Complex exponential A discrete-time signal may be complex-valued. In digital communications complex signals arise naturally. A complex signal may be represented in two forms: z (n) Re z (n) j Im z (n) z (n) e j argz ( n ) ; arg z (n) tan 1 Im z (n) Re z (n) Classification of DT signals Periodic and aperiodic signals: A signal x(n) is said to be periodic if, for some positive real integer N , x ( n) x ( n N ) . Energy signals and power signals: A signal x(n) is said to be an energy signal if its energy is finite. N E lim x(n) . N 2 N P 0 If the average power of a signal is finite then the signal is called power signal. N 1 2 x ( n) . N 2 N 1 N P lim E Simple manipulation of DT signals The time shifting and time reversal operation is shown in Figure (b) and (c) above. Figure (d) shows down-sampling operation and Figure (e) shows up-sampling operation. Mathematically, Down-sampling: f (n) x( Nn) , where N is an integer. f ( n) is formed by taking every N th sample of x(n) . x(n / M ) n 0, M , 2M , otherwise 0 Up-sampling: f (n) . Here there will be M 1 consecutive zeros of f ( n) between consecutive samples of x(n) . Shifting, reversal and time-scaling operations are order-dependent. Therefore, one needs to be careful in evaluating compositions of these functions. Analysis of DT systems T x(n) y(n) T is used to represent a general system. The input signal x(n) is transformed into an output signal y (n) through the transformation T . Classification of systems 1. Linear system: T T y1 (n) and x2 (n) y2 (n) then for a linear system, If x1 (n) T a1 x1 (n) a2 x2 (n) a1 y1 (n) a2 y2 (n) . 2. Time-invariant system: T T y (n n0 ) . If x(n) y(n) then for a time-invariant system, x(n n0 ) 3. BIBO stable system: A system is BIBO stable if and only if whenever the input signal is bounded as x(n) A for all n, the output is also bounded as y(n) B for all n. If this is violated by the input-output pair the system is not stable. 4. Causal system: A system is causal if the output y (n) at any time is independent on any value of x ( n m) for m 0 . 5. Mamoryless system: A system is said to be memoryless if the output at any time n n0 depends only on the input at time n n0 . 6. Invertible system: A system is said to be invertible if the input to the system may be uniquely determined from the output. Example j n cos( n) and ii) x(n) cos(0.125 n) . 17 j N j n n N i) x(n N ) e 16 e 16 cos ; For periodicity, 16 N 2 k1 and 17 N 2 k2 . 17 17 k 17 N 32k1 34k2 1 ; For k1 17 , N 32 17 544 [Ans] k2 16 A. Determine the periodicity of the signals, i) x(n) e 16 So, the period of the signal is, 544. ii) N 2 k 16k . Therefore, N 16. 0.125 B. Find the even and odd part of the signal x(n) nu (n) . Hints: xe (n) [Ans] 1 x(n) x(n) , etc. 2 n C. Is the system y (n) x(n) sin 2 linear? T T y1 (n) and x2 (n) y2 (n) . Then for x(n) a1 x1 (n) a2 x2 (n) , Let, x1 (n) y (n) a1 x1 (n) a2 x2 (n) sin n 2 a1 x1 (n) sin n 2 a2 x2 (n) sin n 2 a1 y1 (n) a2 y2 (n) . As the linearity property holds, the system is linear. D. Is the system y(n) x(n) x(n2 n) causal? For n = -1, y (1) x(1) x(2) . As y (n) depends on the future value of x(n) , the system is noncausal. E. Is the system y(n) e x ( n ) / x(n 1) stable? Let x(n) (n) . Thus, y (n) . Thus the system is not stable. F. Is the system y (n) x(n) x(n 1) invertible? Let, x(n) x1 (n) . Then, y (n) x1 (n) x1 (n 1) . Again let, x(n) x1 (n) C . Then, y (n) x1 (n) x1 (n 1) . In both cases the output is the same. Hence, the system is not invertible. G. Is the system y (n) 2nx(n) time-invariant? T y1 (n) and x1 (n) x(n n0 ) . Then, y1 (n) 2nx1 (n) 2nx(n n0 ) . Let, x1 (n) If the system is time-invariant, y(n n0 ) 2(n n0 ) x1 (n n0 ) . Here, y1 (n) y (n n0 ) . Hence the system is time-variant. [Ans] Response of LTI systems to arbitrary inputs: The convolution sum We denote the response of a system to unit sample sequence as impulse response, h(n). h(n) T [ (n)] For a time-invariant system, h(n k ) T [ (n k )] . y (n) T [ x(n)] = T [ x(k ) (n k )] Now, k y ( n) x(k )T [ (n k )] k k k x ( k ) h( n k ) x ( n ) h ( n ) h ( n ) x ( n ) h ( k ) x ( n k ) The above operation is called convolution operation. Steps to perform convolution: 1. Change the variable from n to k: x(n), h(n) x (k ), h(k ) . 2. Perform the folding operation: h(k ) h(k ) . y (n0 ) 3. Perform the shifting operation: h(k ) h(n0 k ) . 4. Multiplication: x(k )h(n0 k ) vn0 (k ) . 5. Summation: y(n0 ) v n0 x ( k ) h( n k 0 k) (k ) . Example 1. If x(n)=u(n) and h(n) = 0.8nu(n), y(n)=? . y ( n) k n 0.8k u (k ) u (n k ) 0.8k k 0 1 0.8n 1 ; n0 1 0.8 [Ans] 2. x( n) {1, 2,3,1} , h( n) {1, 2,1, 1} , y ( n) ? . x(k ) {1, 2,3,1}, h(k ) {1, 2,1, 1} ; h(k ) {1,1, 2,1} Lower limit of y (n) is 0-1=-1 and the upper limit is 3+2=5. T 1* 1* 2 0 Now, y (1) 1 ; 3 0 1 0 T T 1* 2* 1* 1* 2 1 2 2 y (0) 2 2 4 ; y (1) 1 4 3 8 3 0 3 1 1 0 1 0 T T T 1* 1* 1* 0* 1* 0* 2 1 2 1 2 0 y (2) 1 2 6 1 8 ; y (3) 3 ; y (4) 2 3 2 3 1 3 1 1 1 1 2 1 1 Similarly, y (5) 1 ; y (6) 0 . Thus, y (n) {1, 4,8,8,3, 2, 1} [Ans] 3. Sketch y (n) x(2n 3), and y (n) x(8 3n) for x(n) (6 n)[u (n) u (n 6)] . Properties of convolution 1. Commutative property: x(n) h(n) h(n) x(n) 2. Associative property: {x(n) h1 (n)} h2 (n) x(n) {h1 (n) h2 (n)} 3. Distributive property: x(n) {h1 (n)} h2 (n)} x(n) h1 (n) x(n) h2 (n) # x(n) (n) x(n) and x(n) (n k ) x(n k ) . # if x(n) N1 , h(n) N 2 , then, y(n) N1 N 2 1 . For causal system and causal input the convolution sum becomes, y (n) Some common series n n k 0 k 0 x ( k ) h( n k ) h ( k ) x ( n k ) . Stability of LTI systems y ( n) h(k ) x(n k ) ; Let, x(n) A , then, y (n) h(k ) x(n k ) A k k Therefore, the output y(n) is always bounded if h(k ) is always bounded. k Example: h(n) a nu (n) . S k h( k ) a k k 0 1 ; a 1. 1 a The sum is finite for a 1 . Hence, the system is stable. System described by difference equations Let, a system has h(n) nu (n) . Therefore the response of the system would be, y ( n) k h( k ) x ( n k ) k x ( n k ) y (n 1) k x(n 1 k ) k 0 The first equation may be written as, y (n) x(n) k 0 k x(n k ) . k 1 Putting k=p+1, y(n) x(n) p 0 p 0 p1x(n p 1) x(n) p x(n p 1) x(n) y(n 1) y (n) x(n) y (n 1) . i.e., The above equation is called linear constant coefficient difference equation. The general form of LCCDE is, M N k 0 k 1 y ( n) b( k ) x ( n k ) a ( k ) y ( n k ) (01) where, the coefficients a(k) and b(k) are constants that define the system. if, all a(k ) s are not zero the system is recursive. If a(k)=0 the system is non-recursive. The integer N is called the order of the difference equation or the order of the system. Difference equations provide a way for computing the response of a system, y(n) to an arbitrary input x(n). It is necessary to satisfy a set of initial conditions to solve the above equation. For example, if x(n) begins at n=0, the solution at time n=0 depends on the values of y (1), y (2), , y ( p ) . When the initial conditions are zero, the system is said to be in initial rest. The general solution of an LCCDE system is given as, y (n) yh (n) y p ( n) (02) The homogeneous solution, yh (n) is the response of the system to the initial conditions assuming that the input x(n) 0 . The particular solution is the response of the system to the input x(n) , assuming zero initial conditions. The homogeneous solution may be found by assuming a solution of the form, yh ( n ) z n (03) Substituting in the general equation, z n N a (k ) z nk 0 k 1 N Or, z n N [ z N a (k ) z N k ] 0 , z N a(1) z N 1 a(2) z N 2 a( N ) 0 k 1 The polynomial in equation (04) is called the characteristic polynomial. It has N roots. If all the roots are distinct the homogeneous solution will be of the form, yh (n) A1 z1n A2 z2n AN z Nn If there are m repeated roots, the solution would be, yh (n) ( A11 A12 n A1m n m 1 ) z1n A2 z2n AN z Nn If there are two complex roots, z1 , z2 a jb re j , the solution would be, yh (n) r n ( B1 cos n B2 cos n ) A3 z3n AN z Nn (04) Example: 1. Find yh (n) for the system, y (n) y (n 1) y (n 2) . The characteristic polynomial, z 2 z 1 0 n z 1 5 . 2 n 1 5 1 5 Hence, yh (n) A1 A ; n 0 2 2 2 Let the initial conditions be y (0) 0, y (1) 1 . Applying initial conditions, 1 5 1 5 A1 A2 1 2 2 n n 1 1 5 1 5 yh ( n ) 5 2 2 2. Find yh (n) for the system, y (n) 4 y (n 1) 4 y (n 2) 0 A1 A2 0; The characteristic polynomial, z 2 4 z 4 0; A1 A2 1 5 [Ans] z1 , z2 2, 2 yh (n) ( A11 A12 n)2n [Ans] 3. Find yh (n) for the system, y (n) 2 y (n 1) 2 y (n 2) 0 z1 , z2 1 j1 2e j 3 / 4 The characteristic polynomial, z 2 2 z 2 0; yh ( n ) 2 [ A cos 34n A sin 34n ] n 1 [Ans] 2 --------------------------------------------------------------------------------------------------------------------For the particular solution it is necessary to find the sequence y p (n) that satisfies the difference equation for the given x(n) . Table below lists particular solutions for some common inputs. Example: 1. Find the solution of the difference equation, y (n) 0.25 y (n 2) x(n) for x(n) u ( n) with y (1) 1 and y (2) 0 . y p (n) C1 . Substituting this in the difference equation we get, For x(n) u ( n) , y p (n) 0.25 y p (n 2) 1 , C1 0.25C1 1 ; C1 4 / 3 or, Now, the characteristic polynomial is, z 2 0.25 0 yh (n) A1 (0.5) A2 (0.5) n (01) z 0.5 n The total solution is, y (n) A1 (0.5) n A2 (0.5) n 4 / 3; n0 (02) The total solution only applies to n 0 , therefore, we have to derive an equivalent set of initial conditions and y (1) from the system equation (01). y (0) 0.25 y(2) x(0) 0.25 0 1 1 y (1) 0.25 y (1) x(1) 0.25 1 1 1.25 Using equation (02), 1 A1 A2 4 / 3; 1.25 0.5 A1 0.5 A2 4 / 3 y(n) 0.25(0.5) (1/12)(0.5) 4 / 3; n 0 n n A1 0.25, A2 1/12 [Ans] y (0) 2. Find the response of the system y (n) y (n 1) y (n 2) 0.5 x(n) 0.5 x(n 1) to the input x(n) 0.5n u (n) with initial condition y (1) 0.75 and y (2) 0.25 . z 0.5(1 j 3) e j / 3 . The characteristic equation of the system is, z 2 z 1 0 . Thus, yh (n) A1e j n / 3 A2 e j n / 3 And, y p (n) C1 0.5n u(n); n 0 Substituting this in the difference equation we get, C1 0.5n C1 0.5n 1 C1 0.5n 2 0.5 0.5n 0.5 0.5n 1 C1 0.5 C1 2C1 4C1 0.5 1 ; Therefore, y (n) A1e j n / 3 A2e j n / 3 0.5 0.5 ; n 0 n y(0) y(1) y(2) 0.5 x(0) 0.5 x(1) A1 A2 0.5 1 Applying initial conditions, 1 e j / 3 j / 3 j / 3 0.25 1 y(1) y(0) y(1) 0.5 x(1) 0.5 x(0) A1e A2e 1 j / 3 e 3/ 4 1 A1 0.5 A1 j 2 j / 3 e 3 1 j / 3 A2 A2 0.75 e 3 / 4 2 This yields, y (n) 0.5n 1 3 n (n 1) sin 3 sin 2 3 3 3. Find unit sample response, h(n) of the system described as y (n) [Ans] 3 1 y (n 1) y (n 2) x(n) x(n 1) . 4 8 The impulse response is the response of a system with x(n) (n) and initial rest conditions. The characteristic equation is, z 2 n 3 1 z 0 4 8 1 1 ( z )( z ) 2 4 1 1 z , . 2 4 n 1 1 Hence, y (n) A1 A2 0; n 0 2 4 With initial rest condition, y (1) y (2) 0 , it follows that, 3 1 y (0) y ( 1) y (2) x(0) x(1) 1 A1 A2 4 8 A1 2, A2 3 3 1 3 1 1 1 y (1) y (0) y (1) x(1) x(0) 1 A A 1 2 4 8 4 4 2 4 n n n 1 n 1 1 1 Hence, h(n) 2 3 ; n 0 , or, h(n) 2 3 u (n) [Ans] 2 4 4 2 4. Find the response of the system of problem -3 for the input x(n) u (n) u (n 10) with zero initial condition. Let, s ( n) be the step response of the system. Then, y (n) s (n) s (n 10) . k k n n 1 1 s (n) h(n) u (n) h(k ) 2 3 Now, n0 2 4 k 0 k 0 n 1 n 1 1 (1/ 2) 1 (1/ 4) n n s(n) 2 3 Or, u (n) 2(1/ 2) (1/ 4) u(n) 1 1/ 2 1 1/ 4 Thus, y (n) 2(1/ 2) n (1/ 4) n u (n) 2(1/ 2) n 10 (1/ 4) n 10 u ( n 10) [Ans] 5. Find h(n) of the system described as: y (n) 3 y (n 1) 2 y (n 2) x(n) 3 x(n 1) 2 x(n 2) . h(n) 3h(n 1) 2h(n 2) (n) 3 (n 1) 2 (n 2) z 2 3z 2 0; z1 , z2 1, 2 h(n) C1 C2 2 u (n) C3 (n) n Now, from the difference equation, h(0)=1; h(1)=3+3=6, h(2)=18-2+2=18 (01) Putting these in equation (01) we get, 1 C1 C2 C3 6 C1 2C2 18 C 4C 1 2 C1 6, C2 6, C3 1 h(n) 6 6 2n u ( n) ( n) 6 2 n 1 u ( n) ( n) [Ans] 6. Determine the response y (n) , n 0 , of the system described by the second –order difference equation y (n) 3 y (n 1) 4 y (n 2) x(n) 2 x(n 1) when the input sequence is x(n) 4n u (n) . Here, Ch. Eq.: z 2 3 z 4 0 ( z 1)( z 4) 0 Therefore, z1 , z2 1, 4 . yh (n) c1 (1) c2 (4) . n n As the system root contains one characteristic root, y p (n) c3n(4)n . Putting this in system equation we get, c3n(4) n u (n) 3c3 (n 1)(4) n 1 u (n 1) 4c3 (n 2)(4) n 2 u (n 2) (4) n u (n) 2(4) n 1 u (n 1) For n 2 none of the unit step terms vanish. We can solve c3 for any n 2 . Use n=2 32c3 12c3 24 c3 6 . 5 Thus, the total solution of the difference equation is, y (n) c1 (1) n c2 (4) n Assume that the initial conditions, y (1) y (2) 0 . 6 n 4n . 5 y (0) c1 c2 3 y (1) 4 y ( 2) 1 1 1 26 c1 , and c2 Then, 24 25 25 y (1) c1 4c2 5 3 y (0) 4 y ( 1) 6 9 1 26 6 (1) n (4) n n(4) n ; n 0 Hence, the zero-state response of the system is, y (n) 25 25 5 [Ans] Correlation of D.T. signals The correlation of two sequences is an operation defined by the relation, rxy (l ) ryx (l ) x(k ) y (k l ) x (k l ) y (k ) k k k k y (k ) x(k l ) y (k l ) x (k ) rxy (l ) ryx (l ) The computation of correlation sequence involves the same operation as convolution except for the folding rxy (l ) x(l ) y (l ) . operation, If y (n) x(n) , we have autocorrelation: rxx (l ) x(k ) x(k l ) . k If x(n) and y (n) are causal sequences of length N, (0 n N 1) ,then, N p 1 rxy (l ) x(k ) y (k l ) , where, for l 0 , i l & p 0 and for l 0 , i 0 & p l . k i Applications: Correlation measures the degree to which two signals are similar. This concept is often used in radar, sonar and digital communication. 1. Let, x(n) is the transmitted signal. y (n) is the received signal which is the delayed version of the input signal. Then, y (n) x(n l ) w(n) . Correlation of x(n) and y (n) will be maximum at lag l from where we can measure the distance of a target. 2. In communication system the concept of correlation is used to determine whether the received signal is zero or one. To transmit zero we send x0 ( n) , 0 n L 1 ; To transmit one we send x1 (n) , 0 n L 1 Signal received by the receiver is, y(n) xi (n) w(n); i 0,1 . The receiver compares y (n) with x0 (n) and x1 (n) {pattern available in receiver} to determine the signal that better match y (n) . Example: Determine the correlation between the two sequences x(n) and y (n) given below. x(n) {2, 1,3, 7,1, 2, 3, 7,1, 2, 3} y(n) {1, 1, 2, 2, 4,1, 2,5} Properties of correlation sequences Let, x(n) and y (n) are two finite energy sequences. Now, the energy of the combined sequence, ax(n) y (n l ) : E ax(n) y(n l ) n 2 a 2 rxx (0) ryy (0) 2arxy (l ) 0 rxx (0) rxy (l ) a 0 for any finite value of a . rxy (l ) ryy (0) 1 This equation can be rewritten as: a 1 rxx (0) rxy (l ) is positive semi-definite. r ( l ) r (0) xy yy Thus, the matrix This implies, rxx (0)ryy (0) rxy2 (l ) 0 , For, y (n) x(n) , i.e., rxy (l ) rxx (0)ryy (0) Ex E y . rxx (l ) Ex . Thus autocorrelation attains its maximum value at zero lag. Quadratic Forms A quadratic form ‘q’ in ‘n’ variables x1 , x2 , x1 x2 , x1 x3 , , xn is a linear combination of terms x12 , x22 , , xn2 and cross terms . If n=3, q a11 x12 a22 x22 a33 x32 a12 x1 x2 a21 x2 x1 a13 x1 x3 a31x3 x1 a23 x2 x3 a32 x3 x2 . This sum can be written compactly as a matrix product, q(x) xT Ax , x n , A n n When A AT then, q(x) 0 for all x 0 if and only if A is positive definite (all eigenvalues are positive). In this case ‘q’ is called positive definite. x1 x3 0 x1 x3 0 I A 0 0; x3 x2 0 x3 x2 2 ( x1 x2 ) ( x1 x2 x32 ) 0 ; would be positive if ( x1 x2 x32 ) 0 . Note that, a in the energy equation is an attenuation factor. Therefore the roots of a must not be complex. This implies, rxx (0)ryy (0) rxy2 (l ) 0 . Detection and estimation of periodic signals in noise The period of a signal is first estimated by autocorrelating the noisy signal. The noisy signal is then cross-correlated with a periodic impulse train of period equal to that of the signal. The resulting cross-correlation function is the signal estimate. N p Period of x(n), and N length of signal s ( n ) x ( n) q ( n) ; Let, ( n kN p ) be the periodic impulse train used for autocorrelation. Let N be the number of impulses used for 1 N 1 [ x(n) q(n)] (n kN p j) ; k 0,1, 2, N n j 1 N 1 rs (0) For j=0, [ x(n) q(n)] (n kN p ); k 0,1, 2, N n j 1 rs (0) Or, x(0) q(0) x( N p ) q( N p ) x(2 N p ) q(2 N p ) x( N ) q( N ) N 1 As x(n) is periodic, x(n) x(n kN p ) ; Thus, rs (0) N x(0) q(0) q( N p ) q(2 N p ) q( N ) . N correlation. Or, rs ( j ) 1 rs (0) x(0) N N /Np q(kN k 0 Similarly, for the other values of j, rs ( j ) p ) , The second term is zero. x( j ) , which is the required signal.