Вы находитесь на странице: 1из 0

Curtin University, Australia 1-1

YH Leung (2006, 2007, 2012)


INTRODUCTION


1 What is Digital Signal Processing?

Digital signal processing is concerned with
(i) the digital representation of signals, and
(ii) the use of digital processors to analyze, modify, or extract information from
the signals



A digital signal processor or DSP is the actual device that performs the
processing, often in real-time



Digital signal
Signals can be: continuous or discrete in time
continuous or discrete in amplitude

Fig. 1 Classification of signals

Digital signals are discrete in time and discrete in amplitude

Continuous amplitude Discrete amplitude
Continuous-time
Discrete-time
Analog signal
Digital signal
Curtin University, Australia 1-2
YH Leung (2006, 2007, 2012)
A digital signal can be derived from an analog signal through an analog-to-
digital converter (ADC)


Fig. 2 Analog-to-digital conversion

Sample-and-hold (S/H) often understood to be part of ADC



Three main parameters:
1. Sampling rate w
s

2. Input voltage range of ADC
3. Number of bits or amplitude resolution of ADC

Fig. 2 ADC transfer characteristics



x
c
(t) x[n]
w
s
S/H ADC
000
001
010
011
100
101
110
111
Output
Input (V)
0.5 1.0 -1.0 -0.5
Curtin University, Australia 1-3
YH Leung (2006, 2007, 2012)
For the sake of mathematical tractability, it is often assumed that
1. Sampling rate remains constant with time
2. ADC has unlimited input dynamic range
3. ADC has infinitely fine amplitude resolution
4. ADC is ideal, i.e., is linear, has no DC offsets, has no gain or scale-factor
errors, and has a flat frequency response

That is, we shall focus on discrete-time continuous-amplitude signals



In reality, a digital signal is nothing more than a sequence of numbers
residing inside some digital computational element

Aim of signal processing is to either
(i) transform these numbers in such a way that they exhibit certain
characteristics, or
(ii) extract certain information from these numbers



Fig. 3 Digital Signal Processing

physical
process
transducer
0
1
0
0
1
1
0
1
0100 0001
1
1
1
0






0
1
1
1
0
1
1
0
1
0
1
1
0
0
1
1
1111
Curtin University, Australia 1-4
YH Leung (2006, 2007, 2012)
2 Why Digital Signal Processing?

Pros
Reproducibility
No problems with component tolerances, temperature and aging drifts
Performance identical from implementation to implementation
Guaranteed Accuracy
Accuracy determined only by number of bits used
Predictability
Performance can be predicted exactly from computer simulation
Flexibility
Can be programmed and re-programmed with no or minimal hardware changes
Smaller Size and Lower Cost
Continual advances in digital circuit technology drive size and power
consumption down, for same signal processing function
Superior Performance
Can perform signal processing functions not possible with analog technology
Can correct for non-idealness of the necessary analog parts of a complete
signal processing system


Cons
ADC and DAC Cost
High costs of high speed high precision ADCs and DACs
Design Time, Design Effort and Design Cost
Time to learn digital signal processing techniques and software packages
Time and effort to design high-speed digital circuit boards
High cost of software development packages
Finite Wordlength Effects
Finite wordlengths can degrade performance of signal processing function
Highly non-linear difficult to analyse theoretically
Computation Delay
Unavoidable delay in computation time can create stability problems
Limited Signal Bandwidth
Nyquist sampling theorem
Curtin University, Australia 1-5
YH Leung (2006, 2007, 2012)
3 Application Examples

3.1 Shaping Filters for Digital Radio Transmitter

Problem: Radio spectrum is a precious limited resource
Standards for modern radio systems impose very stringent specifications
on the spectrum of the transmitted radio signal


Example: European Radio Message System (ERMES)
Data rate: 6.25 kbits/sec
RF band: 169.4125169.8125 MHz. 16 channels of 25 kHz bandwidth each
Modulation: 4CPFSK
Symbol frequencies 1.5625 kHz, 4.6875 kHz (3.125 kHz spacing)





Spectrum of transmitted signal determined by shaping filter transfer function
specified by a mask and nominally a 10th order low pass Bessel filter with 3.9
kHz 3dB bandwidth

ERMES also specifies transmitted spectrum must fit inside another mask and
adjacent channel power must be lower than 70 dB. This conflicts with
requirement that shaping filter must have a rise time of about m 88 s

Design difficult to achieve using analog circuits

digital solution


encoder
data bits
{0, 1}
shaping
filter
4-level
PAM
FM modulator
Curtin University, Australia 1-6
YH Leung (2006, 2007, 2012)
3.2 Telephone Echo Cancellation





Problem: Channel characteristics depend on actual connection
Hybrid characteristics vary from hybrid to hybrid


Solution: Adaptive echo canceller
AF is anti-aliasing filter + ADC + adaptive N-tap FIR + DAC + smoothing filter
[ ] s n is Bs speech signal
[ ] e n is Bs echo signal through As hybrid

[ ] e n is estimate of Bs echo signal, as determined by AF



Let [ ] n w be a vector containing coefficients of FIR at time nT, and [ ] n s be
another vector containing last N samples of [ ] s n . Following algorithm will
minimise [ ] r n :
m + = + [ 1] [ ] [ ] [ ] n n r n n w w s
H
y
b
r
i
d
Customer
A
4-wire
circuit
2-wire
circuit
Customer
B
Echo of Bs
speech
Echo of As
speech
Channel
Channel
H
y
b
r
i
d
Mary had
a lamb
Jane cooked
dinner
Mary Jane had cooked
a lamb dinner
H
y
b
r
i
d
H
y
b
r
i
d
Channel
Channel
Customer
A
Customer
B
A
F
A
F
[ ] s n
[ ] e n
[ ] e n
[ ] r n
Curtin University, Australia 1-7
YH Leung (2006, 2007, 2012)
3.3 High Resolution Direction Finding with an Antenna Array



Problem: Direction finding by conventional beamforming limited by width of main
lobe may not be able to resolve two closely spaced (angular) sources


Solution: High resolution techniques, e.g., MUSIC

1. Estimate the signal covariance matrix with
=
=

1
1
[ ] [ ]
N
H
n
n n
N
R x x
where [ ] =
1 2
[ ] [ ] [ ] [ ]
T
L
n x n x n x n x

2. Eigendecompose R and rank its eigenvalues as l l l < < <
1 2 L

Suppose there are - L K uncorrelated incident signals. Then l l
1 K
,
and remaining eigenvalues are significantly larger

3. Take eigenvectors corresponding to l l
1
, ,
K
and form the matrix
[ ] =
1 2 K
E e e e

4. Compute q
q q
=
1
( )
( ) ( )
H H
P
a EE a

for different qs where q ( ) a is the array steering vector
q ( ) P will exhibit peaks at qs corresponding to the directions of arrival of the
incident signals

Try doing this with an analog circuit!
demod demod demod
x
1
(t) x
2
(t) x
L
(t)
. . .
DOA algorithm
^ ^

1
,
2

1
Curtin University, Australia 1-8
YH Leung (2006, 2007, 2012)
4 Topics in Digital Signal Processing

1 Signal Theory
Discrete-time signals and systems
Transform domain representation of signals
Random signals

2 Design of Frequency Shaping Filters
Design of FIR and IIR filters by approximation and optimisation
Multirate digital signal processing

3 Frequency Analysis of Signals
Spectral analysis using the FFT
High resolution methods

4 Optimum Filters
Wiener filters
Kalman filters

5 Adaptive Filters
LMS filters
RLS filters

6 High Order Spectral Analysis

7 Parameter Estimation
Bayesian
Maximum likelihood
Least mean-square
Order statistics




Curtin University, Australia 1-9
YH Leung (2006, 2007, 2012)
5 Reference Books

Textbook

[1] A. V. Oppenheim and R. W. Schafer, Discrete-Time Signal Processing, 3rd ed.,
Prentice-Hall, 2010.




Other Recommended Textbooks

[2] J. G. Proakis and D. G. Manolakis, Digital Signal Processing: Principles,
Algorithms, and Applications, 4th ed., Prentice-Hall, 2007.

[3] S. S. Soliman and M. D. Srinath, Continuous and Discrete Signals and
Systems, 2nd ed., Prentice-Hall, 1998.

[4] E. C. Ifeachor and B. W. Jervis, Digital Signal Processing: A Practical
Approach, 2nd ed., Prentice-Hall, 2002.

[5] R. G. Lyons, Understanding Digital Signal Processing, 3rd ed., Prentice-Hall,
2010.

[6] S. K. Mitra, Digital Signal Processing: A Computer-Based Approach, 3rd ed.,
McGraw-Hill, 2006.




Advanced Textbooks

[7] M. H. Hayes, Statistical Digital Signal Processing and Modeling, Wiley, 1996.

[8] C. W. Therrien, Discrete Random Signals and Statistical Signal Processing,
Prentice-Hall, 1992.

[9] S. Haykin, Adaptive Filter Theory, 4th ed., Prentice-Hall, 2002.

Вам также может понравиться