Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Distributed
Electrical
Systems
Laboratory
cole
Polytechnique
Fdrale
de
Lausanne
-
EPFL
6/3/15
Recall
x(t)
CONTINUOSTO-DISCRETETIME
CONVERSION
x(n)
SIGNAL
WINDOWING
w(n) x(n)
DISCRETE
FOURIER
TRANSFORM
X(k)
SPECTRUM
ANALYSIS
freqeuncy
magnitude
phase
6/3/15
Outline
IntroducLon
IpDFT
formulaLon
OpLmal
selecLon
of
the
IpDFT
parameters
IpDFT-based
synchrophasor
esLmaLon
algorithm
6/3/15
IntroducUon (1)
f = k MAX f
A = X(k MAX )
= X(k MAX )
6/3/15
IntroducUon (2)
( )
( )
6/3/15
IntroducUon
(3)
DFT
interpola7on
In
order
to
improve
the
accuracy
of
this
trivial
synchrophasor
esLmator
there
are
several
opLons:
1. Decrease
the
sampling
frequency
(see
previous
formula)
CONS:
aliasing
may
arise
2. Increase
the
window
length
(i.e.,
improve
the
frequency
resoluLon)
CONS:
higher
number
of
samples
to
be
processed
->
higher
computaLon
Lme
CONS:
in
order
to
reach
high
accuracy
levels
very
long
windows
are
needed
3. DFT
interpolaUon
(see
References)
A
more
exact
esLmaLon
of
the
main
spectrum
tone
locaLon
can
be
given
by
calculaLng
the
abscissa
of
the
maximum
of
an
interpolaLon
curve
of
the
DFT
spectrum
6/3/15
6/3/15
To
be
noLced
that
kpeak
can
also
be
interpreted
as
the
number
of
acquired
cycles
of
the
input
signal.
Therefore
if
kpeak = kMAX
(=0)
-> coherent
sampling
(NO
leakage)
Based
on
the
number
of
DFT
bins
used
to
Perform
the
interpolaLon,
IpDFT
algorithms
may
be
separated
in
2-points,
3-points,
DFT
interpolators.
6/3/15
The
spectrum
of
s(n)
can
be
expressed
in
terms
of
the
posiLve
and
negaLve
images
of
the
main
frequency
tones:
A
A
X( ) = 0 e j W ( 0 ) + 0 e j W ( + 0 )
2
2
being
W()
the
Fourier
transform
of
the
selected
window
funcLon.
Assuming
that
the
eects
of
leakage
are
properly
compensated
by
windowing,
we
can
neglect
the
long-range
spectral
leakage
produced
by
the
negaLve
spectrum
image
and
express
the
2
highest
bins
of
the
DFT
X(kMAX)
and
X(kMAX+), =1
as
a
funcLon
of
the
posiLve
spectrum
image
only:
X ( kMAX + ) X ( k + ) W ( k + 0 )
=
=
=
X ( kMAX )
X ( k )
W ( k 0 )
W ( 0 + ( ) 2 / N 0 ) W (( ) 2 / N )
=
=
W ( 2 / N )
W ( 0 2 / N 0 )
where
the
only
unknown
is
.
0
MAX
MAX
MAX
MAX
6/3/15
10
=
=
=
WR ( 2 / N )
sin ( / N + / N ) sin ( )
/ N + / N sin ( ) 1
X ( k MAX )
we
can
express
delta
as
X(k MAX + )
=
X(k MAX ) + X(k MAX + )
and
the
signal
parameters
as:
f=
( kMAX + ) f
A=
X(k MAX )
= !X(k MAX )
sin( )
(N 1)
N
6/3/15
11
WH ( 2 / N )
2
X ( k MAX )
and
express
delta
as:
2 X(k MAX + ) X(k MAX )
=
X(k MAX ) + X(k MAX + )
and
the
signal
parameters
as:
f=
( kMAX + ) f
A = X(k MAX )
2 1
sin( )
= !X(k MAX )
(N 1)
N
6/3/15
12
Input signal
x(t)
Signal
sampling
x(n)
Signal
buering
x(n), n 2 [0, N
Signal
windowing
w(n) x(n)
DFT
1]
6/3/15
13
Input signal
Hanning window:
2 n
w ( n ) = 0.5 1 cos
, n [ 0, N 1]
N 1
x(t)
Signal
sampling
x(n)
Signal
buering
x(n), n 2 [0, N
Signal
windowing
w(n) x(n)
DFT
1]
14
n 2 [0, N 1]
IpDFT-x(n),
based
synchrophasor
esUmaUon
(2)
Signal
windowing
w(n) x(n)
DFT
N 1
2 X
X(k) =
w(n) x(n) WNkn , k 2 [0, N
B n=0
DFT maximum
identification
{X(k)}, k = kM AX + { 1, 0, +1}
DFT
Interpolation
f, A, '
Synchrophasor
estimation
1]
15
n 2 [0, N 1]
IpDFT-x(n),
based
synchrophasor
esUmaUon
(2)
Signal
windowing
w(n) x(n)
DFT
X(kMAX)
N 1
2 X
X(k) =
w(n) x(n) WNkn , k 2 [0, N
B n=0
DFT maximum
identification
{X(k)}, k = kM AX + { 1, 0, +1}
DFT
Interpolation
f, A, '
Synchrophasor
estimation
1]
X(kMAX+)
16
n 2 [0, N 1]
IpDFT-x(n),
based
synchrophasor
esUmaUon
(2)
Signal
windowing
DFT inteprola7on
w(n) x(n)
DFT
X(kMAX)
N 1
2 X
X(k) =
w(n) x(n) WNkn , k 2 [0, N
B n=0
DFT maximum
identification
{X(k)}, k = kM AX + { 1, 0, +1}
DFT
Interpolation
f, A, '
Synchrophasor
estimation
1]
X(kMAX+)
6/3/15
17
References
1. V.
K.
Jain,
W.
L.
Collins,
and
D.
C.
Davis,
High-accuracy
analog
measurements
via
interpolated
FFT,
InstrumentaLon
and
Measurement,
IEEE
TransacLons
on,
vol.
28,
no.
2,
pp.
113122,
1979.
2. T.
Grandke,
InterpolaLon
algorithms
for
Discrete
Fourier
Transforms
of
weighted
signals,
InstrumentaLon
and
Measurement,
IEEE
TransacLons
on,
vol.
32,
no.
2,
pp.
350355,
1983.Andria
Savino
etc
3. P.
Romano
and
M.
Paolone,
Enhanced
Interpolated-DFT
for
Synchrophasor
EsLmaLon
in
FPGAs:
Theory,
ImplementaLon
and
ValidaLon
of
a
PMU
Prototype,
IEEE
Trans.
On
InstrumentaLon
and
Measurements,
vol.
63,
num.
12,
pp.
2824
-
2836,
Dec.
2014.