EEE 431: Telecommunications I (Spring 2021)

MIDTERM

Date and Time: Friday, April 9th, 2021

Instructor: Tolga M. Duman

Name:

Bilkent ID:

Prob. 1: /7

Prob. 2: /7

Prob. 3: / 11

Total: / 25

• This is closed book, closed notes exam.

Prb. 1 Consider a discrete memoryless source with probability density function (PDF)
   
1 t 1 t
fX (x) = Π + Π
8 4 4 2
is being transmitted through uniform PCM. The sampling rate is 10k samples per second, and the number of
bits per sample is v = 10. (Note: Π(t) is a rectangular pulse of amplitude 1 extending from −1/2 to 1/2.)
a) [1 Point] Determine the minimum bandwidth required to transmit the PCM signal.
b) [3 Points] Assuming that there are no bit errors in transmission, determine the resulting SQNR (in dB).
c) [3 Points] Assume that Gray mapping is employed, and that the resulting bits are transmitted through a
noisy channel with a bit error probability of 10−3 . Since the error probability is low, we can approximately
say that each sample is received correctly, or there is only one bit error (among the 10 bits used for each
sample). That is, we can ignore the case of more than one bit error for each sample. (Note that for Gray
mapping, codewords for adjacent samples differ in only one bit. Also notice that the source PDF is a
constant in each quantization interval.)
- Determine the (conditional) PDF of the quantization error given that there is a bit error in transmission
of a sample.
- Compute the SQNR of the PCM scheme (and express your answer in dB).
Prb. 2 [7 Points] The following four signals are used for digital modulation with a symbol period T . Apply the Gram-
Schmidt orthonormalization procedure in the given order of signals to determine an orthonormal basis. Also
find the vector representations of the four signals. (Note: if you use inspection to solve the problem instead
of applying the Gram-Schmidt procedure in the given order, you will receive at most 4 out of 7 Points.)

"# (!) "* (!) ") (!) "+ (!)

1 1 1 1

! ! ! !
2& & & & & & 2& &
3 3 3 3
 2

Prb. 3 Consider a binary antipodal modulation scheme with transmit pulse p(t) = at (for t ∈ [0, T )), where T is the
bit period and a is a positive constant. The bit “1” is transmitted via p(t), and “0” is transmitted via −p(t).
Assume that the transmitted bits are equally likely, and that the transmission takes place over an AWGN
channel. The received signal is given by r(t) = s(t) + n(t) (for t ∈ [0, T )) where s(t) is the transmitted signal
and n(t) is white Gaussian noise with power spectral density Sn (f ) = N0 /2.
a) [4 Points] Determine the optimal receiver structure. Be specific and clearly label any diagrams you use.
Also determine the average bit error probability (in terms of a, T and N0 ).
b) [3 Points] Assume that another receiver structure is as follows: we integrate the received signal r(t) over
the symbol period, i.e., we obtain

ZT
0
r = r(t)dt,
0

and decide in favor of “1” if r0 > 0 and in favor of “0” if r0 < 0. What is the resulting average bit error
probability? How does this compare with the one in part a?
c) [4 Points] Consider a different receiver which integrates the received signal over two overlapping intervals
(0, 2T /3) and (T /3, T ), and produces
2T
Z /3 ZT
r1 = r(t)dt, r2 = r(t)dt.
0 T /3

- Determine the (conditional) joint PDF of r1 and r2 given that the transmitted bit is 1, i.e., f (r1 , r2 |1).
Also compute f (r1 , r2 |0).
- The ML receiver (based on the two integrator outputs r1 and r2 ) maximizes f (r1 , r2 |m) over m = 0
and 1. Simplify this decision rule as much as you can.