1. Two different signals X 1 ( z ) and X 2 ( z ) are transmitted through independent channels in close proximity so that they are received linearly mixed as Y1 ( z ) AX 1 ( z ) BX 2 ( z ) and Y2 ( z ) AX 1 ( z ) BX 2 ( z ) and modelled as shown in the following Figure 1. The mixing matrix is A D 1, B H 12 ( z ) and C H 21 ( z ) with H 12 ( z ) and H 21 ( z ) unknown FIR transfer functions. The signals Y1 ( z ) and Y2 ( z ) are further processed by a linear system which has a mixing B G12 ( z ) and C G 21 ( z ) matrix A D 1, B G12 ( z ) and C G21 ( z ) with unknown transfer functions, to produce two new signals U 1 ( z ) and U 2 ( z ) . (a) Determine the relationship between the functions H 12 ( z ), H 21 ( z ), G12 ( z ) and G 21 ( z ) so that each of U 1 ( z ) and U 2 ( z ) is a function of only one of the original signals X 1 ( z ) and X 2 ( z ) . Determine the two alternative solutions possible and the associated proportionality transfer functions in terms of the mixing parameters above. (b) It is not known a priori whether the mixing channel transfer functions are minimum phase or not. Comment on possible limitations of one of the alternative solutions. X 1 ( z) X 2 ( z) Y1 ( z ) A C B D Y2 ( z ) Figure 1 Answer The final mixing matrix is given by the following relationship: A B A B 1 B 1 B 1 BC B B C D C D C 1 C 1 C C 1 B C (a) The first possible design is U 1 ( z ) F1 ( z ) X 1 ( z ) and U 2 ( z ) F2 ( z ) X 2 ( z ) B B 0 B B G12 ( z ) H 12 ( z ) C C 0 C C G 21 ( z ) H 21 ( z ) The second possible design is 1 BC 1 BC 1 H 12 ( z ) H 21 ( z ) 1 B C U 1 ( z ) (1 H 12 ( z ) H 21 ( z )) X 1 ( z ) U 2 ( z ) (1 H 12 ( z ) H 21 ( z )) X 2 ( z ) U 1 ( z ) K1 ( z ) X 2 ( z ) and U 2 ( z ) K 2 ( z ) X 1 ( z ) 1 1 G21 ( z ) B H 12 ( z ) 1 1 1 B C 0 B G12 ( z ) C H 21 ( z ) H ( z ) H 21 ( z ) 1 1 C C H 21 ( z ) 12 B B H 12 ( z ) H 12 ( z ) ( H ( z ) H 21 ( z ) 1) U 1 ( z ) 12 X 2 ( z) H 12 ( z ) ( H ( z ) H 21 ( z ) 1) U 2 ( z ) 12 X 1 ( z) H 12 ( z ) 1 BC 0 C (b) In the second case a problem arises if one or both functions H 12 ( z ), H 21 ( z ) have zeros outside the unit circle. In that case the one or both filters G12 ( z ), G 21 ( z ) will be unstable. 2. (a) Show that the analogue transfer function a (1) ,a 0 sa has a lowpass magnitude response with a monotonically decreasing magnitude response with H 1 ( j 0) 1 and H 1 ( j) 0 . Determine the 3-dB cut-off frequency c at which the gain response is 3 dB below the maximum value of 0 dB at 0 . H 1 ( s) (b) Show that the analogue transfer function s (2) ,a0 sa has a highpass magnitude response with a monotonically increasing magnitude response with H 1 ( j 0) 0 and H 1 ( j) 1 . Determine the 3-dB cut-off frequency c at which the gain response is 3 dB below the maximum value of 0 dB at . H 2 ( s) (c) The lowpass transfer function H 1 ( s ) of equation (1) and the highpass transfer function of equation (2) can be expressed in the form 1 1 H 1 (s) A1 (s) A2 (s), H 2 (s) A1 (s) A2 (s) 2 2 where A1 ( s ) and A2 ( s ) are stable analogue allpass transfer functions. Determine A1 ( s ) and A2 ( s ) . Answer (a) H 1 ( s ) a a a2 2 H 1 ( j ) H 1 ( j ) 2 . sa j a a 2 H 1 ( j) 2 is a monotonically decreasing function with H 1 ( j 0) 1 and H 1 ( j) 0 . Let the 3-dB cut2 off frequency be given by c . Hence, H 1 ( j c ) a2 1 c a . 2 2 a c 2 (b) Similarly, H 2 ( s ) j s 2 2 2 H 2 ( j ) H 2 ( j ) 2 . H 2 ( j) is a 2 sa j a a monotonically increasing function with H 2 ( j 0) 0 and H 2 ( j) 1 . Let the 3-dB cut- c2 2 off frequency be given by c . Hence, H 2 ( j c ) (c) H 1 ( s) a 2 1 c a . 2 a 1 ( s a) s 1 ( s a) 1 1 and H 2 ( s) . s a 2 ( s a) s a 2 ( s a) s 1 . Since sa A1 ( s) and A2 ( s) are both allpass transfer functions. Hence, c2 A1 ( j c ) 1 and A2 ( j c ) 1 , A1 (s) 1 and A2 (s) 3. (a) In an audio application the structure in Figure 3 is used as a “reverberator” to reproduce attenuated versions of an impulse applied as input. Determine its transfer function and show that it is allpass. (b) Propose an alternative canonic or non-canonic signal flow graph with one multiplier that realises the transfer function. Explain every step in your answer. (c) Determine the impulse response of the structure knowing that after 800 msecs the impulse response has an absolute value equal to the 1% of its value at the instant n 1 . The signals are sampled at 10 kHz. x (n) y (n) 1 2 z 1 U (z ) Figure 3 Answer (a) Let the output of the intermediate adder be U (z ) . Then, in the z transform domain we can write X ( z) U ( z ) X ( z ) z 1U ( z ) U ( z ) 1 z 1 At the output adder we have X ( z) Y ( z ) X ( z ) (1 2 ) z 1U ( z ) X ( z ) (1 2 ) z 1 1 z 1 or Y ( z) 1 2 z 1 z 1 2 z 1 z 1 H ( z) (1 2 ) z 1 X ( z) 1 z 1 1 z 1 1 z 1 To show that H (z ) is allpass we can write it as z 1 (1 z ) z 1 (1 z 1 ) H ( z) 1 z 1 1 z 1 (Notice that (1 z 1 ) (1 z 1 ) ) H ( z) From H ( z ) z 1 1 Y ( z ) z 1 we can write Y ( z )(1 z 1 ) X ( z )( z 1 ) and hence 1 X ( z ) 1 z Y ( z ) z 1Y ( z ) X ( z )( z 1 ) For a single multiplier realisation we can write the above equation as Y ( z ) z 1 X ( z ) [ z 1Y ( z ) X ( z )] (b) At this stage we can have various signal flow graphs depending on any additional constraints to be taken into consideration. A direct non-canonic realisation is to generate separately the components on the right hand side of equation Y ( z ) z 1 X ( z ) [ z 1Y ( z ) X ( z )] . With the input and the output nodes defined we have for the second term X (z ) Y (z ) + - + z 1 while the first term of the same equation is directly implemented as X (z ) Y (z ) z 1 On combining the two signal flow graphs we obtain + X (z ) Y (z ) + z 1 + + - + z 1 (c) The impulse response h(n) is given by z 1 z 1 1 h(n) 1 1 1 1 z 1 1 z 1 z n n 1 n 1 h ( n ) n0 0 n0 n 1 (1 2 ) h ( n ) 0 n 1 n0 n0 Since the sampling frequency is 10kHz duration of 800 msecs is equivalent to 8000 samples. The impulse response is 1 h(n) 8000 (1 2 ) (1 2 ) 0.9994 100 4. (a) The transfer function of an ideal real coefficient lowpass filter H LP (z ) has cutoff frequency p and impulse response h LP (n) . Show that: (i) that H LP ( z ) is a highpass filter and of cutoff frequency p . (ii) Find an expression for the impulse response hHP (n) of this highpass filter in terms of the impulse response h LP (n) of the original lowpass filter. (b) It is proposed to use a given stable lowpass digital filter transfer function H LP (z ) of impulse response h LP (n) to derive a bandpass transfer function G(z ) centred at 0 according to the formula: G( z ) H LP ( ze j0 ) H LP ( ze j0 ) . (i) Determine the impulse response g (n) of the bandpass filter in terms of the impulse response h(n) of the lowpass filter and the centre frequency 0 . (ii) Show that the signal flow graph in the figure below realises the transfer function G(z ) . Answer (a) (i) Replacing z with z is a z plane rotation by 180 0 . Hence, the passband p p will be rotated to be around the point z 1 e j that corresponds to frequency , i.e., p p . This is a highpass filter response. (ii) The impulse response hHP (n) is the inverse z transform of the transfer function H LP ( z ) . n n n n H HP ( z ) H LP ( z ) hLP (n)( z ) n hLP (n)(1) n z n n [hLP (n)(1) n ]z n hHP (n) hLP (n)(1) n n (b) n (i) G( z ) H LP ( ze j0 ) H LP ( ze j0 ) [hLP (n)( ze j0 ) n hLP (n)( ze j0 ) n ] n n n n n n n [hLP (n)( ze j0 ) n hLP (n)( ze j0 ) n ] [hLP (n)(e j0 ) n hLP (n)(e j0 ) n ]z n [hLP (n)(e j 0 ) n hLP (n)(e j 0 ) n ]z n hLP (n)[(e j 0 ) n (e j 0 ) n ]z n n n n hLP (n)[(e n (ii) j 0 n ) (e j 0 n ) ]z n n 2hLP (n) cos( 0 n) z n n Thus, g (n) 2hLP (n) cos( 0 n) The output from the lowpass filter of the top part of the structure is the convolution between the impulse response of the filter hLP (n) and the signal 2 x(n) cos( 0 n) k which is 2hLP (n k )x(k ) cos(0 k ) . The output from the lowpass filter of the k bottom part of the structure is the convolution between the impulse response of the k filter hLP (n) and the signal 2 x(n) sin( 0 n) which is 2hLP (n k )x(k ) sin(0 k ) . k Therefore, k k y (n) cos( 0 n) 2hLP (n k )x(k ) cos( 0 k ) sin( 0 n) 2hLP (n k )x(k ) sin( 0 k ) k k k k k k 2hLP (n k )x(k ) cos( 0 n) cos( 0 k ) 2hLP (n k )x(k ) sin( 0 n) sin( 0 k ) k 2hLP (n k )x(k )[cos( 0 n) cos( 0 k ) sin( 0 n) sin( 0 k )] k k 2hLP (n k ) cos 0 (n k ) x(k ) g (n) x(n) k Therefore, the above structure realises the filter 5. (a) Prove that the up-sampler and the down-sampler are linear but tine-varying discrete time systems. (b) Prove that down sampling by a factor of M followed by filtering with a filter with transfer function H (z ) is equivalent with filtering first with a filter with transfer function H ( z M ) followed by down sampling by a factor of M . Answer Bookwork