Generalized Linear Phase Quote of the Day The mathematical sciences particularly exhibit order, symmetry, and limitation; and these are the greatest forms of the beautiful. Aristotle Content and Figures are from Discrete-Time Signal Processing, 2e by Oppenheim, Shafer, and Buck, ©1999-2000 Prentice Hall Inc. Linear Phase System • Ideal Delay System Hid ej e j • Magnitude, phase, and group delay H e grdH e Hid e j 1 id j id • Impulse response hid n • If =nd is integer j sinn n hid n n nd • For integer linear phase system delays the input yn xn hid n xn n nd xn nd Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 2 Linear Phase Systems • For non-integer the output is an interpolation of samples • Easiest way of representing is to think of it in continuous hc t t T and Hc j e jT • This representation can be used even if x[n] was not originally derived from a continuous-time signal • The output of the system is yn xnT T • Samples of a time-shifted, band-limited interpolation of the input sequence x[n] • A linear phase system can be thought as H e j H e j e j • A zero-phase system output is delayed by Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 3 Symmetry of Linear Phase Impulse Responses • Linear-phase systems H e j H e j e j =5 • If 2 is integer – Impulse response symmetric h2 n hn =4.5 =4.3 Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 4 Generalized Linear Phase System • Generalized Linear Phase Ae e He j j j j A e j : Real function of and constants • Additive constant in addition to linear term • Has constant group delay d grd H e j arg H e j d • And linear phase of general form arg H e j Copyright (C) 2005 Güner Arslan 0 351M Digital Signal Processing 5 Condition for Generalized Linear Phase • We can write a generalized linear phase system response as H e j A e j e j j A e j cos jA e j sin He hne j n jn hn cosn j hn sinn n n • The phase angle of this system is hn sinn sin n cos hn cosn n • Cross multiply to get necessary condition for generalized linear phase n n hn cosn sin hnsinn cos 0 hncosn sin sinn cos 0 n n n hn sin n hnsin n 0 Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 6 Symmetry of Generalized Linear Phase • Necessary condition for generalized linear phase hn sin n 0 • For =0 or n h2 n hn hn sinn 0 n • For = /2 or 3/2 h2 n hn hn cosn 0 n Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 7 Causal Generalized Linear-Phase System • If the system is causal and generalized linear-phase hM n hn • Since h[n]=0 for n<0 we get hn 0 n 0 and n M • An FIR impulse response of length M+1 is generalized linear phase if they are symmetric • Here M is an even integer Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 8 Type I FIR Linear-Phase System • Type I system is defined with symmetric impulse response hn hM n for 0 n M – M is an even integer • The frequency response can be written as hne He M j jn n0 e jM / 2 M / 2 a n cos n n 0 • Where a0 hM / 2 ak 2hM / 2 k Copyright (C) 2005 Güner Arslan for k 1,2,...,M/2 351M Digital Signal Processing 9 Type II FIR Linear-Phase System • Type I system is defined with symmetric impulse response hn hM n for 0 n M – M is an odd integer • The frequency response can be written as hne He j M jn n0 e jM / 2 M1 / 2 1 bn cos n 2 n 1 • Where bk 2hM 1 / 2 k for k 1,2,..., M 1/2 Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 10 Type III FIR Linear-Phase System • Type I system is defined with symmetric impulse response hn hM n for 0 n M – M is an even integer • The frequency response can be written as hne He M j jn n0 je • Where jM / 2 M / 2 c n sin n n 1 ck 2hM / 2 k for k 1,2,..., M/2 Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 11 Type IV FIR Linear-Phase System • Type I system is defined with symmetric impulse response hn hM n for 0 n M – M is an odd integer • The frequency response can be written as hne He j M jn n0 jM / 2 je M1 / 2 1 dn sin n 2 n1 • Where dk 2hM 1 / 2 k for k 1,2,..., M 1/2 Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 12 Location of Zeros for Symmetric Cases • For type I and II we have z hn hM n Hz zMH z1 • So if z0 is a zero 1/z0 is also a zero of the system • If h[n] is real and z0 is a zero z0* is also a zero • So for real and symmetric h[n] zeros come in sets of four • Special cases where zeros come in pairs – If a zero is on the unit circle reciprocal is equal to conjugate – If a zero is real conjugate is equal to itself • Special cases where a zero come by itself – If z=1 both the reciprocal and conjugate is itself • Particular importance of z=-1 H 1 1 H 1 M – If M is odd implies that H 1 0 – Cannot design high-pass filter with symmetric FIR filter and M odd Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 13 Location of Zeros for Antisymmetric Cases • For type III and IV we have z hn hM n Hz zMH z1 • All properties of symmetric systems holds • Particular importance of both z=+1 and z=-1 – If z=1 H1 H1 H1 0 • Independent from M: odd or even – If z=-1 H 1 1 H 1 M1 • If M+1 is odd implies that H 1 0 Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 14 Typical Zero Locations Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 15 Relation of FIR Linear Phase to Minimum-Phase • In general a linear-phase FIR system is not minimum-phase • We can always write a linear-phase FIR system as Hz Hmin zHuc zHmax z • Where • • • • Hmax z Hmin z1 zMi And Mi is the number of zeros Hmin(z) covers all zeros inside the unit circle Huc(z) covers all zeros on the unit circle Hmax(z) covers all zeros outside the unit circle Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 16 Example • Problem 5.45 Copyright (C) 2005 Güner Arslan 351M Digital Signal Processing 17