3.
Optimum Detection of Binary PAM in Noise
PAM: Pulse Amplitude Modulation
Representation:
h(t)
1
0 T t
where bk represents the kth bit: bk= 0 or 1
h(t) is a rectangular pulse of length T centered at t=T/2.
Advantage of rectangular pulses:
No interference between pulses in the adjacent intervals. 1
Binary PAM in Noise
The matched filter for each pulse h(t-kT) in the interval (k-1)T< t < kT is:
gk(T-t)=h(t-kT)
Consider the received signal: r(t)=s(t)+w(t)
=> The output of the matched filter receiver at the end of the kth symbol
interval is:
+∞
Integrate and-dump detector.
This device simply integrates the received signal over the symbol interval, samples the
output, and then starts for the next interval.
2
BER PERFORMANCE
The receiver role:
must make a decision between two hypotheses:
A simple and efficient decision rule for deciding between the two
hypotheses is to choose H0 that is to choose 0 if y is less than some
threshold γ and choose 1 if not.
Ex: If 1 is transmitted, the Prob. of error is: : 3
BER PERFORMANCE
r(t) = s(t) + w(t) = +w(t)
As symbols are separated in time, consider the detection of the first symbol
(k=0).
T
Matched filter output: Y = ∫ r ( t )dt = S + N
0
T
AT if H1 ( b 0 = 1)
=> Output signal: S = ∫ Ab0dt = Ab0T ; S=
0 0 if H 0 ( b 0 = 0)
T
Output noise: N = ∫ w ( t )dt
0
N is Gaussian; mean E[N]=0;
TT TT T
N N N
var: E[ N ] = ∫ ∫ E[ w ( t ) w ( τ)]dtdτ = ∫ ∫ 0 δ( t − τ)dtdτ = ∫ 0 dτ = 0 T
2
00 00
2 0
2 2
=> Y: N(AT, N0T/2) if H1 and Y: N(0, N0T/2) if H0 4
BER PERFORMANCE
p(y/H1)
The probability of a Type I error ( 1 is transmitted) is the probability that
the output Y falls in the shaded area of the Figure.
Pe1=
5
BER PERFORMANCE
p(y/H0)
Similarly , The probability of a Type II error, is the probability that the
output Y falls in the shaded area of the above Figure.
Pe0 =
6
BER
Over all BER?
Typically: (the transmitted bits are equiprobable)
=>
The two types of error are equal if γ=µ/2 (optimal choice).
In this case :
7
BER
(1)
In terms of Eb/N0?
It was shown that σ2=N0T/2 => σ = N 0T / 2 (2) and µ=AT (3)
As for the signal:
The average energy per bit Eb at the receiver input can be computed from the
first bit as:
=> (4)
8
NONRECTANGULAR PULSE SHAPES
Rectangular pulses have spectral limitations
(Sinc function in the frequency domain)
⇒ Practical pulse shapes :
- Raised-cosine pulses
- Root-raised-cosine pulses
Advantages: - limited in bandwidth while offering zero intersymbol interference (ISI)
- BER performance is the same as with rectangular pulse shaping