DSP: Phase and Group Delay of LTI Systems
Digital Signal Processing
Phase and Group Delay of LTI Systems
D. Richard Brown III
D. Richard Brown III 1/7
DSP: Phase and Group Delay of LTI Systems
Review of Basic Concepts
Recall the frequency response of an LTI system with impulse response h[n]
is defined as
X
H(ej ) = h[n]ejn
n=
and represents the complex gain of an LTI system to the eigenfunction
input x[n] = ejn .
The output of an LTI system with frequency response H(ej ) and input
x[n] X(ej ) is
Y (ej ) = H(ej )X(ej ).
This last expression can be converted to phase/magnitude (polar) form as
|Y (ej )| = |H(ej )| |X(ej )|
Y (ej ) = H(ej ) + X(ej )
D. Richard Brown III 2/7
DSP: Phase and Group Delay of LTI Systems
Phase Delay
Previously, weve shown that an LTI system H(ej ) with input sequence
x[n] = A cos(0 n + ) for all n Z yields the output sequence
y[n] = |H(ej0 )|A cos 0 n + + H(ej0 )
Denote (0 ) = H(ej0 ). Then
y[n] = |H(ej0 )|A cos (0 (n + (0 )/0 ) + )
= |H(ej0 )|A cos (0 (n p (0 )) + )
where p := (0 )/0 is called the phase delay of the LTI system at
frequency 0 .
Remarks:
Note that the units of p (0 ) are samples.
Note that p (0 ) is not necessarily an integer.
The phase delay p (0 ) means that the system effectively delays
sinusoids at 0 by p (0 ) samples.
See Matlab function phasedelay.
D. Richard Brown III 3/7
DSP: Phase and Group Delay of LTI Systems
Linear Phase Systems
Definition
A linear phase system is a system with phase response
() = H(ej ) = c for all and any constant c.
For example, suppose we have an LTI system with impulse response
h[n] = {1, 2, 1}.
We can compute the frequency response
X
H(ej ) = h[n]ejn = 1 + 2ej + 1ej2 = (2 cos() + 2)ej
n=
We see that () = H(ej ) = . This is clearly a linear phase system.
Note the phase delay of a linear phase system is p () = ()/ = c.
In other words, all frequencies are delayed by the same amount of time.
D. Richard Brown III 4/7
DSP: Phase and Group Delay of LTI Systems
Group Delay
Suppose we have an LTI system and a narrowband input sequence
x[n] = A[n] cos(0 n + ). The narrowband assumption means that X() is
nonzero only around = 0 .
To analyze how an LTI system affects this narrowband signal, we take a Taylor
series approximation of the phase response of the LTI system for values of close
to 0 . For values of close to 0 , we have
d()
H(ej ) (0 ) + ( 0 ) = (0 ) ( 0 )g (0 ).
d =0
Similarly, for values of close to 0 , we have
j d()
H(e ) (0 ) + ( + 0 ) = (0 ) ( + 0 )g (0 ).
d =0
where
d()
g (x) :=
d =x
is called the group delay (in samples) at normalized frequency x.
D. Richard Brown III 5/7
DSP: Phase and Group Delay of LTI Systems
Group Delay
() = H(ej )
g (0 )
0 0
g (0 )
D. Richard Brown III 6/7
DSP: Phase and Group Delay of LTI Systems
Remarks
Suppose you have a narrowband modulated signal x[n] = s[n] cos(0 n)
that passes through a system with frequency response H(ej ) with phase
delay p (0 ) and group delay g (0 ) at = 0 . It can be shown that the
output in this case is approximately
y[n] s[n g (0 )] cos (0 (n p (0 ))) .
See Oppenheim & Shafer third edition prob. 5.63 for a detailed derivation.
Phase delay specifies the delay (in samples) of the carrier cos(0 n).
Group delay specifies the delay (in samples) of the envelope s[n].
For a linear phase system, g () = p () = c, i.e. the group delay is
the same as the phase delay.
Group delay is also a measure of the deviation from phase linearity of
a system, i.e. if the group delay varies wildly, then the system has
highly nonlinear phase.
See Matlab function grpdelay.
D. Richard Brown III 7/7
