Вы находитесь на странице: 1из 19

ECE 6337 Homework 1 Fall 2013

1. A box contains 100 capacitors from which 40 are 0.01F with a 100-V voltage rating, 35 are 0.1F
at a rating of 50 V, and 25 are 1.0F and have a 10-V rating. Determine the number of elements in
the following sets:
(a) A = capacitors with capacitance 0.1F
(b) B = capacitors with voltage rating > 5V
(c) C = capacitors with capacitance 0.1F and voltage rating 50V
2. Show that C A if C B and B A.
3. Use Venn diagrams to show that the following identities are true:
(a) (A B) C = C [(A C) (B C)
(b) (A B C) (A B C) = (A B) (B C) (C A)
(c) (A B C)) = A B C
4. Sets A = {1 s 14}, B = {3, 6, 14} and C = {1 < s 9} are defined on a sample space S. State
if each of the following conditions are true or false.
(a) C B
(b) B A
(c) B C =6 O
(d) C B = S
(e) S =6 O
(f) A S =6 O
(g) C A B
5. A die is tossed. Find the probabilities of the events A = odd number shows up, B = number larger
than 3 shows up, A B and A B.
6. A pair of fair dice are thrown. person A wins if the sum of the numbers showing is six or less and
one of the dice shows four. Person B wins if the sum is five or more and one of the dice shows four.
Find:
(a) The probability that A wins.
(b) The probability of B winning
(c) The probability that both A and B win.
7. In three boxes there are capacitors as shown in Table 1. An experiment consists of first randomly
selecting a box, assuming each has the same likelihood of selection, and then selecting a capacitor
from the chosen box.
1

Table 1: Capacitors
number in box
Value (F ) 1
2
3 Totals
0.01
20 95 25
140
0.1
55 35 75
165
1.0
70 80 145
295
Totals
145 210 245
600
(a) What is the probability of selecting a 0.01 F capacitor given that box 2 is selected?
(b) If a 0.01 F capacitor is selected, what is the probability that it came from 3?
8. A company sells high-fidelity amplifiers. It has in stock 100 of 10 W audio power, of which 15% are
defective, 70 of 25 W with 10% defective, 30 of 50 W with 10% defective.
(a) What is the probability that an amplifier sold from the 10-W units is defective?
(b) If each wattage amplifier sells with equal likelihood, what is probability of a randomly selected
unit being 50 W and defective?
(c) What is the probability that a unit randomly selected for sale is defective?
9. Spacecraft are expected to land in a prescribed recovery 80% of the time. Over a period of time, six
spacecraft land.
(a) Find the probability that none land in the prescribed zone.
(b) Find the probability that at least one will land in the prescribed zone.
(c) The landing program is called successul if the probability is 0.9 or more that three or more out
six spacecraft will land in the prescribed zone. Is the program successful?

ECE 6337: Homework 1 Solutions


1. a) 60, b) 100, c) 35.
2. The Venn diagram proves it.

3.

4. a) Fale, b) true, c) false, d) false, e) true, f) true, and g) false.


5. The events are A = {1, 3, 5}, B = {4, 5, 6}, A or B= {1, 3, 4, 5, 6}, A and B = {5}. Assuming a fair die, P(A) =
1/6 + 1/6 + 1/6 = 0.5, P(B) = 0.5, P(A or B) = 5/6 and P(A and B) = 1/6.
6. A) P(A wins) = P(2,4)+P(1,4) + P(4,1) + P(4,2) = 4/36. B) P(B wins) = 11/36. C) P(A and B wins) = P(A
wins) because A is entirely in B.
7. A) P(0.01 F|box 2) = 95/210. B) P(box 3|0.01 F) = P(0.01 F|box 3) P (box 3) / P(0.01 F), with
P(0.01 F) = P(0.01 F| box 1) P(box 1) + P(0.01 F| box 2) P(box 2) + P(0.01 F| box 3) P(box 3). Thus
P(0.01 F|box 2) = 870/5903.

8. A) P(D|10W) = 15/100. B) P(D and 50W) = P(D|50W)P(50W) = (3/30)(30/200) = 3/200. C) P(D) =


P(D|10W)P(10W) + P(D|25W)P(25W) + P(D|50W) P(50W) = (15/100)*(100/200) + (7/70)*(70/200) +
(3/30)*(30/200) = 25/200=0.125
9. This is a Bernoulli trials experiment with N = 6, and p = P(land in recovery zone) = 0.8.
A) P(none in zone) =

= 0.000064

B) P( at least one in zone) = 1 P(none in zone).


C) P(success) = P(3 in zone) + P(4 in zone) + P(5 in zone) + P(6 in zone) =
+
successful.

= 0.983. Yes, the program is

ECE 6337 Homework 2

1. A man matches coin flips with a friend. He wins $2 if coins match and loses $2 if they do not match.
Sketch a sample space showing possible outcomes for this experiment and illustrate how the point
map onto the real line x that defines the values of the random variable X = dollars won on a trial.
Show a second mapping for a random variable Y = dollars won by the friend on a trial.
2. Determine which of the following are valid distribution functions:
(a)
(

GX (x) =

1 ex/2 x 0
0
x<0

(b)

0
x<0
GX (x) = 0.5 + 0.5 sin[(x 1)/2] 0 x < 2

1
x2
(c)
GX (x) =

x
[u(x a) u(x 2a)]
a

3. A random variable X has the density function


1
fX (x) = u(x) exp(x/2).
2
Define events A = 1 < X 3, B = X 2.5 and C = A B. Find the probabilities of events A, B
and C.
4. A random variable X is known to be Gaussian with mean = 1.6 and standard deviation = 0.4. Find:
(a) P (1.4 < X 2.0), and (b) P (0.6 < (X 1.6) 0.6)
5. The lifetime of a system expressed in weeks is a Rayleigh random variable X for which
(

fX (x) =

(x/200)ex
0

2 /400

0x
x<0

(a) What is the probability that system will not last a full week?
(b) What is the probability that the system lifetime will exceed one year?
6. A certain military radar is set up at a remote site with no repair facilities. If the radar is known to
have a mean-time-between-failures (MTBF) of 200 h, find the probability that the radar is still in
operation one week later when picked up for maintenance and repairs.

ECE 6337 Homework 2


Solutions

1. The outcome space is (H,H), (T,T) (both mapping to x = 2) and (H,T) and (T,H) (both mapping
to x = 2). For the friend, (H,H) and (T,T) will map to y = 2) and (H,T) and (T,H) will map to
y = 2).
2. (a) and (b) match all requirements, (c) GX () 6= 1.
3. Realize that fX (x) = 0 for x < 0. Thus P (A) = FX (3) FX (1) = 0.3834. Also P (B) = FX (2.5) =
0.7135 and P (C) = P (1 < x 2.5) = 0.32.
4. (a) P (1.4 < X 2.0) = F ((2.0 1.6)/0.4) F ((1.4 1.6)/0.4) = 0.5328, and (b) P (0.6 <
(X 1.6) 0.6) = P (1.0 < x 2.2) = F ((2.2 1.6)/0.4) F ((1.0 1.6)/0.4) = 0.8664
5. (a) need to know P (x 1) = 1exp(1/400) = 0.0025. (b) P (X 52) = 1[1exp (52)2 /400)] =
0.00116.
6. =1/(MTBF)=1/200, T = 7(24) = 168 hours. P(0 failures in 1 week)=exp(/T ) = 0.4317.

ECE 6337 Homework 3

1. A discrete random variable X has possible values xi = i2 with i = 1, 2, 3, 4, 5 which occur with
= E[X] of X.
probabilities 0.4, 0.25, 0.15, 0.1, and 0.1, respectively. Find the mean value X
2
= aX = 0 and variance X
2. A Gaussian voltage random variable X has a mean value X
= 9. The
voltage X is applied to a square-law, full-wave diode detector with a transfer characteristic Y = 5X 2 .
Find the mean value of the output voltage Y .

3. A certain meter is designed to measure small dc voltages but makes errors because of noise. The
errors are accurately represented as a Gaussian random variable with a mean of zero and a standard
deviation of 103 V . When the dc voltage is disconnected there is a 50% probability that the meter
reading is positive due to noise. When the voltage is present, this probability becomes 25.14%. What
is the dc voltage?
2
= 3, X2 = 11, and X
4. A random variable X has X
= 2. For a new random variable Y = 2X 3,
2
2

find (a) Y , (b) Y , and Y .

5. A random variable X undergoes the transformation Y = a/X, where a is a real number. Find the
density function of Y in terms of fX (x).
= 0.6 and X = 0.8, is transformed to a new random
6. A Gaussian random variable, for which X
variable by the transformation

Y = T (X) =

4 1.0 X <
2 0 X < 1.0
2 1.0 X < 0
4 < X < 1.0

(a) Find the density function of Y .


(b) Find the mean of Y .

ECE 6337 Homework 3


Solutions

1. E[X] =

P5

i=1

xi P (xi ) = 6.85.

2
2. E[Y ] = E[5X 2 ] = 5E[X 2 ] = 5(X
+ E[X]2 ) = 45.

3. The noise alone will have a N (0, 106 ) distribution. If a dc component is added, the distribution
will become N (dc, 106 ). For a standard Gaussian, i.e., N (0, 1), P (X < x) = 1 0.2514 = 0.7486 if
x = 0.67. Thus the dc value is (0.67 + 0) 103 V .
4. (a) E[Y ] = 2E[X] 3 = 9, (b) E[Y 2 ] = E[4X 2 12X + 9] = 89, and c) Y2 = E[Y 2 ] E[Y ]2 = 8.
5. Note that g(x) = y = a/x has a discontinuity at x = 0 and is thus not monotonically increasing or
decreasing. The equation y = a/x has a single solution x1 = a/y thus we should use
fy (y) =

fx (x1 )
|g 0 (x1 )|

with g 0 (x) = a/x2 = y 2 /a, this results in


!

|a|
a
.
fy (y) = 2 fx
y
y
6. (a) Observe that Y is a discrete variable, with P (Y = 4) = P (X < 1) = FX (1). After
normalizing X, this becomes FX ([1 06]/0.8) = 0.0228. Similarly, P (Y = 2) = 0.2038, P (Y =
2) = 0.4649 and P (Y = 4) = 0.3085. Thus fY (y) = 0.0228(y + 4)0.20338(y + 2)0.4649(y
2)0.3085(y 4). (b) E[Y ] = 4 0.0228 2 0.20338 + 2 0.4649 + 4 0.3085 = 1.6650.

ECE 6337
HW # 4
Solutions can be found on next page
1. Two events A and B defined on a sample space S are related to a joint sample space through random
variables X and Y and are defined by A={x1<X x2} and B={ y1<Y y2}. Show a Venn diagram for the sample
space, and indicate by arrows the mappings into appropriate regions of the X-Y plane. Also show the area
corresponding to the event AB.
1
1
2. (a) For fXY(x,y) = 2 (x-1) (y-1) + 2 (x-2) (y-2), find and sketch marginal density functions fX(x),
fY(y).
(b) Show that FXY(x,y) = [1-e-ax-e-ay+e-a(x+y)]u(x)u(y) can be written as a product FX(x)FY(y). Find the
corresponding joint pdf, and by inspection give the marginal distributions and densities.
(c) Find the marginal distributions by appropriate integrations of FXY(x,y).
1
3. A joint density fXY(x,y) is given by 2 for (x2+y2) < r2 and zero otherwise. Find the marginal densities for
r
X and Y. [Hint: For X, draw a circle and deduce the limits in terms of y and r; same for Y with x and r.]
4. A pdf is given as fX(x) and another, independent pdf is fY(y). If a RV Z = X+Y, write the integral
expression that gives the PDF of Z.
b) If the difference Z= X-Y is formed instead of the sum, is the p.d.f. of Z still a convolution of the densities of
X and Y?
5. A pdf is given by fX(x) = rect(x), and another fY(y) is given by rect(y). X and Y are independent. Find the
pdf of Z = X+Y assuming both rectangles are equal in length.
6. A well known theorem in the area of Fourier Transforms is that the transform of the convolution of x(t) and
y(t) is given by the product of their individual transforms:
FT[x(t)*y(t)] = X( )Y( )
Use this to find the FT of the convolution of rect functions in Problem 5, above. What would be the FT of
another convolution with a rect-function?
7. The random variable X has mean 1 and variance 4, and the r.v. Y have mean 2 and variance 1. The
correlation coefficients between X and Y is 0.4. New random variables are defined using V = -X + 2Y and W =
X + 3Y.
Find
(a) the means, (b) the variances, (c) the correlations and (d) the correlation coefficient

VW

of V and W.

ECE 6337
HW # 4 Solutions

1.
1
1
2. (a) For fXY(x,y) = 2 (x-1) (y-1) + 2 (x-2) (y-2), find and sketch marginal density functions fX(x),
fY(y).
[Answer: fX(x) = 1/2 (x-1) + 1/2 x-2); fY(y) = 1/2 (y-1) + 1/2 (y-2).]
(b) Show that FXY(x,y) = [1-e-ax-e-ay+e-a(x+y)]u(x)u(y) can be written as a product FX(x)FY(y). Find the
corresponding joint pdf, and by inspection give the marginal distributions and densities.
[ FXY(x,y) = [1-e-ax]u(x)[1-e-ay]u(y); fXY(x,y) = a e-axu(x) a e-ayu(y); marginals are just the individual terms.]
(c) Find the marginal distributions by appropriate integrations of FXY(x,y). [Since X, Y independent, we
can operate on FX(x) and FY(y) individually, giving a e-axu(x) and a e-ayu(y).]
1
for (x2+y2) < r2 and zero otherwise. Find the marginal densities for
r2
X and Y. [Hint: For X, draw a circle and deduce the limits in terms of y and r; same for Y with x and r.]
3.

A joint density fXY(x,y) is given by


r2-x2

dy
2
2
[ fX(x) = fXY(x,y)dx =
= 2 r2-x2 . and similarly fY(y) = 2 r2-y2
2
r
r
r
-
2
2
- r -x
4. a) A pdf is given as fX(x) and another, independent pdf is fY(y). If a RV Z = X+Y, write the integral

expression that gives the PDF of Z. [fZ(z) = fX(x)fY(z-x)dx or fY(y)fX(z-y)dy ]


-
-
b) If the difference Z= X-Y is formed instead of the sum, is the p.d.f. of Z still a convolution of the densities of
X and Y? [fZ(z) = or f z ( z )

f y ( y) f x ( z

y )dy , this is a correlation integral.]

5. A pdf is given by fX(x) = rect(x), and another fY(y) is given by rect(y). X and Y are independent. Find the
pdf of Z = X+Y. [Answer: fZ(z) = tri(z), use graphically aided convolution as shown in class.]
6. A well known theorem in the area of Fourier Transforms is that the transform of the convolution of x(t) and
y(t) is given by the product of their individual transforms:
FT[x(t)*y(t)] = X( )Y( )
Use this to find the FT of the convolution of rect functions in Problem 5, above. What would be the FT of
another convolution with a rect-function? [Convolving two rect functions is equivalent in the frequency
domain to the product sinc( /2)sinc( /2). Another convolution with a rect functions will result in sinc3( /2).]
7. [(a) E[V] = 3; E[W] = 7. (b)

= 4.8;

2
W

= 17.8. (c) RVW= 22.2; (d)

VW=

0.13.]

ECE 6337
HW # 5
1. If A cos( ot) is sampled randomly in time, the probability density for = ot must be uniform across a
1
range 2, that is, f ( ) = 2 across that range.
(a) If Y is the value of Acos( at an arbitrary point in time, find the pdf fY(y).
(b) If A = 120 2 (an approximate value for household voltage) , find the probability that at any
moment the voltage magnitude is greater than 120.
2. Find the constant K, and find the values of

and m such that K exp(-3x2+2x) =

1
exp[-(x2

m)2/2 2].
3. The random variable X has mean 1 and variance 4, and the r.v. Y have mean 2 and variance 1. The
correlation coefficients between X and Y is 0.4. New random variables are defined using V = -X + 2Y and
W = X + 3Y.
Find
(a) the means, (b) the variances, (c) the correlations and (d) the correlation coefficient VW of V and W.
4. A jointly Gaussian pdf for X and Y has variances 2 and 3, respectively, mean values -1 and 1,
respectively, and correlation coefficient = 0.4. Find the value of the rotation angle required to obtain
new and uncorrelated RVs U and V, and write an expession for the pdf fUV(u,v).
5 A PMF for a random variable X is given by P(X=1) = 1/4, P(X=2) = 1/4, P(X=3) = 1/2, and the same
PMF applies to a RV Y. Find the PMF for a RV Z = X +Y.
6.

Random variables U and V are given by U = X-m and V = Y-mY Show that |
XY

X Y

XY|

1, where

XY=

is the correlation coefficient of X and Y. [Hint: Start with the fact that E[(U-kV)2] 0 for any real

constant k.]

ECE 6337
HW # 5 Solutions
d (y)
1 d
1 d
-1
-1
dy = 2 dycos (y/A) = 2 dycos (y/A) =
1
1
1
2 1-(y/A)2 . Multiply by 2, because 2 regions of , so fY(y) = 1-(y/A)2 . ]
(b) [Change variable from y/A to x and integrate from 1/ 2 to 1 and -1/ 2 to -1. Result: 1/2.]
1. (a) [As usual, the formula is fY(y) = f [ (y)]

2. [For some value of , 3x2-2x+ must equal

(x-m)2
. Hence 2
2 2

= 1/3 and =1/ 6 . Then -2x = -

(2xm/2 2) so m = 1/3. The new term needed, m2/2 2, = 1/3, so K = e1/3


result: fX(x) =

1
is what is needed. Final
/3

1
1
exp{-(1/3)(x-1/3) 2} ]
/3

3. [(a) E[V] = 3; E[W] = 7. (b)

= 4.8;

2
W

= 17.8. (c) RVW= 22.2; (d)

VW=

0.13.]

4. [ From formula we find , then form 2x2 matrix CX with C and S on top row, -S and C on bottom
row, where C = cos( ) and S = sin( ). The covariance matrix for X and Y has (2 0.4 6 ) on top row,
and (0.4 6 3) on bottom row. We perform the calculation [CX-1][CXY][CX] and get a diagonal
matrix representing

= 1.4 and

= 3.6. Mean values of U and V are easily seen to be about -.33

and 1.38. Inserting them into our final result gives the following, where we see U and V are
independent.
fUV(u,v) =

1
(u+.33)2
1
(v-1.38)2
exp[- 2(1.4) ]
exp[- 2(3.6) ]. Note P = 2+3 = 1.4+3.6 = 5, both
1.4 2
3.6 2

cases.]

[PZ(Zk) =

PX(Xi)PY(Zk-Xi) . Thus the first possibility is


i=1
PZ(2)=PX(1)P Y(2-1) = PX(1)PY(1) = 1/16.
Next: PZ(3)=PX(1)PY(3-1)+PX(2)PY(3-2) = 1/16+ 1/16 = 1/8.
Next: PZ(4) =PX(1)PY (4-1)+PX(2)PY(4-2)+PX(3)PY(4-3) =1/8+1/16+1/8 =5/16
Next: PZ(5) =PX(2)PY(5-2)+PX(3)PY(5-3)=1/8+1/8 = 1/4
Next: PZ(6) = PX(3)PY(6-3)=1/4.
(One can visualize graphically with lines, one set reversed, shifted along and the other stationary.)]
6. [Hint: Start with the fact that E[(U-kV)2] 0 for any real constant k.] [Expand E[(U-kV)2] =
2k

2
UV+k

2.

and form a quadratic in k: k2-2k

UV
V

2
U
2
V

2
U -

= k2-2kb+c 0. Complete the square to give

(k-b)2+(c-b2) 0. The first term is zero if k = b, so we must have c-b20, or b2 c. Thus |


and | UV| = | XY| 1. ]

UV|

U V

ECE 6337
HW # 6
1. (a) A random process is WSS with E{x2(t)} = 11. Discuss why RXX( ) = -11 exp{-| |} cannot be its
autocorrelation function.
(b) What about 11 tan -1 ?
2. Determine which of the following functions cannot be valid power spectra:
(a) exp[-( -1)2], (b) 4/(1+ 2+j 6), (c) | /(1+2 + ).
3. (a) Find the autocorrelations RXX(t,t+ ), R YY(t,t+ ), crosscorrelations RXY(t, t+ ) and
RYX(t,t+ ) of the random processes X(t) = A sin( ot + ) and Y(t) = B cos( ot + + /4), where
is a random variable with uniform distribution over a range of 2. Are these processes WSS?
(b) Find the spectral densities and cross-spectral densities of the above.
(c) Find total power by calculation both in -domain and frequency domain. What proportion of
total power is the cross-power portion?
4. A WSS random process X(t) has the autocorrelation RXX( ) = 8 + 8 sinc(
(a) Sketch RXX( ), and find both the DC power and AC power.
(b) Find the PSD and sketch. In what frequency range, in Hz, is most of the power?
(c) Find the covariance matrix for samples of the RP at t, t0.3, t0.6, t0.9.
5. With respect to the previous problem:
(a) In what range of is most of the power? (Use zero crossings of the sinc as the measure.)
(b) Within what frequency range is most of the power
(c) Do your results correlate with the general principle (time-bandwidth product) that the product
of the principal width in the -domain and the principal bandwidth B (positive frequency , Hz)
is approximately 1?
(d) If we change the AC part of the autocorrelation in Prob. 4 so that we have sinc( /2) instead of
sinc( ), how will the first zero-crossings of the sinc be changed? How will the frequency range
be affected?

ECE 6337
HW # 6,
1. (a) not maximum at 0 , (b) not even, not maximum at 0.
2. (a) not even. (b) not real. (c) not even.
A2
B2
3. (a) RXX( ) = 2 cos( o ); R YY( ) = 2 cos( o ), E[X(t)} = E[Y(t)] = 0. Both WSS. RXY( ) =
AB
AB
[sin( o )+cos( o )]; RYX( ) =
[sin( o )-cos( o )].
2 2
2 2
A2
(b) SXX( ) = 2 [ ( + o)+ ( - o)], SYY the same except B2.
AB
AB
SXY( ) =
[(1+j) + o)+(1-j) ( - o)], SYX( )=
[(1-j) + o)+(1+j) ( - o)]
2 2
2 2
A2+B2
(c) RXX(0) + RYY(0) = 2 ,

1
1
A2+B2
2+B2] =
[S
(
)+S
(
)]d
=
[A
XX
YY
2
2
2 .
-

AB
1
1 AB
AB
RXY(0)+RYX(0) =
; 2 [SXY( )+SYX( )]d = 2 [
]=
.
2
2
2
-

The occurrence of negative power stems from the 4 phase shift in the argument of the cosine in Y(t).
In effect Y(t+ ) then consists partially of a term [ sin[( o(t+ )+ )] which correlates negatively with
X(t)! The problem as posed is deceptive, because if if a RV then so is +/4, and would more
properly be approached as having a RV 1 in X(t) and a RV 2 in Y(t). Their sum or difference would
be uniformly distributed on (0,2). Then the cosine terms would drop out of RXY and RYX. SXY and
SYX would be pure imaginary, with zero power contained in the cross terms. That is actually the way
the problem should be worked.
4. (a) DC 8, AC 8.
(b) PSD =
16 ( ) + 8 rect( /2). (Integral would be 16.) All the power is in range |f| < 0.5.
(c) Square matrix and

symmetric with C11 = C22 = C33 = C44 = 8; C12 = C21 = C34 = C43 =

6.86; C13 = C31 = C24 = C42

= 4.04; C14 = C41 = 0.874. Student should show the matrix!

5. (a)
| | < 1.
(b) All the power is in range |f| < 0.5.
(c) f
= 0.5 (2) = 1.
(d) Expanded to width 4, reduced to width 0.25, respectively.

ECE 6337
Homework #7,
1. The input to an LTI system has the autocorrelation RXX( ) = 2e-2| , and it is found that
24
the output PSD is SYY( ) =
. Find the impulse response of the LTI
(4+ 2)(9+ 2)
system.
2. A RP X(t) with autocorrelation RXX( ) = tri( /2) is multiplied by 2 cos( t) to produce
a new random process Y(t). Find and sketch the PSD of Y(t), and calculate both the
power in X(t) and that in Y(t). [Note: X(t) is called a lowpass random process and Y(t)
is called a bandpass random process.]
3. A random process X(t) has an autocorrelation function Rxx( ) = A2 + Be-| |
Where A and B are positive constants. Find the mean value of the response of a system
having an impulse response h(t) = exp(-Wt) u(t) where W is a real positive constant, for
which x(t) is the input.
4. The figure shows the sonar echo system on a submarine.

The system transmits a random noise n(t) to determine the distance to another target
submarine. Distance R is given by v R/2 where v is the speed of sound waves in water

and R is the time it takes the reflected signal to return. Assume that n(t) is a sample
function of an ergodic random process N(t) and T is very large. With these assumptions,
you can approximate the time average of the integral from -T to T as a delayed version of
the ensemble average. Also note that if the correlation peak is being tracked in real time
t, we could let the upper limit T = t = 0, (the present) and the lower limit t = - R, or less..
(a) Find V in terms of a correlation function of N(t).
(b) What delay t will cause V to be maximum?
(c) State in words how the submarine can determine the distance to the target.
(d) What provision must be made by the transmitting submarine to track the target if
the distance between them is changing?
5. The signal x(t) = u(t) exp(-Wt), where W is positive real, is applied to a filter along
with white noise with power density K/2, K>0 and real.
(a) Find the transfer function of the filter matched to x(t) at time t0
(b) Find the filters impulse response
(c) Is there any value t0 of that will make the filter causal?
(d) Find the output maximum signal-to-noise ratio.
6. A Wiener filter (non-causal) is to be designed for AM radio reception for the following
model:
(a) The autocorrelation of the desired signal at the receiver can be modeled as
RSS( ) = A exp(-a| ) cos( o ) ( o is the carrier frequency, >> 1/a)
No
(b) The noise is white, with autocorrelation RNN( ) = 2 ( ).
Find the transfer function of the optimum filter of this type. Note: Your filter function will
have two parts, one for the positive frequency range and the other for the negative frequency
range. Neglect the overlap (cross-product term) in your expression for H( ).

ECE 6337
Homework #7 Solutions

1. Syy() = |H()|2 SXX(SXX() = FT{2e-2|} =

8
3
so |H()|2 =
4+2
9+

3
3
3
3
=
. Causal part is
, so h(t) = 3 e-3tu(t).
3+j
9+2 3+j 3-j

1
2. Y(t) = 2X(t)cos(5t); SYY() = 2 SXX() 2 * [(+5)+
SXX(+5)+SXX(-5). SXX() = FT{tri(/2)} = 2 sinc2(), so SYY()=
2[sinc2(+5)+sinc2(-5)]. See sketch.

A
3. Mean of X(t) is A, so response is Ym = A
e-Wt dt = W .
0

4.
(a)

) (

) . Since T is large,

lim
( ) ( ) = [ ( )) (

) ( )] since n(t) is ergodic. Thus Rnn(R-T)

)] = [ (

(b) Because | Rnn()| Rnn(), V will be maximal if R=T


(c) From part b, V is maximum when receiver delay T equals target delay R. By slowly
varying T until V is maximum, R can then be found from R = vTT being that for
which V is maximum)
(d) Time delay T must be adjustable to match variability in R.

5.
(a) X() =

1
1
so Hopt = X*() =
(arbitrary constant multiplier possible) .
W+j
W-j

(b) hopt(t0-t) = exp[-W(t0-t)]u(t0-t).


(c) Filter is non-causal for any t0.

0
2
1
(d) Maximum SNR = N
e-2Wt dt = N W . SNR increases with decreasing W.

o
o
-

6. Assuming that the noise and signal are uncorrelated, the Wiener filter is given by
( ) =

( )
( ) +
( )

The power spectrum of the noise is N0/2, and the power spectrum of the signal is the
Fourier transform of its ACF, and is given by
( ) =

+(

+( +

where the first term is due to positive frequencies, and the second term to negative
frequencies. Thus the positive frequency part of the Wiener filter is
( ) =
+ 2 [

+(

) ]

+ 2 [

+( +

) ]

and the negative frequency part is


( ) =