Scribd
Upload
3 views7 pages

Causality, Time Invariance, and Linearity

Lecture 3 covers the basic properties of state-space linear systems, focusing on causality, time invariance, and linearity. It discusses how these properties apply to both linear time-varying (LTV) and linear time-invariant (LTI) systems, including the implications for outputs given specific inputs. The lecture also introduces concepts like impulse response and the use of Laplace and Z transforms in analyzing these systems.

Uploaded by

salih zeki
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
3 views7 pages

Causality, Time Invariance, and Linearity

Lecture 3 covers the basic properties of state-space linear systems, focusing on causality, time invariance, and linearity. It discusses how these properties apply to both linear time-varying (LTV) and linear time-invariant (LTI) systems, including the implications for outputs given specific inputs. The lecture also introduces concepts like impulse response and the use of Laplace and Z transforms in analyzing these systems.

Uploaded by

salih zeki
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 7
LECTURE 3 Causality, Time Invariance, and Linearity ‘This lecture introduces a few basic properties of state-space linear systems and some of their direct consequences. The properties in Sections 3.2, 3.3, and 3.5 are direct consequences of causality, time invariance, and linearity. They apply to any system that exhibits these properties, even if such systems are not one of the state-space systems introduced in Lecture 1. For example, they also apply to infinite-dimensional systems. CONTENTS: Basic Properties of LTV/LTI Systems Characterization of All Outputs to a Given Input Impulse Response Laplace and Z Transforms (review) ‘Transfer Function Discrete-Time Case Additional Notes Exercises veer 3.1 Basic PROPERTIES OF LTV/LTI SYSTEMS Notation 1. We recall that u ~» y means that 'y is one of the outputs that corresponds to u.” ps In this section we state three basic properties of LTV and LTI systems. All these properties are simple consequences of results in subsequent lectures, so we will defer their proofs for later. {All state-space systems introduced so far (both LTV and LTI) have the property that the output before some time t does not depend on the input after time ¢ (see Figure 3.1), Such systems are called causal. Property P3.1 (Causality). The state-space system (CLTV) is causal in the sense that1.1 given any two input signals u and i such that a(t) u(t), VOst 0, if ~» y then there exists an output j such that ~~ jvand inputs match on [0, T) then the outputs must ‘match on (0, T) for R(t) = enna WO) =y(), YO 0, if u ~» y then there exists an output j such that i ~+ jand HO) = y(t+T), YEZ0. o Attention! Recall that ii ~> j means that j is one of the outputs corresponding to the input a. In general, the input i has many other outputs (obtained from different conditions) that will not match the time-shifted version of y. Generally, the initial conditions used to obtain y and 7 will not be the same. o Both the LTV and LTI systems have the property that they can be viewed as linear Notation2. Thereis maps from their inputs to appropriate outputs. This justifies the qualifier linear in some abuse inthe LTV and LTI. LTV/LTI terminology. 3 3s Property P3.3 (Linearity). ‘The state-space system (CLTV) is linear in the sense that Note, Inwords:Kfuy given any two input signals u and u2 and scalars a, B € R, ify > yx and ua ~~ Yo» and u, have outputs 91 then au, + Bur ~ ay + By a and ya, respectively, thenay, + By2 must be then ho mest®® Attention! Recall that au + Bu ~ ay: + By means that ays + Byp is one of the tay + Bus, for outputs corresponding to the input au; + Bua, In general, there may be many other appropriate initial outputs (obtained from different initial conditions) that will not be of this form. 1 conditions Notation 2 (LTV and LTI). We use LTV and LTI to denote the continuous-time systems given by the equations (CLTV) and (CLT1), respectively, or the discrete-time systems (DLTV) and (DLTD, respectively. There is some abuse in this terminology because there are many linear time-invariant and time-varying systems that cannot be expressed by these state-space equations, This is the case, for example of infinite- dimensional systems. a 3.2 CHARACTERIZATION OF ALL OUTPUTS TO A GIVEN INPUT Linearity allows one to use a single output yy corresponding to a given input u to construct all remaining outputs corresponding to u. 1. Let y be any other output associated with the given input u. Since we can write the zero input as 0 = u — u, by linearity we conclude that yy = y — yy must Notation, An output necessarily be an output corresponding to the zero input: corresponding tothe zero input sealed « homogeneous response {' Fy we u 0 yyy uy 2, Conversely, suppose that one is given an arbitrary output yz, corresponding to the zero input. Since u = +0, we conclude that y = yy + yh is another 34 LECTURE 3 AL 0) u(t) i Sedtde=1 H(t) Nc fi oA t 7 THd 7 Le Lod (a) Pulse (b) Time-shifted pulse (©) Step approximation FIGURE 3.3. Step approximation to a continuous-time signal, output corresponding to u: um yy = > utO=U yo tyme (a ou The following result summarizes the two prior observations. Theorem 3.1. Let yy be an output corresponding to a given input wu. All outputs corresponding to u can be obtained by Y=yet yh where yy is an output corresponding to the zero input. o This means that to construct all outputs corresponding to u, it is enough to know how to solve the following two problems: 1, Find one particular output corresponding to the input u. 2. Find all outputs corresponding to the zero input. ‘The remainder of this section addresses the first problem. 3.3 IMPULSE RESPONSE Consider a linear SISO system and let 54 denote the unit-area pulse signal in Figure 3.3(a). Using 34, we can write an approximation to an input signal w: (0, 00) —> Ras in Figure 3.3(c): u(t) = Sou(kA)A ba(t— kA), Ve>0. GD = Note. Thisassumeswe For each r > 0,let ga(t, z), f = 0 bean output corresponding to the input 5a(t — t) can choose outputs 8a(t, t) for which the -pw series (3.2) converges. Sa(t — t) ~ galt, 2). Note. The last equality in (33) is ansequence ofthe definition ofthe Riemann integral. It plc assumes that ‘the limit in (3.4) and the integral in 3.3) tothe,» p40 Notation. The output in (35) has the special property that it is equal tozero when u = 0; itis called a forced or _ero-state response. The latter terminology will, become clear in Lecture 4 Note, Equation (35) is generally taken as the definition of impulse response. Note. Soon we will see thatthe impulse response of LTV/LTI systems is unique Therefore one could replace “one can choose the impulse response to satisy” by “the impulse response satisfies.” CAUSALITY, INVARIANCE, AND LINEARITY 35 Because of (3.1) and linearity Us ~ ya(t) = > AulkAdga(t.kA), VE> 0. (3.2) = Moreover, since ua — was A — 0, we conclude that w= yO Lim, ya(0)= fim, J A w(kAdga tka) [ u(rg(t, dr, E20 (3.3) is an output corresponding to u, where g is defined by g(t, t) = jim galt, t)- G4) ‘The function g(t, r) can be viewed as the value of the output at time f, corresponding to an input pulse of zero length but unit area (a Dirac pulse) applied at time r. For MIMO systems this generalizes to the following result. ‘Theorem 3.2 (Impulse response). Consider a continuous-time linear system with k inputs and m outputs. There exists a matrix-valued signal G(t, 2) € R"™* such that for every input u, a corresponding output is given by um y(t) = f G(t, tu()dt, Ve > 0. (3.5) ‘The matrix-valued signal G(t, r) € R™* is called a (continuous-time) impulse re- sponse. Its entry gij(t,t) can be viewed as the ith entry of an output at time corresponding to a pulse of zero length but unit area (a Dirac pulse) applied at the jth input at time r. The impulse response G(t, t) that appears in (3.5) has several important properties that will be explored next. Properties (Impulse response). P34, For causal systems, one can choose the impulse response to satisfy Git.) =0, Vr>t (3.6) 3.5, For time-invariant systems, one can choose the impulse response to satisfy GET. 1+ T)=Glt.r), VAr, TE 0. 7) 36 LECTURE 3 Notation. With some abuse of notation, itis common to write simply Glo, h) = Gl ~ 4), Veh >0. P36. In particular fort = 0,4 =T,h=t+T Glt, 4) =Glh-4,0), Wh 420, which shows that G(tp, fy) is just a function of fy ~ 4, For causal, time-invariant systems, we can write (3.5) as a u~ y(t) =[ Gl Dud = (GW, VED, Gy 0 ) where « denotes the convolution operator. Proof. For simplicity we assume a SISO system P34 P35 By linearity u=0sy= Since the impulse 5(¢ — r) at time t is equal to the zero input for allt in (0, r), because of causality, this impulse must have an output that is zero on (0, t);ie., a(t) =3(t-—1) =0, Yo Ay: ht—1) 7 and j(t)=0, WOst 0, we construct the impulse response g(t, t + T) using a particular output y(t) to the impulse u(t) = d(t—1 = T)attimet + T, asin u(t) = d(t— T) ~ y(t) = g(t, +7). To select the value of the impulse response g(t, 7) corresponding to an im- pulse 8(¢ — r) at time r, we use the fact that the new input ii(t) = 4(¢-*) can be obtained from u(t) = 5(t — t — T) by T time-units shifting, as in i(t) = u(t + T) = 8(t — 1). By time invariance, this means that there must exist an output 7 % such that i(t) = u(t + T) = dt — 1) > HE) = yt + T) = gt + T+ D> Using j to construct g(t, r), we conclude that g(t. t) = W(t) = g(t + T,t +7). CAUSALITY, INVARIANCE, AND LINEARITY 37 By using these outputs to construct the impulse response, we indeed obtain (3.7) P3.6 Equation (3.8) is obtained by using P3.5 to replace G(t, t) by G(t— t) in (3.5). Because of P3.4 we can further replace the oo in the upper integration limit by t, since G(t, t) = G(t ~ r) is equal to zero for t > t. | For causal, time-invariant systems, P3.6 provides a convenient way to compute the forced response. Due to the closed connection between the (time domain) con- volution of signals and the (frequency domain) product of their Laplace transforms, it is especially easy to compute forced responses in the frequency domain. This is explored in the following sections. 3.4 LAPLACE AND Z TRANSFORMS (REVIEW) MATLAB® Hint10, _ Givena continuous-time signal x(t), f > Oits unilateral Laplace transform is given by leplace(.t, 2) symbolically computes 7 eters Lix()] = als [ e“'x(t)dt, s eC. ofr, MATLAB® Hint 11. ‘The Laplace transform of the derivative x(t) of the signal x(t) can be related to £(s) Heplace (Frere) by symbolically computes 7 nents tansormof®, » p71 Lle(0)) = Hs) = 58s) x0), s EC. (9) Given two signals x(t) and y(t), ¢ > 0, the Laplace transform of their convolution is Note5.Theterm—x(0) given by in 39) doesnot appeae inbiatral transforms Ppa £l(x« MO) = ef x(c)y(¢ = r)dr] = £66)9(6). G10) Note. Why does Gu0)hold? > p.41 For discrete-time systems, the Z transform plays a role analogous to the Laplace transform in the study of continuous-time systems. Given a discrete-time signal x(F), MATIAB® Hint 12, € (0, 1, 2...) its (unilateral) Z transform is given by ztrana(P,t, 2) computes he 2 7 transform of ?. (2) = Zia = Soe tx(), ze C. = MATLAB® Hint 13. ‘The Z transform of the time-shifted signal x(t + 1) can be related to £(s) by strana (F,2,¢) peieres rece Zlx(t+ 1D] =2(86)— xO), s eC. (3.11) transform of 8. » p. 72

You might also like