18) |Xk |ejφk ej2πfk t . y(t) = k=−n CHAPTER 1. REPRESENTATION OF FUNCTION CONTENTS 14 A reminder: The de nitions fk = T1k and ωk = 2πfk may be used to express y(t) in terms of Tk (individual period) or ωk (individual angular frequency) in all forms. Also, when fk = k/T , this fact is commonly recognized wherever fk is used. To convert one form to another, one may use the relationship between the coef cients as summarized below. Relation 1 De n e X0 ≡ 0 when A0 and B0 are missing. For 1 ≤ k ≤ n, Ak ∓ jBk , and f−k = −fk .

6 9 = 3 3 . It can be easily veri ed that y(t + To ) = y(t). , y(t + T ) = y(t). Since we have uniform spacing f = fk+1 − fk = 1/T , we may still plot Ak and Bk versus k with the understanding that k is the index of equispaced fk ; of course, one may plot Ak and Bk versus the values of fk if that is desired. 7. REVIEW OF RESULTS AND TECHNIQUES 13 3. A non-commensurate y(t) is not periodic, although all its components are periodic. For example, the function √ y(t) = sin(2πt) + 5 sin(2 3πt) √ is not periodic because f1 = 1 and f2 = 3 are not commensurate.

We do not know how many cycles x ˜(t) has completed over the interval T . Mathematically, the function x ˜(t) interpolating the two samples is no longer unique if the ˜ frequency f is not speci ed. , the frequency we can resolve for x ˜(t) is f˜ = r˜/T = 1/T . When we deal with discrete samples taken from a composite signal, the so-called aliased frequencies are equivalent in the sense that they contribute the same numerical values at the sample points. 5, the signal y(θ) = cos(θ) + 2 cos(3θ) + 3 cos(5θ) cannot be distinguished from x(θ) = 6 cos(θ) based on the two values sampled at θ1 = 0 and θ2 = π, because y(0) = x(0) = 6 and y(π) = x(π) = −6.