The Inverse z-Transform The inverse z transformation can be performed by using a few results from complex variable theory. Suppose that we integrate one (1) from z=0 to z=1: 1 0 1 (1)dz z 1 0 1. 0 The integrals become more involved when the path of integration is in the complex plane. For example, suppose that we wish to integrate around a unit circle in the (complex) z-plane. To do the integration we make a substitution of variables: z=ejq. We have dz = j ejq dq. Im{z} q p/2 qp ze q radius = 1 q 3p/2 q0 jq Re{z} Suppose, first, that we wish integrate one (1) around a semicircle C around the origin from 0 to p. Im{z} z e jq C qp q q0 Re{z} p ( 1 ) dz je d q jq 0 C p j cosq j sin q dq 0 j 0 j 2 2. Suppose that we integrate one (1) around a closed unit circle C around the origin (0 to 2p). Im{z} C z e jq q q0 q 2p Re{z} We have, ( 1 ) dz 2p 0 jq je dq 0. C (The areas under the sine and cosine sum to zero over a full period.) Suppose that we integrate z around a closed unit circle C around the origin (0 to 2p). 2p z dz e je d q jq jq 0 C 2p j e 0 2 jq dq 0. Similarly, if we integrate z2 , z3 or zk (k 0) around the unit circle we get zero. What about integrating z-1? 2p z dz e 1 0 jq jq je dq C j 2p 0 dq 2p j. We do not get zero when we integrate z-1 around the unit circle. What about integrating z-2? 2p z dz e 2 j 2q 0 jq je dq C 2p j e 0 jq dq 0. Similarly, if we integrate z-3 , z-4 or z-k (k>1) around the unit circle we get zero. So integrating z-1 is unique, and z dz 2 p j . 1 C So if we were to integrate a z-transform X ( z) n x [ n ] z , n we get X ( z)dz C n x[n] z n dz 2p jx[1]. C (Only z-1 does not integrate to zero.) In other words the integration “picks-out” the sample x1. We can “pick-out” another sample by adjusting the power of z: X ( z) zdz C n x[n] z n1 2p jx[2]. C (The integral is zero except when –n+1 = -1, or n=2.) Similarly, 2 X ( z ) z dz C n xk z n2 2p jx[3], C and m1 X ( z ) z dz C n xk z nm1 2p jx[m]. C Thus, we have found an expression for the inverse z-transform: x[n] 1 X ( z ) z dz . n 1 2p j C The path of the integral must be carried-out over a region of convergence of X(z). [CROC.] Example: Find the inverse z-transform of X(z)=1 using the complex inversion integral. Solution: Plugging-in X(z)=1, we have x[n] 1 z 2p j C n 1 dz. We know that the integral is equal to zero except when n=0. When n=0, the integral is equal to 2pj. Thus, the final inverse transform is equal to x[n] 1 2p j 2p j 1 when n=0. Hence we have x[n]=d[n]. Example: Find the inverse z-transform of X(z)=z-k using the complex inversion integral. Solution: Plugging-in X(z)=z-k, we have x[n] 1 z 2p j C k n 1 z dz. We know that the integral is equal to zero except when n=0. When n-k=0, or when n=k. Hence we have x[n]=d[n-k]. Example: Find the inverse z-transform of 1 X ( z) 1 1 z using the complex inversion integral. Solution: Let us expand 1/(1-z-1) using the geometric series: 1 1 2 1 z z 1 1 z Solution: Plugging this into the complex inversion integral we have x[n] 1 z 2p j 1 1 2 n 1 z z dz. C For n=0, the only term in 1+z-1+z-2+… which “survives” a non-zero integration is 1. For n=1, the only term in 1+z-1+z-2 which “survives” is z-1. We see that there is exactly one term that “survives” when n0. The resultant integral in each case is 2pj, and the corresponding value for x[n] in each case is one. Thus, x[n] u[n]. Example: Find the inverse z-transform of 1 X ( z) 1 2 (1 z ) using the complex inversion integral. Solution: Using the expansion for 1/(1-z-1), we have 2 2 1 1 2 1 z z 1 1 z 1 2 1 2 z 3z Solution: Plugging this into the complex inversion integral we have x[n] 1 2z 2p j 1 1 2 n 1 3z z dz. C For n=0, the only term in 1+z-1+z-2+… which “survives” a non-zero integration is 1. For n=1, the only term in 1+z-1+z-2 which “survives” is 2z-1. The resultant inverse z-transform is x[n] (n 1)u[n 1] (n 1)u[n]. Example: Find the inverse z-transform of 1 X ( z) 1 1 az using the complex inversion integral. Solution: Expanding 1/(1-az-1) as a geometric series: 1 1 2 2 1 az a z 1 1 az Plugging this into the complex inversion integral we have x[n] 1 az 2p j 1 1 2 2 n 1 a z z dz. C For n=0, the only term in 1+az-1+a2z-2+… which “survives” a non-zero integration is 1. For n=1, the only term in 1+az-1+a2z-2 which “survives” is az-1. The resultant integral in the first case is 2pj, the resultant integral in the second case is 2pja, etc. The inverse z-transform is x[n] a nu[n]. Example: Find the inverse z-transform of 1 1 X ( z) 1 1 1 az 1 bz using the complex inversion integral. Solution: We expand both 1/(1-az-1) and 1/(1-bz-1): 1 1 2 2 1 az a z 1 1 az 1 1 2 2 1 bz b z 1 1 bz The product is 1 1 1 1 1 az 1 bz 1 2 2 1 2 2 1 az a z 1 bz b z 1 2 1 (a b) z (a ab b ) z 2 2 Plugging this into the complex inversion integral, we wind-up with 1 a b x[n] 2 2 a ab b ... (n 0), ( n 1), (n 2), This same problem can be solved in a slightly different manner: 1 X ( z) 1 1 (1 az )(1 bz ) 2 z ( z a )(z b) z z ( z a )(z b) B A z z a z b A B , 1 1 1 az 1 bz where z 1 A 1 z b z a 1 bz , z a and z B za z b 1 1 1 az . z b The inverse z-transform is x[n] Aa Bb u[n]. n n It can be shown algebraically (exercise) that this expression is equivalent to the previous expression for x[n]. In general, to find the inverse z-transform of 1 1 1 X ( z) , 1 1 1 1 az 1 bz 1 cz we perform a partial fraction expansion: A B C X ( z) , 1 1 1 1 az 1 bz 1 cz where 1 A (1 az ) X ( z ) z a 1 B (1 bz ) X ( z ) z b 1 C (1 cz ) X ( z ) etc. z c , , , The resultant inverse z-transform is x[n] Aan Bbn Cc n u[n]. Example: Find the inverse z-transform of z 1 X ( z) e . using the complex inversion integral. Solution: We take the MacLaurin series for ez-1 e z 1 1 2 3 1 z z z . 1 2 1 3! The inverse z-transform follows rather naturally: 1 x[n] . n! We could obtain this result through the complex inversion integral, or we could recognize that x[n]=1/n! corresponds to the coefficients of the ztransform.