Вы находитесь на странице: 1из 3

Communication Technology Laboratory

Prof. Dr. H. Blcskei, Reinhard Heckel


Sternwartstrasse 7
CH-8092 Zrich

Fundamentals of Wireless Communication


Homework 4
Handout date: April 29, 2014

Problem 1 Real World Fading Channels


Use the handout about channel characteristics and common wireless systems to classify the
following wireless systems, as demonstrated in Discussion 7:
1. A GSM900 terminal used by a pedestrian in a built-up urban environment.
2. The same pedestrian, but now using an IS-136 terminal. IS-136 is a U.S. cellular standard,
similar to GSM.
3. An IS-95 terminal used in a car in a suburban environment at typical speeds (50 km/h). IS-95,
also known as CDMAOne, is the second generation cellular system developed by Qualcomm
and used mainly in the U.S.A., South Korea, and other Asian countries.
4. An IEEE 802.11a enabled personal digital assistant (PDA) used at walking speed in an environment where the line-of-sight (LOS) to the base station is blocked.
5. A UMTS terminal operated on board a high-speed train (250 km/h) passing through a rural
environment.
6. Bonus question: Electromagnetic waves are unsuited for communication under water. Instead, acoustic signals are often used. The corresponding channels are quite different from
typical wireless communication channels, because the speed of sound in salt water is approximately c=1500 m/s, and thus much lower than the speed of light; furthermore, time
dispersion is high, with an RMS delay spread of around 10 ms in shallow-water channels. The
bandwidth of communication systems is rather small, on the order of 10 kHz for systems
operating over distances up to 10 km, and these systems often operate at baseband because
path loss is high. Doppler dispersion arises because terminals might be moving, and because
the ocean surface is constantly in motion, especially if winds are strong. Speeds of the ocean
surface and of terminals can reach up to 10 m/s. Classify such a communication system like
the systems described in points 15.
Note: The parameter description of wireless channels reduces all the information contained in
the scattering function to just a few numbers; consequently, the resulting channel characterization
is rather imprecise. The relation between delay spread and coherence bandwidth, and Doppler
spread and coherence time is an approximate one as well, so that some systems cannot be classified
unambiguously.

Problem 2 Selection Combining


The ML receiver for repetition coding with L branches of diversity uses maximum ratio combining
(MRC), where each diversity branch is weighted by the corresponding complex conjugated channel

coefficient, and all branches are added coherently. Hardware implementation of such a receiver
is often costly because a large number of phase rotations need to be performed. Even more challenging and complex is the necessary estimation of all the L channel taps, which is a prerequisite
for MRC. In practice, it is thus often desirable to use a less complex receiver structure. Instead of
estimating and combining all diversity branches, the receiver complexity can be reduced if only the
strongest branch, i.e., the one offering the highest per-branch SNR, is used: this is called selection
combining. It is commonly found in cost-sensitive products, e.g., in IEEE 802.11a/b/g base-stations.
A receiver that performs selection combining checks all the diversity branches in each coherence
time interval of the channel and then uses only the signal received via the diversity branch with
the largest magnitude to detect the transmitted information.
For a receiver that is equipped with L antennas and uses selection combining, compute an
approximation of the high-SNR error probability for BPSK. You can assume that the channels
corresponding to the individual receive antennas are i.i.d. Rayleigh fading. Discuss the differences
and similarities compared to MRC.
Hints: Selection combining is an ad-hoc concept; therefore, do not start from the derivation of
the ML detector, but use the approximate typical error event approach that was explained in class.
Use as well the following high-SNR approximation: Let {Xi }L
i=1 be a set of i.i.d. random variables
x
with PDF fX (x) = e , and define Y = maxi {Xi }. Then, for y 0, the PDF of Y , fY (y), satisfies
fY (y) Ly L1 .

(2.1)

Another way to approach this problem is to use order statistics, a description of which can be found
in any book on probability theory, e.g., the book by Papoulis mentioned on the class website.

Problem 3 Coherent Detection of QPSK


In class and in the discussion session, we computed the error probability for coherent detection of
BPSK over a flat Rayleigh-fading channel. In practice, the goal is to use the channel in the most
efficient way, i.e., we want to achieve a high spectral efficiency (measured in bits/s/Hz). Thus,
signal constellations that use both complex dimensions, also called the inphase and quadrature
components, are commonly used. The simplest complex constellation is quadrature phase-shift
keying (QPSK). Consider QPSK modulation with equiprobable symbols chosen from the set
r
Es
{(1 + j), (1 j), (1 + j), (1 j)}.
A=
2
A pragmatic approach to upper bound the error probability, which also works for more complicated
symbol constellations is to compute the union bound1 of the pairwise error probabilities
X
Pe
P(i j | i) P(i),
i6=j

where P(i j | i) is the error probability of binary detection between symbols i and j when
symbol i is sent.
1

The union bound states that the probability of the union of events is bounded by the sum of the probability of the
individual events, i.e.
!
[
X
P
Ai
P(Ai ).
i

where the Ai denote events.

1. Compute this bound for the discrete-time flat Rayleigh-fading channel model used in class,
where you can assume that the channel taps h[m] are i.i.d.
2. Argue that the error probability decays at least as 1/SNR at high SNR.
3. Draw the QPSK constellation in the complex plane; for a given transmitted symbol, indicate
the error event for the ML detector and the error event used to compute the union bound.
4. On the basis of your diagram, can you come up with an upper bound on the QPSK error
probability that is better than the union bound? Compute its error probability.

Problem 4 Sufficient Statistics (Wozencraft & Jacobs)




Consider the problem of detecting X from Y1 Y2 T , where:
Y1 = X + Z 1 ,
Y2 = Z 1 + Z 2 .
Here, X is equally likely to be

E or E, and Z1 and Z2 are i.i.d. N (0, 2 ), independent of X.

1. Is Y1 a sufficient statistic for the detection problem? Give a rigorous as well as an intuitive
explanation.
Hint: To obtain a rigorous answer, express the detection problem as a vector detection
problem.
2. Derive the ML detector for X given Y1 and Y2 , and compute the probability of error.
3. Consider now the ML detector of X that uses only Y1 . Find the additional energy neededif
anyto achieve the same error probability as in part 2.
Hint: Compare the arguments of the respective Q-functions.

Вам также может понравиться