ESE 547, Digital Signal Processing 1

Digital Signal Processing

ESE 547
ESE 547, Digital Signal Processing 2

Part 1

System Overview
Sampled-Signal Processing Basics
Frequency-Domain and Transforms
Signal Sampling and Reconstruction
ESE 547, Digital Signal Processing 3

1.1 System Overview

• Applications
• Drawbacks
• Advantages
• System Structure
ESE 547, Digital Signal Processing 4

1.1.1 Applications

• Consumer:
– CD

– DAT

– HDTV (Future)

– Speech Recognition

• Military:
– Radar

– Sonar

– Guidance

– Fundamental Technology
ESE 547, Digital Signal Processing 5

Applications (Continued)

• Industrial:
– Control

– Medical Imaging

– Medical Signals (EEG, ECG, EMG)

– Geophysical Exploration

– Nondestructive Testing

– Mechanical Monitoring and Diagnosis

• COMMUNICATIONS
ESE 547, Digital Signal Processing 6

1.1.2 Drawbacks

• Processing Bottlenecks
– A/D

– Processor

• High Initial Cost; justified by

– High performance

– High complexity

– Leverage

– Flexibility
ESE 547, Digital Signal Processing 7

1.1.3 Advantages

• Precision
• Flexibility
• Low incremental Cost
• High capability
ESE 547, Digital Signal Processing 8

1.1.4 Digital Signal Processing System Structure

• Anti-Aliasing Filter
• Sample-and-Hold (or Track-and-Hold)
• A-to-D Converter
• Digital Processor
• D/A converter
• Smoothing Filter

Filters and conversion: extremely important; will cover later.

Conversion process sets a limit on the signal bandwidth

possible – usually < 1/2 sampling frequency.
Present focus on processing algorithms.
ESE 547, Digital Signal Processing 9

1.2 Sampled-Signal Processing Basics

• Input-Output Description and Impulse Response

• FIR & IIR Systems
• Processing Elements and Difference Equations
ESE 547, Digital Signal Processing 10

1.2.1 Input-Output Description

Input: a sequence u = (un ); output y = (yn )

y = B(u)

Linear:

B(u + v) = B(u) + B(v)

B(αu) = αB(u)

for all constants α.

Time-Invariant:

If

(yn ) = B((un ))

then

(yn+k ) = B((un+k ))

for all k .
ESE 547, Digital Signal Processing 11

Input-Output Description (continued):

Fundamental fact: If the processing is linear and

time-invariant then it must be a convolution – so convolution
is the main task of DSP chips.

Convolution is defined by

(yn ) = (un ) ∗ (hn )

where
∞
X
yn = hk un−k
k=−∞

and hn is the impulse response given by:

(hn ) = B((δn ))

where (δn ) is the discrete-time unit (im)pulse:


 1 n=0
δn =
 0 n 6= 0

Causal: hn = 0 for n < 0.

ESE 547, Digital Signal Processing 12

1.2.2 FIR & IIR Systems

Special convolution cases:

1. FIR filtering – Finite Impulse Response, usually causal,

simplest; a direct convolution whose impulse response
is given by the filter coefficients:

yn = h0 un + h1 un−1 + h2 un−2 · · · + hM un−M

M
X
= hi un−i
i=0

2. IIR filtering – Infinite Impulse Response, causal, more

complex than FIR; uses past outputs:

yn = b0 un + b1 un−1 + · · · + bM un−M
−a1 yn−1 − a2 yn−2 − · · · − aN yn−N
M
X N
X
= bi un−i − ak yn−k
i=0 k=1

3. FFT – Fast Fourier Transform: usually noncausal, can

be used for causal, usually FIR
ESE 547, Digital Signal Processing 13

FIR & IIR Systems (continued):

FIR Terminology:

M
X
yn = b0 un + b1 un−1 + · · · + bM un−M = bi un−i
i=0

— Finite Discrete Convolution

— Transversal Filter

— Tapped Delay Line

Other Algorithms:

1. FFT is also widely used for spectral analysis and

spectral estimation

2. Adaptive processing is nonlinear/time-varying

3. Many coding/modulation and decoding/demodulation

and detection techniques are nonlinear.

Note: a DSP chip’s architecture must have fast

multiply-adds, efficient shifting (circular buffers), and
zero-overhead (non-branching) looping; also usually has
bit-reversed addressing.
ESE 547, Digital Signal Processing 14

1.2.3 Processing Elements and Difference Equations

For Linear Time-Invariant processing, the basic processing

elements are:

• Unit Delays
• Constant gains
• Adders
Mathematically, interconnections of these elements give
difference equations.

Schematically, we get block diagrams or signal flow graphs.

Note: z −1 is used to denote a unit delay, even in a

time-domain block diagram or signal flow graph!

Examples: FIR: moving average filter:

yn = 0.5un + 0.5un−1

IIR: (digital) integrator:

yn = un + yn−1

Block Diagrams: note that the FIR has no (directed) loops; if

there are loops, they must have at least one delay.
ESE 547, Digital Signal Processing 15

Elementary sampled signals:

(Recall) unit (im)pulse:


 1 n=0
δn =
 0 n 6= 0

Note: the unit impulse is an ordinary function in DSP. When

simulating continuous-time systems, be careful about its
normalization!

Unit step:

 1 n≥0
Un =
 0 n<0

Check unit step is impulse response of (digital) integrator.

ESE 547, Digital Signal Processing 16

1.3 Frequency-Domain and Transforms

• Sampled Sinusoids
• Frequency, Amplitude, and Phase Responses
• DTFT, Z-transforms and Properties
• Poles and Zeros
ESE 547, Digital Signal Processing 17

1.3.1 Sampled Sinusoids

Continuous-time:

x(t) = A cos(2πf t + φ)

where A is the amplitude, and φ is the phase.

If signal is sampled every T seconds, the sampling instants

are t = nT , and so the sampled (discrete-time) sinusoid is:

xn = A cos(2πf nT + φ)

where A is the amplitude, and φ is the phase.

Amplitude: A (note: not necessarily peak value!)

Phase: φ
Sampling period: T
Sampling frequency: fS = 1/T
Angular sampling frequency: ωS = 2π/T
Note: the sequence (xn ) is unchanged if the frequency f is
changed to f + k/T , for any integer k (aliasing).
ESE 547, Digital Signal Processing 18

Sampled Sinusoids (continued):

Sinusoid in complex form:

gn = Aej(2πf0 nT +φ)
jφ j2πf0 T n


= Ae e

where A is the amplitude, and φ is the phase.

For a decaying sinusoid:

Continuous-time:

g(t) = Aeσ0 t ej(2πf0 t+φ)

Sampled:

gn = Aeσ0 nT ej(2πf0 nT +φ)

= Aejφ z0n

where

z0 = e(σ0 +jω0 )T
= es0 T
ESE 547, Digital Signal Processing 19

Sampled Sinusoids (continued):

Consequence:

The relationship between the (continuous-

time) s-domain and the (discrete-time)
z -domain is given by:

z = esT

Restrict to frequency domain:

z = ej2πf T = ejθ

So the discrete-time angular frequency is

given by
θ = 2πf T = ωT
and the frequency information is on the unit circle.

Also, stability is equivalent to the absence of poles on or

outside the unit circle.
ESE 547, Digital Signal Processing 20

1.3.2 Frequency, Amplitude, and Phase Responses

Suppose we have an FIR system with input un = ej2πf nT .

Then, with ω = 2πf :
jωnT jω(n−1)T jω(n−M )T
yn = b0 e + b1 e + · · · + bM e
jωnT −M jωT
 
−jωT
= e b0 + b1 e + · · · + bM e

jωnT
= e K(ωT )

So if un is a (complex) sampled sinusoid of angular

frequency ω ,

yn = K(ωT )ejωnT
= K(ωT )un

where the complex number K(ωT ) is given by

K(ωT ) = b0 + b1 e−jωT + · · · + bM e−M jωT

ESE 547, Digital Signal Processing 21

Frequency Response (continued)

For an IIR system, if we also assume that the output yn is
given by yn = Kej2πf nT , where K is a complex constant:
KejωnT + a1 Kejω(n−1)T + · · · + aN Kejω(n−N )T

= b0 ejωnT + b1 ejω(n−1)T + · · · + bM ejω(n−M )T

or

jωnT −jωT −N jωT

Ke 1 + a1 e + · · · + aN e
jωnT −jωT −M jωT


=e b0 + b1 e + · · · + bM e

so that

K(ωT ) = H(ejωT )
b0 + b1 e−jωT + · · · + bM e−M jωT
=
1 + a1 e−jωT + · · · + aN e−N jωT
ESE 547, Digital Signal Processing 22

Frequency Response (continued)

For a general convolution, (with ω = 2πf )

∞
X
yn = hk ejω(n−k)T
k=0
∞
X
= ejωnT hk e−jωkT
k=0

= ejωnT H(ejωT )

The function H(ejωT ) is called the frequency response of

the system;

The function A(ωT ) = |H(ejωT )| is called the amplitude

response;

The function φ(ωT ) = 6 H(ejωT ) is called the phase

response.

Note:

• These are periodic with period 2π/T = 2πfS = ωS

• They depend only on the ratio f /fS = ω/ωS
For this reason, digital signal processing often works with a
normalized frequency f /fS .
ESE 547, Digital Signal Processing 23

1.3.3 DTFT and Z-transforms

Discrete-Time Fourier Transform (DTFT):

Based on the formula for the frequency response, the

Discrete-Time Fourier Transform of a (bounded) sequence
(gn ), is defined by:
∞
X
G(θ) = gk e−jkθ
k=−∞

The inverse transform is given by:

Z π
gn = 1/2π G(θ)ejnθ dθ
−π

Note 1: Boundednes means that there is some constant M

such that |gn | ≤ M for all n.
Note 2: If (gn ) is the impulse response of a system, the
frequency response of the system is given in terms of the
DTFT of (gn ) by

G(ej2πf T ) = G(ejθ )|θ=2πf T

ESE 547, Digital Signal Processing 24

DTFT and Z-transforms (continued)

Two-sided Z-transform:

The two-sided Z-transform of a bounded sequence (gn ) is

defined by

G(z) = Z{gn }
∞
X
= gn z −n
n=−∞

= · · · + g−1 z + g0 + g1 z −1 + g2 z −2 + · · ·

for all complex numbers z for which the series converges –

assumed to include the unit circle.
ESE 547, Digital Signal Processing 25

Two-Sided Z-transform properties:

• Linearity
• Shifting: Z{gn−k } = z −k G(z) for all k

• Convolution: If fn and gn are two bounded sequences,

then Z{(fn ) ∗ (gn )} = F (z)G(z)

— The Z-transform transforms convolution into

multiplication.

Examples:

1. an Un with |a| < 1

2. −an U−n−1 with |a| > 1
3. sinusoids
ESE 547, Digital Signal Processing 26

DTFT and Z-transforms (continued)

Relation to frequency response and DTFT:

Frequency response is given by

∞
X
K(2πf T ) = hk e−j2πkf T
k=−∞

so that

K(2πf T ) = G(ej2πf T )
= G(ejθ )|θ=2πf /fS
= G(z)|z=ej2πf /fS

The main use of the two-sided Z-transform is in calculating

the inverse DTFT of functions which are rational in z .

The inverse DTFT of some other types of functions of θ is

found by the direct inverse formula (will see later).
ESE 547, Digital Signal Processing 27

DTFT and Z-transforms (continued)

The one-sided Z-transform of a sequence (gn ) is defined by

G(z) = Z{gn }
∞
X
= gn z −n
n=0

= g0 + g1 z −1 + g2 z −2 + · · ·

for all complex numbers z for which the series converges

(region of convergence, or ROC).

Differs from the two-sided Z-transform and DTFT:

• assumes that (gn ) is one-sided

• does not assume that (gn ) is bounded
• therefore does not necessarily converge on unit circle
• converges in a region of the form {z | |z| > R} for
some R.

• handles initial conditions and transients.

ESE 547, Digital Signal Processing 28

DTFT and Z-transforms (continued)

Relation to frequency response, for a causal, stable system :

As before, frequency response is given by

∞
X
K(2πf T ) = hk e−j2πf T k
k=0

so that

K(2πf T ) = H(ej2πf T )
= H(z)|z=ej2πf T

Here, stable means BIBO stable, which is equivalent to hn

P∞
being absolutely summable, i.e., k=0 |hk | is finite
(stronger than bounded).
ESE 547, Digital Signal Processing 29

DTFT and Z-transforms (continued)

If gn is the impulse response, hn , of a causal system, its

one-sided Z-transform
∞
X
H(z) = Z{hn } = hn z −n
n=0

is called the transfer function of the system

Examples:

1. an Un
2. sinusoids

3. exponentially decaying and growing sinusoids

ESE 547, Digital Signal Processing 30

DTFT and Z-transforms (continued)

(One-sided) Z-transform properties:

• Linearity
• Shifting: Z{gn−k } = z −k G(z) for k ≥0
• Non-Causal Shift: Z{gn+1 } = zG(z) − zg0
• Convolution: If fn = 0 and gn = 0 for n < 0, then
Z{(fn ) ∗ (gn )} = F (z)G(z)
— The Z-transform transforms convolution into
multiplication.

Note 1: The noncausal shifting property enables the

Z-transform to handle initial conditions and transients.

Note 2: Because the signals are one-sided, the convolution

here is given by:
n
X
pn = fk gn−k
k=0
ESE 547, Digital Signal Processing 31

DTFT and Z-transforms (continued)

Inverse one-sided Z-transform:

Theoretical formula (rarely used, actually a contour integral):

π
1
Z
gn = G(Rejθ )(Rejθ )n dθ
2π −π

where R is such that all poles of G(z) are inside the circle
of radius R.

The normal method is to use partial fractions, but the

calculations are simpler if G(z)/z , instead of G(z), is
written in terms of z (not z −1 ) and is then expanded in
partial fractions.

Examples:

1. 1/(1 − az −1 ) with |a| < 1

2. 1/(1 − az −1 ) with |a| > 1
3. real poles – stable

4. real poles – unstable

5. complex poles – stable and unstable

6. comples poles on unit circle – second-order oscillator;

(see p.394)
ESE 547, Digital Signal Processing 32

1.3.4 Poles and Zeros

Terminology (for rational functions with no common factors

between numerator and denominator):

1. Pole: a value of z where H(z) is infinite, i.e., where the

denominator is zero.

2. Simple Pole: a is a simple pole if there is a factor

(z − a) in the denominator, and no factor of (z − a)k
with k > 1.

3. Multiple Pole: a is a multiple pole if there is a factor

(z − a)k in the denominator, with k > 1.
4. Zero: a value of z where H(z) is zero

5. Residue: the coefficient A in the term A/(z − a) in the

partial fraction expansion.
ESE 547, Digital Signal Processing 33

Poles and Zeros (continued)

Notes:

1. All rational functions can be expanded as product of

first-order factors with complex coefficients.

2. All rational functions can be expanded as product of

first- and second-order factors with real coefficients.

3. To have a partial fraction expansion, we must have

H(z) → 0 as z → ∞
4. Partial fraction expansions get more complicated with
multiple poles; the basic formulas can be found by
differentiating the equation
∞
1 X
n −n
−1
= a z
1 − az n=0

5. Knowing the poles gives us a good idea of what the

impulse response will look like.

6. Knowing the poles and using the initial- and final-value

theorems gives a good idea of the step response.

7. The poles and zeros (product expansion) can be used to

give a good idea of the frequency response.
ESE 547, Digital Signal Processing 34

1.4 Signal Sampling and Reconstruction

• Theory
• Antialiasing and Smoothing Filters
• Signal Conversion
ESE 547, Digital Signal Processing 35

1.4.1 Sampling and Reconstruction Theory

Theoretically, sampling is a multiplication operation:

Give a sampling signal s(t) which is periodic with period T ,

and a continuous-time signal g(t), define the sampled
signal gS (t) by
gS (t) = s(t)g(t)
T is called the sampling period and fS = 1/T is the
sampling frequency.

Normally, s(t) will approximate a periodic sequence of

δ -functions, and a further integration step is used to obtain a
specific value.

Note that the sampled function is still a continuous-time

function.
ESE 547, Digital Signal Processing 36

Sampling and Reconstruction Theory (continued)

For ideal sampling, we assume that s(t) is a periodic

sequence of δ -functions:
∞
X
s(t) = δ(t − nT )
n=−∞

and the sampling operation then has two steps:

• Multiply by the sequence of δ -functions:

∞
X
gS (t) = δ(t − nT )g(t)
n=−∞
X∞
= g(nT )δ(t − nT )
n=−∞

• Pick off numbers representing samples

gn = g(nT )
ESE 547, Digital Signal Processing 37

Sampling and Reconstruction Theory (continued)

Effect of sampling in frequency domain:

∞
X
GS (2πjf ) = ck G(2πj(f − kfS ))
k=−∞

where ck are the complex Fourier coefficients of the s(t):

T /2
1
Z
ck = s(t)e−j2πkt/T dt
T −T /2

Here 2πfS = 2π/T = ωS is called the sampling angular

frequency.

Also, fS /2 is sometimes called the folding frequency.

For ideal sampling, we get

∞
X
GS (2πjf ) = 1/T G(2πj(f − kfS ))
k=−∞
ESE 547, Digital Signal Processing 38

Sampling and Reconstruction Theory (continued)

Classical Sampling Criterion:

Must Sample at a rate higher than the Nyquist rate

= 2fM AX
WARNING!1! Nonlinear operations internal to the signal
processing can increase the bandwidth, and require much
higher sampling rates.

WARNING!2! Aliased components are not just a

mathematical abstraction, but very real; you can see them
on an oscilloscope, and hear them.

WARNING!3! Higher sampling rates can also be needed to

reduce phase-shift, for example, in a feedback loop.
ESE 547, Digital Signal Processing 39

Sampling and Reconstruction Theory (continued)

To reconstruct: use lowpass filter

Reconstruction is also possible in other cases:

• Bandpass
• Periodic
In time-domain:
∞
X
gR (t) = g(nT )l(t − nT )
n=−∞

where l(t) is the impulse response of the reconstruction

(smoothing) filter.
ESE 547, Digital Signal Processing 40

Sampling and Reconstruction Theory (continued):

“Ideal” reconstruction: ideal low-pass filter with theoretical

gain T ; unrealizable since

sin(πt/T )
l(t) =
πt/T
= sinc(t/T )

(Note that this vanishes at all sample points except t = 0,

and so gives the correct value at the sample points.)

Actual D/A conversion: zero-order hold followed by

smoothing filter. Frequency response of zero-order hold is
given by

sin(πf /fS )
H(f ) = e−jπf /fS
πf /fS
sin(πx)
= e−jπx
πx
where x is the normalized frequency.

It follows that the zero-order hold has a group delay of T /2,

and has a rough lowpass characteristic with infinite
attenuation at non-zero multiples of the sampling frequency,
and an attenuation of 2/π or about 3.9 dB at fS /2.
ESE 547, Digital Signal Processing 41

1.4.2 Antialiasing (Input) and Smoothing (Output)

Filters:

• Needed from sampling theorem

• To reject spurious signals and noise at input
• To reject aliased components at output
Problem: Need high order analog filter, especially at output

Can alleviate requirements at output:

• Upconvert, filter digitally, high-rate D/A converter,

• Upconvert, high-rate D/A converter, switched-capacitor
filter, final simple smoothing filter.
ESE 547, Digital Signal Processing 42

Antialiasing and Smoothing Filters (continued)

Can also alleviate requirements at input:

• Gentle analog filter, very high sampling rate – low

accuracy

• Use high-frequency, low word-width Digital lowpass filter

• Use DSP to Down-convert to low sampling rate, high
word-width – noise-shaping

• Σ-∆ converters.
ESE 547, Digital Signal Processing 43

1.4.3 Signal Conversion

Track and Hold Amplifier:

• Tracks signal until strobed, and holds constant value

until released.

• Stores charge on capacitor

• Properties
– Aperture

– Acquisition time

– Step

– Droop

– Jitter
ESE 547, Digital Signal Processing 44

Signal Conversion (continued)

A/D Converters:

• Successive Approximation
• Flash
• Sigma-Delta
• Dual-Slope, Counting, usually too slow
ESE 547, Digital Signal Processing 45

Signal Conversion (continued)

D/A Converters:

Switched current source; numerous other types.

Virtually always use zero-order hold.

Zero-order hold amplitude response: “sin(x)/x”

Zero-order hold delay: 1/2 sample period.