-
Notifications
You must be signed in to change notification settings - Fork 1
/
1.tex
80 lines (53 loc) · 2.2 KB
/
1.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
\subsection*{Linearity and Time Invariance; Time-Domain Representations}
$ y_n = b_0 x_n + b_1 x_n + \cdots + b_Mx_{n*m}
- a_1 y_{n-1} \cdots -a_n y_{n-N}$
\subsection*{Recall the simplest lowpass filter}
$ y_n = x_n + x_{n - 1}
\\
(M=1, N =0) \\
(b_0 = 1, b_1 = 1)$
(FIR, order 1)
$h_n = \delta_n + \delta_{n - 1} = [1, 1, 0, \dots]$\\
freq response:\\
$H(w) = \mbox{DTFT}(h) = 1 + e^{-jwT}$\\
phase response:\\
$\Theta(w) = \angle H(w) = -wt/2 = -\pi f T$ \\
phase delay: \\
$P(\omega) = -\theta(\omega)/N = + T / 2$ \\
group delay: \\
$D(\omega) = -2 / ??? \theta(\omega) = T / 2$ \\
phase delay and group delay are a 1/2 sample delay
These characteristics all apply to a Linear Time Invariant (LTI) filter
\subsection*{Recall Test Sinusoid}
$x(n) = e^{j\omega t_n} | \omega = 2 \pi fs / 4 = 2 \pi / 4T \\
=> \omega T = \pi/2 \\
=> x(n) = e^{j\omega n T} = e^{j (\pi/2) n} = j^n$\\
now, \\
$y(n) = x(n) + x(n -1)\\
= j^n + j^{n - 1} = j^n(1 + 1/j)\\
= (1 - j)j^n = (1 - j)x(n)$
$=> G(\pi / 2) \stackrel{?}{=} \sqrt{1^2 + 1^2} = \sqrt{2} (check)\\
T = 1$
$= 2\cos(\omega T/2) | \omega T = \pi / 2 = 2 \cos(\pi / 4) = 2 / \sqrt{2} = \sqrt{2}$
$\theta(\pi / 2) ?= \tan^{-1}(-1 / 1) = - \pi / 4 \\
= -\omega T / 2 = -\pi/2 1/2 = - \pi / 2$
Thus, $H(\pi / 2) = e^{-j \pi/4} \sqrt{2} \\$
$
= [\cos(\pi / 4) - j \sin(\pi / 4) ] \sqrt{2} \\
= (1 - j)
$\\
$y(n) = j^n + j^{n - 1}$\\
$y(1) = j + j^0 = j + 1 = \sqrt{2} \cdot e^{j(pi/2 - pi/4)}$\\
$y(3) = j^2 + j^2 = -j - 1 = \sqrt{2} \cdot e^{3\pi / 2 = \pi / 4}$\\
\subsection*{Phase Delay: $\Theta(\omega) = \mbox{phase shift (rad)}$}
$P(w) \stackrel{\Delta}{=} - \theta(\omega) / w = - \mbox{phase shift} / \mbox{(rad / sec)} \\
= -\theta / 2 \pi 4 = -(\theta / 2 \pi) \mbox{cycle} \cdot \mbox{Period}$
phase shift gives you a fraction of a period
\subsection*{Group Delay}
$D(\omega) \stackrel{\Delta}{=} -d/d\omega \theta(\omega)$
Measure of delay that is local to each frequency zone.
If the phase is going crazy, you just pick a particular frequency and linearize
at that point.
\\\\
Group delay is phase delay at a particular freq\\
Slope of the linearized delay