ECE 598: The Speech Chain Lecture 3: Phasors A Useful One-Slide Idea: Linearity Derivatives are “linear,” meaning that, for any functions f(t) and g(t), x(t) = A f(t) + B g(t) Implies dx/dt = A df/dt + B dg/dt d2x/dt2 = A d2f/dt2 + B d2g/dt2 Example: x(t)=cos(wt-f) dx/dt = –w sin(wt-f) x(t)=Acos(wt-f) dx/dt = –w Asin(wt-f) Review: Spring Mass System Newton’s Second Law: f(t) = m d2x/dt2 Force of a Spring: f(t) = – k x(t) The Spring-Mass System Equation with no external forces: d2x/dt2 = - (k/m) x(t) Solution: Cosine First and Second Derivatives of a Cosine: x(t) = cos(wt-f) dx/dt = –w sin(wt-f) d2x/dt2 = –w2 cos(wt–f) Spring-Mass System: d2x/dt2 = –(k/m) x(t) –w2 cos(wt–f) = – (k/m) cos(wt-f) It only works at the “natural frequency:” w = w0 = √(k/m) Linearity means that you can multiply by “A” here, if you want to, and the same “A” will appear here. The Other Function We Know First and Second Derivatives of an Exponential: x(t) = eat dx/dt = a eat d2x/dt2 = a2 eat Could x(t)= eat also solve the spring-mass system? d2x/dt2 = –(k/m) x(t) a2 eat = – (k/m) eat It only works if: a = √(-k/m) Imaginary Numbers Definition (the part that you memorize): j = √-1 Linearity: 2j = √-4 = √4 √-1 3j = √-9 = √9 √-1 Solution to the spring-mass system: a = √(-k/m) = √-1 √(k/m) = jw0 x(t) = eat = ejw t 0 Complex Exponentials Definition (another bit to memorize): x(t) = ej(wt-f) = cos(wt-f) + j sin(wt-f) Then the “real part” of x(t) is defined to be: Re{x(t)} = cos(wt-f) So x(t)=ejwt and x(t)=cos(wt) are actually exactly the same solution! The “imaginary part,” Im{x(t)}=sin(wt), doesn’t change the solution. If you like, visualize x(t) as a movement in two dimensions: Re(x(t)) is movement in the horizontal direction Im(x(t)) is movement in Buckaroo Banzai’s mysterious 8th dimension. All we really care about is the movement in the horizontal direction; the movement in the Buckaroo Banzai direction is a convenient fiction that just happens to make the math work out. How do you Plot a Complex Exponential? Answer: you can’t! What you CAN do: plot either the real part or the imaginary part x(t) = ej2t Re{x(t)} = cos(2t) Im{x(t)} = sin(2t) 4 3 2 1 0 -10.00 0.79 1.57 2.36 3.14 3.93 4.71 5.50 6.28 -2 -3 -4 3cos(2t) 3cos(2t-pi/2) Special Numbers ejf = cos(f) + j sin(f) 1 = cos(0) + j sin(0) = ej0 j = cos(p/2) + j sin(p/2) = ejp/2 -1 = cos(p) + j sin(p) = ejp (ejp/2)2 = ejp (j)2 = -1 Linearity Again “Real Part” and “Imaginary Part” are linear operators: x(t) = A f(t) + B g(t) Re{x(t)} = A Re{f(t)} + B Re{g(t)} Im{x(t)} = A Im{f(t)} + B Im{g(t)} Example x(t) = 2 ejwt + 3 ej(wt-p/2) Re{x(t)} = 2cos(wt) + 3cos(wt-p/2) Im{x(t)} = 2sin(wt) + 3sin(wt-p/2) Amplitude, Phase, and Frequency of Cosines vs. Exponentials Exponential: x(t) = Aej(wt-f) = A e-jf ejwt Cosine: Re{x(t)} = A cos(wt-f) The Life Story of a Cosine Vocal fold motion is a cosine at (say) w=400p, with amplitude of A=0.001m: x(t) = 0.001 + 0.001 ej400pt m Notice this looks like an exponential, but it’s “really” a cosine. Re{x(t)} = 0.001 + 0.001cos(400pt) Air puffs come through when the vocal folds are open, with a maximum rate of 0.001 m3/second: uGlottis(t)=0.0005+0.0005ej400pt m3/s The same air puffs reach the lips 0.5ms later: uLips(t)=0.0005+0.0005ej400p(t-0.0005) =0.0005 + 0.0005 e-0.2jp ej400pt liter/s Air pressure at the lips is the derivative of u(t), times 0.003 kg/s: pLips(t)= 0 + 0.0003jp e-0.2jp ej400pt = 0.0003p e0.3jp ej400pt Pascals Acoustic wave reaches the listener’s ear 2ms later: pEar(t)= 0.0003p e0.3jp ej400p(t-0.02) = 0.0003p e-j7.7p ej400pt Look Closer: x(t) = 0.001 ej400pt uGlottis(t) = 0.0005 ej400pt uLips(t) = 0.0005 e-0.2jp ej400pt pLips(t) = 0.0003p e0.3jp ej400pt pEar(t) = 0.0003p e-j7.7p ej400pt Amplitude, frequency, phase. Which one stays the same? Phasor Notation x = 0.001 uGlottis = 0.0005 uLips = 0.0005 e-0.2jp pLips = 0.0003p e0.3jp pEar = 0.0003p e-j7.7p Phasor notation: write down only the amplitude and phase, not the frequency. Definition of Phasor Notation A phasor specifies the amplitude and phase of a cosine, but not its frequency. (Written in boldface if possible) x = A ejf To get x(t) back, you look back through the problem definition in order to find w, then write x(t) = Re{ x ejwt } Example Phasor x, u, p Vocal Fold x(t) 10 Complex x(t), u(t), p(t) Real x(t), u(t), p(t) 10 ej400pt 10cos(400pt) Vocal Fold u(t) 5 5 ej400pt 5 cos(400pt) Lips u(t) 5 e-0.2jp 5 e-0.2jp ej400pt 5 cos(400pt-0.2p) Lips p(t) 3p e0.3jp 3p e0.3jp ej400pt 3p cos(400pt+0.3p) Ear p(t) 3p e-j7.7p 3p e-j7.7p ej400pt 3p cos(400pt-7.7p) The Main Purpose of Phasors: They Turn Derivatives into Multiplication x(t) = Re{ x ejwt } dx/dt = Re{ jwx ejwt} d2x/dt2 = Re{ (jw)2x ejwt } = Re{ -w2x ejwt } The Main Purpose of Phasors: They Turn Derivatives into Multiplication In regular notation: v(t) = dx/dt x(t) = A cos(wt-f) v(t) = -w A sin(wt-f) v = -jw Ae-jf In phasor notation: v = -jwx x = Ae-jf In regular notation: d2x/dt2 = -(k/m) x(t) In phasor notation: -w2 x = -(k/m) x Solution: w=√(k/m) !!! Summary Linearity means you can: add two solutions, or scale by a constant. x(t)=eat could solve the spring-mass system, but only if a=√(-k/m) j2 = -1 e-jf = cos(-f) + j sin(-f) We don’t really need the imaginary part; Re{ejwt} = cos(wt) We don’t really need the frequency part either; it is never changed by any linear operation (e.g., scaling, time shift, derivative) Phasor notation encodes just the amplitude and phase of a cosine: x = Ae-jf To get back to the time domain: x(t) = Re{ xejwt }