Sie sind auf Seite 1von 11

Overview of Course

EE 264 covers variations of systems of the form:


A2D x(t) Q T x[n] Digital Filter y[n] D2A y(t)

input : analog, continuous in time and value precise, efficient and reliable digital filter requires discretized and quantized signal in EE 264, we will approximate x[n] = x(nT) (ignore quantization effects)
EE 264: Digital Filtering

1-1

Continuous vs. Discrete


as mentioned above, signals can be either continuous or discrete in time and value
Time Continuous Continuous Value Discrete State Machine Digital Signal Analog Signal Discrete Discrete-time Signal sampled sampled, quantized

EE 264: Digital Filtering

1-2

Applications
Once we have a digital version of a signal, we can perform many functions on it We will focus on filtering, with which we can: Reduce signal bandwidth
anti-aliasing

Extract signal properties


band-pass filter (BPF), frequency selective gain

Remove noise or channel distortion


Wiener noise filters, channel equalization
EE 264: Digital Filtering

1-3

Digital to Analog
In some cases, the desired output may be a digital signal However, we often wish to convert to analog with Digital-to-Analog (D2A) conversion: Frequency-Division Multiplexing (sharing of available frequency band Digital effects for sound Voice Band Data Modems (use analog transmission line) Digital control of plant with analog input
EE 264: Digital Filtering

1-4

Topics
Introduction The Transform Domain (Laplace, Fourier, z-Transform) A2D/D2A, sampling, interpolation Relationships between transforms Simple filters (allpass, minimum-phase, linear-phase) Infinite Impulse response (IIR) filters Finite Impulse response (FIR) filters DFT and relationship to DTFT Wiener Filters Adaptive Filters
EE 264: Digital Filtering

1-5

Notation
x[n] refers to a discrete-time signal x(t) refers to a continuous-time signal x[n] + y[n] refers to pointwise addition of x,y x[n]y[n] refers to pointwise multiplication of x,y c x[n] refers to pointwise multiplication of x by constant c x[n] + c refers to the pointwise addition of x with constant c

EE 264: Digital Filtering

1-6

Time Shifting
Sequences can be shifted in time by adding to/subtracting from indexing variable: y[n] = x[n - k]: y[n] is x delayed by k samples z[n] = x[n + k]: z[n] is x advanced by k samples Given that the first signal 0 is x[n], which could be y[n], z[n]?
0
EE 264: Digital Filtering

1-7

Common Discrete-time Sequences


Impulse:
n=0 "[n]= { 1, n#0 0,

Step:
!

n"0 u[n]= { 1, n<0 0,

Geometric:
!

an "u[n]

Sinusoidal:
!

sin( 2 "n ) k
1-8

EE 264: Digital Filtering

Sifting Property
A useful property of the impulse function:
$

x[n] =

% x[k]"[n # k]
k=#$

A discrete-time signal can be thought of as a sum of weighted (and advanced/delayed) impulses ! Each impulses amplitude corresponds to one value in the discrete-time sequence

EE 264: Digital Filtering

1-9

Discrete-time Systems
Discrete-time systems perform an operation on an input, producing an output We will sometimes use T{ } to represent a system as an operator We have already seen one example, the delay: y[n] = T{x[n]} = x[n - nd] Another example: simple averager:
y[n] = T{x[n]} = x[n] + x[n "1] 2
1-10

EE 264: Digital Filtering

Attributes of Systems
Memoryless: A system is memoryless if the output is a function of the current input only cannot depend on past or future input values neither delay nor simple averager are memoryless example: y[n] = (x[n])2

EE 264: Digital Filtering

1-11

Attributes of Systems
Linear: A system is linear if it preserves scaling and superposition (linear combinations): let T{ } represent the operation performed by the system T{ax1[n] + bx2[n]} = aT{x1[n]} + bT{x2[n]} Is our memoryless example a linear system? y[n] = (x[n])2

EE 264: Digital Filtering

1-12

Attributes of Systems
Time Invariant: A system is time invariant if a shift in input results in the same shift in output: suppose T{x [n]} = y[n] then T{x [n - k]} = y[n - k] Is our example time invariant? y[n] = (x[n])2

EE 264: Digital Filtering

1-13

Attributes of Systems
Causal: A system is causal if its output only depends on current and previous input values If it relies on future values, it is not causal non-causal filters can be useful: how can they be implemented? Stable: A system is (bounded input-bounded output or BIBO) stable if the output is bounded for any input that is bounded by a finite constant: if |x[n]| M1 then |y[n]| M2
EE 264: Digital Filtering

1-14

LTI Systems
We will be most concerned with systems that are both linear and time invariant (LTI):
y[n] = T{x[n]} & $ ) = T ' % x[k]"[n # k]* ( k=#$ +
$

= = =

% x[k]T{"[n # k]} (by linearity)


k=#$ $

% x[k]h [n]
k k=#$ $

(define hk [n] = T{"[n # k]}) (by time invariance - -shift by k)

% x[k]h [n # k]
0 k=#$

= ( x , h )[n] = ( x , h )[n]
EE 264: Digital Filtering

1-15

Impulse Response
This h0[n] (or just h[n]) is the impulse response of the system: If x[n] = [n], then T{x[n]} = h[n] h[n] is the output of the system when a unit impulse is the input What if system is linear, but not time invariant? Is sampling an LTI operation?
EE 264: Digital Filtering

1-16

Impulse Response
If two LTI filters h1, h2, are in cascade, the overall impulse response is h1h2 If two LTI filters are in parallel, the overall impulse response is h1+h2 An LTI filter with finite number of non-zero coefficients is a finite impulse response (FIR) filter An LTI filter with an infinite number of non-zero coefficients is an infinite impulse response (IIR) filter Step response: % %
g[n] = (h " u)[n] =
% %

& h[m]&#[(n $ m) $ k]
m=$% k= 0 % 0 n k= 0 k=$% k=$%

& & h[m]#[(n $ m) $ k] = & h[n $ k] = & h[n + k] = & h[k]


1-17

m=$% k= 0 EE 264: Digital Filtering

Examples
Which of the following are stable, causal, linear, time invariant?
Stable
n

Causal

Linear

TI

Accumulator : y[n] = " x[k]


k= 0 n +N

Finite Avg.: y[n] = ! !


! !
EE 264: Digital Filtering

1 2N +1

# x[k]
k= n"N

Shifter : y[n] = x[n " N] Amp w/ offset : y[n] = ax[n] + b

1-18

Stability of LTI Systems


An LTI system is BIBO stable iff
#

$ h[k] < #
k="#

For example, consider a geometric impulse response: h[n] = anu[n] The system is stable iff # #

$ h[k] = $ a
k="#

<#

And this summation converges for |a| < 1 System only stable if -1 < a < 1
EE 264: Digital Filtering

k="#

1-19

Examples
Is this linear? Is it TI?
M

y[n] = # bm x[n " m]


m= 0

What is the impulse response? Is it IIR, FIR?


!

EE 264: Digital Filtering

1-20

10

Examples
Is this linear? Is it TI?
# x[n] & y[n] = 1 % y[n "1] + ( 2 y[n "1]' $

Why might this system be useful?


!

EE 264: Digital Filtering

1-21

11

Das könnte Ihnen auch gefallen