Problem 1

Consider a continuous-time linear time-invariant system with input signal x(t), impulse response h(t) and
output signal y(t).

(a) Give the relationship for y(t) to x(t) and h(t) involving only operations in the time domain.

(b) Give the relationship for Y(j) to X(j) and H(j) using only operations in the Fourier (frequency) domain.
(c) Give the relationship for Y(s) to X(s) and H(s) using only operations in the Laplace domain.

Solution

(a) y(t) = h(t) ∗ x(t)

(c) Y (s) = H(s)X(s)

Problem 2

(a) Consider sampling modeled as an instantaneous closing and opening of a switch every Ts seconds.
When the sampling switch is open, assume vsampled (t) is zero. Give a time-domain expression for
vsampled (t) in terms of v(t).

(b) Describe the LTI system needed to extract an estimate of v(t) from vsampled (t). The estimate of v(t)
is denoted as v̂(t). Over what frequencies is the estimate of v(t) accurate?

1
Solution
P∞
(a) vsampled (t) = v(t) n=−∞ δ(t − nTS )
(b) The LTI system would extract the spectrum of V(j) from Vsampled (j). The LTI systems would be ideal
lowpass filter with impulse response h(t) = sin(πt/Ts )/(πt/Ts ). The Sampling Theorem says to choose
fs > 2fmax and we can divide by 2 on both sides to obtain fmax < (1/2)fs . The estimate of v(t) is
accurate over –(1/2)fs < f < (1/2)fs .

Problem 3

You are given a causal discrete-time linear time-invariant (LTI) system with unknown impulse response h[n]
to analyze. When the five-sample causal signal x[n] given below is input into the unknown system, the
response y[n] is six samples long and causal, as shown below.

(a) Find h[n].

Solution

(a) Because x[n] and y[n] are finite in length, the length of y[n] is the length of h[n] plus the length x[n]
minus one due to discrete-time convolution. Hence, h[n] has two coefficients. Since x[n] and y[n] are
causal, h[n] is also causal. That is, the convolution of two causal signals would always give a causal
result.
h[n] = h[0]δ[n] + h[1]δ[n − 1] (1)
Since initial conditions are zero (LTI): h[0]=1 and h[1]=-1.

2
 (b)

Problem 4

Consider a discrete-time linear time-invariant (LTI) system with input signal x[n] and output signal y[n]
that is governed by the following second-order difference equation for n ≥ 0:

y[n] = 1.8y[n − 1] − Ky[n − 2] + x[n] (2)

where K is a real-valued constant.

(a) What are the initial conditions of the system and what values should they have?
(b) Derive the transfer function H(z) for the system, which will depend on K.

(c) Give the range of values for K for which the system is bounded-input bounded-output (BIBO) stable.
(d) Describe the possible frequency selectivity (lowpass, highpass, bandpass, bandstop, allpass or notch)
that the system could exhibit for different values of K for which the system is BIBO stable.

Solution

(a) A necessary condition for a system to have LTI properties is that it must be “at rest”. That is, the
initial conditions of the system must be zero. We can find the initial conditions of the system by
computing the first output values y[n] = 1.8y[−1] − Ky[−2] + x[0]. x[-1] and y[-2] are not initial
conditions, y[-1]=0 and y[-2]=0.

(b)
Y (z) = z.8z −1 Y (z) − Kz −2 Y (z) + X(z)
Y (z) z2
H(z) = = 2
X(z) z − 1.8z + K
√
Therefore, H(z) has two zeros at the origin and two poles at 0.9 ± 0.81 − K
(c) Both poles must be inside unit circle for BIBO. Therefore, for K≤ 0.81 : polesarerealvaluedandwithinunitcircle, f orK >
0.81 : thepolesarecomplexvalued.Bysquaringbothsides,(0.9)2 + (K − 0.81) < 1 which means K < 1.
Therefore, 0.8 < K < 1.
(d) For 0.8 < K < 0.81, poles are real-valued between 0.8 and 1.0, not inclusive. Lowpass. At K = 0.81,
there is double pole at z = 0.9, which is also means a lowpass response. As K increases from 0.81 to
1, the pole separation increases and response becomes bandpass.

3
Problem 5

Bluetooth operates in the 2400-2499 MHz unlicensed frequency band. At any given time, Bluetooth will
transmit on one of 79 channels, and each channel is 1 MHz wide. Channel k begins at (2402 + k) MHz
where k = 0, 1, . . . , 78. Bluetooth changes the 1 MHz channel on which it operates 1600 times/second
to avoid interference. A Bluetooth receiver has two subsystems in cascade. The first subsystem involves
continuous-time signal processing blocks and the second subsystem involves discrete-time signal processing
blocks.

(a) The continuous-time signal processing blocks are given below, where r(t) is the received radio frequency
signal. In the plot for R(f), one of the 1 MHz channels is shaded, and its counterpart in negative
frequencies is also shaded. Demodulation produces y(t), whose spectrum Y(f) is below. Let f1 = 2400
MHz and f2 = 2499 MHz. What is the demodulating frequency fc?

(b) The first discrete-time signal processing block is filtering. Design a second-order linear timeinvariant
(LTI) infinite impulse response (IIR) filter to extract channel k from y[n]. Assume the sampling rate
in part (a) is fs = 200 MHz. Give formulas for, and plot, the two poles and two zeros.

Solution

(a) fc = f1 . Analog/RF front end performs sinusoidal amplitude demodulation to shift the positive
frequency band in R(f) to the left by f1 and negative frequency band in R(f) to the right by f1 .

(b) Due to the analog/RF front end in part (a), channel k in y(t) resides between (k+2) MHz and (k+3)
MHz. Center frequency is at (k+2.5) MHz in continuous-time frequency and ωk = 2π k+2.5 200 in discrete-
time frequency. For the second-order discrete-time IIR filter, the poles are at p0 = 0.9ejωk and
p1 = 0.9e−jωk . For small values of k, poles would be close to 0 rad/sample. To avoid strong interaction
between zeros and poles, place the two zeros at z = −1.

Problem 6

When sound waves propagate through air, or when electromagnetic waves propagate through air, the waves
are absorbed, reflected and scattered by objects in the environment. In the transmission of sound waves over
the air in a room from an audio speaker to a microphone, we will model the direct path from the speaker to
the microphone as having zero delay, and a one-bounce path from the speaker to an object and then to the
microphone having delay t1.

4
This single reflection is a type of echo. We model the signal y(t) at the output of the microphone as

y(t) = x(t) − αx(t − t1 ) (3)

where α is a real-valued constant and t1 ≥ 0. We model that system that connects x(t) and y(t) as linear
and time-invariant (LTI).

(a) Derive a formula for the impulse response h(t).

(c) We add an LTI filter at the microphone output to remove as much of the echo as possible. Design the
continuous-time filter by giving its transfer function G(s) in the Laplace domain for cases where α < 0,
α = 0, and α > 0. The filter must be bounded-input bounded-output (BIBO) stable.

Solution

(a) h(t) = δ(t) − αδ(t − t1 )

(c) (There will not be question testing for BIBO in the exam)