2.4.1 Show that the exponential signal sn = Aen is an eigensignal of the time advance operator. What is its eigenvalue?

An operator usually tranforms a signal into a profoundly dierent signal, but eigensignals are not fundamentally changed, they merely acquire a trivial gain. obeys In other words, an eigensignal sn of an operator O sn = sn O where , the eigenvalue, is a constant real number. The time advance operator z is dened as the operation of returning now the value the signal will be next time z sn = sn+1 which usually requires a crystal ball. Of course, for deterministic signals we know what the signals value will be for any time, and there is no problem with predicting its value at time n + 1. However, for stochastic signals, or signals not under our complete control, the time advance operator is not implementable in real time, and for this reason we usually work with the time delay operator z 1 . For deterministic signals it is easy to apply the time advance operator; we need simply replace n with n + 1. z sn = sn+1 = Ae(n+1) = Aen+ = Ae en = e (Aen ) = e sn so sn = Aen is indeed an eigensignal of z and the eigenvalue is e . The real sinusoid sn = A sin(n + ) is the eigensignal of an operator that contains z 1 and z 2 . Can you nd this operator? The question is in principle similar to the rst part, but somewhat more challenging from a technical point of view. We are asked to nd an operator that contains two time delays; we will assume the simplest combination, namely a weighted linear combination. = az 1 + bz 2 O What does this combination do to a signal sn ? sn = (az O 1 + bz 2 )sn = az 1 sn + bz 2 sn = asn1 + bsn2 Now we are interested in applying this operator to a digital sinusoid sn = A sin(n + ) (1)

and nding values a and b such that sn = asn1 + bsn2 that is, for which the following holds. A sin(n + ) = a A sin( (n 1) + ) + b A sin( (n 2) + ) The straight forward way of nding if this is can be true, and for which a and b involves using a lot of trigonometric identities. sn1 = A sin( (n 1) + ) = A sin((n + ) ) = A sin(n + ) cos( ) A cos(n + ) sin( ) sn2 = A sin( (n 2) + ) = A sin((n + ) 2 ) = A sin(n + ) cos(2 ) A cos(n + ) sin(2 ) = A sin(n + )(2 cos2 ( ) 1) A cos(n + )(2 sin( ) cos( )) = 2 cos( ) [A sin(n + ) cos( ) A cos(n + ) sin( )] A sin(n + ) Now, the expression in the square brackets on the last line is precisely what we found for sn1 and the nal term is sn itself. So we have found sn2 = 2 cos( )sn1 sn or sn = 2 cos( )sn1 sn2 (3) (2)

and by comparing this to 2 we see that indeed all digital sinusoids obey a second order dierence equation, and that a = 2 cos( ) and b = 1. Using precisely the same manipulations, the sinusoids equation is found for analog sinusoids on page 249 of the book. The above proof is straightforward, but somewhat tiring. There are several ways to simplify the algebra. One way is to realize that equation 1 can be rewritten sn+1 = asn + bsn1 without changing its meaning. This time we will assume = 0 in order to further simplify. A sin( (n + 1)) = a A sin(n) + b A sin( (n 1)) Now the left hand side is A sin(n) cos( ) + A cos(n) sin( )

and the right hand side is only slight more complex. a A sin(n) + b (A sin(n) cos( ) A cos(n) sin( )) By comparing the coecient of cos(n) on both sides we nd b = 1, and then comparing the coecients of sin(n) leads us directly to a = 2 cos( ). Another way of going about this is to use complex exponentials. As usual, this signicantly simplies the algebra as we dont need dicult trigonometric identities. ein = aei(n1) + bei(n2) = aei + be2i ein For this to hold, but note that trivially (ei + ei )ei e2i = 1 so a = (ei + ei ) = 2 cos( ) and b = 1. Our last solution will be a bit more elegant. Observe what the time advance and time delay operators do to a generic sinusoid. z +1 sin(n + ) = sin(n + ) cos( ) + cos(n + ) sin( ) z 1 sin(n + ) = sin(n + ) cos( ) cos(n + ) sin( ) We can dene an operator to be the average of the time delay and time 1 = 1 ( advance operators, that is, +z +1 ). Based on the above equations, 2 z when this operator is applied to a generic sinusoid, the second terms above cancel. sin(n + ) = cos( ) sin(n + ) (4) with eigenvalue Therefore sn = sin(n + ) is an eigensignal of the operator cos( )! The solution to the exercise is now immediate. aei + be2i = 1