
270
PROCEEDINGS
OF
THE
IEEE,
MARCH
1973
B.
Intersymbol Interference
In digital transmission through analog channels, we fre-
quently encounter the following situation. The input se-
quence
u,
discrete-time and discrete-valued
as
in the shift-
register model, is used to modulate some continuous waveform
which is transmitted through
a
channel and then sampled.
Ideally, samples
zk
would equal the corresponding
Uk,
or some
simple function thereof; in fact, however, the samples
,zk
are
perturbed both by noise and by neighboring inputs
Up.
The
latter effect is called intersymbol interference. Sometimes
intersymbol interference is introduced deliberately for pur-
poses
of
spectral shaping, in so-called partial-response sys-
tems.
In such cases the output samples can often be modeled
as
zk
=
yk
+
nk
where
yk
is
a
deterministic function of
a
finite number of in-
puts, say,
yk
=f(Uk,
* *
.
,
Uk-y),
and
nk
is
a
white Gaussian
noise sequence. This is precisely Fig.
2.
To
be still more specific, in pulse-amplitude modulation
(PAM)
the signal sequence
y
may be taken
as
the convolu-
tion
of
the input sequence
u
with some discrete-time channel
impulse-response sequence
(ho, hl,
*
*
):
yk
=
hi24k-i.
i
If
h,
=O
for
i>v
(finite impulse response), then we obtain our
shift-register model.
An
illustration
of
such
a
model in which
intersymbol interference spans three time units
(v=
2)
ap-
pears in Fig.
4.
It was shown in
[29]
that even problems where time is
actually continuous-Le., the received signal
r(t)
has the form
r(t)
=
Ukh(1
-
KT)
+
n(1)
for some impulse response
h(t),
signaling interval
T,
and reali-
zation
n(t)
of
a
white Gaussian noise process-can be reduced
without
loss
of optimality to the aforementioned discrete-time
form (via
a
"whitened matched filter").
K
k=O
C.
Continuous-Phase
FSK
This example is cited not for its practical importance, but
because, first, it leads to
a
simple model we shall later use in
an example, and, second, it shows how the
VA
may lead to
fresh insight even in the most traditional situations.
In FSK,
a
digital input sequence
u
selects one of
m
fre-
quencies (if
z&
is m-ary) in each signaling interval of length
T;
that is, the transmitted signal
q(t)
is
where
O(Uk)
is the frequency selected by
Uk,
and
ek
is some
phase angle. It is desirable for reasons both
of
spectral shap-
ing and
of
modulator simplicity that the phase be continuous
at
the transition interval; that
is,
that
w(Uk-1)kT
+
ek-1
E
w(uk)kT
+
ek
modulo
27r.
This is called continuous-phase FSK.
The continuity
of
the phase introduces memory into the
modulation process; i.e., it makes the signal actually trans-
mitted in the kth interval dependent
on
previous signals. To
take the simplest possible case ("deviation ratio"
=a),
let
the
JEi-J-E+
Yh
+
h0
Fig.
4.
Model
of
PAM
system subject to intersymbol
interference and white Gaussian noise.
"k
L9-jw-g
SIGNAL
+
"lk
Fig.
5.
Model
for
binary continuous-phase
FSK
with deviation ratio
3
and coherent detection in white Gaussian noise.
input sequence
u
be binary and let
w(0)
and
w(1)
be chosen
so
that
w(0)
goes through an integer number of cycles in
T
seconds and
w(1)
through an odd half-integer number; i.e.,
w(O)T=Oandw(l)T=7rmodulo27r.
Thenif
eo=O,
O1=Oorr,
according to whether
uo
equals zero
or
one, and similarly
&=O
or
r,
according to whether an even or odd number of
ones has been transmitted.
Here we have
a
two-state process, with
X
=
(0,
7r
1.
The
transmitted signal
yk
is
a
function
of
both the current input
uk
and the state
xk:
yk
=
cos
[W(Uk)t
+
Xk]
=
cos
cos
w(Uk)f,
KT
t
<
(R
+
1)T.
Since transitions
&
=
(xk+l,
%$
are one-to-one functions
of
the
current state
xk
and input
Uk,
we may alternately regard
yk
as
being determined by
(k.
If we take
qo(t)
&os
w(0)t
and
vl(t)
&os
w(1)t
as
bases
of
the signal space, we may write
yk
=
yOk'?O(l)
+
ylkvl(t)
where the coordinates
(yo, ylk)
are given by
I
((0,
-I),
if
Uk
=
1,
Xk
=
Finally, if the received signal
((t)
is
q(t)
plus white Gaussian
noise
v(t),
then by correlating the received signal against both
qo(t)
and
ql(t)
in each signal interval (coherent detection), we
may arrive without
loss
of information
at
a
discrete-time out-
put signal
zk
=
(ZW,
Zlk)
=
(yak,
ylk)
+
(nokt
flu)
where
no
and
n1
are independent equal-variance white Gaus-
sian noise sequences. This model appears in Fig.
5,
where the
signal generator generates
(y~,
ya)
according to the aforemen-
tioned rules.
D.
Text Recognition
\Ve include this example to show that the
VA
is not limited
to digital communication. In optical-character-recognition
(OCR)
readers, individual characters are scanned, salient
Authorized licensed use limited to: Universitatsbibliothek der TU Wien. Downloaded on March 03,2010 at 07:23:58 EST from IEEE Xplore. Restrictions apply.