Homework

Problem 1 A communication system operates over an AWGN channel with white noise (power spectral density: N0 /2). The system communicates one information bit (0/1) to the receiver by sending a waveform according to the following scheme. To communicate a 0, the transmitter sends nothing, i.e., s(t) = 0 for 0 t T. To communicate a 1, the transmitter ips a (fair) coin. If the coin shows head, the transmitter sends the waveform sa (t), and if the coin shows tail, it sends the waveform sb (t) during 0 t T . Suppose the transmitted bit is equally likely to be a 0 or 1. Also assume that sa (t) and sb (t) have the same energy E:
T 0

s2 (t)dt = a

T 0

s2 (t)dt = E. b

Determine the optimal receiver and draw a block diagram. Discuss possible approximations. Problem 2 We have QPSK, where each signal carries two bits. The signals are A sin(2f0 t 2i ), 0 t T M si (t) = 0, elsewhere for i = 1, 2, 3, 4 with f0 T = 1. The phase angles are thus and 3 . 4 4 The transmission is disturbed by additive white gaussian noise, and the detection is done using a correlator receiver followed by an ML detector. We wish to increase the data rate by 50% by letting each signal carry three bits, using 8-PSK, with phase angles , 3 , 5 and 7 . How much do 8 8 8 8 we need to increase the amplitude A in order to maintain the symbol error probability?

Problem 3 Let C be a binary error correcting code with parity check matrix

H=

1 0 0 0

0 1 0 0

0 0 1 0

0 0 0 1

0 1 1 1

1 0 1 1

a1 a2 a3 a4

Determine all possible binary vectors (a1 , a2 , a3 , a4 )T that maximizes the minimum distance of the code. Also, determine the length, dimension and minimum distance of the code. Problem 4 A discrete source has the alphabet A = {X, Y, Z}. An estimation of the statistics of the source follows here. Pr{XX} = 54/110, Pr{Y X} = 0, Pr{ZX} = 6/110, Pr{XY } = 6/110, Pr{Y Y } = 24/110, Pr{ZY } = 0, Pr{XZ} = 0, Pr{Y Z} = 6/110, Pr{ZZ} = 14/110.

1. Determine Pr{X}, Pr{Y } and Pr{Z}. Also, construct a Human code for this source, where the symbols are encoded one by one. 2. Construct a Human code for this source, where the symbols are encoded two by two, based on Pr{X}, Pr{Y } and Pr{Z}, under the assumption that consecutive symbols are independent. 3. Construct a Human code for this source, where the symbols are encoded two by two, based on the probabilities given above. 4. Determine the expected codeword lengths for the three cases. Comments? Problem 5 Consider a pulse s(t) = sinc(at)sinc(bt) where a b. 1. Sketch the spectrum S(f ) of the pulse. 2. Suppose that the pulse is to be used over an ideal real baseband channel with one-sided bandwidth 400 Hz. Determine a and b such that the pulse is Nyquist for 4-PAM at 1200 b/s and exactly lls the channel bandwidth.

3. Now, suppose that the pulse is to be used over a passband channel spanning the frequency band 2.42.42 GHz. Assuming that we use 64-QAM at 60 Mb/s, determine a and b so that the pulse is Nyquist and exactly lls the channel bandwidth. 4. Sketch an argument showing that the magnitude of the transmitted waveform in the preceding settings is always nite. Problem 6 For a time-discrete signal, consider a sample modeled as a stochastic variable X. The signal is supposed to be uniformly quantized. Let Q be the quantization distorsion of that quantization. Determine the smallest number 2 of quantization steps such that Q/X is at most 104 for the following cases. 1. X is uniformly distributed between A and A. 2. We do not know fX (x) exactly, only that we have fX (x) = 0 for |x| A, 2 and X = A2 /12. . Problem 7 The Gaussian disturbance N in Figure I, has mean 0. When two equally probable messages are transmitted using the signals in Figure II, we get the error probability Pe = 0.01 with an ML detector. 1. What is the variance of the noise? 2. Determine the error probability if the channel in Figure I is used to transmit three equally probable messages using the signals in Figure III? 3. Determine the error probability if the channel in Figure I is used to transmit four equally probable messages using the signals in Figure IV? 4. How are the answers in (b) and (c) aected if the noise is known to have E{N } = 1 instead of 0? N
6

?  - + 

X 3

Fig. I.
6

s0 r -2

s1 r s 2 Fig. II.
6

s0 r -4

4 Fig. III.

rs1

s r2 -

s0 r -4

s2 r 4 Fig. IV.

rs1

s r3 -

Problem 8 Let C be the binary (15, 5) cyclic code with parity-check polynomial h(X) = (1 + X)(1 + X + X 4 ). a) Show that C is capable of correcting any combination of three or fewer errors. b) Devise a simple error-trapping decoder for C. c) Which error-patterns of weight at most 3 can not be decoded by this decoder? Problem 9 A time-continuous stochastic process with power-spectral density RX (f ) = R0 , |f | < B, 0, elsewhere,

is sampled with sampling frequency fs = 4B. The sampled signal is ltered by a time-discrete lter with frequency response H[] for which we have |H[]|2 = 1 2||, || < 1 , 2 |H[ + m]|2 , m integer.

Determine the quadratic mean of the output. Problem 10 A memoryless source produces the symbols A, B, C and D with the symbol probabilities P r{A} = 0.4, P r{B} = 0.3, P r{C} = 0.2 and P r{D} = 0.1. a) Construct a Human code that encodes one symbol at a time, and determine the average codeword length per source symbol. b) Extend the source by encoding two symbols at a time. The new source symbols are now AA, AB, AC, AD, BA, BB, BC, BD, CA, CB, CC, CD, DA, DB, DC and DD. Construct a Human code for the extended source, and determine the average codeword length per source symbol from the original source. Problem 11 Let A be n n matrix, and let 1 , 2 , . . . , n be its eigenvalues (counting multiplicities). Show that 4

1. trace(A) = 1 + 2 + + n ; 2. det(A) = 1 2 n ; Problem 12 Find characteristic polynomials, eigenvalues and eigenvectors of the following matrix 1 3 3 3 5 3 . 3 3 1 Problem 13 Compute eigenvalues and eigenvectors of the rotation matrix cos sin sin cos .

Problem 14 Let s be the largest singular value of matrix A, and let be the eigenvalue of A with largest absolute value. Show that || s. Problem 15 Show that the rank of a matrix is the number of its non-zero singular values (counting multiplicities). . Problem 16 Find singular value decomposition of the matrix A= Use this SVD to nd: 1. max 2. min
x =1 x =1

2 3 0 2

Ax and the vectors where the maximum is attained; Ax and the vectors where the minimum is attained.

Problem 17 Find the matrices of the orthogonal projections in R3 onto the planes dened by the equations x1 = 2x2 + x3 and x2 = x1 3x3 . Problem 18 Let X N (0, 1) and let Y = 2X. 1. Show that X and Y are jointly Gaussian. 2. Find the covariance matrix cov((X, Y )T ). 5

Problem 19 Let X = (X1 , X2 , X3 ) be a multinomial random variable with a probability density function fX (x) = n! p x1 p x2 p x3 , x1 !x2 !x3 ! 1 2 3

where x = (x1 , x2 , x3 ), n = x1 + x2 + x3 and p1 + p2 + p3 = 1. Calculate the conditional probability P [X2 |X1 + X2 = y].
2 Problem 20 If Xi N (mi , i ), i = 1, 2, . . . , n are independent random variables, show that X = (X1 , X2 , . . . , Xn )T is a Gaussian random vector, i.e. that for any coecients ci , the random variable n ci Xi is scalar Gaussian i=1 random variable. Calculate the mean and the variance of this random variable.

Problem 21 Let X = (X1 , X2 , X3 )T be a random vector with mean = E[X] = (5, 1, 4)T . The covariance matrix of X is 5 2 1 cov(X) = 2 5 0 . 1 0 4 Calculate the mean and variance of Y = AT X + B, where A = (1, 2, 1)T and B = 3. Problem 22 Consider the continuous-time system with input random process X(t) and output process Y (t):
+10

Y (t) = 4
10

X(t s)ds.

Assume that the input X(t) is a wide-sense stationary process with power spectral density SXX (w) = 5 for < w < . Find the psd SY Y (w) and the autocorrelation function RY Y (w) of the output process. Problem 23 The m random variables X1 , X2 , . . . , Xm and the n random variables Y1 , Y2 , . . . , Yn are all statistically independent exponentially distributed random variables with pdf fXi (v) = fYj (v) = 6 ev , if v 0, 0, if v < 0.

Find the probability density function of the random variable Z= X1 + X 2 + + Xm = Y1 + Y2 + + Yn

m i=1 Xi . n j=1 Yj

Problem 24 Let X = (X1 , X2 , . . . , Xn )T , where X1 , X2 , . . . , Xn are independent Gaussian random variables with expected values zero and equal variance 2 and let
n n n

Y = bT X =
i=1

bi Xi and Z = XT AX =
i=1 j=1

aij Xi Xj .

1. Calculate the joint characteristic function Y Z (1 , 2 ) = E{j1 Y + j2 Z}. 2. Show that Y and Z are statistically independent if Ab = 0. 3. Prove that the estimates m= and 1 n X 1 + X2 + + Xn Xi = n n i=1
n 1 v= (Xi m)2 n 1 i=1

are statistically independent under the conditions of this problem. Problem 25 The stochastic process X(t) is stationary and Gaussian with expected value zero, autocovariance function CX ( ) and spectral density SX (). In terms of CX ( ) nd the autocovariance function CY ( ) of the output Y (t) = X 2 (t) of a quadratic rectier to which X(t) is applied. In terms of SX () nd the spectral density SY () of Y (t). Problem 26 A square matrix A is called nilpotent if Ak = 0 for some k. Prove that if A is nilpotent, then 0 is the only eigenvalue of A. Problem 27 Let A be a Hermitian matrix, i.e. A = A and let 1 2 n be its eigenvalues taken in the decreasing order. Show that k =
E:dim E=k xE,||x||=1

max

min

(Ax, x) =

F :codimF =k1 xF,||x||=1

min

max (Ax, x).

Problems about complex numbers 1. Find all solutions of z 5 = 6i. 2. Find the real part of (cos 0.7 + i sin 0.7)53 . 3. Find all complex numbers z, in Cartesian (rectangular) form such that (z 1)4 = 1. 4. Write ( 3 + i)50 in polar and in Cartesian form. 5. Find all fth roots of 32. 6. Write the following in Cartesian form a + ib, where a and b are real and simplied as much as possible: 1 1 + 1+i 1i (b) e2+i/3 (a) 7. Write all solutions of z 3 = 8i in polar and Cartesian form, simplied as much as possible. 8. Find all complex solutions of the equation z 5 = 1 + i. 9. Find the imaginary part of 2+i . 3i

10. Find the angle between 0 and 2 that is an argument of (1 i)2011 . 11. Find all z such that eiz = 3i. 12. Write (1 i)100 as a + ib with a and b real numbers and simplify your answer. 13. Find the real part of e(5+12i)x , where x is real, and simplify your answer. 14. Find all solutions to z 6 = 8 and plot them in the complex plane. 15. Evaluate sin n . n! n=0 sin n converge? If it converges, what does it 2n n=0

16. For what does converge to?