Linear Infinite State Systems Last time we • Looked at systems with an infinite number of states (Reals N) • Examined property of linearity • Formulated (A, B, C, D) system definition • Derived equation to calculate state response Today we will • Characterize linear discrete-time systems by their impulse response • Formulate zero-input output response using impulse response • Compute impulse response for example systems • See examples of FIR and IIR systems EECS 20 Chapter 5 Part 2 1 Zero-State and Zero-Input Responses Last time, we came up with equations that give us s(n) and y(n): n-1 s(n) = An s(0) + y(n) = C An s(0) + n-1 Σ An-1-m B x(m) m=0 Σ C An-1-m B x(m) + D x(n) m=0 We gave names to the terms in these equations: Zero-input state response: An s(0) Zero-input output response: C An s(0) n-1 Zero-state state response: Zero-state output response: EECS 20 Chapter 5 Part 2 Σ An-1-m B x(m) m=0 n-1 Σ C An-1-m B x(m) + D x(n) m=0 2 Impulse Function The general system response can also be described by its response to a particular input signal called the impulse: : Integers {0, 1} n Integers, (n) = { 1 if n = 0 0 if n ≠ 0 This function is also known as the Kronecker Delta function. EECS 20 Chapter 5 Part 2 3 Impulse Response When we apply the impulse function as the input x to a SISO system with an initial state of zero, we denote the resulting output y by h. This special output, h, is called the impulse response. Using the formula for the zero-state output response, n-1 y(n) = Σ C An-1-m B x(m) + D x(n) m=0 We see that when the function is substituted for x, h(0) = D h(1) = CB h(n) = CAn-1B for n > 0 EECS 20 Chapter 5 Part 2 h(2) = CAB h(0) = D h(3) = CA2B h(n) = 0 for n < 0 4 General Response Using Impulse Response Looking again at the zero-state output response, n-1 y(n) = Σ C An-1-m B x(m) + D x(n) m=0 we see that we can write a general output y(n) using the impulse response h: n-1 y(n) = = Σ h(n-m) x(m) + h(0) x(n) m=0 n Σ h(n-m) x(m) Since h(n) = 0 for n < 0, ∞ y(n) = Σ h(n-m) x(m) m=-∞ m=0 EECS 20 Chapter 5 Part 2 5 Convolution This operation being performed on h and x is called convolution, and it is denoted by ∞ y(n) = Σ h(n-m) x(m) m=-∞ y=hx The convolution operation is commutative: hx=xh ∞ ∞ m=-∞ k=-∞ Σ h(n-m) x(m) = Σ h(k) x(n-k) The convolution operation has useful properties in the frequency domain, covered in later chapters. That is one of the reasons we use the impulse response characterization. EECS 20 Chapter 5 Part 2 6 Example: FIR Consider the 7-day moving average: 6 Σ x(n-k) The impulse response has finite duration (h(n) = 0 for all n > 6). k=0 If a system has an impulse Find the impulse response. response with finite duration, we call it a finite impulse response 6 (FIR) system. h(n) = 1/7 Σ (n-k) y(n) = 1/7 k=0 The (n-k) will equal 1 for k = n, 0 otherwise. We can have k = n for n between 0 and 6. h(n) = 1/7 for 0 ≤ n ≤ 6 h(n) = 0 otherwise EECS 20 Chapter 5 Part 2 7 Example: IIR Consider the oscillator system, and find the impulse response: cos () A= sin () 0 B= 1 -sin () cos () C=1 0 D=0 We showed that h(n) = CAn-1B for n > 0 h(0) = D We can also show that An = cos (n) -sin (n) sin (n) cos (n) h(n) = -sin( (n-1) ) for n>0 h(n) = 0 otherwise EECS 20 Chapter 5 Part 2 h(n) = 0 for n < 0 The impulse response does not die out; there is no N for which h(n) = 0 for n>N. We say these systems have infinite impulse response (IIR). 8 Example: Unstable Consider the bank interest system s(n+1) = (1+r)s(n) + x(n) y(n) = s(n) Find the impulse response h. Here, the impulse response has an interesting interpretation: it is the yearly value of a bank account when $1 is deposited in year 0, and nothing is deposited thereafter. We may show that h(n) = (1+r)n Here, the impulse response does not die out; it is an IIR system. In addition, h(n) increases with n, without bound. If a system produces an unbounded output for a bounded input, we say the system is unstable. EECS 20 Chapter 5 Part 2 9 MIMO Systems Since the impulse function takes values in Reals, the impulse response h is technically defined only for single-input systems. We may define a matrix-valued function h with the same values h(n) = CAn-1B for n > 0 h(0) = D h(n) = 0 for n < 0 and the general output y will be given by the same convolution ∞ y(n) = Σ h(n-m) x(m) m=-∞ y=hx So we can still have an h with the same properties for general MIMO systems, but we won’t call it an impulse response. EECS 20 Chapter 5 Part 2 10