EE 359: Wireless Communications - Fall 2003

SECOND EXAM

This exam is open book and notes, and calculators and your reader are needed. Graded exams and solutions
will be available within the next week. Good luck!

Problem 1 (20 points): Short Problems.

(a) Equalizer Tracking: Why does an equalizer that tracks the channel during data transmission
still need to train periodically? Name 2 benefits of tracking.
(b) Narrowband Interference Effects: Describe qualitatively what effect a narrowband interfer-
ence signal has on an OFDM signal, a direct sequence spread spectrum signal, and a frequency
hopped spread spectrum signal.
(c) OFDM Design: Consider a channel with maximum delay spread TM = 40µsec. Is this likely
to be an indoor or outdoor channel and why? Suppose you build an OFDM system for this
channel with a total system bandwidth of 640 KHz. How many subchannels are needed to
obtain approximately flat fading on each subchannel? How many samples long does the cyclic
prefix need to be to insure no ISI between FFT blocks, assuming the subchannel symbol time
equals the inverse of the subchannel bandwidth?

Problem 2 (30 points): Capacity vs. Diversity in MIMO with Adaptive Modulation.
This problem illustrates how multiple antennas can be used for either diversity gain or capacity gain,
and how adaptive modulation can be used with either technique. Consider a 2x2 MIMO system that
decomposes into two parallel AWGN channels:

y1 = α1 x1 + n1 .

y2 = α2 x2 + n2 .
Assume α1 = 1, α2 = 3, the total power to be divided between the two transmit antennas is 100 mW, the
spectral noise density of n1 and n2 is N0 = 10−10 W, the bandwidth of the MIMO system (and hence each
of the associated parallel channels) is 10 MHz, and the symbol time xi , i = 1, 2 is T = 1/B.

(a) We first use adaptive modulation on each of the two parallel channels. Find the optimal power
allocation and the corresponding maximum constellation size M that can be sent over each
channel for a target Pb = .2e−1.5γ/(M −1) = 10−3 . Assume no restrictions on M . What is the
total data rate (sum of rates on each of the parallel channels) associated with this technique?
(b) Let us now use the parallel channels for diversity gain. Assume that we send the same symbol
over each channel with transmit power c1 P over the channel with gain α1 and c2 P over the
channel with gain α2 for some c1 and c2 with c1 + c2 = 1. The received SNR on channel 1 will
thus be c1 α12 P/(N0 B) and on channel 2 it will be c2 α22 P/(N0 B). Assuming we use MRC to
 combine the output of the two receive antennas, find the SNR of the combiner output. Based
on this expression, find the optimal values for c1 and c2 to maximize this output SNR. Finally,
compute the output SNR for these optimal ci values, and determine the maximum constellation
size M that can be sent over this diversity channel with Pb = .2e−1.5γ/(M −1) = 10−3 and the
corresponding data rate.
(c) Which technique leads to better performance for this system: using antennas for capacity gain
or for diversity gain and why?

Problem 3 (30 points): Spread Spectrum and RAKE Receivers

This problem illustrates the benefits of RAKE receivers and the optimal choice of multipath components
for combining when the receiver complexity is limited. Consider a multipath channel with impulse response

h(t) = α0 δ(t) + α1 δ(t − τ1 ) + α2 δ(t − τ2 ).

The αi are Rayleigh fading coefficients, but their expected power varies due to shadowing such that
E[α02 ] = 5 with probability .5 and 10 with probability .5, E[α12 ] = 0 with probability .5 and 20 with
probability .5, and E[α22 ] = 5 with probability .75 and 10 with probability .25 (all units are linear). The
transmit power and noise power are such that a spread spectrum receiver locked to the ith multipath
component will have an SNR of αi2 in the absence of the other multipath components.

(a) Assuming maximal linear codes, a bit time Tb , and a spread spectrum receiver locked to the LOS
signal component (with zero delay and gain α0 ), for what values of τ1 and τ2 , 0 ≤ τ1 ≤ τ2 < Tb
will their corresponding multipath components be attenuated by −1/N , where N is the number
of chips per bit.

For the rest of the problem assume spreading codes with autocorrelation equal to a delta function.

(b) What is the outage probability of DPSK modulation at an instantaneous Pb = 10−3 for a single
branch spread spectrum receiver locked to the LOS path.
(c) What is the outage probability of DPSK modulation at an instantaneous Pb = 10−3 for a 3-
branch RAKE receiver where each branch is locked to one of the multipath components and SC
is used to combine the paths.
(d) Suppose receiver complexity is limited such that only a 2-branch RAKE with SC can be built.
Find which two multipath components the RAKE should lock to in order to minimize the outage
probability of DPSK modulation at Pb = 10−3 and find this minimum outage probability.
Problem 4 (20 points): Equalizers and Noise Enhancement
This problem illustrates the noise enhancement of zero-forcing equalizers, and how this enhancement
can be mitigated using an MMSE approach. Consider a frequency-selective fading channel with baseband
frequency response 

 1 0 ≤ f < 10KHz



 1/2 10KHz ≤ f < 20KHz

 1/3 20KHz ≤ f < 30KHz
H(f ) =

 1/4 30KHz ≤ f < 40KHz



 1/5 40KHz ≤ f < 50KHz


0 else
The frequency response is symmetric in positive and negative frequencies. Assume an AWGN channel with
noise PSD N0 = 10−9 .

(a) Find the analog zero-forcing equalizer for this channel.

(b) Find the total noise power at the output of the equalizer from part (a).
(c) Assume a MMSE analog equalizer of the form Heq (f ) = 1/(H(f ) + α). Find the total noise
power at the output of this equalizer for an AWGN input with PSD N0 for α = .5 and for α = 1.
(d) Describe qualitatively two effects on a signal that is transmitted over channel H(f ) and then
passed through the MMSE equalizer Heq (f ) = 1/(H(f ) + α) with α > 0. What design consid-
erations should go into the choice of α?
(e) What happens to the total noise power for the MMSE equalizer in part (c) as α → ∞? What
is the disadvantage of letting α → ∞ in this equalizer design?