch.
The e
s
timate
d pa
ram
e
ters by the P
LL
unde
r
steady
state i
s
sh
o
w
n i
n
Ta
ble
2 for two
different
pole lo
cation
s of the ob
se
rver with ‘
a
’ = 0.5 an
d
1. The re
sult
s show that a
s
‘
a
’ red
u
ces, the
accuracy
of the e
s
timated
paramet
e
r
s i
n
cr
ea
se
s. Thi
s
o
c
c
u
r
s
with
a co
mpr
o
mi
se o
n
s
pee
d
of
estimation.
Figure 13. Simulation resp
onse of PLL on estimatio
n
of V
d
, drift frequen
cy and p
hase error fo
r
freque
ncy ch
ane in ste
p
s
of 10Hz from
40Hz to 70
Hz
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 1, April 2016 : 49 – 60
58
Figure 14(a).
S
imulation re
spo
n
se of PLL on
estimation of
V
d
, drift in fre
quen
cy and p
hase
error when a
step chan
ge i
n
pha
se an
gl
e of
40
o
c
curs at t=0.2s
Figure 14(b).
Simulation re
spo
n
se of PLL on
estimation of
V
d
, drift in fre
quen
cy and p
h
ase
error when a
step chan
ge i
n
pha
se an
gl
e of 40
occurs at t=0.
2s
Figure 15(a).
Simulation re
spo
n
se of PLL
on estimatio
n
of V
d
, drift in
freque
ncy an
d
pha
se erro
r when a ha
rmo
n
ic ri
ch si
gnal
is
applie
d to co
mposite o
b
se
rver ba
se
d PLL
for DC
offset of 0.5V
Figure 15(b).
Simulation re
spo
n
se of PLL on
estimation of
V
d
, drift in fre
q
uen
c
y and p
hase
error when a
harm
oni
c rich
signal i
s
appl
i
ed to
comp
osite o
b
s
erve
r ba
se
d PLL for DC offset of
0.5V
Table 2. Error in Estimated Paramete
rs o
f
Co
mpo
s
ite Observe
r
Based PLL in Pre
s
en
ce of
Unmo
delle
d Harmoni
cs
Pole location w
i
th ‘a‘= 0.5
Pole location w
i
th ‘a’ = 1
Peak Phase Erro
r
0.006
0.015
Peak Freque
nc
y
Error
0.446mHz
1mHz
Magnitude Error
0.33%
1.5%
4.4. Perform
ance o
f
the
PLL
w
i
th a Slo
w
ly
Drifting Frequen
c
y
The voltage
across a
re
ctifier type ca
pacitive no
nli
near l
o
ad
(3
6
|| 2200
F) wh
en
con
n
e
c
ted to a grid of red
u
ce
d voltage
90V
p-p
through a 4mH in
ducto
r is u
s
e
d
to validate the
perfo
rman
ce
of the PLL. The third h
a
rm
onic
co
m
pon
ent is predo
m
inant (3.4%
)
alon
g with
DC
(0.7%) a
nd fundam
ental (100%) in thi
s
gri
d
voltag
e having T
H
D of 3.9%. The e
s
timate
d
magnitud
e
, freque
ncy a
n
d
pha
seof the
gridvoltag
e
with a
simpl
e
ob
serve
r
and
comp
osi
t
e
observe
r based PLL are
sho
w
n in Fig
u
re 16 (a) a
n
d (b)
re
spe
c
tively. When
the actual g
r
id
freque
ncy i
s
drifting sl
owly
from 50
Hz, t
he e
s
ti
mated
freque
ncy fro
m
the simpl
e
observe
r
ba
se
d
PLL is fo
und
to have ri
pple
s
an
d e
s
timat
ed fre
quen
cy
by the co
mpo
s
ite ob
se
rver
modelle
d wit
h
m=0, 1 a
nd 3
in the PLL i
s
50.04
Hz. As t
he PLL i
s
u
s
e
d
to estimate
the frequ
en
cy by measuri
n
g
voltage from the grid, this e
s
timator
coul
d be
modell
e
d with the req
u
ired n
u
mb
er of blocks.
Evaluation Warning : The document was created with Spire.PDF for Python.