THE MATCHED FILTER

matched
filter is linear
A a
filter
designed provide to output maximum
sang for given transmitted signals It)
a

corrupted by additive white noise net) with .

PSD Nok The . set is

signal ) confined
to OETET

signal
.

let the sit )

plus noise att
is
input to an LTI
fillet having
impulse as shown below followed by
response htt) , a

sampler .

sett -

⑦g¥¥T¥¥
htt)
.
yep
-

Now ,
we derive the matched filter
which
impulse response
at its
hits ,
maximizes
SNR output 's .
.

the
fillet outputs
(t ) Ect) It )] * htt)
y = th

= s * h #I + net) * htt)
-
x '
←
Ys Ct ) t Ynet)
⇐

output signal component is

Ys =
SHI * hit I
 7- fits HD = SCH Hlf)

f! e' Hdf
'"
or
Ysltl = sit) Hlf )

att-tysuy-ffslfltkfjei.tt#I
The output noise
component is
htt ) hit)
yal t) *
=

of Jatt ) Sno ft) NI l 12

'

PSD Htt)
= =

Average ofp
or
power
variance of the noise term
Effort D=
noise ,
=

NEITH Hitt
at the output
Now the
output SNR is
EI
,

III. =

Cy 's HIT F-
2

| !! SH ) Hlf ) e' ft d f
f hit

#
fool Hft ) Pdf

Applying
slates
the
Cauchy Schwarz
-

inequality ,
which

I faff , CHACH df Tsf!If Ifl Pdf! If Cfl Tdf

fi (f)
't

the
equality holds
if
=
Kfz Cf )
K is an arbitrary real number and * denotes complex
conjugate -
If we Ad
? f , (f) = H (f) & fzcf) = S (f) eJ2TfT
we write

ehtftdf kfagltkftfdf.FI/sefsehTfTpdfEfo9HCfITdf f
can

I foot Hisef)
-

f) Pdf

substituting into (SIN )o we

get
-

¥1 s
HHTfdf foolish Pdf

¥f¥f
,

S
IT falls Ift Pdf

¥! I
2t÷
Es
-1,9 sets Rdf
Energy of
=
s

(F) o
, may
=

and
[Scf) ejzltft]
*

H (f) =
k

Iff
H k Eff ) e- J2TfT
For real 't
Sff )
a
signal Cf ) =
,
s

I 2T ft
Sff ) e-
-

Hlf ) = k

htt ) = K
f! Sff) e- j2TfTeJ2Tftdf
 t.la?sff1eJ2TfCT-
t)
= df

f → f
eJ2Tf ( T
=kf! ) off
t
htt ) Sff)
-

hItl=Ks(T-t#
-

Thus , the impulse response of the optimum filler ,

except fer the scale

factor K ,
is matched to
the -

input signal .

k¥1 In
deriving the matched filter the
only assumption
about the
with
noise hit) is that it is
In
stationary and while

zero mean 4 PSD MHz .

other words
,
no

assumption was made on the statistics of the

noise htt ) .

Notes

Note I A
filter with
impulse response htt) Sitt) ,
-

where Sct ) is confined to Etat is called

-

the matched
filter to the Signal set) .
The

posse of htt) SG
Yess
t) to the
signal set
-

-
-

YH) =
lot ) htt
SK -

2) de

=
lots (e) ( s T t -12) de
-

time autocorrelation
of
which is
the basically
signal set .
the
Example :

#or the
given signal ,
the
impulse
response of the matched fillet is

htt ) =
SG -

t) =
sit) as shown

÷
"

it
output
÷i is
it
the
yet )
Y# = s * htt ) =

which is below :
plotted
 set t)
-
-
-

htt) yet)

¥÷¥¥
,

AF - - - -
- -

Note that the matched fillet ofp ( auto -

correlation ) function is an even

feuction of t which attains a

peak at t T
-

Now Mani mum SNR is

,

¥1 =
2¥ .
.
E -

f Tdt

=/ ET
'
Edt
¥i÷
-

Exampling find of
the the RC output
felter and determine maximum the

& NR )o .
if input
the a
rectangular
duration T 4 Amplitude A
is

pulse of .

compare the result with matched fillet

sit ,

Finny
is

↳
- o

"

7 t
set '
=c
Htt
t ¥4
 The
frequency the
response and
impulse
fillet
response of
are

Hwi
YETI titre
- -
-
-

IRC
feet
&
hit ) =
uit )

YH)
D

the
filter output is
given by

yet)

the maximum
t
value
of yet ) occurs
at T
-

Y H) = A Cl - ETIRe )

Average output noise

power
 I

F- Crites ] -

-
t
21T

-
)
N
# IHCWIRDO

Note
=

thus
RC
1447122
output sa
,q=
ECNUTD
=
A2(te9
Holy RC

sure 4A2÷(teTK
IRC +
substituting
into sun ,
we have

- xx
MATCHED FILTER TYPE DEMODULATOR

halt) -
-
d G- t) ,
oats
= 0 ,
otherwise

→
y-4-0 r,

received on

¥
;

→0t# rn

at t -7
sample ,

The outputs of these filters are

Ykct ) = Ict) * halt) K -

112 -
-

at

fotr.cc/hklt-z)d2
 Yklt =
fotrtedk ( T-
t -12 ) de
at t
sample T
-

Lat ) Ls =/!rt) Ok (2) d2 his

-

K= 1,2 .
- N