A different signal set from the previous two uses the bit to
affect the frequency of a carrier sinusoid.
s
0
0=A
p
T
tsin2π
f
0
t
s
0
0
A
p
T
t
2
π
f
0
t
(1)
s
1
0=A
p
T
tsin2π
f
1
t
s
1
0
A
p
T
t
2
π
f
1
t
(2)
The frequencies
f
0
f
0
,
f
1
f
1
are usually harmonically related to the bit
interval. In the depicted example,
f
0
=3T
f
0
3
T
and
f
1
=4T
f
1
4
T
. This signal set is known as
frequency shift keying or FSK. As can be seen
from the transmitted signal for our example bit stream (Figure 2),
the transitions at bit interval boundaries are smoother than those of BPSK.
To determine the bandwidth required by this signal set, we again
consider the alternating bit stream. Think of it as two signals added
together: The first comprised of the signal
s
0
t
s
0
t
, the zero signal,
s
0
t
s
0
t
, zero, etc., and the
second having the same structure but interleaved with the first
(Figure 3).
Each component can be thought of as a fixed-frequency sinusoid
multiplied by a square wave of period
12T
1
2
T
that
alternates between one and zero. This baseband square wave has the
same Fourier spectrum as our BPSK example, but with the addition of
the constant term
c
0
c
0
. This
quantity's presence changes the number of Fourier series terms
required for the 90% bandwidth: Now we need only include the zero and
first harmonics to achieve it. The bandwidth thus equals, with
f
0
<
f
1
f
0
f
1
,
f
1
+12T−(
f
0
−12T)=
f
1
−
f
0
+1T
f
1
1
2
T
f
0
1
2
T
f
1
f
0
1
T
. This bandwidth is larger than
the BPSK bandwidth.
Show that the bandwidth for FSK is always larger. What is
the smallest it can be?
We want to show that
f
1
−
f
0
+1T>32T
f
1
f
0
1
T
3
2
T
,
which means that
f
1
−
f
0
>12T
f
1
f
0
1
2
T
must be true. As we have
f
0
=
k
0
T
f
0
k
0
T
,
f
1
=
k
1
T
f
1
k
1
T
,
the smallest difference between the two frequencies is
1T
1
T
, giving a transmission
bandwidth of
2T
2
T
or
2R
2
R
in
terms of data rate.