ESE 531: Digital Signal Processing
Lec 8: February 12th, 2019 Sampling and Reconstruction
Penn ESE 531 Spring 2019 - Khanna
Lecture Outline
! Review " Ideal sampling " Frequency response of sampled signal " Reconstruction " Anti-aliasing filtering
! DT processing of CT signals " Impulse Invariance
! CT processing of DT signals (why??)
2 Penn ESE 531 Spring 2019 - Khanna
Last Time…
Sampling, Frequency Response of Sampled Signal, Reconstruction, Anti-aliasing filtering
Penn ESE 531 Spring 2019 - Khanna 3
DSP System
4 Penn ESE 531 Spring 2019 - Khanna
Ideal Sampling Model
5 Penn ESE 531 Spring 2019 - Khanna
Frequency Domain Analysis
! How is x[n] related to xs(t) in frequency domain?
6 Penn ESE 531 Spring 2019 - Khanna
Frequency Domain Analysis
7 Penn ESE 531 Spring 2019 - Khanna
Ωs
2⋅T = π
ω =ΩT
X (e jω )
ω
Frequency Domain Analysis
8 Penn ESE 531 Spring 2019 - Khanna
Ωs
2⋅T = π
ω =ΩT
X (e jω )
ω
Reconstruction of Bandlimited Signals
! Nyquist Sampling Theorem: Suppose xc(t) is bandlimited. I.e.
! If Ωs≥2ΩN, then xc(t) can be uniquely determined from its samples x[n]=xc(nT)
! Bandlimitedness is the key to uniqueness
9 Penn ESE 531 Spring 2019 - Khanna
Mulitiple signals go through the samples, but only one is
bandlimited within our sampling band
Reconstruction in Frequency Domain
10 Penn ESE 531 Spring 2019 - Khanna
Reconstruction in Time Domain
11 Penn ESE 531 Spring 2019 - Khanna
*
= The sum of “sincs”
gives xr(t) # unique signal that is
bandlimited by sampling bandwidth
Aliasing
! If ΩN>Ωs/2, xr(t) an aliased version of xc(t)
12 Penn ESE 531 Spring 2019 - Khanna
Anti-Aliasing Filter
13 Penn ESE 531 Spring 2019 - Khanna
1
-ΩN ΩN
XC ( jΩ)X LP ( jΩ)
ΩN
ΩS/2
XS ( jΩ)1/T
1
-ΩN ΩN
XC ( jΩ)
ΩS/2
ΩN
ΩS/2
XS ( jΩ)1/T
Reconstruction in Frequency Domain
14 Penn ESE 531 Spring 2019 - Khanna
Different T?
Signal Processing
! Use theory of sampling (C/D) and reconstruction (D/C) to implement signal processing
! Two cases: " Discrete-time processing of continuous-time signals
15 Penn ESE 531 Spring 2019 - Khanna
x[n] y[n]
Signal Processing
! Use theory of sampling (C/D) and reconstruction (D/C) to implement signal processing
! Two cases: " Discrete-time processing of continuous-time signals
" Continuous-time processing of discrete-time signals
16 Penn ESE 531 Spring 2019 - Khanna
x[n] y[n]
Discrete-Time Processing of Continuous Time
17 Penn ESE 531 Spring 2019 - Khanna
x[n] y[n]
T T
Discrete-Time Processing of Continuous Time
18 Penn ESE 531 Spring 2019 - Khanna
x[n] y[n]
X (e jω ) = 1T
Xc jωT−2πkT
⎛
⎝⎜
⎞
⎠⎟
⎡
⎣⎢
⎤
⎦⎥
k=−∞
∞
∑
T T
Ωs
2⋅T = π
X (e jω )
ω
Discrete-Time Processing of Continuous Time
19 Penn ESE 531 Spring 2019 - Khanna
x[n] y[n]
X (e jω ) = 1T
Xc jωT−2πkT
⎛
⎝⎜
⎞
⎠⎟
⎡
⎣⎢
⎤
⎦⎥
k=−∞
∞
∑
T T
Sum of scaled shifted sincs
yr (t) = y[n] sin[π (t − nT ) /T ]π (t − nT ) /Tn=−∞
∞
∑
Discrete-Time Processing of Continuous Time
20 Penn ESE 531 Spring 2019 - Khanna
x[n] y[n]
X (e jω ) = 1T
Xc jωT−2πkT
⎛
⎝⎜
⎞
⎠⎟
⎡
⎣⎢
⎤
⎦⎥
k=−∞
∞
∑
T T
yr (t) = y[n] sin[π (t − nT ) /T ]π (t − nT ) /Tn=−∞
∞
∑
! If h[n]/H(ejω) is LTI " Is the whole system from xc(t)#yc(t) LTI?
Discrete-Time Processing of Continuous Time
21 Penn ESE 531 Spring 2019 - Khanna
x[n] y[n]
X (e jω ) = 1T
Xc jωT−2πkT
⎛
⎝⎜
⎞
⎠⎟
⎡
⎣⎢
⎤
⎦⎥
k=−∞
∞
∑
T T
yr (t) = y[n] sin[π (t − nT ) /T ]π (t − nT ) /Tn=−∞
∞
∑
! If xc(t) is bandlimited by Ωs/T=π/T, then, Yr ( jΩ)
Xc ( jΩ)= Heff ( jΩ) =
H (e jω )ω=ΩT
Ω <Ωs /T
0 else
⎧⎨⎪
⎩⎪
Example 1
! Consider the following system
! Where
! What is the effective frequency response of the system? What happens to a signal bandlimited by ΩN?
22 Penn ESE 531 Spring 2019 - Khanna
H (e jω ) =1 ω <ωc
0 ωc < ω ≤ π
⎧
⎨⎪
⎩⎪
T T
Example 2
! DT implementation of an ideal CT bandlimited differentiator
! The ideal CT differentiator is defined by
! With corresponding
23 Penn ESE 531 Spring 2019 - Khanna
yC (t) =ddt[xC (t)]
HC ( jΩ) = jΩ
T T
Impulse Invariance
! Want to implement continuous-time system…
24 Penn ESE 531 Spring 2019 - Khanna
Impulse Invariance
! Want to implement continuous-time system…
! …in discrete-time
25 Penn ESE 531 Spring 2019 - Khanna
Impulse Invariance
! With Hc(jΩ) bandlimited, choose
! With the further requirement that T be chosen such that
26 Penn ESE 531 Spring 2019 - Khanna
H (e jω ) = Hc ( jΩ)Ω=
ωT
, ω < π
Hc ( jΩ) = 0, Ω ≥ π /T
Impulse Invariance
! With Hc(jΩ) bandlimited, choose
! With the further requirement that T be chosen such that
27 Penn ESE 531 Spring 2019 - Khanna
Hc ( jΩ) = 0, Ω ≥ π /T
h[n]=Thc (nT )
H (e jω ) = Hc ( jΩ)Ω=
ωT
, ω < π
Impulse Invariance
! Let,
28 Penn ESE 531 Spring 2019 - Khanna
h[n]= hc (nT )
Impulse Invariance
! Let, ! If sampling at Nyquist Rate then
29 Penn ESE 531 Spring 2019 - Khanna
h[n]= hc (nT )
H (e jω ) = 1T
Hc jωT−2πkT
⎛
⎝⎜
⎞
⎠⎟
⎡
⎣⎢
⎤
⎦⎥
k=−∞
∞
∑
Impulse Invariance
! Let, ! If sampling at Nyquist Rate then
30 Penn ESE 531 Spring 2019 - Khanna
h[n]= hc (nT )
H (e jω ) = 1T
Hc jωT−2πkT
⎛
⎝⎜
⎞
⎠⎟
⎡
⎣⎢
⎤
⎦⎥
k=−∞
∞
∑
Hc ( jΩ) = 0, Ω ≥ π /T
H (e jω ) = 1THc j
ωT
⎛
⎝⎜
⎞
⎠⎟, |ω|<π
Impulse Invariance
! Let, ! If sampling at Nyquist Rate then
31 Penn ESE 531 Spring 2019 - Khanna
h[n]= hc (nT )
H (e jω ) = 1T
Hc jωT−2πkT
⎛
⎝⎜
⎞
⎠⎟
⎡
⎣⎢
⎤
⎦⎥
k=−∞
∞
∑
Hc ( jΩ) = 0, Ω ≥ π /T
H (e jω ) = 1THc j
ωT
⎛
⎝⎜
⎞
⎠⎟, |ω|<π
T
T
T
Impulse Invariance
! Want to implement continuous-time system in discrete-time
32 Penn ESE 531 Spring 2019 - Khanna
h[n]=Thc (nT )
Example 3: DT Lowpass Filter
! We wish to implement a lowpass filter with cutoff frequency Ωc on continuous time signal in discrete time with the following system
33 Penn ESE 531 Spring 2019 - Khanna
Continuous-Time Processing of Discrete-Time
! Useful to interpret DT systems with no simple interpretation in discrete time
34 Penn ESE 531 Spring 2019 - Khanna
Is the effective H(ejω) LTI?
Continuous-Time Processing of Discrete-Time
35 Penn ESE 531 Spring 2019 - Khanna
Xc ( jΩ) =TX (e jΩT ) Ω < π /T
0 else
⎧⎨⎪
⎩⎪
Continuous-Time Processing of Discrete-Time
36 Penn ESE 531 Spring 2019 - Khanna
Xc ( jΩ) =TX (e jΩT ) Ω < π /T
0 else
⎧⎨⎪
⎩⎪
Yc ( jΩ) = Hc ( jΩ)Xc ( jΩ)Also bandlimited
Continuous-Time Processing of Discrete-Time
37 Penn ESE 531 Spring 2019 - Khanna
Yc ( jΩ) = Hc ( jΩ)Xc ( jΩ)
Y (e jω ) = 1T
Yc j Ω− kΩs( )⎡⎣
⎤⎦
k=−∞
∞
∑Ω=ω /T
=1TYc ( jΩ)
Ω=ω /T
Continuous-Time Processing of Discrete-Time
38 Penn ESE 531 Spring 2019 - Khanna
Yc ( jΩ) = Hc ( jΩ)Xc ( jΩ)
Y (e jω ) = 1T
Yc j Ω− kΩs( )⎡⎣
⎤⎦
k=−∞
∞
∑Ω=ω /T
=1TYc ( jΩ)
Ω=ω /T
Continuous-Time Processing of Discrete-Time
39 Penn ESE 531 Spring 2019 - Khanna
Yc ( jΩ) = Hc ( jΩ)Xc ( jΩ) Y (e jω ) = 1TYc ( jΩ)
Ω=ω /T
Continuous-Time Processing of Discrete-Time
40 Penn ESE 531 Spring 2019 - Khanna
Yc ( jΩ) = Hc ( jΩ)Xc ( jΩ) Y (e jω ) = 1TYc ( jΩ)
Ω=ω /T
Y (e jω ) = 1THc ( jΩ) Ω=ω /T Xc ( jΩ) Ω=ω /T
=1THc ( jΩ) Ω=ω /T (TX (e
jω ))
Continuous-Time Processing of Discrete-Time
41 Penn ESE 531 Spring 2019 - Khanna
Yc ( jΩ) = Hc ( jΩ)Xc ( jΩ) Y (e jω ) = 1TYc ( jΩ)
Ω=ω /T
Y (e jω ) = 1THc ( jΩ) Ω=ω /T Xc ( jΩ) Ω=ω /T
=1THc ( jΩ) Ω=ω /T (TX (e
jω )) ω < π
H (e jω )
Example
42 Penn ESE 531 Spring 2019 - Khanna
Example: Non-integer Delay
! What is the time domain operation when Δ is non-integer? I.e Δ=1/2
43 Penn ESE 531 Spring 2019 - Khanna
δ[n]↔1
δ[n− nd ]↔ e− jωnd
Example: Non-integer Delay
! What is the time domain operation when Δ is non-integer? I.e Δ=1/2
44 Penn ESE 531 Spring 2019 - Khanna
delay of ΔT in continuous time
Example: Non-integer Delay
! What is the time domain operation when Δ is non-integer? I.e Δ=1/2
45 Penn ESE 531 Spring 2019 - Khanna
delay of ΔT in continuous time
Example: Non-integer Delay
46 Penn ESE 531 Spring 2019 - Khanna
Example: Non-integer Delay
47 Penn ESE 531 Spring 2019 - Khanna
Example: Non-integer Delay
48 Penn ESE 531 Spring 2019 - Khanna
Example: Non-integer Delay
! The block diagram is for interpretation/analysis only
49 Penn ESE 531 Spring 2019 - Khanna
yc (t) = xc (t −TΔ)
Example: Non-integer Delay
! The block diagram is for interpretation/analysis only
50 Penn ESE 531 Spring 2019 - Khanna
yc (t) = xc (t −TΔ) y[n]= yc (nT ) = xc (nT −TΔ)
= x[k]k∑ sinc t − kT −TΔ
T
⎛
⎝⎜
⎞
⎠⎟t=nT
= x[k]k∑ sinc n− k −Δ( )
Example: Non-integer Delay
! The block diagram is for interpretation/analysis only
51 Penn ESE 531 Spring 2019 - Khanna
yc (t) = xc (t −TΔ) y[n]= yc (nT ) = xc (nT −TΔ)
= x[k]k∑ sinc t − kT −TΔ
T
⎛
⎝⎜
⎞
⎠⎟t=nT
= x[k]k∑ sinc n− k −Δ( )
! My delay system has an impulse response of a sinc with a continuous time delay
52 Penn ESE 531 Spring 2019 - Khanna
Example: Non-integer Delay
! My delay system has an impulse response of a sinc with a continuous time delay
53 Penn ESE 531 Spring 2019 - Khanna
Example: Non-integer Delay
! My delay system has an impulse response of a sinc with a continuous time delay
54 Penn ESE 531 Spring 2019 - Khanna
Example: Non-integer Delay
! My delay system has an impulse response of a sinc with a continuous time delay
55 Penn ESE 531 Spring 2019 - Khanna
Example: Non-integer Delay
Big Ideas
! Sampling and reconstruction " Rely on bandlimitedness for unique reconstruction
! CT processing of DT " Effectively LTI if no aliasing
! DT processing of CT " Always LTI " Useful for interpretation
! Changing the sampling rates next time " Upsampling, downsampling
56 Penn ESE 531 Spring 2019 - Khanna
Admin
! HW 3 due Sunday
57 Penn ESE 531 Spring 2019 - Khanna