Digital Communications

Laboratory

Erlangen – January 2018

Ⓧc Institute for Digital Communication, Friedrich-Alexander-Universit¨at Erlangen-
Nu¨rnberg
 Contents
Course Information 1
0.1 General Information.............................................................................................1
0.2 Lab Guidelines.............................................................................................................1

1 3
Digital Transmission of Data

1.1 Introduction, Background, and Motivation..........................................................3

1.2 Purpose.................................................................................................................4
1.3 Lab Environment.................................................................................................4
1.3.1 Transmitter...............................................................................................4
1.3.2 Receiver...........................................................................................................7
1.4 Lab Exercises.............................................................................................................10
1.4.1 Signal Generation at the Transmitter.................................................10
1.4.2 (Coherent) Receivers for Pulse Amplitude Modulation.........................18
1.4.3 Transmission over the AWGN Channel.................................................18

2 28
Implementation of Transmitter and Receiver in MATLAB

2.1 Introduction, Background, and Motivation........................................................29

2.2 Purpose...............................................................................................................30
2.3 Lab Environment................................................................................................30
2.3.1 Oversampling factor...............................................................................31
2.3.2 Transmitter.............................................................................................31
2.3.3 Channel..........................................................................................................32
 2.3.4 Receiver.........................................................................................................33
2.4 Lab Exercises.............................................................................................................34
2.4.1 Transmitter.............................................................................................34
2.4.2 Channel..........................................................................................................40
2.4.3 Receiver.........................................................................................................42
2.4.4 BER calculation......................................................................................46
ii Contents

3 Variants of PAM-Transmission Schemes 58

3.1 Introduction, Background, and Motivation........................................................59

3.2 Purpose...............................................................................................................60
3.3 Lab Environment................................................................................................60
3.4 Lab Exercises.............................................................................................................61
3.4.1 Basic Pulse Shape..................................................................................61
3.4.2 Offset-QAM...................................................................................................66
3.4.3 Gaussian Minimum Shift-Keying.................................................................70
3.4.4 “Carrierless“ Amplitude and Phase Modulation....................................76

4 OFDM 78

4.1 Introduction, Background, and Motivation........................................................79

4.1.1 Orthogonal Frequency-Division Multiplexing.......................................79
4.1.2 Bit Loading.............................................................................................81
4.2 Purpose...............................................................................................................81
4.3 Lab Environment................................................................................................81
4.4 Lab Exercises.............................................................................................................82
4.4.1 OFDM Transmitter.................................................................................82
4.4.2 OFDM Receiver............................................................................................86
4.4.3 Bit Loading.............................................................................................88

5 Signal Space Representation 90

5.1 Introduction........................................................................................................91
5.2 Purpose...............................................................................................................91
5.3 Lab Environment................................................................................................92
5.4 Signal Space Representation..............................................................................92
5.4.1 Orthogonality..........................................................................................93
5.4.2 Orthogonalization..........................................................................................94
5.5 Lab Exercises.............................................................................................................98
5.5.1 Transmission with signal elements.........................................................98
5.5.2 Gram-Schmidt Procedure.....................................................................106
5.5.3 Frequency Shift Keying.......................................................................108

6 Signal Processing in MIMO Systems 112

6.1 Introduction, Background, and Motivation......................................................113

6.2 Lab Environment..............................................................................................115
6.3 Lab Exercises...........................................................................................................116
6.3.1 System Model..............................................................................................116
Contents iii

6.3.2 SISO.............................................................................................................118
6.3.3 SIMO...........................................................................................................120
6.3.4 MIMO..........................................................................................................126
0.1 General Information

Instructor: Ebrahim Amiri

Class time: room R4.10

Dates: Lab #1: —

Lab #2: —

Lab #3: —

Lab #4: —

Lab #5: —

Lab #6: —

Prerequisites: Digital Communications (DiCo in WS)

Credits 2.5 ECTS credits (= work load of 75 hours time)

Reference: Johannes Huber, Digital Communications, lecture

notes

Lab work: 6 lab exercises

Homework: 6 homework sets due in class before lab work

Grading: on passed/not passed basis

0.2 Lab Guidelines

In the Lab
There are six labs in this course. Students are required to complete each lab during
their assigned time slot (length: approx. 5 hours) on-site. In the lab students work in
2 Contents

pairs. At least one teaching assistant will be present to help the students and to check
their work.

The work in the lab is to solve the problems marked by an L in this manual. An example
lab exercise is shown in L-0.1.

Lab Exercise L-0.1

In the first lab solve all the lab exercises marked by L-1.X (with X∈ N).

Homework

Students are required to come prepared with the pre-lab reading work. In addition, each
of the seven labs has a set of homework problems marked by an H. For example in H-0.1
your first homework is given.

Homework H-0.1

Carefully read the instructions for Lab 1 and solve all the homework problems
marked by H-1.X!

Collaboration on solving the homework problems is encouraged. Homework assignments

(one per pair) should be turned in to the teaching assistant in charge no later than the
beginning of the assigned time slot. Your assignments should be written neatly, should
contain clear and complete solutions, and should be stapled if consisting of multiple
sheets.

Grading policy

Grading is done on a passed/not passed basis. In order to pass, at least fair results in all
homework sets and lab exercises have to be achieved.
 Project 1
Digital Transmission of
Data

1.1 Introduction, Background, and Motivation

In this first experiment you will study the basics of digital data transmission. Today,
digital transmission techniques are widely used in many common applications, e.g.
mobile telephony, digital video broadcasting etc. Some of the advantages of digital
transmission techniques over analog schemes like amplitude modulation and frequency
modulation are a much better power efficiency and a tremendous flexibility. The focus of
this experiment is on the most popular technique for digital data transmission, so-called
pulse amplitude modulation (PAM).

The experiment comprises two main parts: first, the structure of a generic digital PAM
transmitter is analyzed. Second, the receiver of such a system is studied. The latter
requires to establish an entire digital transmission system consisting of transmitter, chan-
nel, and receiver. Detailed descriptions and discussions of the optimal design of digi-
tal PAM transmission systems are part of any textbook on digital communications, e.g.
[Pro00, Hay00]. Here, in the context of this lab course, we use the system model and
the notation introduced in the lecture notes to “Digital Communications” by Prof. Huber
[Hub11a].
4 1. Digital Transmission of Data

1.2 Purpose

The aim of this experiment is to demonstrate the effectiveness and efficiency of digital
transmission schemes. You will learn how to design a transmitter for digital pulse am-
plitude transmission and how to represent the digital data by complex coefficients. You
should understand the design rules and constraints for the basic pulse shape in the time
domain and the frequency domain and know how to develop the optimal receiver for
digital transmission systems. You will get to know the effects of some transmission
disturbances on the received signal.

1.3 Lab Environment

In this experiment you work with a MATLAB-based simulation environment which can
model several digital transmission techniques. The lab computers are equipped with
powerful analog-to-digital converters (ADC) and digital-to-analog converters (DAC). In
the first part of the experiment, just the DAC are used to generate physical signals from
the software-based digital transmission model. That is, the digital transmission schemes
selected and simulated in the software are analyzed based on “real” voltages and currents.

In the second part, two of the lab computers are combined into a complete digital trans-
mission system comprising transmitter, channel, and receiver. One of the computers acts
as the transmitter, i.e., the MATLAB-generated signals are D/A converted and transmit-
ted to the receiver using patch cables (coaxial cables with BNC connectors). The other
computer acts as the receiver; the physical signal is A/D converted and then again fed
into a MATLAB-based receiver structure. The (AWGN) channel is modeled by an electronic
noise generator.

The physical signals can be analyzed using the Tektronix oscilloscopes or the Agilent
multimeters.

1.3.1 Transmitter

In order to simulate the transmitter of a digital transmission system the respective

MAT- LAB program has to be started. This done by starting MATLAB and then typing

experiment1_transmitter
1.3. Lab 5

Figure 1.1: MATLAB GUI of PAM transmitter.

q[l] am[k] s(t) RF sRF(t)

Source M g(t) ↑
Baseband

Figure 1.2: Block diagram of transmitter structure of a digital transmission system.

into the command window. Then, you should see a MATLAB GUI as depicted in Fig. 1.1.

The transmitter is built from several individual blocks as depicted in Fig. 1.2. These
blocks are subsequently described in some more detail.

1.3.1.1 Data Source

The data source emits a pseudo-random sequence of binary symbols q[l] ∈ {0, 1}. Zeros
and ones occur with equal probabilities.

1.3.1.2 Binary Labeling (Bit Mapping)

As we focus on digital PAM, the binary labeling in block M assigns binary m-tuples to
(complex) amplitude coefficients am[k].
6 1. Digital Transmission of Data

The system supports several different sets of amplitude coefficients, i.e., signal constella-
tions. In Table 1.1 a list of the implemented signal constellations and constellation sizes
is given.

Table 1.1: Supported signal constellations

Size of Constellation
Type 2 4 8 16 32 64
ASK (unipolar) × × × × × ×
ASK (bipolar) × × × × × ×
× × × × × ×
PSK
× × ×
QAM ×
STAR
×
CROSS

1.3.1.3 Pulse Shaping

The transition from the discrete-time domain signal to the (usually still complex-valued)
continuous-time domain signal (complex baseband) in a digital transmission system is
performed by the pulse shaping. In PAM transmission the basic pulse shape g(t) is
simply weighted with the (complex) amplitude coefficients am[k]. The (spectral) shape of
the pulse significantly influences the spectrum of the RF signal; the waveform of the
pulse shape in time-domain domain is crucial for a transmission not suffering from
inherent interference.

In the simulation environment used in this experiment, the basic pulse shape is still
located in the digital domain. That is, the continuous-time signal is simulated by an over-
sampled discrete signal.

The equivalent complex baseband signal can be measured at the output of the D/A
converters if the respective check box is set. The inphase component is at channel A, the
quadrature component at channel B.
 1.3. Lab 7

1.3.1.4 RF Modulation

The radio-frequency signal is generated from the baseband signal by multiplying with
sinusoidal waveforms with the carrier frequency fc. In the simulation environment this
step is also implemented in MATLAB, i.e., even the modulation into a radio-frequency
signal is software-based. The physical signal which can be then analyzed at channel A of
the D/A converter is the result of a simple D/A conversion of an entirely digitally
modeled block diagram. The second output at channel B is a synchronization signal
required for the combination of two computers into an entire transmission system.

1.3.2 Receiver

The receiver can be started by typing

experiment1_receiver
into the MATLAB command window. Please make sure, that you do not run the trans-
mitter and the receiver in parallel using two instances of MATLAB. The MATLAB GUI
for the receiver is shown in Fig. 1.3.
The incoming signal at the computer acting as receiver is first A/D converted and then
processed in an entirely MATLAB-based receiver structure which is sketched in Fig. 1.4.
The individual parts of this block diagram are briefly explained below.

1.3.2.1 RF/Baseband

First, the incoming noisy radio-frequency signal eRF(t) is demodulated, i.e., a baseband
signal e(t) is generated by multiplying with the carrier and low-pass filtering.

1.3.2.2 Filtering

In the baseband, the next step at the receiver is a (matched) filter to reduce the noise
and optimize the SNR for the sampling. The filters impulse response can be chosen from
a given selection of pulse shapes.

1.3.2.3 Sampling

After the input filter, the pseudo-continuous signal a(t) is sampled in order to obtain a
stream of discrete symbols d[k] on the symbol raster of Ts.
8 1. Digital Transmission of Data

Figure 1.3: MATLAB GUI of PAM receiver.

kTs
eRF(t) RF e(t) a(t)
Matched Filter d[k] aˆ[k]
Decision Device qˆ[l]
−1
↓
Baseband M

Figure 1.4: Block diagram of receiver structure of a digital transmission system.

1.3.2.4 Decision Device

The time-discrete samples d[k] are then fed into the decision device to obtain estimates
aˆ[k] on the initial coefficients am[k] .
Bibliography 9

1.3.2.5 Demapping

The last step is to retrieve the binary data stream from the estimates aˆ[k]. M−1 is
the inverse operation to M and yields the estimated binary data symbols qˆ[l].

1.3.2.6 Disturbed Received Signals

At the receiver, we can model the effects of several problems which are likely to occur
on the channel. The impact of frequency offsets, phase offset, and mismatched sampling
instances can be tested on the received signal.

1.3.2.7 Error Counters

In order to assess the entire transmission scheme consisting of transmitter, channel, and
receiver, the bit error introduced on the channel or by the effects mentioned above can
be counted. The GUI offers an overall error counter and moving average values, resulting
from window sizes of 1 s and 10 s. For reliable BER diagrams, the latter is preferably
used.

Readings for Lab 1

[Hay00] Simon Haykin, Communication systems, 4th ed., John Wiley & Sons, New
York, NY, USA, 2000.

[Hub11a] Johannes B. Huber, Digital Communications, Lecture Notes, Erlangen, October

2011.

[Hub11b] , Nachrichtentechnische Systeme, Vorlesungsskript, Erlangen, Oktober

2011.

[Pro00] John G. Proakis, Digital communications., 4th ed., McGraw-Hill, New York,
NY, USA, 2000.
1 1. Digital Transmission of

1.4 Lab Exercises

1.4.1 Signal Generation at the Transmitter

First, we concentrate on the generation of the pulse-amplitude-modulated signal at the

transmitter side. This is done in the equivalent complex baseband, i.e., the radio-
frequency signal is not considered in these first steps.

In order to illustrate and analyze the properties of the signal constellations, a “rectangular
pulse shape” has to be selected in the respective drop-down menu.

1.4.1.1 Signal Constellations (at the Transmitter)

Homework H-1.1

Calculate (by hand and analytically) the peak-to-average power ratio (PAPR) of
the equivalent complex baseband signals for the following signal constellations: 2-
ASK (unipolar/bipolar), 8-ASK (unipolar/bipolar), 16-QAM, and 8-PSK.

How are the PAPR and the crest factor related?

Note: The peak-to-average power ratio is defined as

maxm{Pm}
PAPR = E{Pm} ,

Pm (m = 1, . . . , M ) being the power of the m-th signal point.

Hint: Use the graphical representation of the signal constellations in the ECB to
determine the PAPR. The squared distance of a signal point from the origin is
proportional to its power Pm.

Lab Exercise L-1.1

Display the signal constellations from Tab. 1.1 on the oscilloscope (X-Y mode).
1.4. Lab 1
1 1. Digital Transmission of

Lab Exercise L-1.2

Verify your analytical results from H-1.1, i.e., measure the PAPR of the physical
signals of the signal constellations listed in H-1.1.

Note, you can measure voltages (peak, amplitudes, etc.) using the oscilloscope
and RMS voltages using the Agilent multimeter. You cannot directly measure
the power. In addition you can just measure these quantities in one of the two
quadrature components.

Lab Exercise L-1.3

Compare the results for unipolar and bipolar ASK. How do you rate the two tech-
niques in terms of the PAPR? Give some advantages for both variants of ASK.
1.4. Lab 1
1 1. Digital Transmission of

1.4.1.2 Pulse Shaping

The properties of the transmit signal are–apart from the signal constellation–mainly de-
termined by the employed basic pulse shape g(t). In the following, you will study several
basic pulse shapes with respect to their suitability to be used in digital transmission
scheme. In particular, the spectra of the pulses shall be analyzed (baseband transmis-
sion).

Lab Exercise L-1.4

Study the time-domain inphase component of the equivalent complex baseband sig-
nal. First, use a unipolar 2-ASK signal constellation. Compare the signals resulting
from a rectangular pulse shape and a cosine-roll-off pulse shape, respectively.

Repeat these steps for a bipolar 2-ASK, a 4-QAM, a bipolar 4-ASK, and a 16-QAM
signal constellation.

Lab Exercise L-1.5

Compare the spectra of the ECB signals of a bipolar 2-ASK using a rectangular
pulse shape and a cosine-roll-off pulse shape, respectively.

Homework H-1.2

How are the spectra of the radio-frequency and the baseband signal related?

Lab Exercise L-1.6

Verify your results from H-1.2.

1.4. Lab 1
1 1. Digital Transmission of

1.4.1.3 The Radio-Frequency Signal

After the weighting of the basic pulse shape g(t) with the channel coefficient am[k], the
baseband signal has to be transformed into a radio frequency signal centered around the
carrier frequency fc.

Lab Exercise L-1.7

Display and study the radio frequency signal of the following signal constellations
in both, time domain and frequency domain: 2-ASK (unipolar/bipolar), 4-QAM,
and 16-QAM.

First use a rectangular pulse shape and then a cosine-roll-off pulse shape.

Lab Exercise L-1.8

How does the size of the signal constellation M affect the spectra of the radio
signals (Note, the data rate 1/Tb is assumed to be constant)?
1.4. Lab 1
 1 1. Digital Transmission of

1.4.2 (Coherent) Receivers for Pulse Amplitude Modulation

In the following, we study the receiver used in PAM transmission. In particular, the filter
at the receiver input is analyzed.

Homework H-1.3

How is the optimal filter at the receiver called? What is the optimization criterion
in its derivation?
The combination of the basic pulse shape at the transmitter and the receive filter
has to fulfill the so-called Nyquist criterion. Give the definitions of this criterion is
both, time and frequency domain and sketch them for an exemplary pulse shape.

1.4.3 Transmission over the AWGN Channel

For the remaining experiments two of the lab systems have to be combined. One system
acts as the transmitter, the other one acts as the receiver in a transmission scenario. An
illustration for the practical implementation of this combination is given in Fig. 1.5.

Transmitter Noise Receiver Oscilloscope

Generator

A/D D/A A/D D/A

AB ABAB AB

AWGN Channel

Figure 1.5: Transmission system using two of the lab computers.

1.4.3.1 Signal Constellations at the Receiver

In the following section, we finally want to study the signal constellations at the receiver
side. This goal needs some preliminary steps which are described in the following.
1.4. Lab 1
2 1. Digital Transmission of

√
Select a 4-QAM signal constellation and the cos pulse with a roll-off factor of 0.94 at
the transmitter. Study the inphase and the quadrature component of the demodulated
(from RF) time-domain signal at the receiver.

Lab Exercise L-1.9

Describe the effect you can observe regarding the amplitude and the phase of the
signal. How can you compensate for these effects?

Lab Exercise L-1.10

Activate the appropriate matched filter and study the resulting eye patterns.
Iden- tify the best time instance for sampling.

Lab Exercise L-1.11

Now activate the sample and hold check-box and describe the resulting signal. What
happens, if you vary the sampling time instance?
1.4. Lab 2
2 1. Digital Transmission of

Lab Exercise L-1.12

Use the oscilloscope (in X-Y mode) to display the phasors of the sampled received
signal for 2-ASK (unipolar/bipolar), 4-QAM, 16-QAM, 4-PSK, and 8-PSK.

Lab Exercise L-1.13

How does the additive white Gaussian noise affect the signal constellations? Use a
4-QAM constellation and varying noise powers to demonstrate the effect.

Lab Exercise L-1.14

How do frequency offsets and phase offsets affect the signal constellation? Use a
unipolar 2-ASK and a 4-PSK signal constellation and vary the respective
parameters in the receiver GUI.

Lab Exercise L-1.15

How does a mismatched sampling affect the signal constellation at the receiver?
Use again a 4-QAM signal constellation to study the effects.
1.4. Lab 2
2 1. Digital Transmission of

1.4.3.2 Measuring the Bit Error Ratio

Lab Exercise L-1.16

Determine the bit error ratio over Eb/N0 (range: 0 dB to 15 dB in steps of 2 dB)
for the following modulations: 2-ASK unipolar, 2-ASK bipolar, 4-PSK, 8-PSK, 16-
QAM. Use the diagram on page 26 to illustrate your results in a single plot.

Lab Exercise L-1.17

Compare the modulation schemes given in L-1.16 in terms of spectral efficiency

and power efficiency for a bit error ratio of BER = 10−4). Use the diagram on
page 27 for an illustration.
1.4. Lab 2
 2
0
10

−1
10

−2
10
Figure 1.6: Diagram for L-

Bitfehlerra

−3
10

−4
10

1
Digital Transmission of
−5
10

−6
10
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
10*log (E /N ) [dB]
10 b 0
 1.
5

Lab
4.5

3.5
Figure 1.7: Diagram for L-

3
spektrale Effizienz

2.5

1.5

0.5

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
10*log (E /N ) [dB]
10 b 0

2
2 2. Implementation of Transmitter and Receiver in
 Project 2
Implementation of
Transmitter and Receiver
in MATLAB

2.1 Introduction, Background, and Motivation

In lab 1 a whole PAM system with different signal constellations has been analysed. The
effects of some transmission disturbances on the received signal have been measured
using oscilloscopes and multimeters and should be understood at this point. Furthermore,
the advantages and disadvantages of each constallation should have become clear.

In this experiment, a complete digital communication system shall be implemented,

where the whole system shall work on one computer using MATLAB without any
additional hardware components. The script data ’./lab2/simulation.m’ will be the
framework of this simulation, where all the following MATLAB functions will be called
from. The missing MATLAB functions for transmitter, channel and receiver shall be
implemented step by step by the student. To evaluate the performance of the simulation
setup, in addition, a function calculating the bit error rate (BER) shall be
implemented. At the end of this lab, the BER-Eb/N0-curves if different transmission
settings shall be visualized and compared to the well-known curves from the lectures.
3 2. Implementation of Transmitter and Receiver in

2.2 Purpose
The aim of this experiment is to get more into detail of a digital communication system.
By building the code of the different steps of transmitter, channel and receiver, the
student will get a better insight into their functionalities.

2.3 Lab Environment

%% Simulation Parameters
%PAM_type = '4QAM';
PAM_type = '8ASKbipolar';

GrayMappingOn = 0; % 0 = off, 1 = on
EbN0_dB = 10;
numberOfBits = 12e4;
f_b = 1e3; % symbol/baud rate
oversamplingFactor = 4; % should be at least 4
f_c = 5e3; % carrier frequency; should be ca. 5 times f_s
f_s = f_c*oversamplingFactor;

%% Creation of a random bit stream

traBits = round(rand(1,numberOfBits));

%% Transmitter
traSignal = transmitter(bitBits, PAM_type, ...
GrayMappingOn, f_b, f_c, f_s);

%% Channel
recSignal = channel(traSignal, EbN0_dB, M, f_s, f_b);

%% Receiver
recBits = receiver(recSignal, PAM_type, ...
GrayMappingOn, f_b, f_c, f_s);

%% Calculate BER
BER = calculateBER(traBits, recBits)
2.3. Lab 3

q[l] am[k] s(t) RF sRF(t)

Source M g(t) ↑
Baseband

Figure 2.1: transmitter structure of a digital transmission system.

The code of the simulation script (’./lab2/simulation.m’) is listed in this section. The
code starts with setting the parameters. Especially the parameter oversamplingFactor
is important. It will be described in subsection 2.3.1. Then, a random bit vector is created
with the bit length set before. Subsequently the function calls of transmitter, channel
and receiver are listed. The last part calculates the BER.

2.3.1 Oversampling factor

Transmitting data on a physical channel means to make use of D/A and A/D converter.
So, the signals are not only time discrete but also time continuous. MATLAB works on
vectors and matrices and cannot handle time continous signals. Even so, simulating a
time continous signal is also possible in MATLAB: The continuous signal is simply
approximated by working with an oversampled signal. By default, the parameter
oversamplingFactor is set to 4.

Remark: In the following, the time continous signal is described as a function of the time
continous value t, i.e. sRF(t). In the context of MATLAB the signal should be a function
of a time discrete value, i.e. sRF[t]. Using a oversamplingFactor of 4 means, that the
system sampling rate f_s is 4· f_c. Thus, the values sRF [t] can be seen as sampled values
at sampling rate f_s.

2.3.2 Transmitter

The block diagram of the transmitter is the same as in experiment 1, see Fig. 2.1. Please
have a closer look at the sections describing the transmitter units in experiment 1. In the
following, Bit Mapping using Gray Mapping will be described in detail.

The job of Bit Mapping (block M in Fig. 2.1) is to map a M binary m-tuple to (complex)
amplitude coefficients am[k]. This can be done in a natural way or in a more sophisticated
3 2. Implementation of Transmitter and Receiver in

Gray 000 001 011 010 110 111 101 100

Natural 000 001 010 011 100 101 110 111

-7 -5 -3 -1 +1 +3 +5 +7

Figure 2.2: Constellation diagram for bipolar 8-ASK

0011 0111 1011 1111 0100 0101 1101 1100

+3j +3j

0000 0100 1000 1100 0000 0001 1001 1000

-3j -3j
(a) Natural Mapping (b) Gray Code

Figure 2.3: Constellation diagram for 16-QAM

way: the Gray Code. The resulting constellation diagrams are shown in Fig. 2.2 for
(bipolar) 8-ASK and in Fig. 2.3 for 16-QAM.
If a symbol was detected at receiver-side which was not the transmitted one, the proba-
bility of detecting the neighbor of the transmitted symbol is much higher than detecting
the others. By using the Natural Mapping neighboring signal points might differ in more
than one bit value. This is not the case by using the Gray Code. There, detecting the
direct neighbor yields to only one bit error in the whole m tuple.

2.3.3 Channel

The channel is kept simple in this experiment. The signal is just degraded by additive
white Gaussian noise (AWGN) which has a constant power spectral density with the two-
sided power spetral density denoted as N0/2. So, the block diagram is also simple, see
Fig. 2.4. Adding the noise can be described with:

eRF(t) = sRF(t) + n(t) (2.1)

2.3. Lab 3

n(t)

sRF (t) eRF (t)

Figure 2.4: AWGN channel

kTs
eRF(t) RF e(t) a(t)
Matched Filter d[k] aˆ[k]
Decision Device qˆ[l]
−1
↓
Baseband M

Figure 2.5: receiver structure of a digital transmission system.

2.3.4 Receiver

In a consistent way, the block diagram of the receiver is also the same as in experiment
1, see Fig. 2.5. Please have a closer look at the sections describing the receiver units in
experiment 1. In this experiment, no Matched Filter is used, since a rectangular has been
used to perform pulse shaping.
3 2. Implementation of Transmitter and Receiver in

2.4 Lab Exercises

2.4.1 Transmitter

Lab Exercise L-2.1

Copy the file ’./lab2/simulation.m’ into a subfolder ’working’.

Lab Exercise L-2.2

Create the MATLAB file transmitter.m in the working folder with the following head:

function traSignal = transmitter(bitvector, ...

This function will be filled in the following step by step.
PAM_type, f_b, f_c, f_s)

2.4.1.1 Mapping

Homework H-2.1

The bit stream 10 11 00 01 00 11 is given. Map these bits onto complex PAM con-
stellation symbols using bipolar 8-ASK and 16-QAM, first with Natural Mapping,
then using Gray Code (see Fig. 2.2 and 2.3). Write the results in following table.

Natural Mapping Gray Code

bipolar
8-ASK
16-QAM

Lab Exercise L-2.3

Create the MATLAB file mapBitsToSymbol.m in the working folder with the
follow- ing head:
2.4. Lab 3
3 2. Implementation of Transmitter and Receiver in

function PAM_symbols = mapBitsToSymbols(bitvector, PAM_type)

Write the code which maps the bits inside the vector bitvector onto complex symbols
according to input string PAM_type using Natural Mapping. Possible values for
the input variable PAM_type can be ’8ASKbipolar’ and ’4QAM’. Add the
function call into the MATLAB file transmitter.m.

Lab Exercise L-2.4

Create the MATLAB file mapBitsToSymbol_Gray.m in the working folder with the
following head:

function PAM_symbols = mapBitsToSymbols_Gray(bitvector,

Write the code equivalent to the function mapBitsToSymbol.m. It shall map the
bits onto complex symbols, but with Gray Code. Add the call of this function to
the function transmitter. During the later simulation run, take care, that one of
these two mapping functions is commented out.

2.4.1.2 Pulse shaping

Lab Exercise L-2.5

The MATLAB file pulseShape.m is already in the working folder and ready to use.
Add the following function call into the MATLAB file transmitter.m.

ecb_signal = pulseShape(PAM_symbols, f_b, f_s)

2.4.1.3 Modulation

Homework H-2.2

Write down the block diagram of the modulation of the complex PAM symbols to
get the high frequency signal with carrier frequency f_c (last block in Fig. 2.1). A
2.4. Lab 3
3 2. Implementation of Transmitter and Receiver in

simple filter with an rectangular impulse answer is used for pulse shaping. Complete
the following equation:

sRF(t) =

Homework H-2.3

Given the following ECB signal:


+3 − 5i for 0 ≤ t < T

−1 − 1i for T ≤ t < 2T
s(t) = 
+7 + 3i for 2T ≤ t < 3T

0 else

Calculate the HF signal after modulation for 0 ≤ t < 3T

Lab Exercise L-2.6

Create the MATLAB file modulate.m in the working folder with the following head:

function hf_signal = modulate(ecb_signal, f_c, f_s)

Write the code which takes the ecb signal and modulates it onto the carrier frequency
f_c using sine and cosine. Add the function call into the MATLAB file transmitter.m.
2.4. Lab 3
4 2. Implementation of Transmitter and Receiver in

2.4.2 Channel

Homework H-2.4

Given a signal vector x in MATLAB, write a pseudo code calculating the mean power
of this signal.

Homework H-2.5

Given the calculated mean power S¯ of the transmit signal sRF (t), the symbol
rate f_c and the modulation constellation size M . How can the energy per
information bit Eb be calculated?

Homework H-2.6

A MATLAB Gaussian noise vector shall be added to the analogue transmit signal
given these the two-sided power spectral density N0/2 and the system sampling rate
f_s. How is the mean power of this noise vector calculated given these parameters

Lab Exercise L-2.7

Create the MATLAB file channel.m in the working folder with the following head

function recSignal = channel(traSignal, EbN0_dB, M, f_s, f_b)

Write the code which adds the Gaussian noise given the parameters of the function
head.
2.4. Lab 4
4 2. Implementation of Transmitter and Receiver in

2.4.3 Receiver

Lab Exercise L-2.8

Create the MATLAB file receiver.m in the working folder with the following head:

function recBits = receiver(recSignal, PAM_type, ...

This function will be filled step by step. It is called in the MATLAB script simula-
f_b, f_c, f_s)
tion.m.

2.4.3.1 Demodulation

Homework H-2.7

Write down the block diagram of the demodulation of the received signal to get the
complex baseband signal e(t) (first block in Fig. 2.5).

Lab Exercise L-2.9

Create the MATLAB file demodulate.m in the working folder with the following head:

function demodSignal = demodulate(recSignal, f_c, f_s)

Write the code which takes the hf signal and demodulates it into ecb domain. Take
care that both signal parts (real and imaginary part) are taken into account. Add
the call of this function to the function receiver.

2.4.3.2 Matched Filter / Downsampling

Lab Exercise L-2.10

The MATLAB files MatchedFilter.m and Downsample.m are already in the

working folder and ready to use. So, there is no need to take care of correct
synchronisation. Add the following function call into the MATLAB file receiver.m.
2.4. Lab 4
4 2. Implementation of Transmitter and Receiver in

filtered_signal = MatchedFilter(ecb_signal, f_b, f_s);

PAM_symbols = Downsample(filtered_signal, f_s/f_b);

2.4.3.3 Decision Device / Demapping

Homework H-2.8

The received PAM symbols are degraded by noise. What are the decision boundaries
for 16-QAM in the AWGN case? Fill them into Fig. 2.3 using dotted lines.

In the following, the transmitted bit stream is 00 01 11 10 10 11. Consequently, the

transmitted 16-QAM symbol sequence using Natural Mapping is {−3 − 1j; 3 + 1j; 1 + 3j}.
The transmitted 16-QAM symbol sequence with Gray Code is {−1 − 3j; 3 + 1j; 1 − 1j}.
At the receiver-side, the received 16-QAM symbol sequence is degraded by additive
white Gaussian noise. The complex noise part can be described by the sequence {+1.5-
0.7j; 0.2-1.1j; -1.1+0.5j} which is added to the transmitted symbols.

Homework H-2.9

Calculate the estimated 16-QAM symbols aˆ[k] for both received signal sequences!

Homework H-2.10

Demap both sequences to bit streams using the decision boundaries defined in the
homework and calculate the bit error rate! What is better in this case: Natural
Mapping or Gray Code? Why is it better?

Lab Exercise L-2.11

Create the Matlab file mapSymbolsToBits.m in the working folder with the following
head:

function bitvector = mapSymbolsToBits(PAM_symbols, PAM_type)

2.4. Lab 4
4 2. Implementation of Transmitter and Receiver in

Write the code which first maps the estimated PAM points aˆ[k] to the nearest
possibly transmitted point and second map these values to bits again according
to input string PAM_type and with Natural Mapping. Possible values for the
PAM_type can be ’4QAM’ and ’8ASKbipolar’. Add the call of this function to the
function receiver.

Lab Exercise L-2.12

Create the Matlab file mapSymbolsToBits_Gray.m in the working folder with the
following head:

function bitvector = mapSymbolsToBits_Gray(PAM_symbols,

Write the code equivalent to the demapping without Gray Code in the function
receiver. It shall map the received PAM points to the nearest possibly transmitted
point and map these values to bits again according to input string PAM_type. Add
the call of this function to the function receiver.

2.4.4 BER calculation

Homework H-2.11

Having the transmitted bit stream and the received bit stream. How can the bit
error rate be calculated?

Homework H-2.12

How many bits have to be simulated to calculate reliable value for a bit error rate
of less than 10−6?

Lab Exercise L-2.13

Create the Matlab file calculateBER.m in the working folder with the following
head:
2.4. Lab 4
4 2. Implementation of Transmitter and Receiver in

function BER = calculateBER(traBits, recBits)

Write the code which calculates the bit error of the communication system.

Lab Exercise L-2.14

Run the simulation script with different parameters such as number of bits, PAM
type, etc. and check if everything works as you expect.

Up to now, the simulation script is ready to run one simulation. However, to compare
different settings, it is common to simulate with different Eb/N0 values and save the
BER in a vector. Then, BER / Eb/N0 curves can be plotted for the selected settings.
Therefore, the simulation script has to be adapted to the form on the following page.
2.4. Lab 4
5 2. Implementation of Transmitter and Receiver in

%% LOOP PARAMETERS
EbN0_dB_Vector = 0:1:15;
BER_Vector = zeros(1,length(EbN0_dB_Vector));

%% Simulation Parameters
%PAM_type = '4QAM';
PAM_type = '8ASKbipolar';

GrayMappingOn = 0; % 0 = off, 1 = on
%EbN0_dB = 10; % this is now commented out
numberOfBits = 12e4;
f_b = 1e3; % symbol/baud rate
oversamplingFactor = 4; % should be at least 4
f_c = 5e3; % carrier frequency; should be ca. 5 times f_s
f_s = f_c*oversamplingFactor;

%% Creation of a random bit stream

traBits = round(rand(1,numberOfBits));

%% LOOP START
curLoop = 0;
for EbN0_dB = EbN0_dB_Vector
curLoop = curLoop + 1;

%% Transmitter
traSignal = transmitter(traBits, PAM_type, ...
GrayMappingOn, f_b, f_c, f_s);

%% Channel
recSignal = channel(traSignal, EbN0_dB, M, f_s, f_b);

%% Receiver
recBits = receiver(recSignal, PAM_type, ...
GrayMappingOn, f_b, f_c, f_s);

%% Calculate BER
BER = calculateBER(traBits, recBits)

%% LOOP END
BER_Vector(curLoop) = BER;
end
2.4. Lab 5
5 2. Implementation of Transmitter and Receiver in

Lab Exercise L-2.15

Add the simulation loop lines in your simulation script and run the simulation for
the two PAM types and the configuration with and without Gray mapping. Do not
forget to change the name the specific BER vector at the end of the code for the
plots.

Lab Exercise L-2.16

Plot the four BER curves in Matlab using the functions ’figure’, ’semilogy(x,y)’ and
’hold on’. You can label the figure with the function xlabel, ylabel, legend, title.
Discuss the curves in your group and then with your supervisor.

Lab Exercise L-2.17

Fill the values into the diagram on page 56 and compare your curves with the curves
from theory on page 52.

Figure 2.6: Bit error rates (BER) of different PAM schemes.

2.4. Lab 5
5 2. Implementation of Transmitter and Receiver in

.
2.4. Lab 5
 5
0
10

−1
10

−2

2. Implementation of Transmitter and Receiver in

10
Bitfehlerr

−3
10

−4
10

−5
10

−6
10
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
10*log (E /N ) [dB]
10 b 0
2.4. Lab 5
5 3. Variants of PAM-Transmission
 Project 3
Variants of
PAM-Transmission
Schemes

3.1 Introduction, Background, and Motivation

The third experiment will be about variants of PAM transmission. You will encounter
different well-known and widely applied schemes of PAM-transmission and explore and
compare their properties.

In the first experiment you already studied the building blocks of a standard PAM-
transmission scheme and their function and properties. You further studied the building
blocks in the second experiment, when you implemented them yourselves. In this exper-
iment you will find the same building blocks, but of course some modifications will be
done to reach certain goals. These goals may be to optimize spectral efficency, reduce
complexity, or other desirable properties. You will design your own basic pulse shapes
and investigate their impact on the transmission system. You will also learn about some
well-known variants of PAM transmission schemes and their connection and motivations.
6 3. Variants of PAM-Transmission

3.2 Purpose

This experiment is supposed to provide an overview of variants of PAM and show the
plurality of it. Furthermore the connections between the variants shall be explored.

3.3 Lab Environment

The lab environment is the same as in experiment 1.

3.4. Lab 6

3.4 Lab Exercises

Figure 3.1 depicts a standard PAM transmission scheme, which is the foundation of all
variants found in this experiment. One fundamental building block is the basic pulse
shape and its influence is oftentimes underestimated.

Figure 3.1: Standard PAM transmission scheme

3.4.1 Basic Pulse Shape

To understand the influence of the basic pulse shape we want to examine the impact of
different basic pulse shapes on the standard PAM system. In experiment 2 the basic pulse
shape was always set to be rectangular.

Homework H-3.1

What property do basic pulse shapes usually possess if intersymbol interference-free

transmission is desired? State the name and explain what this property means.

Homework H-3.2

Sketch the following basic pulse shapes in time domain: Rectangular, cosine with
roll-off slope and one pulse of your choice, which does not fulfill the property from
homework 1. How does the cosine pulse change for different roll-off slope factors
α. For the pulse shape of your choice also provide a formula.
6 3. Variants of PAM-Transmission

Lab Exercise L-3.1

In the MATLAB -file “Eigener Impuls” you can define your own impulse. “T” stands
for the symbol interval and “t” for the time variable. You can also use variable
“Alpha” if you want to define a roll-off slope. Please make sure your impulse is
defined in the intervall from “0” to “T”. Compare the impulse shape the transmitter
shows to your sketch you prepared.

Lab Exercise L-3.2

Compare the spectra at the receiver, while transmitting 8ASK (bipolar) signals
using different pulse shapes and roll-off slope factors α = 0, 0.3, 0.6, 1. (Use the
built in MATLAB function at the receiver). Measure the bandwidth of the signals.

Lab Exercise L-3.3

Compare the eye-patterns at the receiver, while transmitting 8ASK (bipolar) signals
using different pulse shapes and different roll-off slope factors α. (Use the built in
MATLAB function at the receiver)

Lab Exercise L-3.4

Determine the bit error ratio versus Eb/N0 (range: 0 dB to 15 dB in steps of 2 dB)
for 8PSK using the following basic pulse shapes: rectangular, raised cosine with
α = 0.3 and your own one. Use the diagram on page 64 to illustrate your results
in a single plot.
3.4. Lab 6
 6
0
10

−1
10

−2
10
Figure 3.2: Diagram for L-

Bitfehlerra

−3
10

3 Variants of PAM-Transmission
−4
10

−5
10

−6
10
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
10*log (E /N ) [dB]
10 b 0
3.4. Lab 6
6 3. Variants of PAM-Transmission

3.4.2 Offset-QAM

In order to minimize variations of the envelope of the transmit signal Offset-QAM has
been invented. In order to achieve this property the quadrature component of the signal
is delayed by half a symbol duration (T/2) as depicted in Figure 3.3.

√
2 cos(2πf t)
0

Re {·} g(t)
am[k] sRF(t)

Im {·} T/2 g(t)

√
− 2 sin(2πf0t)

Figure 3.3: Offset-QAM transmission scheme

Homework H-3.3

Why is it useful to minimize variations of the envelope of the transmit signal?

Lab Exercise L-3.5

Compare the phasors of a 4QAM transmission scheme with and without offset
(time- delay) at the transmitter as well as at the receiver. At the transmitter you can
check “Offset” on and off. At the receiver use MATLAB to display the phasor.

Lab Exercise L-3.6

How do different basic pulse shapes with different parameters α influence the pha-
sor?
3.4. Lab 6
6 3. Variants of PAM-Transmission

Lab Exercise L-3.7

Using the possibility of defining your own impulse define a basic pulse shape of
positive sinusodial shape, more precisely a sinus from “0” to “π” fitted into the
intervall from “0” to “T”. Transmit Offset-4QAM using this basic pulse shape and
display the phasor of the signal. What shape does it have?

Lab Exercise L-3.8

What crest factor does the transmit signal possess?

Lab Exercise L-3.9

What is the name of this well-known transmission scheme you just put together?
3.4. Lab 6
7 3. Variants of PAM-Transmission

3.4.3 Gaussian Minimum Shift-Keying

For the following exercises use the basic pulse shape “MSK”. As MSK does possess a
constant envelope of the transmit signal, all the transmitted information is contained in
the phase of the signal and not in the amplitude. Therefore MSK can also be interpreted
as binary Frequency-Shift-Keying (FSK).

Homework H-3.4

In order to revisit this equivalence between MSK and binary FSK please fill out the
solution sheet for the given sequence. (You will also find an example in the DiCo
script)

Lab Exercise L-3.10

At the receiver compare the spectrum of a standard 4QAM signal and an MSK
signal.

As you may have realized the changes of frequency in MSK/binary FSK are very abrupt.
A softer slope in the frequency over time would result in a smaller spectrum. In order
to reach that goal a filter can be applied, which softens the abrupt rectangular changes
in frequency into Gaussian sloped ones. The name of this scheme is “Gaussian Minimum
Shift-Keying” (GMSK).

• MSK:
0.25
T  f (t) −

−0.25
0 1 2 3 4 5 6
t/Tb −−>

• GMSK:
0.25
T  f (t) −

−0.25
0 1 2 3 4 5 6
t/Tb −−>
3.4. Lab 7
7 3. Variants of PAM-Transmission

a[k] +1 + j +1 + j −1 − j +1 − j −1 − j +1 − j I
Q I Q I Q I Q I Q I
b[k] Q

Re{s(t)}

t
1 T 3
2T 2T

Im{s(t)}

t
Phasor

π
π
2

Phase ϕ(t) 0 t
— 2π

−π

1
2T

Instaneous
0 t
frequency
dϕ(t)
2πdt
1
− 2T

k 0 1 2 3 4 5 6 7 8 9 10 11

c[k]

Figure 3.4: Solution sheet for exercise 4

3.4. Lab 7
7 3. Variants of PAM-Transmission

Modulation:

MSK:

GMSK:

with

−(f/B ) b·ln2 2/2

H G(f ) = e

GSM: Bb · Tb = 0, 3
Parameter: Bb · T b

Lab Exercise L-3.11

Use the basic pulse shape GMSK. Display the phasor of the transmit signal and
compare it to the phasor of MSK.

Lab Exercise L-3.12

What influence does the parameter Bb · Tb have on the bandwidth and the phasor
of the received signal? How does the eye-pattern change?

Lab Exercise L-3.13

Compare standard 4QAM, MSK and GMSK in terms of spectral efficiency as well
as the opening of the eyepatterns.
3.4. Lab 7
7 3. Variants of PAM-Transmission

3.4.4 “Carrierless“ Amplitude and Phase Modulation

Instead of first filtering the signal with the basic pulse shape g(t) and then mixing it into
the RF-band using carrier frequency f0 one can also combine these two blocks by not
defining G(f ) in the baseband but shifting it to a higher frequency (=carrier frequency).
If we wanted to use in-phase and quadrature components we would have to use two
orthogonal transmit pulses, one for each component. Therefore we will just use bipolar
ASK (onedimensional).

Lab Exercise L-3.14

Define a basic pulse shape for transmission in the CAP scheme. Make sure you
check the ”CAP“ check-box to tell the transmitter you are now using CAP.

Lab Exercise L-3.15

How would the orthogonal basic pulse shape for the quadrature component look
like?

Lab Exercise L-3.16

Compare the spectra, eye-patterns and transmission errors for 8ASK using CAP
and standard transmission.
3.4. Lab 7
7 4.
 Project 4
OFDM

4.1 Introduction, Background, and Motivation

In many (digital) transmission schemes the channel cannot be modeled by a simple

AWGN channel but the channel is also dispersive. This means that the transfer function
of the channel is not only a scalar as for the AWGN channel, but depends on the
frequency. The channel introduced intersymbol interference. Orthogonal frequency-
division multiplexing (OFDM) is a well-known and popular technique to cope with these
frequency-selective channels.

4.1.1 Orthogonal Frequency-Division Multiplexing

For this lab, we follow the description and definition of OFDM as given in [Hub11]. The
basic idea of OFDM can be visualized by splitting the frequency-selective transfer
function of the dispersive channel into a number of quasi-constant subbands. These
subchannels are then non-dispersive and can be used in parallel. They do not introduce
any intersymbol interference.

The basic pulse shapes used for the parallel transmission over these subchannels have to
be orthogonal in the frequency domain. At the receiver a bank of matched filters prevents
interference between the individual subchannels. This general approach to OFDM is is
shown in Fig. 4.1.
8 4.

a1[ks] s1(t)

a2[ks] s2(t)

aD[ks] sD(t)

Figure 4.1: Basic principle of OFDM

For the practical implementation of an OFDM system, a different approach is preferred

over the individual parallel modulators. Instead, the inverse discrete Fourier transform
(IDFT) is use to transform blocks of D symbols from the frequency domain to the time
domain. These symbols are then modulated by a single basic pulse shape g(t). At the
receiver the matched filter wrt. g(t) is applied an then the block of D receive symbols is
transformed again into the frequency domain using a DFT.

The individual blocks generated at the transmitter are separated by a cyclic prefix to
avoid interblock interference. In Fig. 4.2 this block-based description of OFDM is
illustrated. Finally, the OFDM transmission over independent parallel subchannels can be
described by just using individual scaling factors per channel.

Figure 4.2: Transmission over independent subchannels

4.2. Purpose 81

4.1.2 Bit Loading

The varying scaling factors λi per subchannel lead lead to varying signal-to-noise
ratios over the subchannels. If these different SNRs are known at the transmitter, so-
called bit loading can be used. That is, the data rate per subchannel is adapted to the
subchannel conditions.

One of the potential approaches to adapt the individual data rates is “rate allocation
according to the channel capacity“. R = ΣD ld(1 + 1 ) A detailed explanation of this
N ∗γ
bit loading algorithm will be given in class. i
i

4.2 Purpose

The aim of this experiment is to demonstrate the OFDM and a simple bit loading al-
gorithm. You should understand why OFDM is widely used in dispersive channels. In
addition you should understand why bit loading can useful in OFDM systems.

4.3 Lab Environment

The lab environment is identical to that of Chapter 2.

Readings for Lab 4

[Hub11] Johannes B. Huber, Digital Communications, Lecture Notes, Erlangen, October

2011.
8 4.

4.4 Lab Exercises

In this Lab we will extend the simulation script from Chapter 2. The aim is a simulation
of an OFDM system.
The first change in the transmission scheme is the dispersive channel.

Lab Exercise L-4.1

There is a function dispersive_channel() which models the dispersive channel.

Add the frequency-selective channel to your transmission chain before the addition
of the white noise. How does the bit error ratio change?

Homework H-4.1

How can you observe the influence length of the dispersive channel.

In this Lab we use an all-digital representation of the transmission scenario. Thus, the
influence length of the channel can not be measured in seconds, but in samples only.

Lab Exercise L-4.2

Determine the length of the impulse response (in samples) of the given dispersive
channel.

4.4.1 OFDM Transmitter

We now want to transmit the data using an OFDM system.

Homework H-4.2

Draw the block Diagram of a general OFDM transmitter.

Homework H-4.3

The OFDM system shall be designed to operate in the base band (like in DSL)?
Are there any constraints on the signal?
4.4. Lab 8
8 4.

In order to construct the OFDM transmitter, it is reasonable to split it into smaller

blocks.

Lab Exercise L-4.3

Save the file old transmitter.m as ofdm_transmitter.m. The bitstream has to be

split in D = 16 substreams. Use the given function map_bits_qam(bitvector, M)
for the mapping for each substream on symbols.

Lab Exercise L-4.4

Use the inverse discrete Fourier transform to generate the OFDM signal out of the
symbols for each subchannel.

Keep in mind that the system should transmit in the base band.

Lab Exercise L-4.5

Proof that your signal is real valued.

Lab Exercise L-4.6

Add the cyclic prefix. The length of the cyclic prefix should be optimized for the
given dispersive channel.

Next, we analyze the transmit signal.

Lab Exercise L-4.7

In order to analyze the OFDM transmit signal, we generate a signal resulting from
1,200,000 Bits. Use the function write_to_card(Block) to generate an output of
the D/A converter. Now analyze the OFDM signal in time and frequency domain
using the oscilloscope and spectrum analyser.
4.4. Lab 8
8 4.

Lab Exercise L-4.8

Study the distorted signal after the dispersive channel and the AWGN with the
oscilloscope and spectrum analyser.

4.4.2 OFDM Receiver

Homework H-4.4

Draw the block diagram of a generic OFDM receiver.

Lab Exercise L-4.9

Save the file old receiver.m as ofdm_receiver.m.

Lab Exercise L-4.10

Remove the cyclic prefix from the signal.

Lab Exercise L-4.11

Use the Fourier transform to generate the OFDM signal out of the symbols for each
subchannel. Keep in mind that the system should transmit in the base band.

Lab Exercise L-4.12

In order to equalize the effect of the scaling factors in the subchannels, we apply a
so- called frequency domain equalization (FEQ). The FEQ is provided in the
function feq(Symbols).

Lab Exercise L-4.13

The bitstream has to be split up for the 16 individual subchannel. Use the given
function demap_bits_qam(Symbols, M) for the mapping of individual stream.
4.4. Lab 8
8 4.

After the transmission the BER can be calculated.

Lab Exercise L-4.14

Modify the function calculateBER.m to match the needs for your OFDM system
and save it as the new file calculateBER_OFDM.m.

Homework H-4.5

The dispersive channel has an influence on the SNR. How can the SNR be computed
at the receiver? How can it be calculated per subchannel?

Lab Exercise L-4.15

Write a function to compute the SNRper subchannel and plot the SNR.

Lab Exercise L-4.16

Plot the SNRper subchannel.

Now you know the SNR per subchannel. It is possible to use bit loading to achieve higher
data rates if you use modulations of a higher order in subchannels with an high SNR.

4.4.3 Bit Loading

Lab Exercise L-4.17

Extend your OFDM transmitter and receiver to support bit loading, i.e., varying
data rates per subchannel. Determine the data rates of the 16 subchannel by hand!

Lab Exercise L-4.18

Write a block of the loaded signal to the D/A converter and analyze the signal in
time and frequency domain.
4.4. Lab 8
9 5. Signal Space
 Project 5
Signal Space
Representation
5.1 Introduction
This chapter will introduce the implementation of modulation schemes with signal ele-
ments. Here we will focus on memory-less modulation schemes such as Frequency Shift
Keying (FSK). Those modulation schemes with inherent encoding, such as Continuous
Phase Modulation (CPM) are not considered.

5.2 Purpose
In this exercise the student will extend his existing transmission scheme to use a set
of signal elements. Thereby the implementation of other modulation schemes, such as
Frequency Modulation (FSK) becomes easy. It will become obvious that a memory-less
modulator is used to connect the signal numbers – an abstract mathematical description
of information – with signals that can be transported over a physical channel. A matched
filter (MF) receiver (also known as correlation or vector receiver) will demodulate the
transmitted signal which is degraded by additive noise due to the AWGN channel. We
will see that with the use of signal elements the set of modulator and MF receiver can
easily be described by a single set of signal elements. We will also learn that with the use
an orthogonalization algorithm the complexity of the receiver may be reduced.
92 5. Signal Space

5.3 Lab Environment

The lab environment is equal to those of the prior exercises. We will extend existing
implementations to include the features of signal space representation. In order to later
distinguish your implementations of different lab courses we will copy existing files
instead of editing them.

5.4 Signal Space Representation

In general transmission schemes contain several blocks such as an encoder, a mapper,

and a modulator. We will focus on uncoded transmission so we can neglect the encoder.
Basically the modulator is memory-less and contains a set of – not necessarily orthogonal
– signal elements (Fig. 5.1).

Memory-less modulator

s0(t)
m[k] T
M
q[l] s1(t) s(t)

s2(t)

s3(t)

Figure 5.1: Block diagram of a transmitter. The binary data sequence q[k] is mapped to
signal numbers m[k] ∈ {0, 3} which selects the signal elements.

Instead of weighting your basic pulse shape g(t) with ±1 as it is done for binary PAM
transmission, we now have a signal space S with a set of signal elements of a duration
T from which the corresponding signal element is selected for each time instance. The
resulting transmit signal can be written as
+∞
Σ
s(t)
sm[k](t − kT ) (5.1)
= k

=−∞
5.4. Signal Space 9

where m ∈ {0, · · · , (M − 1)} describes the signal number at time instance k and sm[k](t)
is the corresponding signal element.

5.4.1 Orthogonality

An important condition is that signal elements sm(t) must fulfill the temporal orthogo-
nality condition [Hub11]

+∞ s (t + kT ) · s ∗ (t) dt = δ , ∀i, l ∈ {0, 1, . . . , (M − 1)},

E∫il1 i l 0k (5.2)
−∞

with Eil being the energy of the cross correlation of the signal elements si(t) and sl(t)
and δij being the Kronecker-symbol:

1for i = j
δij = (5.3)
0 for i j.

In general the signal elements are not limited in time but we will focus on time-limited
signals which are by definition temporally orthogonal.
Although this temporal orthogonality is fulfilled the signal elements might not be mutual
orthogonal. If the double orthogonality condition is also fulfilled the signal elements are
also mutually orthogonal! Throughout this lab course signal elements are described by
si(t) with i ∈ {0, . . . , (M − 1)} unless they fulfill the double orthogonality. Else signal
elements are described by gi(t) and are orthonormal basis vectors of the signal space, i.e.

1 ∗
∫ +∞gl (t + kT ) · gl (t) dt = δll · δ0k , ∀i, l ∈ {0, 1, . . . , (M − 1)}. (5.4)
E g−∞

Each of these continuous-time signals si(t) or gi(t) will be represented as a vector in

discrete-time k with s[k] = s(kTs). Each signal element can be represented by a vector
with T/Ts elements (T/Ts must be an integer) and which contains one symbol. Thus the
quotient T/Ts describes the number of samples per symbol.
The orthogonality condition must then be described in discrete-time giving
+∞
1
Σ g [k] · g∗[k] = δ · δ , ∀i, l ∈ {0, 1, . . . , (M − 1)}. (5.5)
l l ll
g
E
94 5. Signal Space

If all signal elements or basis vectors of the signal space are summarized into a single
vector we get
d ef
g(t) = [g0 (t), g1 (t), . . . ,Mg (t)] (continuous-time) (5.6)
d ef
g[k] = [g0 [k], 1g [k], . . . ,Mg [k]] . (dicrete-time) (5.7)

In case of discrete-time the resulting matrix g[k] has the dimension

Number of signal elements × (T/Ts). (5.8)

The double orthogonality condition can then be tested with:

+∞
1
Σ g[k] · g∗[k] = I δ (5.9)
M 0
g
E

5.4.2 Orthogonalization

Although in most digital transmission scheme the signal elements are not mutually or-
thogonal the same signal space can be represented using orthonormal basis vectors.

The goal is to find basis vectors gi(t)/gi[k] (which are in most cases not unique for a
given set of signal elements) and its linear factors si,l so that each signal element can be
represented as a linear combinations of weighted basis vectors [Hub11]:
D−1
Σ
si[k] = sl,i · gl[k] (5.10)
l=0

Introducing a matrix-vector notation, we have

s[k] = [s0[k], . . . , sD−1[k]] , (5.11)

g[k] = [g0[k], . . . , gD−1[k]] and  (5.12)
s · · · s0,(M −1)
 0,0 
.
(5.13)
S .. sl,i . 
 s(D−1),0 · · · s1,(M −1)

which results in

s[k] = g[k] · S. (5.14)

5.4. Signal Space 9

At this point it should be clear that there might suffice D ≤ M orthonormal basis vectors
gi(t) to span the signal space.
This representation is a linear transformation of basis vectors into signal elements. This
process can be reversed. The orthonormal basis vectors can be represented as linear
combinations of orthogonalization factors gi,l and the signal elements themselves:
M −1
Σ
gl[k] = gi,l si[k]. (5.15)
i=0

To find a set of orthonormal basis functions we will focus on the so called Gram-Schmidt
Procedure (GSP) which is described by a recursive algorithm [Hub11] and calculates the
• Linear Factors (linear factor matrix S) and the
• Orthonormal Basis Function gi(t)/gi[k]
A set of M row vectors s0[k], s1[k], . . . , sM−1[k] representing the signal elements over time
k is given. The goal is to find a set of orthonormal basis vectors g0[k], g1[k], . . . , gD−1[k]
with unknown dimensions D ≤ M and linear factors sl,i such that
D−1
Σ
si [k] = sl,i · gl [k]. (5.16)
l=0

The recursive algorithm works as follows:

1. Calculate the first basis vector using the first signal element vector
1
g0[k] = s0[k]
rΣ s0[k] · s∗[k]0
k
Number of basis vectors determined so far: l = 1

Linear factors for s0:

s
Σ
s0,0 = s0 [k] · s0∗ [k], sl,0 = 0 for l = 1, 2, . . . , D − 1 (5.17)
k

rΣ 
s0 [k] · s 0 [k]k · · ·
∗

S= (5.18)
 ..
0 .
.

 . 
0
g[k] = h 1 i
s0,0 s0[k], · · · (5.19)
Dimension D still undetermined.
96 5. Signal Space

2. Moving on to the next vector si[k] (i = 1, 2, M − 1):

a) Linear factors regarding the l basis vectors computed so far:

Σ
sl,i = si [k] · gl∗ [k] (5.20)
k

s0,0 s0,1 s0,2 ···
..
0 s1,1 s1,2 . 
.. (5.21)
S= 0 0 s2,2 . 

0 0 ..
 . 
0
. . . .
h i
1 1
g[k] = s
s0[k] , s
s0[k] + 1
s1[k] , · · ·
s
(5.22)

b) Now test if current signal element can already be represented with the existing
basis vectors:
Σ l−1
∆i = si [k] · s∗i [k] − Σ 2
|sl,i | (5.23)
` k ˛¸ `l ˛ ¸ x
ener gy of the s ignal
energy of th e ex is ting basis
x vectors
ele ment

i. ∆i = 0: The current signal element sl[k] can fully be represented by a linear

combination of existing basis vectors. No new basis vector is needed.

sl,i = 0 für l = l, l + 1, . . . , D − 1

ii. ∆i > 0: There is energy left, that cannot be represented by a linear

combination of basis vectors. Thus a new basis vector has to be
introduced.

1 !
l−1
gl[k] = √ Σsl,i gl [k] (5.24)
si[k] −
∆ i
l=0

Determine the new linear factor for si[k]:

√
sl,i = ∆i, sl,i = 0 for l = l + 1, l + 2, . . . , D − 1 (5.25)

Increment number of basis vectors: l → l + 1

3. As long as i < M − 1, repeat step 2 with vector si+1[k] (recursion)

4. Orthogonalization finished. Dimensionality: D = l
Bibliography 97

The result is a set of new orthonormal basis vectors which describes the identical signal
space as the original set of signal elements, and a set of linear factors for each basis.

The above description uses an “incremental” order to select the next signal element in the
progress. Alternatively one can choose that signal element that has the maximum energy
∆i that cannot be represented as linear combination of basis vectors. This modification
is later on called “sorted-by-energy”.

Readings for Lab 5

[Hub11] Johannes B. Huber, Digital Communications, Lecture Notes, Erlangen, October

2011.

[Kam08] Karl-Dirk Kammeyer, Nachrichtenübertragung, 4 ed., B. G. Teubner, Stuttgart,

March 2008.

[Pro00] John G. Proakis, Digital communications., 4th ed., McGraw-Hill, New York,
NY, USA, 2000.
9 5. Signal Space

5.5 Lab Exercises

5.5.1 Transmission with signal elements

We define the signal elements as depicted in Fig. 5.2. There are three different transmission
schemes with signal elements s1,i(t), s2,i(t) and s3,i(t), i ∈ {0, . . . , 3} defined. Each set
of signal elements describe its own transmission scheme connecting signal numbers to
waveforms.

On the other hand one can interpret the set of signal elements as a multiplexing scheme
where multiple users (in our case four users) can transmit data at the same time using
one individual signal element.

Homework H-5.1

Assign the signal elements to the following multiplexing schemes

• Time Division Multiplex

• Frequency Division Multiplex

• Code Division Multiplex

and justify your decision!

We now neglect the ability to implement a multiplexing scheme using signal elements
and focus on the so call point-to-point transmission. This means, one transmitter with
a set of signal elements and one receiver which also knows the signal elements. By
selecting a signal element at each time instant the transmitter can communicate a
sequence of information symbols.

Lab Exercise L-5.1

Implement a transmitter that uses the signal elements that describes a code multi-
plexing transmission scheme! Copy your existing implementations of a transmitter
and extend it so that you can use it in the following form:

signalElements = generateSignalElements(type, overSamplingFactor);

transmittedSignal = transmitter_SE(bitStream, ...
signalElements);
5.5. Lab 99
1 5. Signal Space

s1,0(t) s1,1(t) s1,2(t) s1,3(t)

0 t
T
t
T
t
T
t
T
0.51 0.51 0.51 0.51
−1
঩ ঩ ঩ ঩
{s2,0(t)} {s2,1(t)} {s2,2(t)} {s2,3(t)}

1
t t t t
T T T T
0 0.51 0.51 0.51
0.51
−1 s3,1(t) s3,2(t) s3,3(t)
s3,0(t)

1
t t
t t
0 T T T T
0.51 0.51 0.51 0.51
−1

Figure 5.2: Three different transmissions schemes and its signal elements

The parameter type is a string that describes which set of signal elements are used.
Note that these signal elements are generated by a separate function and stored in
the variable signalElements which are then used in the transmitter. Proceed as
follows:

1. Familiarize yourself with the functions bi2de and de2bi and use these within
your transmitter and receiver (later).
2. Write a function generateSignalElements. Generate your set of signal elements
depending on the parameter type!
3. Take a look at the signal elements (real and imaginary party, absolute value). Use
an oversamplingFactor of 128 for testing purpose.
4. In the transmitter, map the sequence of binary data to the signal numbers! (use
natural labeling)
5. Make sure the mapping can address all signal elements!
6. Generate the transmit signal s(t) by concatenation!
5.5. Lab 10
1 5. Signal Space

Homework H-5.2

Determine the sequence of signal numbers for the following data sequence which
uses the signal numbers s1,0(t), . . . s1,3(t)!

s(t)

0 T
t

1 2 3 4 5 6 7 8 9 10
−1

Lab Exercise L-5.2

Generate the given transmit signal with your implementation (ECB-domain).

Homework H-5.3

Using your knowledge in signal space representation, what signal elements exist for
a 4-ary PAM with a rectangular pulse-shaping? And how many possibilities exist
to get from binary data streams to signal numbers? Also

• sketch all signal elements in one diagram!

• give the equation to generate each signal element!

• determine the number of orthonormal basis functions are needed to represent all
signal elements? (No calculation necessary!)
5.5. Lab 10
1 5. Signal Space

5.5.1.1 Vector Receiver

Our transmission system now contains a transmitter which selects for each time instant t
a signal element of a M dimensional set of signal elements S = {s0(t), s1(t), . . . , sM−1(t)}.
By concatenating signal elements we generated the transmit signal which is transmitted
over an AWGN channel.
A coherent Maximum Likelihood detection can be achieved by a correlation receiver as
depicted in Fig. 5.3. Here, the received signal r(t) is uses as an input signal for a vector
receiver, or matched filter bank which consists of the matched filter for each signal
element in the signal space S. This vector receiver correlates the received signal r(t) with
each signal element si(t). After sampling, we get a correlation vector d[k] which
constitutes a sufficient statistics on the receiver input signal with respect to the
transmitted sequence of signal numbers m[k]. The decision for a signal number is then
simply made by picking the maximum in the correlation vector.
kT + TV
1 d1[k]
Eg s 1(T V − t)
kT + TV
1 2 d2[k]
Eg s (T V − t)
r(t) d[k]
.

kT + TV
1 dD[k]
Eg s D (T V − t)

Figure 5.3: Correlation Receiver (also: Vector Receiver) [Hub11]

Lab Exercise L-5.3

Implement a vector receiver using the signal elements defined in exercise L-5.1!
We recommend to first implement the filter bank and the correlation using the
MATLAB table with the signal elements. Take a look at the results of this correlations
before implementing the sampling. After locating the maximum value in the
correlation vector compare the sequence of signal numbers with those at the
transmitter!

Lab Exercise L-5.4

Implement a function
5.5. Lab 10
1 5. Signal Space

ber = function SigSpaceTrans(binData, sigElements, SNR)

which performs a complete transmission using predefined signal elements. Proceed

as follows:
1. Use your transmitter to generate the transmit signal s(t) and the signal elements!
2. Perform a transmission over an AWGN channel using the predefined noise variance!
3. Use the noisy signal as input signal for the vector receiver!
4. Sample the signal and decide for the most likely signal numbers!
5. Reverse the mapping from signal numbers to a binary sequence!
6. Count the number of bit errors made!

Lab Exercise L-5.5

Implement a simulation of the transmission for different Es and plot the resulting
N
error probabilities over the SNR! Use a data block with a length of at least 105.
Reduce the oversamplingFactor to around 10.

5.5.2 Gram-Schmidt Procedure

We now consider the Gram-Schmidt Procedure to orthogonalize the signal elements.

Therefor we use the MATLAB-table which contains the signal elements.

Lab Exercise L-5.6

Implement the Gram-Schmidt Procedure in MATLAB!

Homework H-5.4

Use the Gram-Schmidt Procedure to orthogonalize the following set of signal ele-
ments:

a = [ 1, 1, 1, 1 ]
b = [ −1, −1, 1, −1 ]
c = [ 2, 2, 0, 2 ]
5.5. Lab 10
1 5. Signal Space

Lab Exercise L-5.7

Compare the result of your homework with the result of your implementation!

Lab Exercise L-5.8

Orthogonalize the signal elements from exercise 1! Sketch the resulting

orthonormal signal elements gi(t)! How many basis functions are necessary?

Lab Exercise L-5.9

Discuss how to describe the original signal elements si(t) using the orthonormal
basis vectors gi(t)! Rebuild the original signal elements in MATLAB.

Homework H-5.5

Describe in your own words how the receiver in task 5.4 has to be modified so that
a matched filter bank with less filters can be used!

Homework H-5.6

If the orthogonal basis do not contain the zero-vector, which basis vector would you
use for your reduced complexity receiver and which would you neglect? Does the
option “incremental” and “sorted-by-energy” influence the decisions?

5.5.3 Frequency Shift Keying

In a M -ary frequency shift keying modulation (FSK) [Kam08, Pro00] scheme the signal
elements are defined as
√
si(t) = Eg /T ej(2πi h t/T +ϕ0 ) for t ∈ [0, T (5.26)
)
0 for t ∈/ [0, T )
5.5. Lab 10
1 5. Signal Space

For the time-discrete version it holds t = kTs:

Eg/T ej(2πi h (kTs)/T +ϕ0) for k 0, 1, . . . , (T/Ts 1)

si[k] =  √
0 ∈ { −
else (5.27)
}
with a modulation index h ∈ R which describes the maximum phase shift and the symbol
numbers i = 1, 2, . . . , M . The transmit signal s(t) can then again be described as a
concatenation of signal elements.

Homework H-5.7

Determine the modulation index h and the alphabet size M of an FSK scheme
which uses the signal elements s2,0(t),. . . ,s2,3(t) in Fig. 5.2 (real value of s2,i(t)
plotted)!

Lab Exercise L-5.10

Generate a FSK signal elements for M = 4 and h = 1 ! Plot them and describe the
4
result!

Lab Exercise L-5.11

Determine the orthonormal basis functions for the signal space in exercise 10! De-
scribe the resulting basis vectors!

Homework H-5.8

Which sets of parameters in FSK result in orthogonal signal elements?

5.5. Lab 11
11 6. Signal Processing in MIMO
 Project 6
Signal Processing in
MIMO Systems

6.1 Introduction, Background, and Motivation

Modern digital communication systems employ multiple antennas at the transmitter and
the receiver side. Two basic principles can be distinguished which constitute the benefits
of multiple-input/multiple-output (MIMO) systems compared to single antenna systems:
On the one hand the multiplexing gain, i.e., the increased data rate due to
transmission of independent data streams over multiple transmit antennas. On the
other hand, the possibility to observe several independent copies of the transmit signal,
e.g., through the use of multiple receive antennas, and thus increased robustness to
shadowing and noise due to diversity.

In this lab, an introduction to signal processing for MIMO communications is given,

motivating the potentials and the benefits of the techniques enabled through the use of
multiple antennas at transmitter and/or receiver.

To this end, we consider the discrete-time equivalent model of digital pulse-amplitude

modulation transmitted over a flat-fading MIMO channel with AWGN as depicted below.
11 6. Signal Processing in MIMO

h1,1 n1
x1 h2,1 y1

hNR,1
n2
x2 y2

n NR
xNT yNR

Aspects such as pulse-shaping, up-/down-conversion to and from the carrier frequency, as

well as matched filtering are hidden in this equivalent model, and implicitly assumed to
op- erate perfectly synchronized. With these assumptions, adopting a compact
vector/matrix- notation we define the transmit signal as

 
x= . (6.1)
 .x1 
 
x NT
and the received signal as

 
y= . = Hx + n (6.2)
 .y1 
 
y NR
where n is a (NR × 1)-vector collecting the additive white Gaussian noise samples on each
receive antenna and H denotes the (NR × NT)-dimensional MIMO channel matrix given by
 
h . . . h1,N
 1,1 
T

.
.. (6.3)
.H =  .
hN ,1 . . . hN .,N 
R R T

The transmit symbols xi, i = 1, . . . , NT, are taken from an arbitrary complex-valued
signal constellation (here, we will only consider square QAM constellations).

We begin with studying the special case of single-input/multiple-output (SIMO)

trans- mission, i.e., the transmitter communicates a single data stream using a single
transmit antenna (NT = 1) to a receiver which is equiped with multiple receive antennas
(NR). The results are compared to a conventional single-input/single-output (SISO)
system. Next,
6.2. Lab Environment 115

we consider the dual case of multiple-input/single-output (MISO) transmission (NT ≥ 1,

NR = 1) and briefly review a technique called space-time codes. We will then extend
our considerations to include multiple antennas at transmitter and receiver side and
im- plement and discuss two different (receiver-side) equalization strategies for this
MIMO system. To this end, we first implement a generic system model of MIMO
transmission using QAM.
The notation and organization of this lab closely follows [?]; the corresponding course
and its lecture notes are not required for preparation of this lab.

Noteworthy, multi-antenna systems are only one, but probably the most intuitive example
for MIMO systems; other communication systems, as, e.g., code-division multiple-access
(CDMA) systems, can be modeled in a similar way. The studied principles of this lab are
thus applicable in a broad class of digital communication systems.

6.2 Lab Environment

The MIMO transmission system considered in this lab is implemented in MATLAB. Dif-
ferent from previous labs, all simulations operate in the discrete-time equivalent system
of digital transmission. Pulse shaping and modulation is not required; functions of
previous labs are not reused.

Readings for Lab 6

11 6. Signal Processing in MIMO

6.3 Lab Exercises

6.3.1 System Model

Homework H-6.1

Give an equation for the variance of the transmit symbols for the case of a M -
ary square-QAM constellation. Assume that the signal points are taken from the
grid

A ⊆ {±1 ± j, ±3 ± j, ±1 ± 3j, ±3 ± 3j, . . . } . (6.4)

Lab Exercise L-6.1

Implement a function

x = GetQAM(r, c, M)
which generates an r × c matrix x containing randomly drawn M -ary QAM
symbols taken from A.
Generate a vector x of reasonable length and determine the variance of the QAM
symbols.
Compare this result with the analytical one from the lecture notes.

Lab Exercise L-6.2

Implement a function

a_hat = QuantQAM(z, M)
which quantizes the elements of the matrix z to M -ary QAM symbols.
Use the MATLAB-function round and note the finite size of the constellation.

In this lab we will only consider the following channel model:

• the elements hm,n of the MIMO channel matrix H are independent zero-mean
unit- variance complex-Gaussian distributed
6.3. Lab 11
11 6. Signal Processing in MIMO

• the elements of the additive noise vector are independent zero-mean complex-
Gaussian distributed with equal variance σ2.
n

Homework H-6.2

Interprete the assumptions of this channel model and briefly explain the reasoning
behind.

Homework H-6.3

Give simple methods, which generate a random channel realization H and a

real- ization of the noise vector n according to this model in MATLAB.

In this lab, we restrict ourselves to the performance measure symbol error rate (SER).

Homework H-6.4

Give a simple method to calculate the SER in MATLAB given the QAM transmit
symbols and its estimates.

We now have implemented three major building blocks of a generic MIMO transmission:
the transmitter generating a sequence of QAM symbols transmitted over multiple anten-
nas, the MIMO AWGN channel, and the decision device quantizing the equalized
symbols into QAM symbols.

6.3.2 SISO

As a reference, we first implement a conventional SISO system (NR = NT = 1), as

depicted below.
h n
x y

As in the generic MIMO scenario we assume the fading coefficient h to be complex-

Gaussian distributed with zero mean and unit variance.
6.3. Lab 11
12 6. Signal Processing in MIMO

Homework H-6.5

State the name of the distribution of the magnitude of the fading coefficient, i.e.,
the channel gain, and give a formula for its probability density function (pdf).

Lab Exercise L-6.3

Write a simulation script which generates and transmits a sequence of QAM

symbols according to this setup.
The results shall be averaged over a large number of channel realizations. Include
a for-loop in order to simulate the SER for different SNRs and plot the resulting
SER-vs.-SNR curve. Take care that the SNR is correctly set. Measure the SER for
4-QAM and an SNR range of −10, 5, . . . , 50.

Note: Several (L) symbols can be transmitted over a single channel realization
simultaneously by constructing a (1 × L)-dimensional transmit vector.

Lab Exercise L-6.4

Estimate the pdf of the channel gain (using the built-in MATLAB-function hist)
and compare it to the analytical results from problem 5.

6.3.3 SIMO

We now consider the case of transmission of a single data stream (transmit symbols x
with variance σ2) using a single transmit antenna (NT = 1) to a receiver equiped with
x
multiple (NR) receive antennas.

Grouping the receive symbols into a vector, we have

y = hx + n (6.5)

Linearly combining the signals observed at the antennas as depicted below (combining
coefficients gn) offers to significantly increase robustness to noise (diversity gain); there
is no multiplexing gain compared to the SISO system.
6.3. Lab 12
12 6. Signal Processing in MIMO

h1,1 n1 g1
x1 h2,1 y1 z
hNR,1 n2 g2
y2

nNR gNR
yNR

Assuming channel state information at the receiver side, the optimum (w.r.t end-to-
end SNR) combining strategy for this SIMO setup is maximum ratio combining (MRC)
which sets gn = h∗n .

Homework H-6.6

Write the signal processing of MRC compactly in vector/matrix-notation and give

an interpretation of the underlying principle.

Lab Exercise L-6.5

Extend your simulation setup to incorporate multiple antennas at receiver side and
implement MRC. Measure the SER for 4-QAM, NR = 2, and an SNR range of
−10, 5, . . . , 50.
For comparison, the SER of both schemes SISO and SIMO shall be recorded.

Note that the combined receive signal has to be scaled to match the expected input
range of the decision device.

Lab Exercise L-6.6

Which effect is observed for an increasing number of receive antennas.

Run simulations for NR = 4, and 8 and also for different QAM constellations to
verify your observations.

Using MRC, the equivalent effective channel z = ǁhǁx can be introduced, i.e., the effective
channel gain calculates to ǁhǁ2 .
6.3. Lab 12
12 6. Signal Processing in MIMO

Homework H-6.7

State the name of the distribution of the effective the channel gain and give its
mean and variance.

Lab Exercise L-6.7

Estimate the pdf of the channel gain (using the built-in MATLAB-function hist)
and compare it to the analytical results and the pdf of the channel gain of SISO
transmission.

A simpler scheme to exploit the benefits of multiple antennas at the receiver is so-called
antenna selection (AS), as depicted below.
h1,1 n1
x1 h2,1 y1 z
hNR,1 n2
y2

nNR
yNR

In AS, for each time step, the best antenna is selected; its receive symbol is used as a
decision variable.

Homework H-6.8

Explain the advantages of AS over MRC with respect to hardware implementation

complexity.
Which antenna should be selected in each time step?

Lab Exercise L-6.8

Extend your simulation script to incorporate AS and compare the resulting perfor-
mance with MRC and the SISO system for different values of NR.
Which effect do you observe?
6.3. Lab 12
12 6. Signal Processing in MIMO

6.3.4 MIMO

In the final part of this lab, we investigate systems employing multiple antennas at both
transmitter and receiver side. Compared to the previous systems, such schemes offer the
potential to have both, a multiplexing gain (increased data rate through the transmission
of independent data symbols on the antennas for each time step) and a diversity gain
(increased robustness to noise).
In vector/matrix-notation, the MIMO receive signal can be written as

y = Hx + n (6.6)

Similar to equalization of inter-symbol interference in digital communication over disper-

sive channels, the straight-forward approach for equalization of the interference induced
by the use of the same transmit medium (time/frequency) is so-called zero-forcing linear
equalization (ZF-LE). In case of square systems (NT = NR) the receive signal is processed
with the inverse of the MIMO channel matrix, yielding

z = H−1y (6.7)

In case of non-square systems, the left-pseudo inverse (Moore-Penrose inverse) has to be

employed, yielding

−1
z = HHH HHy (6.8)

Lab Exercise L-6.9

Write a new simulation script which implements MIMO transmission with ZF-LE
and run simulations for 4-QAM using NT = NR = 4.
Which diversity order is obtained using this simple equalization strategy?
Note: Several symbols can be transmitted over a single channel realization simulta-
neously, by letting x represent a matrix of dimension NT × number of channel uses.

Similar to equalization of inter-symbol interference in digital communication over dis-

persive channels, linear equalization is also possible in a minimum mean-squared error
(MMSE) sense. In this case, we have
2 −1
z= HHH + σ2n I HHy (6.9)
σx
6.3. Lab 12
12 6. Signal Processing in MIMO

Lab Exercise L-6.10

Extend your simulation script incorporating MMSE-LE and compare the results
with ZF-LE for different parameter settings.

The optimum detection scheme for MIMO transmission is maximum-likelihood

detection. As the receive signal y is corrupted by AWGN, this is equivalent to finding the
noise-free signal point with minimum Euclidean distance to the receive signal point, i.e.,
2
xML = argminx˜ ǁy − Hx˜ǁ (6.10)

There are various algorithms to solve the ML detection problem with moderate compu-
tational complexity. Here, we only consider the simple strategy of a full search over all
candidate signal points.

Homework H-6.9

Calculate and discuss the number of candidate signal points for M -QAM transmit-
ted over NT transmit antennas.

Lab Exercise L-6.11

Implement a separate function for ML detection based on the full-search approach

and include it in your simulation script.
Compare the results to ZF/MMSE-LE for NT = NR = 2 and 4-QAM.
6.3. Lab 12