
86-11Spectrum Analysis and Correlation
at the beginning and end of the signal and, therefore, reduces its eective signal duration. is eect is
re
ected
b
y
t
he
e
quivalent
n
oise
b
andwidth
(
ENBW)
d
ened
a
s
[
19]
ENBW =
=
−
=
−
∑
w
w
n
n
N
n
n
N
2
0
1
0
1
2
(86.21)
For
a re
ctangular
wi
ndow
EN
BW
= 1.0 an
d
fo
r
th
e
Bl
ackman–Harris
wi
ndow
(E
quation
86.20
),
th
e
co
r-
responding
v
alue
i
s
2.0.
e
m
ajority
o
f
o
ther
w
indow
s
hapes
h
ave
i
ntermediate
v
alues
o
f
E
NBW
[
19]
.
86.4.2 Effects of “Zero Padding”
Most FFT algorithms operate with N = 2
q
samples, the choice of q is many times critical. Since frequency
re
solution
is in
versely
pr
oportional
to N, in ma
ny
ci
rcumstances