Minimum Distance One-Shot Receiver

Minimum Distance One-Shot Receiver

Receiver Design for PAM

Govind Sharma

Department of Electrical Engineering

Indian Institute of Technology, Kanpur

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Isolated Pulse or One-Shot Receiver

We concentrate on effect of noise and so assume that there is no

ISI. We assume that only one pulse has been transmitted. We
design receiver based on minimum distance criterion.

r̃ (t) = ah̃(t) + ñ(t)

where a ∈ A is transmitted symbol, h̃(t) is lowpass equivalent

receive pulse and ñ(t) is lowpass equivalent noise. In baseband
case all these are real and complex in passband case.

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Minimum Distance Criterion

Distance between two complex waveforms is defined as

sZ
∞
d(x (t), y (t)) = kx (t) − y (t)k = |x (t) − y (t)|2 dt
−∞

Receiver chooses a symbol that minimizes distance between

received signal and noise free signal.
Z ∞
â = arg min |r̃ (t) − ah̃(t)|2 dt
a∈A −∞

Minimum distance receiver has to calculate M = |A| integrals to

arrive at a decision.

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Correlation Receiver

Z ∞
J= |r̃ (t) − ah̃(t)|2 dt
−∞
Z ∞  Z ∞ 
= |r̃ (t)|2 dt − 2Re a∗ r̃ (t)h̃∗ (t)dt
−∞ −∞
Z ∞
+ |a|2 |h̃(t)|2 dt
−∞
= Er − 2Re {a∗ y } + |a|2 Eh

where Z ∞
y= r̃ (t)h̃∗ (t)dt
−∞

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Correlation Receiver

y can be seen as correlation or inner-product between received

waveform and receive pulse.
R∞
r̃ (t) × −∞ dt
y

h̃∗ (t)

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Matched Filter

y (t) t=0
r̃ (t) h̃∗ (−t) y (0)

Z ∞
y (t) = r̃ (τ )f (t − τ )dτ
−∞
Z ∞
y (0) = r̃ (τ )f (−τ )dτ
−∞

If f (t) = h̃∗ (−t) then y (0) is same as correlator output.

A filter with impulse response h̃∗ (−t) is said to be matched to
h̃(t). The above structure is called sampled matched filter.

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Minimum Distance Cost Function

J = Er − 2Re {a∗ y } + |a|2 Eh

Er ∗ y
   
2
= Eh − 2Re a + |a|
Eh Eh
2
|y |2

y 
= Eh  − a − + Er
Eh Eh

Minimum distance receiver becomes

2
 
y
â = arg min  − a
a∈A Eh

In absence of noise y /Eh will be equal to the transmitted symbol.

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Passband one shot receiver

t=0
r (t) u(f ) × h̃∗ (−t) â
r̃ (t) y (t) Eh−1
Slicer
√
2e −j2πfc t

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Properties of the Matched Filter

If we use a receive filter with impulse response f (t) then output is

Z ∞
y (0) = r̃ (τ )f (−τ )dτ
−∞
Z ∞ Z ∞
= ah̃(τ )f (−τ )dτ + ñ(τ )f (−τ )dτ
−∞ −∞
= S +N

where S is signal part and N is noise part. If we assume that noise

is white with PSD N0 /2 then E[|N|2 ] is Ef N0 /2. For passband
case we have to assume that bandwidth of filter is smaller than fc .

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Properties of the Matched Filter

We define signal to noise ratio as

E[|S|2 ]
SNR =
E[|N|2 ]
2
2EA  ∞
Z

= h̃(τ )f (−τ )dτ 
Ef N0 −∞
 

By Cauchy Schwartz inequality SNR is maximum when

f (−t) = c h̃∗ (t). We can take c = 1 as it affects signal and noise
equally. Thus we get

SNR ≤ 2EA Eh /N0

This is known as matched filter bound.

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Properties of the Matched Filter

Frequency response of the Matched filter is H̃ ∗ (f ) and so overall

pulse shape is P(f ) = |H̃(f )|2 . If h̃(t) is causal the matched filter
will be anticausal. So we have to introduce a delay.

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Matched Filter and ISI

To avoid ISI the output of the matched filter should satisfy Nyquist
criterion ∞  
1 X k 2

Sh̃ (f ) =  H̃ f − =1
T k=−∞  T 

For example H̃ (f ) with square root of raised cosine pulse satisfies

this criterion. Sh̃ (f ) is called folded spectrum of the received pulse.

G. Sharma EE321A
 Minimum Distance One-Shot Receiver

Example

Example 10.1
Raised cosine pulse has non-negative real valued symmetric
frequency response that satisfies Nyquist criterion of zero ISI.
Hence we can take
q
H(f ) = F (f ) = PRC (f )
C(f) = 1

If channel is flat then g(t) = f (t). The pulse shape is called square
root raised cosine pulse and can be well approximated in practice.

G. Sharma EE321A