Beruflich Dokumente
Kultur Dokumente
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
August 2, 2011
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Overview
Random Variables Random Processes Noise Signal to Noise ratio Transmission and Filtering Digital Communications. Formatting analog information.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Random Variables
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Example: S= { Tail, Head}, Sx = {0, 1} A function that maps S into Sx is called as a Random Variable,
denoted by X (.)
Sx is countable: Discrete Random Variable One to one mapping, Many to one mapping Sx is uncountable or innite: Continuous Random Variable
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
X (s2 )
Tail
x1 1
s1
Head
x2 0
S X {1,0}
X (s1)
Sample Space S
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
P[sj ]
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
px(x)
10
15 x
20
25
30
P[a X b] =
a
fX (x)dx
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
fX (x) 0
PMF must sum to 1
M
fX [xi ] = 1
i=1
fX [xi ] = 1
i=1
fX (x)dx = 1
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Expectation
Expectation E [X ], Mean
E [X ] =
i
xi fX [xi ]
=
Expectation is linear
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
nth moment
Expectation E [X n ].
E [X ] =
i
xin fX [xi ]
x n fX (x)dx
Variance, 2
(X E [X ])2 fX (x)dx
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
HH TH TT HT
1
x
S XY
S
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
X (si ) x = i S Y (si ) yi Two random variables dened on the same sample space are Jointly distributed. 1 0 0 1 SX ,Y = , , , 0 1 0 1 These are discrete events, so joint PMF
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Joint PMF
For a single RV, PMF: fX [xi ] = P[X (s) = xi ] For a multiple RV, Joint PMF:
0 fX ,Y [xi , yj ] 1
Nx Ny
fX ,Y [xi , yj ] = 1
i=1 j=1
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Joint distributions
r =1
Sample Space S
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Joint distributions
r =1
(r , )
Sample Space S
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Random Processes
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Random Variables
X (s2 )
Tail
x1 1
s1
Head
x2 0
S X {1,0}
X (s1)
Sample Space S
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
x1 = (x[0], x[1], . . . )
The random process is a mapping from S which is a set of innite
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Random process
1.5 X[n,s1] = x1[n] 1
x1[n]
0.5 0 0
10 n
12
14
16
18
20
x2[n]
0.5 0 0
10 n
12
14
16
18
20
x3[n]
0.5 0 0
10 n
12
14
16
18
20
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Random process
A sample space or ensemble composed of functions of time is called
s1
s2
s3
Sample Space
x 2 (t )
x 2 (t k )
x3 (t )
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Random process
Is an ensemble composed of functions of time. For a single sample point s1 we get a single function of time x1 (t). If time is held constant t1 we get a random variable. For a given random process, the mean value of X (t) at arbitrary
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
E [X (t)] =
process
In this case it is possible to dene the time average of a signal
1 x(t) = lim T T
T 2
x(t)dt.
T 2
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Ergodic Process
E [X (t)] = x(t) = x
2
X (t) x
2 = (x(t) x )2 = x
Variance
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Autocorrelation
Autocovariance
Cx (t1 , t2 ) = E [{X (t1 ) x (t1 )}{X (t2 ) x (t2 )}] = RX (t1 , t2 ) x (t1 )x (t2 )
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Crosscorrelation
Cross covariance
CXY (t1 , t2 ) = E [{X (t1 ) x (t1 )}{Y (t2 ) y (t2 )}] = RXY (t1 , t2 ) x (t1 )y (t2 )
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Operations
The two random processes are said to be uncorrelated if
CXY (t1 , t2 ) = RXY (t1 , t2 ) x (t1 )y (t2 ) = 0 RXY (t1 , t2 ) = x (t1 )y (t2 ) RXY (t1 , t2 ) = E [X (t1 )]E [Y (t2 )].
For a complex Stochastic Process, autocorrelation is
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
fX (t1 + ),X (t2 + ),...,X (tn + ) (x1 x2 . . . xn ) = fX (t1 ),X (t2 ),...,X (tn ) (x1 x2 . . . xn )
The joint p.d.f of random variables obtained by observing the
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
So is IID random process stationary? Yes, Marginal PDF is the same for each RV. Therefore
fX (t1 + ),X (t2 + ),...,X (tn + ) (x1 x2 . . . xn ) = fX (t1 ),X (t2 ),...,X (tn ) (x1 x2 . . . xn )
Any process whose mean or variances change with time is NOT
stationary.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
fX ( ) = fX (0) (x)
Since the PDF does not depend on a particular time, so should not
mean/expectation. Therefore
E [X (t)] =
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
t1 , t2
t1 , t2
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Covariance
Cx (t1 , t2 ) = E [{X (t1 ) x (t1 )}{X (t2 ) x (t2 )}] = RX (t1 , t2 ) x (t1 )x (t2 )
If it is WSS,
CX (t1 , t2 ) = RX (t2 t1 ) 2 x
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
To summarise
A second order stationary process can be wide-sense stationary but
independent of time and the autocorrelation function depends on the time dierence.
A random process is strictly stationary when the statistics do not
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
RX ( ) = E [X (t + )X (t)] t
Mean square value of the process can be obtained by = 0.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Properties
Autocorrelation is an even function
RX ( ) = RX ( )
100
80
60
autocorrelation
40
20
20
40 100
80
60
40
20
0 shift
20
40
60
80
100
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Properties of correlation
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Joint stationary
If for all , RXY ( ) = 0 we say X(t) and Y(t) are orthogonal If for all , CXY ( ) = 0 we say X(t) and Y(t) are uncorrelated.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Example
where A and fc are constants and is a random variable uniformly distributed over the interval [, ] and independent of X (t). Obtain the Rx ( ) and PSD.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Homework
stationary process as X1 (t) = X (t) cos(2fc t + ) X2 (t) = X (t) sin(2fc t + ) where fc is a carrier frequency and is a random variable uniformly distributed over the interval [0, 2] and independent of X (t). Obtain the cross correlation.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
X (t) h1 (t)
V (t)
Y (t) h2 (t)
Z (t)
Rvz ( ) = v ( ) z( )
= h1 ( ) x( ) h2 ( ) y ( ) Rvz ( ) = h1 ( ) h2 ( ) Rxy ( ) Gvz (f ) = H1 (f )H2 (f )Gxy (f )
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Quadrature Filters
shifts the phase of its input by some number of degrees, A cos(2f0 t) 90 phase shifter A cos(2f0 t 90 )
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Quadrature Filters
The transfer function of a Quadrature lter can be written as
HQ (f ) = jsgn(f ) = j = +j
Hence,
HQ ( f )
f >0 f <0
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Quadrature Filters
If we want to write the impulse response,
F[sgn(t)] = F[
1 jf
hQ (t) =
1 . t
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
g ( ) d. t
The integral formula is known as the Hilbert transform and g (t) and
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
transform of g (t).
The inverse Hilbert transform is given by
g (t) =
g ( ) d. t
orthogonal:
g (t) (t)dt = 0. g
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Noise
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Thermal Noise
Produced by random motion of charged particles in conducting
media.
In 1928, Johnson & Nyquist studied noise observed in resistors
GV (f ) =
V 2 /Hz
R is the resistance (), T is the temperature (K), k = 1.38 1023 J/K is Boltzmanns constant, h = 6.62 1034 J s is Plancks constant.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Thermal Noise
Gv (f ) = 2RkT 1
h|f | 2kT
|f | <<
kT h
GV (f ) = 2RkT
V 2 /Hz.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
RW ( ) =
N0 ( ) 2 N0 GW (f ) = 2
RW ( ) = 0 t = 0
Hence, any two samples separated in time are uncorrelated. However, a theoretical problem is that white noise has innite
average power.
Filtered white noise is called coloured noise.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Coloured Noise
N0 2
through an LTI
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Signal-to-Noise Ratio
Y (t) = S(t) + N(t) S(t) is the noise-free signal (or process) and N(t) is the noise.
We usually assume that: the noise is ergodic and zero-mean, the noise is independent of the signal.
RV
PMF
Cont. RV
Operations
MRV
Joint Dist.
Process
Operations
Types
Quad lts
Noise
DC
DC
Signal-to-Noise Ratio
E [|Y (t)|2 ] = PS + PN
so the signal and noise power add also in the received signal. An important quantity is the signal-to-noise ratio (SNR):