ECE4305: Software-Defined Radio Systems and Analysis

Laboratory 3: Receiver Structure & Waveform Synthesis

of a Transmitter and a Receiver
C-Term 2011

Objective
This laboratory will cover basic receiver structures and implementations. We will also show how
to construct a series of orthonormal basis functions that can be combined to produce a wide range
of signal waveforms. Then, we will study a specific digital transceiver implementation based on
the multicarrier transmission concept called orthogonal frequency division multiplexing (OFDM). In
the experimental part of this laboratory, you will implement two different receiver structures and
then observe their performance during over the air transmission. This is followed by a Simulink
implementation of an OFDM communication system. Finally, the open-ended design problem will
focus on bi-directional communication.

Contents
1 Theoretical Preparation 3
1.1 Gram-Schmidt Orthogonalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Optimal Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Maximum Likelihood Detection in an AWGN Channel . . . . . . . . . . . . . 4
1.2.2 Maximum A Posteriori Detection . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Basic Receiver Realizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 Matched Filter Realization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.2 Correlator Realization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Multicarrier Transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.1 Orthogonal Frequency Division Multiplexing . . . . . . . . . . . . . . . . . . . 9
1.4.2 Dispersive Channel Environment . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.3 OFDM with Cyclic Prefix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Suggested Readings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 Software Implementation 16
2.1 Observation Vector Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Maximum-likelihood Decoder Implementation . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Correlator Realization of a Receiver in Simulink . . . . . . . . . . . . . . . . . . . . . 18
2.4 Multicarrier Modulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4.1 MATLAB Design of Multicarrier Transmission . . . . . . . . . . . . . . . . . . 20
2.4.2 Simulink Design of Multicarrier Transmission . . . . . . . . . . . . . . . . . . 20

1
3 USRP2 Hardware Implementation 23
3.1 Eye Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.1 Discrete-Time Eye Diagram Scope . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Matched Filter Observation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

4 Open-ended Design Problem: Duplex Communication 26

4.1 Duplex Communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Half-duplex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3 Time Division Duplexing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.4 Hints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

5 Lab Report Preparation & Submission Instructions 29

2
1 Theoretical Preparation
Signals can be represented as either waveforms or vectors. Waveform representations are defined by
the Fourier series of a signal, or the sum of sines and cosines that make a particular shape. As a
vector, a signal is represented as series of orthonormal vectors. This part of the lab will show how
to derive these vectors and their basis functions. It will then show two receiver designs and their
implementations.

1.1 Gram-Schmidt Orthogonalization

In mathematics, particularly linear algebra and numerical analysis, the Gram-Schmidt process is a
method for creating an orthonormal set of vectors in an inner product space such as the Euclidean
space Rn . The Gram-Schmidt process takes a finite, linearly independent set S = {S1 (t), S2 (t), ...,
SM (t)} and generates an orthogonal set S ′ = {φ1 (t), φ2 (t), ..., φi(t)} that spans the same subspace
of Rn as S.

To derive a set of orthogonal basis functions φ1 (t), φ2 (t), ..., φi (t) from set of energy signals denoted
by S1 (t), S2 (t), ..., SM (t) we will use the following functions:
i−1
X
gi (t) = Si (t) − Sij φj (t), (1)
j=1

where: Z T
Sij = Si (t)φj dt, j = 1, 2, ..., i − 1. (2)
0
Using these two functions, we define our basis function φi as:
gi (t)
φi = q R , i = 1, 2, ..., N. (3)
T 2
0
gi (t)dt

In Equation (1), we are developing our basis functions by subtracting the projections of the other
waveforms on Si from Si . Note that Sij is the projection of Sj onto Si and is defined in Equation (2),
while Equation (3) is normalizing gi by its energy. Consequently, in several circumstances it may be
much simpler to compute this graphically than to work out the integral. Note that the order in which
you conduct the Gram-Schmidt process with respect to the selection of the waveforms Si will yield
different sets of orthonormal basis functions.

1.2 Optimal Detection

Detection theory, or signal detection theory, is used in order to discern between signal and noise.
Using this theory, we can explain how changing the decision threshold will affect the ability to discern
between two or more scenarios, often exposing how adapted the system is to the task, purpose or goal
at which it is aimed.

When a signal is received, the probability of correct detection:

P (correct) = MAX → P (m̂i = mi ), (4)

3
and the probability of an error:
P (error) = MIN → P (m̂i 6= mi ). (5)
Recall Bayes’ Rule which gives the conditional probability of an event and is stated by:
\
P (A|B)P (B) = P (A B) = P (B|A)P (A)

when all components are expressed as probabilities, or:

p(x, y) = p(x|y)p(y) = p(y|x)p(x)
when all components are expressed as PDFs, or:
P (Si |r = p) = p(p) = p(p|Si )P (Si)
when all components are expressed as a mixture. These expression result in the rule:
p(p|Si ) · P (Si)
P (Si |r = p) = . (6)
p(p)
Recalling the rule for the optimal detector given in Equation (4), using the mixed form of Bayes’
Rule, we may define the optimal detector as [1]:
p(p|Si) · P (Si )
max P (Si |r = p) = max . (7)
Si Si p(p)
Since p(p) does not depend upon Si , we can simplify the optimal detector to:
max p(p|Si )P (Si ). (8)
Si

1.2.1 Maximum Likelihood Detection in an AWGN Channel

Maximum likelihood detection is a popular statistical method employed for fitting a statistical model
to data, and for identifying model parameters. In general, for a fixed set of data and underlying prob-
ability model, the method of maximum likelihood selects values of the model parameters that produce
the distribution that are most likely to have resulted in the observed data (i.e., the parameters that
maximize the likelihood function).

To develop the maximum-likelihood detection algorithm, let us consider an example. The signal
transmitted in each signal interval is binary PAM. Hence, there are two possible transmitted signals
√
corresponding to the signal points s1 = −s2 = εb , where εb is the energy per bit. The output of the
matched-filter or correlation demodulator for binary PAM in the kth signal interval may be express
as:
√
rk = ± εb + nk , (9)
where nk is a zero-mean Gaussian random variable with variance σn 2 = N0 /2. Consequently, the
conditional PDFs for the two possible transmitted signals are:
√
(rk − εb )2
 
1
p(rk |s1 ) = √ exp − 2
,
2πσn 2σ√n
(10)
(rk + εb )2
 
1
p(rk |s2 ) = √ exp − .
2πσn 2σn 2

4
Suppose we observe the sequence of matched-filter outputs r1 , r2 , ..., rk . Since the channel noise is
assumed to be white and Gaussian, it follows that E(nk nj ) = 0, k 6= j. Hence, the noise sequence
n1 , n2 , ..., nk is also white. Consequently, for any given transmitted sequence s(m) , the joint PDF of
r1 , r2 , ..., rk may be expressed as a product of K marginal PDFs, i.e.:
K
(m)
Y
(m)
p(r1 , r2 , ..., rK |s ) = p(rk |sk )
k=1
K
" #
(m)
Y 1 (rk − sk )2
= √ exp −
2πσn 2σn 2
k=1
" K #

1
K X (rk − s(m) )2
k
= √ exp − 2
, (11)
2πσn k=1
2σn

√ √
where either sk = εb or sk = − εb . Then, given the received sequence r1 , r2 , ..., rK at out-
put of the matched filter or correlation demodulator, the detector determines the sequence s(m) =
(m) (m) (m)
s1 , s2 , ..., sK that maximizes the conditional PDF p(r1 , r2 , ..., rK |s(m) ). Such a detector is called
the maximum-likelihood (ML) detector.

By taking the logarithm of Equation (11) and neglecting the terms that are independent of p(r1 , r2 , ..., rK ),
we find that an equivalent ML detector selects the sequence s(m) that minimizes the Euclidean distance
metric, namely:
K
(m)
X
(m)
D(r, s ) = (rk − sk )2 . (12)
k=1

For more information about maximum likelihood detection, please refer to Section 5.6 of the course
textbook [4].

1.2.2 Maximum A Posteriori Detection

In contrast to the maximum-likelihood detector for detecting the transmitted information, we now de-
scribe a detector that makes symbol-by-symbol decisions based on the computation of the maximum
a posteriori probability (MAP) for each detected symbol. Hence, this detector is optimum in the sense
that it minimize the probability of a symbol error. The detection algorithm that is presented below
is due to [3], who developed it as a detection algorithm for channels with intersymbol interference,
i.e., channels with memory.

We illustrate the algorithm in the context of detecting a PAM signal with M possible levels. Suppose
that it is desired to detect the information symbol transmitted in the kth signal interval, and let
r1 , r2 , ..., rk+D be the observed received sequence, where D is the delay parameter which is chosen to
exceed the signal memory, i.e., D ≥ L, where L is the inherent memory in the signal. On the basis of
the received sequence, we compute the posterior probabilities:
P (s(k) = Am |rk+D , rk+D−1 , ..., r1 ), (13)

for the M possible symbol values, and choose the symbol with the largest probability. Since we have:
p(rk+D , ..., r1 |s(k) = Am )P (s(k) = Am )
P (s(k) = Am |rk+D , rk+D−1 , ..., r1 ) = , (14)
p(rk+D , rk+D−1, ..., r1 )

5
and since the denominator is common for all M probabilities, the MAP criterion is equivalent to
choosing the value of s(k) that maximize the numerator of Equation (14). Thus, the criterion for
deciding on the transmitted symbol s(k) is:
s̃(k) = arg{max p(rk+D , ..., r1 |s(k) = Am )P (s(k) = Am )}. (15)
When the symbols are equally probable, the probability P (s(k) = Am ) may be dropped from the
computation.

The algorithm for computing the probabilities in Equation (15) recursively begins with the first symbol
s(1) . In general, the recursive algorithm for detecting the symbol s(k) is as follows: Upon reception of
rk+D , ..., r2 , r1 , we compute:
s̃(k) = arg{max p(rk+D , ..., r1 |s(k) )P (s(k) )}
X X
= arg{max ··· pk (sk+D , ..., s(k+1) , s(k) )}, (16)
s(k+D) s(k+1)

where by definition:
pk (sk+D , ..., s(k+1) , s(k) )
X
= p(rk+D |s(k+D) , ..., s(k+D−L))P (s(k+D) ) pk−1 (s(k−1+D) , ..., s(k−1) ). (17)
s(k−1)

Thus, the recursive nature of the algorithm is established by the Equations (16) and (17).

1.3 Basic Receiver Realizations

The fundamental challenge of digital communications is recovering what was transmitted after it has
passed through a channel and been corrupted by noise. The first receiver structure we will examine
is based on filtering the received signal with a static filter that maximizes the SNR of the channel.
This receiver structure is ideal because it maximizes SNR. From Laboratory 2, it was shown that this
minimizes the bit error rate. The major disadvantage to matched filtering is that it requires a priori
knowledge of what was transmitted.

1.3.1 Matched Filter Realization

Often in digital transmissions, data fields are preceded by headers. These headers are often standard-
ized and can be anticipated. If you know what was transmitted (such as the header, for example),
you can design your receiver as a filter that exactly matches the transmission pulse shape. This shape
maximizes the signal-to-noise ratio (SNR) and yields the best decoding results. When designing a
receiver, we are interested in the optimal detection rule, or the receiver that yields the correct result
more than any other receiver would.

Now that the basis for our detection rule has been outlined, define a signal x(t) as:
x(t) = g(t) + w(t), 0 ≤ t ≤ T, (18)
where g(t) is a pulse signal and w(t) is additive noise with µ = 0 and P SD = N20 . We are assuming
that the receiver knows the possible waveforms that g(t) could take. The function of the receiver is
to detect the pulse signal g(t) in an optimum manner, given the received signal x(t).

6
 h1 (t )

C
E1 / 2 H
O
h 2
(t ) O
r(t)
S m(t)
E
E2 / 2
M
A
X
h N
(t )

EN / 2

Figure 1: Matched Filter Realization. A matched filter is obtained by correlating a known signal,
or template, with an unknown signal to detect the presence of the template in the unknown signal.
This is equivalent to convolving the unknown signal with a conjugated time-reversed version of the
template (cross-correlation).

The first step is to take the inverse Fourier Transform of H(f )G(f ), which is a convolution of h(t)
and g(t): Z ∞
g0 (t) = H(f )G(f )ej2πf t df. (19)
−∞
Next, we find the energy: Z ∞
2
|g0(t)| = | H(f )G(f )ej2πf t df |2. (20)
−∞
Furthermore, the peak pulse SNR is defined as:
|g0 (T )|2
η= , (21)
E{n2 (t)}
where |g0 (t)|2 is the instantaneous power of the output signal and E(n2 )(t) is the average power of the
output noise. Since w(t) is Gaussian, we know its power spectral density (PSD) to be N20 . Applying
the definition of η, we get: R∞
| −∞ H(f )G(f )ej2πf T df |2
η= N0 ∞
R . (22)
2 −∞
|H(f )|2df
We see that we need to solve for H(f ) such that it yields the largest possible η. Applying Schwarz’s
Inequality, it can be shown that Hopt = kG∗ (f )e−j2πf T which is equal to:

Hopt = kg(T − t). (23)

This result yields insight into the term “matched filter.” We are convolving the time flipped and
time-shifted version of the transmitted pulse with the transmitted pulse itself. This attempt to match

7
the pulses is also maximizing the SNR.

Consider the realization shown in Figure 1, where hi (t) = Si (T − t). Our decision rule here, derived
from Bayes’ procedure is given as:
max(E{P (Si|r = p)}. (24)
i

For more information about matched filter, please refer to Section 5.1.4 of the course textbook [4].

1.3.2 Correlator Realization

A matched filter realization assumes knowledge of the transmitted data. A receiver based upon a
correlator structure loosens this assumption by instead assuming only a knowledge of the waveforms.

Consider again the received signal given in (18). Since we possess a knowledge of the signal set and
we want to know which Si was transmitted, we can correlate r(t) with Si (t). If we normalize the
result by the energy with respect to Ei , we will have a good approximation of how close r(t) is to Si .
This decision rule can be expressed mathematically as:
Z T
Ei
max( r(t)Si (t)dt − ). (25)
i 0 2

(K 1)T

( ) dT
KT
C
DUMP H
S 1 (t ) E1 / 2
(K 1)T O
( ) dT O
r(t)
KT S m(t)
E
DUMP E2 / 2
S 2 (t )
M
A
(K 1)T X
( ) dT
KT

DUMP
S N (t ) EN / 2

Figure 2: Correlator realization, which does not assume knowledge of the transmitted data.

1.4 Multicarrier Transmission

Instead of having one center frequency, fc , Multi-Carrier Modulation (MCM) multiplexes serial input
data into several parallel streams and transmits them over independent sub-carriers. These sub-
carriers can be individually modulated and manipulated allowing for optimization with respect to

8
the channel. This is especially true for a wireline communications system where the channel is well
defined.
Why MCM? What are the disadvantages of transmitting at a high data rate over a single carrier?
What happens if part of your channel is severely attenuated? MCM offers a different approach to
these problems. By sub-dividing the transmission over many carriers and tuning the modulation
schemes and power levels of each carrier, a multi-carrier transmission scheme becomes quite robust
to fast-fading channels and narrowband interference.
There are some important distinctions between OFDM in wired communications systems versus wire-
less communications systems. We will be focusing on wireless communications, but is important to
note that certain assumptions made here are not necessarily optimal solutions a wireline implemen-
tation.

1.4.1 Orthogonal Frequency Division Multiplexing

Orthogonal Frequency Division Multiplexing (OFDM) is an efficient type of multicarrier modulation,
which employs the discrete Fourier transform (DFT) and inverse DFT (IDFT) to modulate and de-
modulate the data streams. The set-up of an OFDM system is presented in Figure 3. A high-speed
digital input, x[m], is demultiplexed into N subcarriers using a commutator. The data on each sub-
carrier is then modulated into an M-QAM symbol, which maps a group of log2 (M) bits at a time.
For subcarrier k we will rearrange ak [ℓ] and bk [ℓ] into real and imaginary components such that the
output of the “modulator” block is pk [ℓ] = ak [ℓ] + jbk [ℓ]. In order for the output of the IDFT block
to be real, given N subcarriers we must use a 2N-point IDFT, where terminals k = 0 and k = N are
“don’t care” inputs. For the subcarriers 1 ≤ k ≤ N − 1, the inputs are pk [ℓ] = ak [ℓ] + jbk [ℓ], while for
the subcarriers N + 1 ≤ k ≤ 2N − 1, the inputs are pk [ℓ] = a2N −k [ℓ] + jb2N −k [ℓ].

The IDFT is then performed, yielding

2N −1
1 X
s[2ℓN + n] = pk [ℓ]ej(2πnk/2N ) , (26)
2N k=0

where this time 2N consecutive samples of s[n] constitute an OFDM symbol, which is a sum of N
different QAM symbols.

This results in the data being modulated on several subchannels. This is achieved by multiplying
each data stream by a sin(Nx)/ sin(x), several of which are shown in Figure 4.

The subcarriers are then multiplexed together using a commutator, forming the signal s[n], and
transmitted to the receiver. Once at the receiver, the signal is demultiplexed into 2N subcarriers of
data, ŝ[n], using a commutator and a 2N-point DFT, defined as
2N
X −1
p̄k [ℓ] = ŝ[2ℓN + n]e−j(2πnk/2N ) , (27)
n=0

is applied to the inputs, yielding the estimates of pk [ℓ], p̄k [ℓ]. The output of the equalizer, p̂k [ℓ], then
passed through a demodulator and the result multiplexed together using a commutator, yielding the
reconstructed high-speed bit stream, x̂[m].

9
 dk[m] pk[l]
s[n]
x[m]
Mod IDFT2N

Channel
N 2N

CP
d^k[m] ^
pk[l] pk[l]

^
x[m]
Demod Equalizer DFT2N ^
s[n]

N 2N 2N

Figure 3: Overall schematic of an Orthogonal Frequency Division Multiplexing System

8
Frequency Content

0
0 2 4 6
Frequency (rad)

Figure 4: Characteristics of Orthogonal Frequency Division Multiplexing: Frequency response of

OFDM subcarriers.

10
1.4.2 Dispersive Channel Environment
Until now, we have only considered an OFDM system operating under ideal conditions (i.e. the trans-
mitter is connected directly to the receiver without any introduced distortion). Now we will cover
channel distortion due to dispersive propagation and how it is compensated for in OFDM system.

From our high school physics courses, we learned about wave propagation and how they combine
constructively and destructively. The exact same principles hold in high speed data transmission.
For example, in a wireless communication system, the transmitter emanates radiation in all directions
(unless the antenna is “directional”, in which case, the energy is focused at a particular azimuth). In
an open environment, like a barren farm field, the energy would continue to propagate until some of
it reaches the receiver antenna. As for the rest of the energy, it continues on until it dissipates.

In an indoor environment, as depicted in Figure 5(c), the situation is different. The line-of-sight com-
ponent (if it exists), p1 , arrives at the receiver antenna first, just like in the open field case. However,
the rest of the energy does not simply dissipate. Rather, the energy is reflected by the walls and
other objects in the room. Some of these reflections, such as p2 and p3 , will make their way to the
receiver antenna, although not with the same phase or amplitude. All these received components are
functions of several parameters, including their overall distance between the transmitter and receiver
antennas as well as the number of reflections. At the receiver, these components are just copies of
the same transmitted signal, but with different arrival times, amplitudes, and phases. Therefore, one
can view the channel as an impulse response that is being convolved with the transmitted signal.
In the open field case, the channel impulse response (CIR) would be a delta, since no other copies
would be received by the receiver antenna. On the other hand, an indoor environment would have
a several copies intercepted at the receiver antenna, an thus its CIR would be similar to the exam-
ple in Figure 5(a). The corresponding frequency response of the example CIR is shown in Figure 5(b).

In an xDSL environment, the same principles can be applied to the wireline environment. The trans-
mitted signal is sent across a network of telephone wires, with numerous junctions, bridging taps, and
connections to other customer appliances (e.g. telephones, xDSL modems). If the impedances are not
matched well in the network, reflections occur and will reach the devices connected to the network,
including the desired receiver.

With the introduction of the CIR, new problems arise in our implementation which need to be
addressed. In Section 1.4.3, we will look at how to undo the smearing effect the CIR has on the
transmitted signal.

1.4.3 OFDM with Cyclic Prefix

The CIR can be modelled as a finite impulse response filter that is convolved with a sampled version
of the transmitted signal. As a result, the CIR smears past samples onto current samples, which are
smeared onto future samples. The effect of this smearing causes distortion of the transmitted signal,
increasing the aggregate BER of the system and resulting in a loss in performance.

Although equalizers can be designed to undo the effects of the channel, there is a trade-off between
complexity and distortion minimization that is associated with the choice of an equalizer. In partic-
ular, the distortion due to the smearing of a previous OFDM symbol onto a successive symbol is a

11
 1 6
0.8 4

Magnitude (db)
2

Amplitude
0.6
0
0.4 −2

0.2 −4
−6
0
0 1 2 3 4 0 1 2 3
Time Frequency (rad/sec)
(a) Impulse response (b) Frequency response

p2
Rx

p1

Tx

p3

(c) The process by which dispersive propagation arises

Figure 5: Example of a channel response due to dispersive propagation

difficult problem. One simple solution is to put a few “dummy” samples between the symbols in order
to capture the intersymbol smearing effect. The most popular choice for these K dummy samples are
the last K samples of the current OFDM symbol. The dummy samples in this case are known as a
cyclic prefix, as shown in Figure 6(a).

Therefore, when the OFDM symbols with cyclic prefixes are passed through the channel, the smearing
from the previous symbols are captured by the cyclic prefixes, as shown in Figure 6(b). As a result,
the symbols only experience smearing of samples from within their own symbol.

At the receiver, the cyclic prefix is removed, as shown in Figure 6(c), and the OFDM symbols proceed
with demodulation and equalization.

Despite the usefulness of the cyclic prefix, there are several disadvantages. First, the length of
the cyclic prefix must be sufficient to capture the effects of the CIR. If not, the cyclic prefix fail to
prevent distortion introduced from other symbols. The second disadvantage is the amount of overhead
introduced by the cyclic prefix. By adding more samples to buffer the symbols, we must send more
information across the channel to the receiver. This means to get the same throughput as a system

12
without the cyclic prefix, we must transmit at a higher data rate.

}
}
CP SymbolM-1 CP SymbolM CP SymbolM+1

(a) Add cyclic prefix to an OFDM symbol

h[n] h[n] h[n]

CP SymbolM-1 CP SymbolM CP SymbolM+1

(b) Smearing by channel h(n) from previous symbol into cyclic prefix

CP CP CP

SymbolM-1 SymbolM SymbolM+1

(c) Removal of cyclic prefix

Figure 6: The process of adding, smearing capturing, and removal of a cyclic prefix

1.5 Suggested Readings

Although this laboratory handout provides some information about the fundamentals of digital mod-
ulation schemes, the reader is encouraged to review the material from the following references in order
to gain further insight on these topics.

• Overview of Maximum Likelihood Detection

– Section 5.6 in [4]

• Maximum Likelihood Detection in Synchronization

– Section 7.8 and 8.8 in [4]

• Introduction to Simulation Techniques on Correlation Receiver

– Section 2.4 and 3.1 in [5]

• Overview of Multicarrier Data Transmission

– Section 2.1 in [6]

13
1.6 Problems
1. Carry out the Gram-Schmidt Orthogonalization procedure of the signals in Figure 7 in the order
s3 (t), s1 (t), s4 (t), s2 (t) and thus obtain a set of orthonormal functions {fm (t)}. Then, determine
the vector representation of the signals {sn (t)} by using the orthonormal functions {fm (t)}.
Also, determine the signal energies.

s1 (t ) s3 (t )

1 1

t t
1

s2 (t ) s4 (t )

t t
1 1

Figure 7: Four signal waveforms for the Gram-Schmidt Orthogonalization procedure.

A
2. Consider the signal: s(t) = T
tcos(ωc t) for 0 ≤ t ≤ T

(a) Determine the impulse response of the matched filter for the signal.
(b) Determine the output of the matched filter at t = T .
(c) Suppose the signal s(t) is passed through a correlator that correlates the input s(t) with
s(t). Determine the value of the correlator output at t = T . Compare your result with
that in part (b).

3. A certain digital baseband modulation scheme uses the pulse shown in Figure 8 to represent
binary symbol “1” and the negative of this pulse to represent binary “0”. Derive the formula
for the probability of error incurred by the maximum likelihood detection procedure applied to
this form of signaling over an AWGN channel.

4. Exercise 5.23 from course textbook [4].

5. Exercise 5.24 from course textbook [4].

14
Figure 8: Pulse representing binary “1”.

15
2 Software Implementation
2.1 Observation Vector Construction
Referring to Figure 9, this part of the laboratory experiment will involve the construction of observa-
tion vector X, that will be used for a correlator-based detector. This section employs the orthonormal
functions {fm (t)} specified in Problem 1 of Section 1.6. This bank of correlators operates on the signal
x(t) to produce the observation vector X.

T
X dt x1
0
f1(t)
T
X dt x2
0
x(t) f2(t) Observation vector, X
. . .
. . .
. . .
T
X dt xN
0
fN(t)

Figure 9: A bank of correlators operating on an input signal x(t) to produce the observation vector
X, that will be used for a correlator-based detector.

Although the design presented here is for continuous time signals, we need to employ a discrete time
approximation of 1000 samples per second in order for this to work in MATLAB. Also assume perfect
synchronization between the received signal and the communications system.

Open correlator.m available from the course website. The code has has been partially implemented,
and you will need to finish the rest of the code with the following steps and plot the results.

• Define the four equiprobable signals s1 (t), s2 (t), s3 (t), s4 (t) shown in Figure 7. Since you need
to choose one of the four signals randomly, a good way is to use switch and case in MATLAB.
Plot a randomly generated stream of these waveforms generated in MATLAB.

• Define the orthonormal functions {fm (t)} that you have derived in Problem 1 of Section 1.6.
This step is similar to the previous step, just make sure that the vector for si (t) and fm (t) has
the same length.

• Create an AWGN channel that will be used to add white Gaussian noise to the transmitted
signal consisting of the four waveforms. You can use randn in MATLAB to define the noise.
Plot the time domain representations of the input and output signals for the channel for several
different noise variances. Explain how the noise could potentially impair the successful decoding
of the intercepted signal at the receiver.

16
 • Implement an “integrate-and-dump” block in order to obtain the observation vector X. Since
we are employing discrete time approximation for continuous time signals, integration could be
realized by sum in MATLAB.

• Test your implementation with a transmission of duration 300 seconds, randomly consisting of
one of the four equiprobable signals shown in Figure 7 of duration T=3 seconds each, with zero
mean white Gaussian noise of variance 0.5 added.

• Plot each of the elements of the observation vector X using the stem command in MATLAB.

2.2 Maximum-likelihood Decoder Implementation

Referring to Figure 10, this part of the laboratory experiment involves the implementation of the
signal transmission decoder subsystem in the form of a maximum-likelihood (ML) decoder in order
to accompany the subsystem described in Section 2.1. This decoder will operate on the observation
vector X from Section 2.1 to produce an estimate m̂ of the transmitted symbol mi in a way that
would minimize the average probability of symbol error.

XTs1 +
X Accumulator
-
s1 0.5E1
T
Accumulator X s2 +
X
-
X s2 Select Largest
0.5E2
. . largest ^
m
. .
. .
XTsM +
X Accumulator
-
sM 0.5EM

Figure 10: A maximum-likelihood (ML) decoder system operating on the observation vector X from
Section 2.1 to produce an estimate m̂ of the transmitted symbol mi to minimize the average probability
of symbol error.

Download and open decoder.m from the course website. Although this code is incomplete, it will
serve as a starting point for the rest of this experiment.

• Define the four signal vectors that you have derived in Problem 1 of Section 1.6. Each signal
vector should be a 1 × 3 vector.

• Calculate the energy of each signal vector defined in the previous step. Ei = 3j=1 Eij2 , i =
P
1, 2, 3, 4, where Eij is the component of each signal vector.

• Accumulator and subtraction as shown in Figure 10. Accumulator is realized by sum in MAT-
LAB. Plot the output of the accumulator for each of the branches. Provide your observations
and explain.

17
 • Select the largest branch to produce an estimate m̂. You can use max in MATLAB to select the
largest branch.

• Remove the energy subtraction from each of the branches and repeat your experiment. Do you
get the same results as before? What would be required for the results with and without energy
subtraction to be equivalent?
Now try out the following parts with parameters in this structure:
• In Section 2.1, do not integrate the entire period T , integrate until 0.75T . Plot both the estimate
m̂ and the actual transmitted symbol mi and compare it with the original plot.

• In Section 2.2, do not subtract the energy of si (t) from each branch. Plot both the estimate m̂
and the actual transmitted symbol mi and compare it with the original plot.

• Combine you implementation from Sections 2.1 and 2.2. Change the parameters of AWGN
channel in Section 2.1 to make the signal to noise ratio (SNR) range from 10−5 to 10−1 . Compare
the m̂ in Section 2.2 with the input x in Section 2.1 to get the bit error rate. Plot a BER curve.
Note: In order to get an accurate BER curve, you may need a large number of transmitted
signals, especailly when the SNR value is high.

2.3 Correlator Realization of a Receiver in Simulink

In this part of the laboratory, you will be responsible for implementing a correlator realization of the
receiver structure shown in Figure 2 using the signal waveforms shown in Figure 7. Consequently,
you will be constructing a transmitter that is capable of randomly generating s1 (t), s2 (t), s3 (t), and
s4 (t), devising an additive white Gaussian noise (AWGN) channel, and a four-branch receiver using
correlation in each branch to help determine which symbol was transmitted every T = 1 second.
Assume perfect synchronization in this system.
• Using integrate and dump.mdl, which is also shown in Figure 11, run the experiment for
different pulse durations T (the default is T = 1 second). Note that the integrate-and-dump
block accumulates samples according to the signal possessing the most samples per unit time.
In this case, the Gaussian random number generator is producing 100 samples per 1 second
duration. Explain the end-to-end operation of this basic Simulink model, from the binary PAM
transmitter all the way across to the output of the integrate-and-dump block. Use time domain
plots to help illustrate your explanation.

• Create a new Simulink model document that will contain your design for the four-branch
correlator-based receiver. Implement a Simulink transmitter that can generate each of the
four waveforms shown in Figure 7. Ensure that each waveform is equally likely to occur. Plot
the output of this transmitter to show the randomly generated sequence of waveforms s1 (t),
s2 (t), s3 (t), and s4 (t).

• Now pass this signal stream through the AWGN channel model in order to corrupt the signal
with noise. Vary the amount of noise introduced to the signal stream and observe the shape of
the signal and whether it matches the clean signal stream. How much noise must be added to
make the signal stream unrecognizable?

• To create the receiver, you will need to implement the following steps:

18
 AWGN Integrate-and-Dump
Binary PAM Transmitter
Channel Receiver

Figure 11: A simple Simulink model of a digital transceiver employing an integrate-and-dump block
at the receiver.

– For each branch, you will need to repeat si (t) indefinitely such that it aligns with each
corresponding symbol period of the corrupted intercepted signal r(t). One approach is to
use the repeating sequence block in the basic Simulink sources blockset.
– Multiply the repeating si (t) waveforms in each of the branches with the intercepted signal
r(t). Take the output of this multiplication and input it into the integrate-and-dump block.
What did you set as the number of samples to be integrated?
– Take the output of the integrate-and-dump block and subtract from it the energy of the
corresponding waveform si (t). Since the receiver knows about all the waveforms ahead of
time, simply calculate manually the symbol energy ahead of time and use the constant
source block and the addition block to subtract off the energy from the integrated signal
stream. What do you notice about the relative energy levels for each of the branches?
– Finally, implement the decision making block that selects the maximum value at each time
instant T = 1 seconds. Plot the output of the decision making process, indicating which
of the four waveforms were selected. Why is there a delay in the decoded signal relative to
the originally transmitted signal?

• Calculate the bit error rate of this receiver design for a range of noise variance values. Plot the
BER curve of this model down to 10−3 .

• Suppose you removed the energy subtraction for all the branches. Does the BER performance
of the system change?

19
2.4 Multicarrier Modulation
2.4.1 MATLAB Design of Multicarrier Transmission
Implement Figure 3 according to the following hints. The size of the input/output should be a
variable.

• Implement multiplex and demultiplex blocks using reshape.

• Implement IFFT and FFT blocks using ifft and fft.

• Each subcarrier equalizer should be a singel complex value that is the opposite of the channel
frequency attenuation corresponding to the center of the subcarrier.

• Modulation and demodulation should be implemented on subcarrier based.

Implement the channel filter, given as an input the channel impulse response, and add the noise
generator. Test your implementation with the following parameters:

h1 = [ 1 0 0 0 0 0 0 0 ]
h2 = [ 1 0.1 0.0001 0 0 0 ] (28)
h3 = [ 1 0.3 0.4 0.32 0.2 0.1 0.05 0.06 0.02 0.009 ]

with µ = 0 and σ 2 = 0.001. Hint: Use filter command in MATLAB

Implement the cyclic prefix add and remove blocks of the OFDM system. Make the length of the
cyclic prefix an input variable.

With the system completely implemented, we can now study how it performs in a variety of condi-
tions. The performance measure commonly used in digital transmission systems is the BER. Given
an SNR value and channel conditions, the BER is the fraction of bits that are errors to the total
number of bits transmitted. The only question is “how many bits need to be transmitted in order to
obtain a reliable BER?” The rule-of-thumb answer is that one continues transmitting bits until 100
errors have been reached. [2]

With your implementation, plot the BER curves for the three channel realizations in Equation (28)
between BER values of 10−2 and 10−4. In particular, simulate your system when it uses N = 16,
32, and 64 subcarriers and employs a cyclic prefix of length N/4. The high speed input x [m] is a
uniformly distributed. Plot the BER results.

2.4.2 Simulink Design of Multicarrier Transmission

Open ofdm.mdl available from the course website. The example OFDM transmitter and receiver has
already been implemented and provided as shown in Figure 12. You will need to finish the rest of the
model with the following steps and play around with the parameters.

• Get the AWGN block from Simulink Library Browser. Connect the transmitter and receiver via
AWGN block.

20
 Figure 12: OFDM transmitter and receiver implementation in Simulink.

• Connect Spectrum Scope after the AWGN channel. This allows you to observe how the spectrum
looks like after the OFDM transmitter and channel.

• Connect the transmitter and receiver to the inputs of Error Rate Calculation block. This
allows you to examine how well the receiver performs with different fading characteristics and
generate BER curve for varying SNR values.

• Change the SNR value of the AWGN channel and observe the effect in the frequency domain
via Spectrum Scope. Plot your observation with several different SNR values and conclude how
the specturm change due to the channel.

• Open the OFDM Transmitter and OFDM Receiver subsystems, as shown in Figure 13 and Fig-
ure 14, change the number of subcarriers, and observe the effect in the frequency domain via
Spectrum Scope. Plot your observation with several different number of subcarriers and con-
clude how the specturm change due to subcarrier number.

21
Figure 13: OFDM transmitter implementation in Simulink, where the subcarriers are connected by a
Matrix Concatenator.

Figure 14: OFDM receiver implementation in Simulink, where the cyclic prefix is removed and the
transmitted data is retrieved.

22
3 USRP2 Hardware Implementation
A matched filter is a theoretical framework and not the name of a specific type of filter. It is an
ideal filter which processes a received signal to minimize the effect of noise. Hence, it maximizes the
signal to noise ratio (SNR) of the filtered signal. It happens that an optimum filter does not exit for
each signal shape transmitted and it is a function only of the transmitted pulse shape. Because of its
direct relationship to the transmitted pulse shape, it is called a matched filter. The two most common
matched filters are integrate and dump and root raised cosine. We have already studied integrate and
dump in Section 2.3, so we will focus on root raised cosine in this section.

In this section, we use two models to illustrates a typical setup in which a transmitter uses a square
root raised cosine filter to perform pulse shaping and the corresponding receiver uses a square root
raised cosine filter as a matched filter. The receiver plots an eye diagram from the filtered received
signal. Based on this diagram, we can learn the effect of matched filter in a communication system.

3.1 Eye Diagram

In telecommunication, an eye diagram, also known as an eye pattern, is an oscilloscope display in
which a digital data signal from a receiver is repetitively sampled and applied to the vertical input,
while the data rate is used to trigger the horizontal sweep. It is so called because, for several types
of coding, the pattern looks like a series of eyes between a pair of rails.

Several system performance measures can be derived by analyzing the display. If the signals are too
long, too short, poorly synchronized with the system clock, too high, too low, too noisy, or too slow
to change, or have too much undershoot or overshoot, this can be observed from the eye diagram.
For example, in Figure 15, DA is the range of amplitude differences of the zero crossings and is a
measure of distortion caused by intersymbol interference (ISI), JT is the range of amplitude differences
of the zero crossing and is a measure of the timing jitter, MN is a measure of noise margin, and ST
is measure of sensitivity-to-timing error. In general, the most frequent usage of the eye pattern is for
qualitatively assessing the extent of the ISI. As the eye closes, the ISI increases; as the eye opens, the
ISI decreases.

3.1.1 Discrete-Time Eye Diagram Scope

The Discrete-Time Eye Diagram Scope block displays multiple traces of a modulated signal to pro-
duce an eye diagram. You can use the block to reveal the modulation characteristics of the signal, such
as pulse shaping or channel distortions. This is a most important block that will be used in this section.

An open eye pattern corresponds to minimal signal distortion. Distortion of the signal waveform due
to intersymbol interference and noise appears as closure of the eye pattern.

3.2 Matched Filter Observation

In this section, you can use the DBPSK transmitter model on the transmitter side. On the receiver
side, download and open rcfilter receiver.mdl from course website, as shown in Figure 16. The
top level of this model looks similar to the DBPSK receiver model, however, its enabled subsystem

23
Figure 15: A typical eye pattern. The width of the opening indicates the time over which sampling
for detection might be performed. The optimum sampling time corresponds to the maxmum eye
opening, yielding the greatest protection against noise. If there were no filtering in the system then
the system would look like a box rather than an eye.

contains the information that we are interested in.

Figure 16: A model that illustrates a typical receiver setup, in which the receiver uses a square root
raised cosine filter as a matched filter corresponding to the square root raised cosine filter on the
transmitter.

Go into the matched filter subsystem, as shown in Figure 17, where there is a Raised Cosine
Receive Filter and a Discrete-Time Eye Diagram Scope. Perform the following tasks and plot
your observations:

• Compare Raised Cosine Receive Filter and Raised Cosine Transmit Filter on the trans-
mitter side, see whether they match each other.

24
Figure 17: The subsystem of the receiver model, in which the receiver plots an eye diagram from
the filtered received signal. Based on this diagram, we can learn the effect of matched filter in a
communication system.

• Change the parameters (group delay, rolloff factor and etc.) of Raised Cosine Receive Filter
to make it match the Raised Cosine Transmit Filter, how does the eye diagram look like?
1

• Change the parameters of Raised Cosine Receive Filter to make it mismatch the Raised
Cosine Transmit Filter, how does the eye diagram look like?

• Delete Raised Cosine Receive Filter block, as shown in Figure 18, how does the eye diagram
look like?

When the eye diagram has two widely opened “eyes”, it indicate appropriate instants at which to
sample the filtered signal before demodulating. It also indicates the absence of intersymbol interference
at the sampling instants of the received waveform. A large signal-to-noise ratio in the channel will
produce a low-noise eye diagram. You can construct a Simulink-only model using AWGN Channel
block and then change the SNR parameter in the AWGN Channel block to see how the eyes in the
diagram change.

Figure 18: The subsystem of the receiver model without a Raised Cosine Receive Filter.

1
To get better observation, you need to use ‘Autoscale’ of eye diagram.

25
4 Open-ended Design Problem: Duplex Communication
4.1 Duplex Communication
A duplex communication system is a system composed of two connected parties or devices that can
communicate with one another in both directions. Duplex systems are often employed in many com-
munications networks, either to allow for a communication “two-way street” between two connected
parties or to provide a “reverse path” for the monitoring and remote adjustment of equipment in the
field.

Systems that do not need the duplex capability use instead simplex communication. These include
broadcast systems, where one station transmits and the others just “listen”. Several examples of
communication systems employing simplex communications include television broadcasting and FM
radio transmissions.

Thus far, the models that we have been used in this course all use simplex communication, such as the
DBPSK transmitter and receiver studied in previous laboratory experiments. Since we are using the
XCVR2450 transceiver daughterboard, our USRP2 can be either a transmitter or a receiver. However,
it cannot be a transmitter and a receiver at the same time. Consequently, in this open-ended design
problem, you will need to implement a system that is capable of doing half-duplex communication.

4.2 Half-duplex
A half-duplex (HDX) system provides communication in both directions, but only one direction at
a time (not simultaneously). Typically, once a party begins receiving a signal, it must wait for the
transmitter to stop transmitting, before replying.

An example of a half-duplex system is a two-party system such as a “walkie-talkie” style two-way

radio, wherein one must use “Over” or another previously-designated command to indicate the end
of transmission, and ensure that only one party transmits at a time, because both parties transmit
and receive on the same frequency.

A good analogy for a half-duplex system would be a one-lane road with traffic controllers at each end.
Traffic can flow in both directions, but only one direction at a time, regulated by the traffic controllers.

There are several different ways to control the traffic in a half-duplex communication system, and we
suggest you to use time division duplexing (TDD).

4.3 Time Division Duplexing

Time division duplexing (TDD) refers to a transmission scheme that allows an asymmetric flow for
uplink and downlink transmission which is more suited to data transmission. In a time division duplex
system, a common carrier is shared between the uplink and downlink, the resource being switched in
time. Users are allocated one or more timeslots for uplink and downlink transmission, as shown in
Figure 19.

26
 Station Station
A B
t1
mes
sag
e
t2
t3

ssage
me
t4
t5
mes
sag
e
t6
t7
ge
ssa
me
t8
time time

Figure 19: A half-duplex communication system using time division duplexing. A common carrier is
shared between Station A and Station B, the resource being switched in time.

For radio systems that aren’t moving quickly, an advantage of TDD is that the uplink and downlink
radio paths are likely to be very similar. This means that techniques such as beamforming work well
with TDD systems.

Examples of time division duplexing systems are:

• UMTS 3G supplementary air interfaces TD-CDMA for indoor mobile telecommunications.

• DECT wireless telephony.

• Half-duplex packet mode networks based on carrier sense multiple access, for example 2-wire
or hubbed Ethernet, Wireless local area networks and Bluetooth, can be considered as Time
Division Duplex systems.

• IEEE 802.16 WiMAX.

4.4 Hints
Suppose you have a radio station A and a radio station B, you are required to perform the following
tasks in time sequence: Station A transmits “Hello world” to Station B, Station B transmits “Hello
world” to Station A, Station A transmits “Hello world” to Station B, Station B transmits “Hello
world” to Station A, and etc., , as shown in Figure 19.

• At the beginning, you may want to try something simple and see whether it works. For example,
you can transmit ‘1010......’ instead of “Hello world”, such that frame synchronization is not
necessary. You can test one cycle only, that is Station A transmits “Hello world” to Station B
and Station B transmits “Hello world” to Station A.

27
• In order to make it an automatic communication system, you need to write a MATLAB code
for Station A and another MATLAB code for Station B, such that all you need to do is to click
the “Run” button on two stations at the beginning of simulation.

• A MATLAB function called sim allows you to run your Simulink model from MATLAB code.

• A MATLAB function called pause allows you to specify the interval between two actions in your
MATLAB code. This function is necessary due to the different computation ability of PCs. You
want both of them get ready before continuing to the next step. It is suggested that you have
at least 20 secondes buffer time.

• In order to specify the simulation time of your model, go to Simulation → Configuration

Parameters → Solver tab, change Stop time.

• For example, a very simple piece of code is as follows: sim(’tx’); pause(10); sim(’rx’);
According to this code, the radio station first runs a Simulink model called ‘tx’, then it pauses
for 10 seconds, and in the end it runs a Simulink model called ‘rx’.

28
5 Lab Report Preparation & Submission Instructions
Include all your answers, results, and source code in a lab report formatted as follows:

• Cover page: includes course number, laboratory title, names and student numbers of team,
submission date

• Table of contents

• Pre-lab

• Responses to laboratory questions and explanation of observations

• Responses to open-ended design problem

• Source code

Please include images and outputs wherever possible, as well as insights on your laboratory.

Each group is to submit a single report electronically (in PDF format not exceeding 2MB) to
[email protected] by scheduled due date. Reports that do not meet these specifications will be
returned without review.

References
[1] John A. Gubner. Probability and Random Processes for Electrical and Computer Engineers. Cam-
bridge University Press, 2006.

[2] Michel C. Jeruchim, Philip Balaban, and K. Sam Shanmugan. Simulation of Communication
Systems: Modeling, Methodology and Techniques. Springer, 2nd edition, October 2000.

[3] K.Abend and B.D.Fritchman. Statistical detection for communication channels with intersymbol
interference. In Proceedings of IEEE, volume 58, pages 779–785, May 1970.

[4] Michael Rice. Digital Communications: A Discrete-Time Approach. Pearson/Prentice Hall, Upper
Saddle River, NJ, USA, 2009.

[5] Dennis Silage. Digital Communication Systems using MATLAB and Simulink. Bookstand Pub-
lishing, Gilroy, CA, 2009.

[6] Alexander M. Wyglinski. Physical Layer Loading Algorithms for Indoor Wireless Multicarrier
Systems. PhD thesis, McGill University, Montreal, QC, Canada, November 2004.

29