0% found this document useful (0 votes)
91 views55 pages

Discrete Time Systems - Properties

This document contains the two important lectures of digital signal processing, which when downloaded will give a thorough knowledge of the topics in digital signal processing.

Uploaded by

King
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
91 views55 pages

Discrete Time Systems - Properties

This document contains the two important lectures of digital signal processing, which when downloaded will give a thorough knowledge of the topics in digital signal processing.

Uploaded by

King
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 55

Discrete Time Systems -

Properties
Discrete Time Systems
• Both Ip and Op are discrete signals.
• Common System Properties:
– Static vs. dynamic
– Time-invariant vs. time-variant
– Linear vs. nonlinear
– Causal vs. non-causal
– Stable vs. unstable systems
Static vs. dynamic

• Static system (a.k.a. memoryless): the output at time n depends


only on the input sample at time n; otherwise system is dynamic.
• 𝑦(𝑛) =
𝑇 [𝑥(𝑛 − 𝑁); 𝑥(𝑛 − 𝑁 + 1); 𝑥(𝑛 − 1); 𝑥(𝑛); 𝑥(𝑛 +
1) ; 𝑥(𝑛 + 𝑀 − 1); 𝑥(𝑛 + 𝑀)]; 𝑁; 𝑀 > 0
• For N = M =0, 𝑦(𝑛) = 𝑇 [𝑥(𝑛)], the system is static.
Examples
• 𝑦(𝑛) = 𝐴 𝑥(𝑛), 𝐴 ≠ 0
• 𝑦 𝑛 = 𝐴 𝑥 𝑛 + 𝐵, 𝐴, B ≠ 0
• 𝑦(𝑛) = 𝑥(𝑛) cos(n − 1)
• 𝑦(𝑛) = 𝑥(−𝑛)
• 𝑦(𝑛) = 𝑥(𝑛 + 1)
• 𝑦 𝑛 = (𝑛 − 1)𝑥(𝑛2)

• Do you know of any such system?

If I increase the duration of class beyond regular time…..


Time-invariant vs. time-variant
• IF TI, 𝑥 𝑛 → 𝑦 𝑛 ⟹ 𝑥 𝑛 − 𝑘 → 𝑦 (𝑛 − 𝑘 )
– for every input 𝑥 (𝑛) and every time shift 𝑘
• 𝑦(𝑛) = 𝑥(−𝑛)?

• If you perform an experiment today, it should lead to


same results tomorrow as well.
Time-invariant vs. time-variant
• 𝑦(𝑛) = 𝑥(−𝑛)
• Time variant

• If n is inside brackets # x[n + const], then the


system is TI .
• A system that computes average should be time
invariant.

• x[kn], where k is not one, x[n2] are?


Linear vs. Nonlinear Systems
• Follows superposition principle
• Governed by :
• 𝑇 [𝑎1 𝑥1 (𝑛) + 𝑎2 𝑥2 (𝑛)] = 𝑎1 𝑇 [𝑥1 (𝑛)] +
𝑎2 𝑇 [𝑥2 (𝑛)]
– for any arbitrary input sequences 𝑥1 (𝑛) and 𝑥2 (𝑛), and
any arbitrary constants a1 and a2
– y(n) = nx(n)
– Y(n) = x(n2)
– Y(n) = x2(n)
– Y(n) = ex(n)
• Revision of signals and systems

• Basic concepts of systems, notations


• Direct form I & II realizations
• Crosscorrelation, relation with convolution
• I/p Op relations using Crosscorrelation
• Determines periodicity of a given signal, time
lag etc
TI system, find linearity?
Discrete signal representation, ^ represents the origin,
otherwise default is the left most element for origin
• x1(n) = {1,0,2} {0,1,2}
• x2(n) = {0,0,3} {0,1,0,2}
• x3(n) = {0,0,0,1} {1,2^,1}
Linear system, find TI
• x1(n) = {-1,2^,1} {1,2^,-1,0,1}
• x2(n) = {1,-1^,-1} {-1,1^,0,2}
• x3(n) = {0,1^,1} {1^,2,1}
Find impulse response of LTI system {1,3,3,1}  {1,4,6,4,1},
system is under zero initial condition or relaxed.
Causal vs. non-causal
• Causal system: output of system at any time n
depends only on present and past inputs
• System is causal if;
𝑦 (𝑛) = 𝑇 [𝑥 (𝑛), 𝑥 (𝑛 − 1), 𝑥 (𝑛 − 2), . . . ]∀𝑛
𝑦 𝑛 = 𝑥(𝑛2 )
Stability
• Bounded Input-Bounded output (BIBO) Stable: every
bounded input produces a bounded output
• System is BIBO stable if
• 𝑥 𝑛 ≤ 𝑀𝑥 < ∞ → 𝑦 𝑛 ≤ 𝑀𝑦 < ∞ ∀𝑛
FIR vs IIR
• Finite vs. Infinite Impulse Response
• For causal LTI systems, h(n) = 0 for n < 0.
• Finite impulse response (FIR):
• 𝑦 (𝑛) = 𝑛𝑘=0 ℎ(𝑘 )𝑥 (𝑛 − 𝑘 )
• Infinite impulse response (IIR): 𝑦 (𝑛) =
𝑛
𝑘=0 ℎ(𝑘 )𝑥 (𝑛 − 𝑘 )
• How to realize these systems?
• Implementation :Two classes
• Finite impulse response (FIR): 𝑦 (𝑛) =
𝑛
𝑘=0 ℎ(𝑘 )𝑥 (𝑛 − 𝑘 )
• These are Nonrecursive systems
• In infinite impulse response (IIR):
𝑦 (𝑛) = 𝑛𝑘=0 ℎ(𝑘 )𝑥 (𝑛 − 𝑘 )recursive
systems
System Realization
• FIR systems can be readily implemented by their
convolution summation (involves additions,
multiplications, and a finite number of memory
locations).

• IIR systems, however, cannot be practically


implemented by convolution as this requires infinite
memory locations, multiplications, and additions.

• There is a practical and computationally efficient


means of implementing all FIR and a family of IIR
systems that makes use of difference equations
LCCDE
• Lets take a recursive equation
• y(n) = ay(n-1) + x(n)
• y(1) = ay(0) + x(1) = a2y(-1) + ax(0) +x(1)
• y(2) = ay(1) + x(2) = a3y(-1) + a2x(0) + ax(1) +
x(2)

𝑛 𝑘
• Y(n) = an+1y(-1) + 𝑘=0 𝑎 𝑥 𝑛 − 𝑘 ,𝑛 ≥ 0
Zero state response
• Lets take y(-1) = 0 ie the system is relaxed or
zero initial conditions.
• Only Second term remains and is called zero-
state response.
• Zero state also refers the fact that system is at
zero state, where state also has a connotation
of memory.
• Also called as forced response as it is forced by
an input.
Zero input response
• Suppose that the system is non-relaxed ie y(-
1) is non-zero, and the input x(n) = 0.
• Then the output of the system is called zero
input response.
• First term is zero input response.
• Also called natural or free response.
• Nth order LCCDE
• 𝑁 𝑘=0 𝑘 𝑎 𝑦(𝑛 − 𝑘) = 𝑀
𝑘=0 𝑏𝑘 𝑥(𝑛 − 𝑘)
• Initial
conditions: 𝑦(−1); 𝑦(−2); 𝑦(−3); 𝑦(−𝑁).
• Solution – y(n) = yh(n) + yp(n)
Homogenous solution
• 𝑁𝑘=0 𝑎𝑘 𝑦 𝑛 − 𝑘 = 0
• Put yh(n)=𝜆𝑛
• 𝑁𝑘=0 𝑎𝑘 𝜆
𝑛−𝑘 = 0

• 𝜆𝑛−𝑁 𝜆𝑁 + 𝑎1𝜆𝑁−1 + 𝑎2𝜆𝑁−2 + ⋯ + 𝑎𝑁 − 1𝜆 + 𝑎𝑁 − 𝑁 =


0
• yh(n) = C1𝜆1𝑛 + C2𝜆2𝑛 +…+ CN𝜆𝑁𝑁 𝑛
Example
• Determine the homogenous solution of the system
described by the first order difference equation y(n) + ay(n-
1) = x(n)
• 𝜆𝑛 + 𝑎 𝜆𝑛−1 = 0
• yh(n)=C 𝜆𝑛 , C is determined by initial conditions

• If we want to determine zero-input response, then x(n) = 0,


and particular solution is zero in this case.
• yh(0)=C
• y(0)=-ay(-1) = yh(0) = C
• yzi(n) = -ay(-1) 𝜆𝑛 = (-a)n+1y(-1), n>=0
Example
• Determine the zero input response of the system described
by the homogenous second order difference equation y(n) –
3y(n-1) – 4y(n-2) = 0.
• Yh(n) = C1 𝜆1𝑛 +C2 𝜆2𝑛
• Y(0) = 3y(-1) + 4y(-2) = C1 + C2
• Y(1) = 13y(-1) + 12y(-2) = -C1+4C2
• Given y(-1) and y(-2), we can find C1 and C2, which gives zero
input response.
Particular solution
• First order difference equation with input
x(n) = u(n).
• Yp(n) = Ku(n)
• Ku(n) + aKu(n-1) = u(n)
• For n>=1 Particular solution is of same
form as Input
• K+aK = 1 X(n) = Au(n), then yp(n) = Ku(n)
X(n) = Amnu(n), then yp(n) = Kmnu(n)
• K = 1/(a+1) X(n)= Acos wn, then
yp(n) = K1coswn+K2sin wn
X(n) = δ(n), yp(n) = 0
Find particular solution
• y(n) = 5/6 y(n-1) – 1/6y(n-2) + x(n), x(n) = 2n,
n>=0
y(n) + .2y(n-1) = x(n)
a = [1 0 .2]; b = [1 1]; %%%coefficients of DE
x =[1 1 1 1 1 1 ]; %%%input
zi = [1]; %%initial condition y(-1)
xic = filtic(b,a,zi); %%%generating filter initial condition for
command filter
y = filter(b,a,x,xic); %filtering while taking care of initial
conditions
plot(y)

%x = zeros(1,10); % for zero input response as input = 0


%zi = 0; for zero state response as initial conditions = 0
Total solution
• y(n) + ay(n-1) = x(n), x(n) = u(n)
• yp(n) = 1/(1+a)u(n)
• yh(n) = C(-a)n
• Suppose we want to obtain zero state response,
y(-1) = 0
• Y(0) +ay(-1) = 1
• Whereas, y(0) = C + 1/(1+a) = 1
• Thus C = a/(1+a)
• Y(n) = {1 – (-a)n+1}/(1+a)
Total solution contd.
• y(n) – 3y(n-1) – 4y(n-2) = x(n) + 2x(n-1), x(n) =
4nu(n), initial conditions y(-1) and y(-2).

• Y(n) = C1(-1)n + C2(4)n + yp(n)

• Yp(n) = Kn4nu(n)
• Determine y(n), n>=0 of the system y(n) –
4y(n-1) + 4y(n-2) = x(n) – x(n-1)
when input is x(n) = (-1)nu(n), and the system
is relaxed ie zero initial conditions.
Impulse response of LTI system using
zero state response
• Compute homogenous solution as particular solution is
zero.

• In other words, we compute only zero state response .

• Start with homogenous solution and then find the


coefficients, imposing zero initial conditions

• Here particular solution is zero and total solution is the


homogenous solution only.
• Determine impulse response and step
response for y(n) = 0.6y(n-1) -0.08y(n-2) + x(n)

• Impose three conditions


• -start with homogenous response
• -zero initial conditions
• -input is impulse function
Impulse response using matlab
• Let us assume a system y(n) = x(n) + x(n-1),
with zero initial conditions.
• h(n) is computed by substituting x(n) =
delta(n)
• %%%using matlab
• b = [1 1]; a = [1]; x = [1 0 0 0 0 0 0 0]; zi = 0;
• h = filter(b,a,x);
• n = 1:numel(x); stem(n,h)
• y(n) − y(n − 1) + 0.9y(n − 2) = x(n); ∀n

• Calculate and plot the impulse response h(n)


at n = −20, . . . , 100.
• Calculate and plot the unit step response s(n)
at n = −20, . . . , 100.
• Is the system specified by h(n) stable?
Basic building blocks
• Adder, Multiplier and Unit Delay, Unit Advance
• Consider a 5-point local averager:
1 𝑛
• 𝑦 𝑛 = 𝑘=𝑛−4 𝑥 𝑘 , 𝑛 = 0, 1, 2 …
5
• Impulse response of system
1 𝑛
• ℎ 𝑛 = 𝑘=𝑛−4 𝛿 𝑘 , 𝑛 = 0, 1, 2 …
5
• FIR implementation
• Memory requirements stay constant; only
need to store 5 values (4 last + present). Fixed
number of adders required.
• In general, length M FIR impulse response requires
• M multipliers, M-1 addition and M-1 memory
elements
IIR system realization
• IIR
• 𝑦 𝑛 =
𝑛
𝑘=0 𝑥 𝑘 , 𝑛 = 0, 1, 2 … 𝑓𝑜𝑟 𝑦 −1 = 0
• The impulse response is given by:
• ℎ 𝑛 = 𝑛𝑘=0 𝛿 𝑘 , 𝑛 = 0, 1, 2 …
• IIR memory requirements seem to grow with
increasing n.
• 𝑦 𝑛 =
𝑛−1
𝑘=0 𝑥 𝑘 + 𝑥 𝑛 = 𝑦 𝑛 − 1 + 𝑥(𝑛)
Direct Form I vs. Direct Form II
Realizations
𝑁 𝑀
• 𝑘=0 𝑎𝑘 𝑦(𝑛 − 𝑘) = 𝑘=0 𝑏𝑘 𝑥(𝑛 − 𝑘)

𝑁 𝑀
• 𝑦 𝑛 =− 𝑘=1 𝑎𝑘 𝑦 𝑛−𝑘 + 𝑘=0 𝑏𝑘 𝑥(𝑛 − 𝑘)

𝑀
• Output 1: 𝑣 𝑛 = 𝑘=0 𝑏𝑘 𝑥(𝑛 − 𝑘) (nonrecursive)

𝑁
• Output 2: 𝑦 𝑛 = − 𝑘=1 𝑎𝑘 𝑦 𝑛 − 𝑘 + 𝑣(𝑛)
(recursive)
Direct Form I IIR Filter
Implementation

• Requires: M + N + 1 multiplications, M + N additions, M + N memory locations


• Uses separate delays for input and output
Direct Form II IIR Filter
Implementation
• Requires: M + N + 1 multiplications, M + N
additions, M + N memory locations
Direct Form II IIR Filter
Implementation
• Requires: ??? multiplications, ?? additions, ??
memory
Direct Form II IIR Filter
Implementation
• Requires: M + N + 1 multiplications, M + N
additions, max(M , N ) memory
• Last lecture –
– Basics of TI, stable, causal systems
– Notion of FIR and IIR impulse responses
– Difference equations and LTI
– Direct form I and II realization

• Cross and Auto-correlations


• Periodicity, Echo, impulse response
• Z-transform
Cross correlation (Matched filter)
• Correlation of Discrete-Time Signals:
• Consider a class of problems related to radar

𝑦 𝑛 = 𝑎𝑥(𝑛 – 𝐷) + 𝑤(𝑛)
w(n) = noise
x(n) = transmitted signal
D = round trip delay
a = attenuation
Crosscorrelation
• Define crosscorrelation for energy signals as –
 
rxy (l )   x(n) y(n  l )   x(n  l ) y(n),
n   n  
l  0,1,2,...
(Put 𝑛 = 𝑙 + 𝑚)

• 𝑟𝑦𝑥 (𝑙) is thus the folded version of 𝑟𝑥𝑦(𝑙)


around 𝑙 = 0

ryx (l )   y(n) x(n  l ), l  0,1,2,... ⟹ rxy (l )  ryx (l )
n  
Example – anu(n), 0<a<1
• The cross-correlation machine and convolution
machine are identical, except that in the
correlation machine this flip doesn't take place,
and the samples run in the normal direction.

rxy (l )  x(l )  y (l )


Auto-correlation
• For autocorrelation, we thus have:
rxx (l )  rxx (0)  E x
• This means that autocorrelation of a signal attains its
maximum value at zero lag (makes sense as we expect the
signal to match itself perfectly at zero lag).
𝑟𝑥𝑥 (𝑙) 𝑟𝑥𝑥 (𝑙)
• Normalized autocorrelation 𝜌𝑥𝑥 = =
𝑟𝑥𝑥 (0) 𝐸𝑥
𝑟𝑥𝑦 (𝑙)
• Normalized crosscorrelation 𝜌𝑥𝑦 (𝑙) = =
𝑟𝑥𝑥 (0)𝑟𝑦𝑦 (0)
𝑟𝑥𝑦 (𝑙)
𝐸𝑥 𝐸𝑦
Correlation of Periodic Sequences:

• Let x(n) and y(n) be periodic with period N


signals. Their crosscorrelation is defined
as:
1 N 1
rxy (l )   x(n) y (n  l )
N n 0
Periodic signals corrupted with noise
• 𝑦 𝑛 = 𝑥 𝑛 + 𝑤(𝑛)
• 𝑟𝑦𝑦 𝑙 = 𝑟𝑥𝑥 𝑙 +𝑟𝑤𝑤 𝑙
• Autocorrelation of noise is usually a spike at l
= 0 due to assumption of white noise.
• Apart from peaks at l=0, there will be peaks at
integer multiples of N.
• ECG and PPG signals are periodic and period
can be extracted using this way.
• %%% Generate periodic signal an test for perdiodicity
• nx = [-50:50];
• x = cos(pi*nx); %periodic with period N = 2
• z=x;
• y = x;
• % [y,ny] = sigshift(x,nx,2); % obtain x(n-2)
• ny = nx;
• w = .15*randn(1,length(y)); nw = ny; % generate w(n)
• [y,ny] = sigadd(y,ny,w,nw); % obtain y(n) = x(n) + w(n)
• [y,ny] = sigfold(y,ny); % obtain y(-n)
• [rxy,nrxy] = conv_m(x,nx,y,ny); % cross-corrlation
• subplot(3,1,1);stem(nx,z)
• subplot(3,1,2);stem(nx,y)
• subplot(3,1,3);stem(nrxy,rxy./length(nx))
IP-OP relations for LTI Systems
• 𝑦 𝑛 = 𝑥 𝑛 ∗ ℎ(𝑛)

ryx (l )  y (l )  x(l )  h(l )  x(l )  x(l )


 h(l )  rxx (l )
• As an illustration, if x(n) is white sequence, auto-
correlation would be similar to delta function.
• If this is given as an input to LIT, the output is its
impulse response.
Use white noise as input to determine
impulse response
• nx = [-50:50];
• ny = nx;
• w = .15*randn(1,length(nx));w=w/norm(w); nw = ny; %
generate w(n)
• h = [1, 2];
• [y,ny] = sigfold(w,nw); % obtain y(-n)
• [rxx,nrxx] = conv_m(w,nw,y,ny); % auto-correlation, rxx
• [ryx,nryx] = conv_m(h,1:numel(h),rxx,nrxx); % h*rxx
• subplot(3,1,1);stem(nrxx,rxx) % plot of rxx
• subplot(3,1,2);stem(nryx,ryx) % plot of ryx = h*rxx
• subplot(3,1,3);stem(1:numel(h),h) % plot of h, should be
same as that of ryx
Prove

ryy (l )  y (l )  y (l )  rhh (l )  rxx (l )


In this lecture, we learned about:
• Representations of discrete time signals and common basic DT signals
• Manipulation and representations/diagrams of DT systems
• Various classification of DT signals:
• Periodic vs. non-periodic, symmetric vs. anti-symmetric
• Classifications of DT systems:
– Static vs. dynamic, time-invariant vs. time-variant, linear vs. non-linear, causal vs.
– non-causal, stable vs. non-stable, FIR vs. IIR
• LTI systems and their representation
• Convolution for determining response to arbitrary inputs
• Cross-correlation

You might also like