An experimental study of OFDM in a software defined ...1322/fulltext.pdf · The second technique is...

An Experimental Study of OFDM in a Software Defined

Acoustic Testbed

Rameez Ahmed, Northeastern University, Boston MA

E-mail: [email protected]

Advisor: Milica Stojanovic, Northeastern University, Boston MA

E-mail: [email protected]

December 2010

Abstract

Orthogonal Frequency Division Multiplexing (OFDM) is being considered for high speed

communication over wireless (acoustic) underwater channels which are characterized by fre-

quency selectivity and strong motion-induced Doppler distortion. Receiver algorithms and

adaptive power control are the focus of this thesis. Two types of receiver algorithms are

considered: those based on coherent detection and those based on differentially coherent

detection. While coherent detection theoretically offers better performance, it relies on ac-

curate channel estimation and phase tracking, which are challenging in conditions of fast

time variations. Coherent detection then suffers, making differentially coherent detection a

preferred choice in many situations. To partially offset for the channel variation, adaptive

power control based on receiver feedback is considered. Proof of concept of various com-

munication techniques is usually left to experimental testing because of the lack of widely

accepted statistical models for underwater acoustic channels. Real data recordings from a

recent Mobile Acoustic Communication Experiment (MACE-2010) are used to compare the

performance of adaptive receiver algorithms. In addition, an in-air acoustic test-bed has been

developed as a part of this thesis. Such an approach is deemed to provide a more realistic

vision of the channel conditions than a simplified computer simulation, while overcoming the

cost and difficulty of at-sea deployments. Processing of the in-air acoustic data confirms the

similarity with underwater conditions. Moreover, in-air tests enable on-line demonstration

of adaptive power control.

1

Acknowledgments

Foremost, I would like to thank my advisor Prof. Milica Stojanovic. She has been a teacher,

mentor, and more than that, a very good friend. Her ideas have always been a source of

inspiration for me. I thoroughly enjoyed my learning experience in the course ’Special Top-

ics in Digital Communication’ that she conducted. She has been extremely supportive and

guided me in all my endeavors. I would also like to thank Dr.Ranga Narayanaswami for his

positive feedback and his constant association with my work here at Northeastern.

I thank my parents Rasheed and Rabia, and my brother Razeen, for their constant sup-

port and inspiration. They have always inspired me to excel in everything I do, and provided

me with the right opportunities. My masters in the US would have been impossible without

their motivation and support.

I also thank my lab mates Baosheng, Srinivas, Ashish and Jordi. I have thoroughly

enjoyed their company in the lab. Their company provided a positive energy in the work

place. I would like to thank Jordi specifically for helping with various experiments in the

lab. I would also like to thank all my lab mates at CDSP, Joan Pratt and Faith Crisley for

the their help and guidance.

I would also like to thank my friends Sindhu Ghanta and Foram Thakkar, for spending

sleepless nights with me editing and proof reading this thesis. I would also like to thank my

roommate and friend, Ram, and all my other friends in Boston for all their support.

Last but not the least, I like to thank Dunkin Donuts for all the caffeine.

2

Contents

Abstract 1

1 Introduction 10

1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.1.1 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Underwater Channel 13

2.1 Acoustics for Underwater Communication . . . . . . . . . . . . . . . . . . . 13

2.2 Acoustic Propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.2.1 Attenuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.2.2 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.2.3 Propagation Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.2.4 Multipath . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2.5 Doppler Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3 Resource Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3.1 The AN Product and the SNR . . . . . . . . . . . . . . . . . . . . . . 18

2.3.2 Optimal Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.3.3 Transmission Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3 System Design 20

3.1 Basic OFDM Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.1.1 Orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.1.2 Modulation Using Fast Fourier Transform (FFT) . . . . . . . . . . . 22

3.1.3 Guard Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.1.4 Equalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.1.5 Advantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.1.6 Disadvantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3

CONTENTS

4 OFDM Modulator and Demodulator 25

4.1 OFDM Transmitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4.1.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4.1.2 Transmitter Implementation . . . . . . . . . . . . . . . . . . . . . . . 26

4.2 OFDM Receiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

4.2.1 FFT Demodulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

4.2.2 Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.3 Performance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.3.1 MSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

4.3.2 Bit and Symbol Error Rates . . . . . . . . . . . . . . . . . . . . . . . 41

5 In-Air Acoustic Testing 43

5.1 Hardware Used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5.1.1 EDIROL FA-101 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5.1.2 Transmitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5.1.3 Receiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5.1.4 Laptop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5.2 Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5.2.1 OFDM Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6 At Sea Testing 59

6.1 Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

6.1.1 Transmitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

6.1.2 Receiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

6.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

7 Online Processing 67

7.1 Online Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

7.2 Transceivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

7.3 Power Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

7.3.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

7.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

7.4 Text Messaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

7.4.1 Network Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

7.4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4

CONTENTS

8 Future Work and Conclusion 75

5

List of Figures

2.1 Absorption coefficient as a function of frequency. . . . . . . . . . . . . . . . . 14

2.2 Power spectral density of ambient noise N(f)[dBreµPa2] . . . . . . . . . . . 16

2.3 Frequency dependent part of SNR ∼ 1/A(l, f)N(f), k = 1.5 is used for A(l, f). 18

2.4 Optimal frequency plotted as a function of distance. . . . . . . . . . . . . . . 19

3.1 Bandwidth utilization of OFDM. . . . . . . . . . . . . . . . . . . . . . . . . 21

3.2 Transmitter implementation using IFFT. . . . . . . . . . . . . . . . . . . . . 23

4.1 Transmitter block diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4.2 Scrambling using matrix interleaving. . . . . . . . . . . . . . . . . . . . . . . 27

4.3 PSK constellations. QPSK on the left and 8PSK on the right. . . . . . . . . 29

4.4 Frequency interleaving to avoid errors. . . . . . . . . . . . . . . . . . . . . . 29

4.5 Synchronization preamble. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

4.6 Autocorrelation of synchronization preamble. . . . . . . . . . . . . . . . . . . 31

4.7 Transmitted OFDM signal, . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.8 OFDM signal showing the multipath arrivals. . . . . . . . . . . . . . . . . . 31

4.9 Receiver block diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

4.10 Top figure shows the synchronization preamble, center figure shows the re-

ceived sequence, bottom figure shows the cross correlation. . . . . . . . . . . 33

4.11 Received constellation for a 8PSK sequence. . . . . . . . . . . . . . . . . . . 34

4.12 Constellation phase tracking is performed. . . . . . . . . . . . . . . . . . . . 36

4.13 Estimated channel frequency response. . . . . . . . . . . . . . . . . . . . . . 38

4.14 Channel time-domain estimates. . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.15 Block Diagram of Partial FFT. . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.1 Transmitter: conventional multimedia speaker . . . . . . . . . . . . . . . . . 45

5.2 Transmitter frequency response. . . . . . . . . . . . . . . . . . . . . . . . . . 45

5.3 Capacitive microphone. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

6

LIST OF FIGURES

5.4 Deployment in lab. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5.5 Transmitter set-up. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5.6 Receiver set-up. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5.7 Channel time domain estimates. . . . . . . . . . . . . . . . . . . . . . . . . . 48

5.8 Plots of received constellation after detection, phase estimates and Doppler

factor estimates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.9 Bit error rate with coherent detection K = 1024QPSK. . . . . . . . . . . . . 49

5.10 Channel estimates for first block, frequency and time . . . . . . . . . . . . . 50

5.11 Phase estimates for to-and-fro motion. . . . . . . . . . . . . . . . . . . . . . 51

5.12 Bit error rate of received sequence. . . . . . . . . . . . . . . . . . . . . . . . 51

5.13 BER of decoded sequence with errors reduced. . . . . . . . . . . . . . . . . . 52

5.14 Phase estimates for fast motion. . . . . . . . . . . . . . . . . . . . . . . . . . 52

5.15 BER with differential detection for fast to-and-fro motion. . . . . . . . . . . 53

5.16 BER using coherent detection for fast motion. . . . . . . . . . . . . . . . . . 53

5.17 BER using differential detection for fast motion. . . . . . . . . . . . . . . . . 54

5.18 Differentially coherent results using Partial FFT. . . . . . . . . . . . . . . . 55

5.19 Receiver diversity using two receivers. No motion. . . . . . . . . . . . . . . . 57

5.20 Receiver diversity using two receivers. Fast Motion. . . . . . . . . . . . . . . 58

6.1 System deployment during MACE 2010. . . . . . . . . . . . . . . . . . . . . 60

6.2 Left: Horizontal transducer orientation. Right: Right vertical transducer

orientation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

6.3 Transmitted signal and frequency spectrum. . . . . . . . . . . . . . . . . . . 62

6.4 Received signal at Buoy and its frequency spectrum. . . . . . . . . . . . . . 63

6.5 Synchronization of preamble. . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

6.6 Received constellation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

6.7 Constellation after Differential Detection. . . . . . . . . . . . . . . . . . . . . 65

6.8 BER for K=256 and K=1024. . . . . . . . . . . . . . . . . . . . . . . . . . . 65

6.9 MSE for K=256 and K=1024. . . . . . . . . . . . . . . . . . . . . . . . . . . 66

7.1 Transceiver set-up. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

7.2 Block diagram of power control algorithm. . . . . . . . . . . . . . . . . . . . 69

7.3 Measured SNR at receives plotted against distance. . . . . . . . . . . . . . . 70

7.4 Case: static. Top: plot of power level for each transmission. Bottom: received

SNR for the transmission for the given power level. . . . . . . . . . . . . . . 71

7

LIST OF FIGURES

7.5 Case: Motion. Top: plot of power level for each transmission. Bottom:

Received SNR for the transmission for the given power level. . . . . . . . . . 72

7.6 Token ring network protocol. . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

7.7 Text messaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

8

List of Tables

5.1 OFDM signal parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5.2 Actual channel calculations. . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

6.1 OFDM signal parameters for MACE 2010. . . . . . . . . . . . . . . . . . . . 61

7.1 OFDM signal parameters for text messaging. . . . . . . . . . . . . . . . . . . 73

9

Chapter 1

Introduction

1.1 Background

In the recent years underwater wireless communication has found its application in many

fields; speech transmission between divers, remote control in oil rigs, ocean studies, wireless

sensor networks, to name a few. The need for better wireless underwater communication

systems is best explained by quoting the Titanic example. The passenger liner that sank in

1912 was discovered only after 73 years with a robot designed by the Woods Hole Oceano-

graphic Institute. The robot was connected by wires to a base ship which supressed its

mobility. The cables were not only expensive but also restricted the motion of the robot.

Underwater acoustic communications is a rapidly growing field of research and engineer-

ing as the applications which were once exclusive only to the military are now extending its

horizons into the commercial field. The successful transmission of these signals would enable

the elimination of physical connection to gather data at the sea beds and the unobstructed

operation of autonomous underwater vehicles.

Underwater communication is not a new field of research. The roots of underwater com-

munication can be dated all the way back to the times of Leonardo Da Vinci who used a

long tube submerged in the water to discover distant ships. The first underwater telephone

was developed in 1945 during the World War II to communicate with submarines. This de-

vice used a single-sideband (SSB) suppressed carrier amplitude modulation in the frequency

range of 8 kHz–11 kHz and was capable of sending acoustic signals over distances of sev-

eral kilometers [1]. The early efforts of digital video transmission was achieved by Japanese

scientists who showed a crab crawling at a depth of 1000m in an ocean trench [2]. With

the advancement of VLSI design and the availability of compact DSP’s, implementation of

10

Introduction

complex signal processing algorithms became possible.

In the recent years, significant advancements have been made in this field with respect to

the range of operation and the data throughput. Some commercial solutions for underwater

communications include the development of digital broadband underwater acoustic modems

for offshore oil field applications, environmental monitoring or AUV commandband control.

1.1.1 Outline

The main objective of this project has been to investigate OFDM as an application for under-

water acoustic communication where there is a strong presence of motion-induced Doppler

distortion. Two OFDM detection techniques have been considered and a comparison study

has been performed.The first technique is based on coherent detection which requires adap-

tive channel estimation and phase tracking [3]. The algorithm has a block by block depen-

dency and relies on accurate channel prediction. In conditions where fast varying channels

are present and accurate channel estimation is not possible the performance of this tech-

nique has been found to be sub-optimal. The second technique is based on the differentially

coherent detection which does not rely on channel estimation and hence is suitable for fast

varying channels [4].

To partially offset for the channel variations, adaptive power control based on receiver

feedback has been explored as a part of this project. Signal-to-noise ratio (SNR) at the

receiver is fed back to the transmitter on a duplex channel to perform delayed feedback

adaptive power control.

Because of the lack of widely accepted statistical channel models for the underwater

channel, proof-of-concept for various communication techniques is usually left to the exper-

imental tests. However, underwater tests are time consuming and costly. As part of this

thesis an in-air acoustic testbed was developed. Such an approach is expected to provide a

more realistic scaled version of the acoustic channel than simulations. The testbed was de-

veloped based on the EDIROL FA-101 external sound card and using conventional speakers

and transmitter and microphone as receivers.

The testbed was used to evaluate the performance of the different receiver algorithms.

The advantage of having such a testbed is the perfect knowledge of the channel which

can be verified experimentally. Text messaging has been introduced as an application to

demonstrate the working of the testbed using the above mentioned detection algorithms.

11

Introduction

Real data recordings from the Mobile Acoustic Communication Experiment(MACE 2010)

has been used to compare the performance of the algorithms and verify the similarity between

the underwater and in-air channel.

12

Chapter 2

Underwater Channel

2.1 Acoustics for Underwater Communication

There are a few means for wireless communication underwater. Radio waves of extra low fre-

quency (30 Hz–300 Hz) are the only waves that can propagate any distance in sea water. But

such low frequencies require large antennas and high transmission power. The other common

means is optical waves. Even though optical waves do no suffer much attenuation, they are

considerably affected by scattering and hence only lasers of extreme intensity can propagate

in water. Acoustics remains the best known solution for wireless underwater communication.

The underwater channel is one of the most challenging channels for reasons such as

frequency dependent attenuation, distance dependent bandwidth, and accentuated Doppler

effect which is non uniform in the signal bandwidth. This is in addition to the common

background noise which is predominant and non-negligible, is frequency dependent and site

dependent.

2.2 Acoustic Propagation

2.2.1 Attenuation

For a distance l and a frequency f , the attenuation in an underwater acoustic channel is

given as [5]

A(l, f) = A0lka(f)l (2.1)

where k is the spreading factor, which describes the geometry of propagation (typically

1.5 is used for practical spreading), and a(f) is the absorption coefficient. Equation 2.1 is

13

Underwater Channel

Figure 2.1: Absorption coefficient as a function of frequency.

expressed in dB:

10logA(l, f) = k · 10logl + l · 10loga(f) (2.2)

The absorption coefficient for frequencies above a few hundred Hz can be expressed

empirically, using the Thorp’s formula [5] which gives a(f) in dB/km for f in kHz as:

10loga(f) = 0.11f 2

1 + f 2+ 44

f 2

4100 + f 2+ 2.75 · 10−4f 2 + 0.003 (2.3)

For lower frequencies, the following formula may be used:

10loga(f) = 0.002 + 0.11f 2

1 + f 2+ 0.011f 2 (2.4)

The absorption coefficient is the major factor that limits the maximal usable frequency of

an underwater system. As it rapidly increases with frequency, the path loss will also increase

(see Figure 2.1), and therefore only the frequencies below a threshold may be used when

deploying an underwater communication link.

14

Underwater Channel

2.2.2 Noise

The ambient noise in the ocean can be modeled using four sources: turbulence, shipping,

waves, and thermal noise. Gaussian statistics and a continuous power spectral density de-

scribe the major sources or ambient noise. The following empirical formulae give the p.s.d.

of the four noise components in dB re µ Pa per Hz as a function of frequency in kHz [6]:

10logNt(f) = 17− 30logf

10logNs(f) = 40 + 20(s− 0.5) + 26logf − 60log(f + 0.03)

10logNw(f) = 50 + 7.5w1/2 + 20logf − 40log(f + 0.4)

10logNth(f) = −15 + 20logf (2.5)

Turbulence noise influences only the very low frequency region, f < 10 Hz. Noise caused

by distant shipping is dominant in the frequency region 10 Hz – 100 kHz, and it is modeled

through the shipping activity factor s, whose value ranges between 0 and 1 for low and high

activity, respectively.Surface motion, caused by wind-driven waves (w is the wind speed in

m/s) is the major factor contributing to the noise in the frequency region 100 Hz – 100 kHz

(which is the operating region used by the majority of acoustic systems). Thermal noise

becomes dominant for f > 100 kHz.

The overall power spectral density of the noise is given by

N(f) = Nt(f) +Ns(f) +Nw(f) +Nth(f) (2.6)

Figure 2.2 illustrates the power spectral density N(f) of the ambient noise for the case

of no wind (solid line) and wind at 10m/s with different shipping activity factors.

The noise increases at the low frequency range, thus limiting the useful acoustic bandwidth

from below. Due to the linear decayment of the noise p.s.d. on the logarithmic scale (in a

certain frequency range), the following approximation may then be useful:

10logN(f) ≈ N1 − ηlogf (2.7)

The approximation is shown in Figure 2.2 with N1 = 50dBreµPa and η = 18dB/decade.

2.2.3 Propagation Delay

The experienced delays in an underwater acoustic communication link are much higher than

in an open-air link. The nominal speed of sound in water is 1,500 m/s, which is much

lower than the speed of electromagnetic waves in air (3×108 m/s). This fact causes long

propagation delays, which becomes a major complication for the application of feedback to

15

Underwater Channel

Figure 2.2: Power spectral density of ambient noise N(f)[dBreµPa2]

correct for the channel distortions. As an example, typical propagation delays in acoustic

underwater links can be on the order of several seconds, while the measured coherence time in

an underwater channel can be on the order of milliseconds. In contrast with the propagation

delays in underwater channels, the open-air radio propagation delay is typically of the order

of microseconds.

2.2.4 Multipath

Multipath propagation is one of the common problems in communications through underwa-

ter acoustic links. This propagation phenomenon results in communication signals reaching

the receiving hydrophones by two or more paths. At the receiver, due to the presence of

multiple paths, more than one pulse will be received, and each one will arrive at different

times. The channel impulse response will be expressed by:

h(τ, t) =N∑

p=1

hp(t)δ(τ − τp(t)) (2.8)

where the channel taps, hp, arriving at τp, can be described by an amplitude component

ρp and a phase shift φp.

Underwater multipath can be caused either by reflection or refraction of the acoustic waves.

Reflection of the acoustic waves occurs when the wave bounces either at the surface or the

16

Underwater Channel

bottom and reaches the receiver. Refraction of the communication waves is a typical phe-

nomenon in deep water links, where the speed of sound changes at different depths.

The distortions caused by multipath must be equalized in the receiver. In addition, ways

to avoid intersymbol interference (ISI) must be designed in order to correctly demodulate

and detect the transmitted data. Later on in Chap. 4 a receiver algorithm for underwater

acoustic channels that deals with multipath effects will be explained.

2.2.5 Doppler Effect

The Doppler effect is caused by the relative motion of the transmitter-receiver pair, and it

causes a shift in the frequency components of the transmitted signal. The frequency shift is

mainly described by the factor vr/c, where vr is the relative velocity between transmitter and

receiver, and c is the signal propagation speed (the speed of sound underwater in this case).

In underwater environments c is approximately 1500m/s which is much lower in comparison

with radio communication where c is 3×108m/s. This gives a Doppler factor 5 times order of

magnitude greater than radio communication, and so the Doppler effect cannot be ignored.

In addition, the fact that underwater systems are wideband causes different Doppler shifts

for different frequency components of the transmitted signal. This is typically known as

frequency spreading.

It is of key importance that underwater acoustic systems deal with non-uniform Doppler

effect. As an example, a very high Doppler effect correction in a multicarrier system could

cause inter-carrier interference (ICI), which happens when some distortion due to other

subcarriers’ information is present in a selected channel [7].

2.3 Resource Allocation

Taking into account the physical models of acoustic propagation loss and ambient noise,

the optimal frequency allocation for communication signals can be calculated. Considering

optimal signal energy allocation, such frequency band is defined so that the channel capacity

is maximized [8; 9].

The results that are assessed suggest that, despite the fact that frequency spectrum is

not yet been regulated by the Federal Commission of Communications (FCC) for underwa-

ter acoustic communications, the possibilities in terms of usable frequency bands are not

numerous, due to acoustic path propagation and noise characteristics.

17

Underwater Channel

2.3.1 The AN Product and the SNR

The narrow-band signal to noise ratio (SNR) observed at a receiver over a distance l when

the transmitted signal is a tone of frequency f and power P is given by

SNR(l, f) =P (f)/A(l, f)

N(f)∆f=

S(f)

N(f)A(l, f)(2.9)

where ∆f is a narrow band around the frequency f , and S(f) is the power spectral density

of the transmitted communication signal. Directivity indices and losses other than the path

loss are not counted. The AN product, A(l, f)N(f), determines the frequency-dependent

part of the SNR. The inverse of the AN product is illustrated in Figure 2.3.

Figure 2.3: Frequency dependent part of SNR ∼ 1/A(l, f)N(f), k = 1.5 is used for A(l, f).

2.3.2 Optimal Frequency

Observing the inverse of the AN product, 1/A(l, f)N(f) in Figure 2.3 , it can be concluded

that for each distance l there exists an optimal frequency f0(l) for which the maximal narrow-

band SNR is obtained at the receiver. The optimal frequency is plotted in Figure 2.4 as a

function of transmitter receiver distance.

When implementing a communication system, some transmission bandwidth around

f0(l) is chosen. The transmission power is adjusted so as to achieve the desired SNR level

18

Underwater Channel

throughout the selected frequency band. Practically, the response of the transducers and

hydrophones must be taken into account and the optimal transmission frequency may vary.

Figure 2.4: Optimal frequency plotted as a function of distance.

2.3.3 Transmission Power

Once the transmission bandwidth is set, the transmission power P (l) can be adjusted to

achieve a desired narrow-band SNR level corresponding to the bandwidth B(l). If we de-

note by Sl(f) the p.s.d. of the transmitted signal chosen for the distance l, then the total

transmitted power is

P (l) =

∫

B3dB(l)

Sl(f)df = SNR0B(l)

∫

B3dB(l)N(f)df

∫

B3dB(l)A−1(l, f)df

(2.10)

where the transmitted signal p.s.d. is considered constant in the signal bandwidth.

19

Chapter 3

System Design

Multi-carrier modulation is an attractive alternative to single-carrier broadband modulation

on channels with frequency-selective distortion. Research in the area of underwater acoustic

communications over the past several years has resulted in demonstrating a different type of

bandwidth-efficient modulation and detection method, which uses multiple carriers instead

of a single carrier. In its basic form, this method is known as Orthogonal Frequency Division

Multiplexing (OFDM)[10; 11].

Rectangular pulse shaping combined with multi-carrier modulation and detection can be

easily implemented using the Fast Fourier transform, which enables easy channel equalization

in the frequency domain, thereby eliminating the need for potentially complex time-domain

equalization of a single-carrier system. For this reason OFDM has found application in a

number of systems, including the wire-line digital subscriber loops (DSL), wireless digital

audio and video broadcast (DAB, DVB) systems, and wireless LAN (IEEE 802.11)[3]. It

is also considered for the fourth generation cellular systems, and ultra-wideband (UWB)

wireless communications in general.

3.1 Basic OFDM Principle

The primary motive of transmitting the data on multiple carriers is to reduce intersymbol

interference and, thus, eliminate the performance degradation that occurs in single carrier

modulation. Multicarrier modulation is an approach to design a bandwidth efficient digital

communication system in the presence of channel distortion, by sub-dividing the available

channel bandwidth into a number of subchannels, such that each channel is nearly ideal.

Dividing the available channel bandwidth into subbands of relatively narrow width would

result in the channel transfer function being constant inside each subband, eliminating the

20

System Design

need for complex time-domain channel equalization.

OFDM is a frequency-division multiplexing scheme utilized as a digital multi-carrier

modulation method. A large number of closely-spaced orthogonal subcarriers are used to

carry data. The data is divided into several parallel data streams or channels, one for each

subcarrier. Each subcarrier is modulated with a conventional modulation scheme (such as

Quadrature Amplitude Modulation -QAM- or Phase Shift Keying -PSK-) at a low symbol

rate, maintaining total data rates similar to conventional single-carrier modulation schemes

in the same bandwidth. Figure 3.1 shows the utilization of the available bandwidth for a 7

sub-carrier OFDM signal.

Figure 3.1: Bandwidth utilization of OFDM.

3.1.1 Orthogonality

OFDM is a special type of multicarrier modulation in which the subcarriers of the corre-

sponding subchannels are mutually orthogonal. For a OFDM system with K subcarriers and

a total available bandwidth B, the subcarrier separation ∆f = B/K. If this subcarrier sep-

aration ∆f is small enough, the channel frequency response C(f) essentially constant across

each subband. With each subband a sinusoidal carrier signal is associated which is of the

form sk(t) = cos2πfkt, where fk is the center frequency of the kth subchannel. By selecting

the symbol rate 1/T in each of the subchannels to be equal to the frequency separation ∆f ,

the subcarriers are orthogonal over the symbol interval, independent of the relative phase

between the subcarriers [10]. That is,

∫ T

0

cos(2πfkt+ φk)cos(2πfjt+ φj) = 0; k 6= j (3.1)

21

System Design

One of the most important advantages of orthogonality between subcarriers is that it

allows high spectral efficiency, as almost the full available frequency band can be utilized.

A disadvantage that results from the use of orthogonality is the need for highly accurate

frequency synchronization between the transmitter and the receiver. The frequency devi-

ation that OFDM systems can tolerate is very small, as the subcarriers will no longer be

orthogonal, causing intercarrier interference, or cross-talk between subcarriers.

Frequency offsets are typically caused by Doppler shifts due to motion, or mismatched

transmitter and receiver oscillators. While Doppler shift alone may be compensated for by

the receiver, multipath arrivals with independent Doppler distortions makes the correction

more difficult.

3.1.2 Modulation Using Fast Fourier Transform (FFT)

Due to the orthogonality of OFDM subcarriers, the modulator and demodulator can be

efficiently implemented using the FFT algorithm on the receiver side, and the inverse FFT,

or IFFT, on the transmitter side. On the transmitter side, the IFFT of a signal U(k), where

k denotes the frequency component index, is

u(l) =1

K

K−1∑

k=0

U(k)ej2πkl/K , l = 0, ...K − 1 (3.2)

where K designates the number of frequency components, and u(l) is the resulting sam-

pled signal, which is formed by the sum of the modulated frequency components U(k) (at

their corresponding digital frequency k/K) (see Figure 3.2). To retrieve again the digital

frequency components, the inverse equation must be used:

U(k) =K−1∑

l=0

u(l)e−j2πkl/K , k = 0, ...K − 1 (3.3)

which corresponds to the K-point FFT of U(k).

3.1.3 Guard Time

One key principle of OFDM is that since low symbol rate modulation schemes (i.e., where

the symbols are relatively long compared to the channel time characteristics) suffer less from

intersymbol interference caused by multipath propagation, it is advantageous to transmit

a number of low-rate streams in parallel instead of a single high-rate stream. Since the

duration of each symbol is long, it is feasible to insert a guard interval between the OFDM

22

System Design

Figure 3.2: Transmitter implementation using IFFT.

symbols, thus eliminating the intersymbol interference. The guard interval also eliminates

the need for a pulse-shaping filter, and it reduces the sensitivity to time synchronization

problems [2].

In a cyclic prefix OFDM, the guard interval consists of the end of the OFDM symbol

copied into the guard interval. The reason that the guard interval consists of a copy of the

end of the OFDM symbol is because it allows the linear convolution of a frequency selective

multipath channel to be modeled as a circular convolution which in turn may be transfered

to the frequency domain using a discrete Fourier transform.

3.1.4 Equalization

The effects of the channel conditions, for example fading caused by multipath propagation,

can be considered as constant (flat) over an OFDM sub-channel if the sub-channel is suf-

ficiently narrow-band. This makes equalization far simpler at the receiver in OFDM in

comparison to conventional single-carrier modulation. The equalizer only has to multiply

each detected sub-carrier (each Fourier coefficient) by a constant complex number [12].

If a differential detection and differential modulation (such as DPSK or DQPSK) is ap-

plied to the subcarriers, equalization can be completely eliminated, since these non-coherent

schemes are insensitive to slowly changing amplitude and phase distortion.

3.1.5 Advantages

The main advantage of the OFDM modulation scheme in terms of practical implementation

is that it enables channel equalization in the frequency domain, thus eliminating the need

23

System Design

for potentially complex time-domain equalizers.

OFDM modulation techniques have been used both in wired and wireless systems due to

its advantages. Among them, the following must be mentioned:

• Simple and effective channel equalization in the frequency domain.

• High spectral efficiency.

• Robustness against inter-symbol interference and fading caused by the multipath chan-

nel.

• Efficient implementation using the FFT, avoiding the need for complex subchannel

filters.

• Easy equalization in the frequency domain.

3.1.6 Disadvantages

The major disadvantages of OFDM are:

• Sensitivity to frequency offsets.

• High Peak to Average Ratio (PAR), with a subsequent difficulty to optimize the trans-

mission power.

The major difficulty in applying OFDM to an underwater acoustic channel is the signal’s

sensitivity to frequency offsets, which imposes strict synchronization requirements. Motion-

induced Doppler effect in an acoustic channel creates a frequency offset that is not uniform

across the signal bandwidth. This fact is in stark contrast to the frequency distortion in radio

systems, and, hence, many of the existing synchronization methods cannot be used directly.

Instead, dedicated methods have to be designed. Such methods have been proposed over

the past several years, and demonstrated good performance in initial trials with real data

transmitted over a few kilometers at bit rates on the order of 10 kbps within comparable

acoustic bandwidths[3].

24

Chapter 4

OFDM Modulator and Demodulator

The system implementation has been discussed in this chapter. The transmitter set-up is

explained in detail followed by the receiver algorithms .

4.1 OFDM Transmitter

4.1.1 System Model

The transmitter system model consists of a typical OFDM system with K subcarriers. The

input data stream is serial-to-parallel converted into K streams dk(n) where k = 0, 1, ...K−1

[3]. This data stream is then used to generate the signal

uk(t) =∑

n

dk(n)ejk∆ω(t−nT ′)g(t− nT ′) (4.1)

where,

• The signal g(t) is a rectangular pulse in time with unit amplitude and duration T

• T ′ = T +Tg, where Tg is the guard interval, which is longer than the multipath spread

• ∆ω = 2π∆f , where ∆f = 1/T is the carrier spacing

The signals uk(t) are added and shifted in frequency to give the modulated signal

s(t) = Re{

K−1∑

k=0

uk(t)ejω0t} (4.2)

where

• f0 is the lowest carrier frequency

25


• fk = f0 + k∆f denoted the kth subcarrier frequency

• the symbol rate is R = K/(T + Tg)

• and the signal bandwidth is given as B = K∆f

The system block diagram is shown in Figure 4.1.

Figure 4.1: Transmitter block diagram

4.1.2 Transmitter Implementation

The different transmitter implementation are discussed in this section. The most important

parts of the transmitter include the symbol mapping, frequency interleaving, synchronization

and guard time and upconversion.

Scrambler

Scrambling or randomizing binary sequences eliminates the correlation between bits of data.

This process eliminates the occurance of continuous zeros or ones in the bit stream. After

the completion of scrambling, the sequence of bits will appear random with a similar amount

of zeros and ones.

Matrix interleaving is used to scramble the data bits. Matrix interleaving permutes bits

by filling a matrix by rows and emptying it by columns. The data bits are used to fill a

matrix row by row. The output of the interleaving is obtained by returning the contents of

the matrix column by column. This is shown in Figure 4.2.

26


Figure 4.2: Scrambling using matrix interleaving.

Forward Error Correction (FEC) Coding

The long packet delays due to the low speed of propagation in underwater environments

suggest the avoidance of packet retransmissions due to errors. Given the conditions of the

channel, the inclusion of Forward Error Correcting (FEC) codes is strongly necessary given

the conditions of the channel, as they provide error correction capability and avoid such

critical circumstances [13].

FEC coding is a system of error control for data transmission, whereby the transmitter

adds redundant data to its messages, so that a specific amount of random errors in the

received bit sequences can successfully be corrected by the decoder. To make best use of the

FEC coder, the possible error bursts due to momentary bad channel conditions or highly at-

tenuated frequency bands must be minimized, and that is why the system must also include

an interleaver block.

The channel coding field is a very active one, as there is a two-fold approach when de-

signing an appropriate coding technique: first, it has to provide the desired error correction

capability, and second, it has to provide enough coding and decoding speed so as not to

affect the transmission timing. There are different types of coders depending on the imple-

mentation:

1. Block codes: this family of codes work on fixed-size blocks (packets) of bits or symbols

of predetermined size. Practical block codes can generally be decoded in polynomial

time to their block length.

27


2. Convolutional codes: this family of codes work on bit or symbol streams of arbitrary

length. They are most often decoded with the Viterbi algorithm, though other algo-

rithms are sometimes used. Viterbi decoding allows asymptotically optimal decoding

efficiency with increasing length of the convolutional code, but at the expense of ex-

ponentially increasing complexity.

Some of the most popular FEC include the BCH (Bose, Ray-Chaudhuricodes and Hoc-

quenghem code), which are known by its decoding ease, the Low Density Parity Check

(LDPC) codes, which are highly-efficient block codes, and the Turbo codes, that combine

multiple convolutional codes and an interleaver to produce a block code.

The type of error correcting code (ECC) that has been chosen is the BCH code, due

to the following facts: its coding and decoding efficiency, the fact that it has been used in

previous successful OFDM system implementations for underwater environments, and the

implementation ease as it comes in a MATLAB built-in package.

The BCH codes is polynomial codes over a finite field with a particularly chosen gen-

erator polynomial from which the codewords are produced. These codes are multiple error

correcting codes and a generalization of the Hamming codes.

The codewords are formed by taking the remainder after dividing a polynomial represent-

ing the information bits by a generator polynomial. The generator polynomial is selected to

give the code its characteristics. All codewords are multiples of the generator polynomial.

Symbol Mapping

Phase shift keying has been found to be an suitable modulation method for this project.

Phase-shift keying (PSK) is a digital modulation scheme that conveys data by changing,

or modulating, the phase of a reference signal (the carrier wave). In particular Quadrature

Phase Shift Keying (QPSK), and 8PSK has been used. QPSK uses four points on the

constellation diagram, equispaced around a circle. With four phases, QPSK can encode

two bits per symbol, shown in the diagram with gray coding to minimize the bit error rate

(BER). Figure 4.3 shows the constellation of a QPSK signal.

Frequency Interleaving

Frequency interleaving is performed to optimize any channel coding technique that is used

by interleaving the mapped QPSK symbols across the K subcarriers. The advantage of

28


Figure 4.3: PSK constellations. QPSK on the left and 8PSK on the right.

such an interleaving is shown in Figure 4.4. As it can be seen clearly from the image,

frequency interleaving avoids the occurance of errors in the same code word of a channel

coding technique and thereby increasing the overall Bit Error Rate (BER) of the system.

Figure 4.4: Frequency interleaving to avoid errors.

Synchronization and Guard Time

Time synchronization in an OFDM system is extremely important in order to precisely

locate the starting of an OFDM block. In the current implementation for a proper informa-

29


tion retrieval, a pseudo-random sequence based synchronization preamble is used for time

synchronization. The synchronization preamble is shown in Figure 4.5 [12; 14; 15].

0 0.005 0.01 0.015 0.02 0.025−1

−0.8

−0.6

−0.4

−0.2

0

0.2

0.4

0.6

0.8

1

Time(s)

Val

ue

Synchronization Preamble

Figure 4.5: Synchronization preamble.

The synchronization preamble has very good auto correlation properties. The synchro-

nization preamble has an autocorrelation function that has a peak only at the middle. It

has zero correlation with shifter versions of its own sequence. This is clearly illustrated in

Figure 4.6.

Another important consideration for an OFDM signal is the guard time. This is impor-

tant because it has to absorb the multipath propagation sure to channel characteristics. Due

to the propagation delay there can be multiple arrivals even after 10ms of the arrival of the

main one. The ability to avoid ISI which is caused by the overlap of one symbol with the last

arrival of the previous symbol is determined by the guard interval Tg. Typically a guard time

of several milliseconds is chosen in order the overcome the coherence time of the underwater

channel which is approximately 1− 3ms. Figure 4.7 shows the transmitted signal with zero

padding in the guard interval. Figure 4.8 shows the received signal in which the multipath

propagation effects can be seen.

4.2 OFDM Receiver

The OFDM receiver block diagram is shown in Figure 4.9.

The receiving hardware records the OFDM signals with the specified sampling frequency.

This process is followed by a synchronization in time. The received synchronization preamble

30


0 500 1000 1500 2000 2500−400

−300

−200

−100

0

100

200

300

400

500

600

Samples

Val

ue

Autocorrelation

Figure 4.6: Autocorrelation of synchronization preamble.

Figure 4.7: Transmitted OFDM signal,

Figure 4.8: OFDM signal showing the multipath arrivals.

31


Figure 4.9: Receiver block diagram.

and its cross correlation are shown in Figure 4.10. Once the signal is properly synchronized

in time the downshifting in frequency is performed to bring the signal to the baseband. The

downshifting is performed after synchronization in time to retrieve the exact starting point

of the signal [16].

4.2.1 FFT Demodulation

At this stage, the OFDM blocks have been obtained. These OFDM blocks are then passed

through a FFT to get the raw constellation points.

The FFT algorithm is used to retrieve the subcarriers’ received symbols before chan-

nel treatment, by using the complementary method to the IFFT modulator. The applied

equation is [10]

y(k) =Ns−1∑

l=0

vr(l)e−j2πkl/Ns , k = 0, ..., Ns − 1 (4.3)

where Ns is the number of samples during a symbol time, and vr(l) is the received sam-

pled signal. An efficient implementation is possible by computing the Ns-point FFT of the

received signal vr(l).

As it has been previously mentioned the signal is oversampled by a factor γ, which means

that the retrieved signal after FFT demodulation includes the K subcarriers received sym-

bols plus (γ−1)K appended zeros (which may not be zeros at the receiver due to the channel

response and noise). That is why, to retrieve the useful part of the signal y(k), the first K

values are chosen and passed to the post-FFT detection algorithms.

32


Figure 4.10: Top figure shows the synchronization preamble, center figure shows the received

sequence, bottom figure shows the cross correlation.

At this stage the constellation points are scattered due to the noise and the channel effects.

The detection algorithm then performs the task of detecting the transmitted symbols.

4.2.2 Detection

As a part of this thesis, two detection techniques have been studied. The first technique

is based on the coherent detection which relies on the calculation of the channel and the

phase using an adaptive algorithm to neutralize the effects of the Doppler distortion and the

multipath channel. The second technique is the differentially coherent detection which takes

advantage of a differentially encoded data symbols among the subcarriers at the transmitter.

The differentially coherent detection is a better solution in conditions where accurate channel

estimations are not possible.

Coherent Detector

The coherent algorithm used to perform the detection of the received OFDM blocks is

explained in [3]. It is an adaptive algorithm for OFDM signal detection on Doppler-distorted,

time-varying multipath channels. It focuses on a low complexity post-FFT signal processing

33


Figure 4.11: Received constellation for a 8PSK sequence.

using adaptive channel estimation over the blocks. Non-uniform Doppler compensation

across subbands is performed using a single adaptively estimated parameter representing

the Doppler rate for each block.

Algorithm Description

After the application of the FFT algorithm, the received symbol values for each subcarrier

k in the block n, yk(n) is obtained. Once these values are assessed, the algorithm performs

the following steps.

1. Assessment of a first data estimate dk1(n) using MMSE combining of the different

receivers for each subcarrier k in the block n. To do so, it uses the phase offset θk(n−1)

and the channel frequency-domain coefficient Hk(n − 1) calculated in the previous

block, along with a normalization parameter γk(n − 1) according to the amplitude of

the channel coefficient. Figure 4.11 shows the received 8PSK constellation

dk1(n) = γk(n− 1)H′

k(n− 1)yk(n)e−jθk(n−1) (4.4)

2. Calculation of an angular offset ∆θk(n) between the assessed dk1(n), and either the

pilot symbols that were transmitted in the block, or a second estimate that corrects

the block-to-block motion by using the Doppler coefficient from the last block. The

performed steps are:

(a) (if no pilots are included) Retrieval of the estimate dk2(n), given by the following

equation:

34


dk2(n) = γk(n− 1)H′

k(n− 1)yk(n)e−jθk(n) = dk1(n)e

−ja(n−1)ωkT′

(4.5)

(b) Calculation of the reference for the angular offset measurement, dk(n). Con-

sidering that dk(n) are the transmitted symbols, and K(n) is the pool of pilot

subcarriers, it is given by

dk(n) =

{

dk(n) if k ∈ K(n)

decision[dk2(n)] otherwise(4.6)

(c) Finally, the angular offset is assessed by evaluating the following scalar product:

∆θk(n) = 〈dk1(n)d∗

k(n)〉 (4.7)

3. Assessment of the Doppler coefficient a(n) from the angular offsets ∆θk(n) by calculat-

ing the mean over the considered channels (being K the total number of subcarriers).

a(n) =1

K

K−1∑

k=0

∆θk(n)

ωkT ′(4.8)

Alternatively, the implementation can use only the pilot channels for phase tracking

purposes, so that the dk2(n) estimate do not have to be computed, and the previous

mean, a(n), is performed over the pilot subcarriers.

4. Calculation of the phase offset θk(n) for each subcarrier using the Doppler coefficient

a(n).

θk(n) = θk(n− 1) + a(n)ωkT′ (4.9)

5. Retrieval of a second more accurate data estimate, dk(n), using MMSE combining

of the different receivers for each subcarrier k in the block. This estimate includes

the correction of the possible tilt (See Figure 4.12) of the constellation due to the

motion-induced Doppler distortion.

dk(n) = γk(n− 1)H′

k(n− 1)yk(n)e−jθk(n) = dk1(n)e

−ja(n)ωkT′

(4.10)

6. The symbols are detected in blocks of multiple codewords so that the data bits can be

directly retrieved and some errors corrected.

35


Figure 4.12: Constellation phase tracking is performed.

dk(n) =

{

dk(n) if k ∈ K(n)

decision[dk(n)] otherwise(4.11)

7. Calculation of the channel estimates Hk(n) as explained in the next section.

8. Assessment of the normalization parameters γk(n) from the received values yk(n), the

assessed phase offsets θk(n), and the second set of data estimates dk(n). A high SNR

regime is supposed, so the noise variance σ2z is omitted.

γk(n) =(

σ2z + H′

k(n)Hk(n))

−1

≈(

H′

k(n)Hk(n))

−1

(4.12)

The algorithm is initialized using Hk(1) = yk(1)d∗

k(1), a(1) = 0 and θk(1) = 0. For a

correct performance, the receiver must know the transmitted symbols dk(1) to accurately

initialize the channel estimates, so a training sequence is transmitted in the first OFDM

block.

Channel Estimation Algorithm

A low-complexity channel estimation algorithm explained in [14] is used in the time domain.

Sparsing of the channel impulse response leads to an improved performance, reducing the

error variance. In this case, instead of implementing the complete algorithm, which includes

the selection of the number of samples L ≤ K of the impulse response of the channel, the

full-sized impulse response is used by choosing L = K.

36


The channel estimates to be used in the next block must be calculated after the assess-

ment of the detected symbols using the coherent detector algorithm. To do so, the channel

estimation [15] technique performs the following steps (experimental underwater channel

examples are shown in the referenced figures):

1. Calculation of the first frequency-domain channel estimates from the received values

yk(n), the phase offsets θk(n) and the set of data decisions and pilots dk(n)

Xk(n) = yk(n)e−jθk(n)d∗k(n), k = 0, ..., K − 1 (4.13)

2. Application of the IFFT algorithm to get the first time-domain channel estimates xl(n)

from Xk(n)

xl(n) =1

K

K−1∑

k=0

Xk(n)e+j2πkl/K , l = 0, ..., K − 1 (4.14)

3. Application of the sparsing by simply truncating the time-domain channel coefficients.

A threshold αc is defined so that every time-domain channel coefficient below this

value is set to zero. After the truncation is performed, the new time-domain channel

response xl(n) is taken.

4. Adaptive channel estimation to calculate the time-domain channel estimates applying

the following algorithm :

hl(n+ 1) = λhl(n) + (1− λ)xl(n) (4.15)

5. Application of the FFT algorithm to get the final frequency-domain channel estimates

used in the next block:

Hk(n) =K−1∑

l=0

hl(n)e−j2πkl/K , k = 0, ..., K − 1 (4.16)

Figure 4.13 and Figure 4.14 show the frequency response and time response of the

estimated channel.

Differentially Coherent Detection

Transmitter The differential detector algorithm requires differential encoding of the trans-

mitted constellation. That is, given the subcarrier mapped symbols for each block, bk(n),

37


Figure 4.13: Estimated channel frequency response.

Figure 4.14: Channel time-domain estimates.

38


the transmitter calculates the differentially encoded symbols that must be transmitted for

each subcarrier, dk(n), as

dk(n) = dk−1(n)bk(n), k = 0, ..., K − 1 (4.17)

where the first subcarrier symbol is initialized as d−1(n) = 1.

After the differential coding is performed, the dk(n) symbols are passed to the IFFT

modulator.

Receiver The receiver gets the subcarriers’ complex points in the OFDM blocks contained

in the received signal, once the FFT demodulation is performed. The received values, yk(n),

can be represented as

yk(n) = Hk(n)dk(n)ejθk(n) (4.18)

whereHk(n) represents the channel effect (a complex value for each block n and subcarrier

k, containing the amplitude attenuation and the phase shift), and θk(n) represents the phase

shift due to the motion-induced Doppler effect [4; 17].

As it can be easily perceived from the previous equation, if the assumption that the

channel estimates, Hk(n), and the phase estimates, θk(n), are similar between adjacent

subcarriers holds, the mapped symbols, bk(n), can be extracted by applying the following

operation:

yk(n)

yk−1(n)=

Hk(n)dk(n)ejθk(n)

Hk−1(n)dk−1(n)ejθk−1(n)≈

dk(n)

dk−1(n)= bk(n) (4.19)

A couple of last comments must be made about the differential detector regarding the

data rate of a differential detection transmission. The fact that there is one subcarrier in

each OFDM block used to initialize the algorithm directly implies that there is one subcarrier

unused in the OFDM transmission. In the practical implementation, there is no reduction

in data rate for the implemented system as a pilot is relocated on that first subcarrier to

avoid the loss in data rate.

Differential Detection with Partial FFT

Doppler distortion causes ICI which prevents the use of differentially coherent detection in

OFSM systems. Most of the approaches considered in literature are based of post-FFT

signal processing. In this thesis, an approach which is based [4] on pre-FFT processing is

researched. The goal is to exploit the information about the channel’s time-variation before

its lost in the process of demodulation. Several non-overlapping segments of an OFDM block

39


are combined in order to perform adaptive match filtering in an optimal manner [18]. In this

case, several FFTs would be required which leads to a small increase in the computational

complexity. While this process does not entirely eliminate the ICI, it sufficiently reduces ICI

so that differential detection can be preformed [19].

The received signal r(t) is down converted by the lowest carrier frequency f0 to obtain

v(t). While the conventional demodulator produces one output yk per carrier, a partial de-

modulator yields M outputs per carrier as shown in the Figure 4.15. Each FFT block is of

the same size as that of a conventional demodulator but it operates on a vector of windowed

signals. The M outputs can then be combined using equation 4.20

yk =M∑

m=1

p∗k,myk,m = p′

kyk (4.20)

Using equation 4.20, differential detection yeilds the decision variable

bk =p

′

kyk

p′

k−1yk−1

(4.21)

The combiner weights are adaptively obtained so as to minimize the MSE i.e E|e(k)|2

where,

ek = bk − bk (4.22)

The combiner weights can now be updated by using a stochastic gradient algorithm [20]

with a step size µ given by equation 4.23

pk+1 = pk + µgk (4.23)

where the gradient of the squared error is given by

gk =1

y2k−1

[yk.yk−1 − yk.yk−1] (4.24)

The results obtained by using these algorithms are discussed in the next chapter

4.3 Performance Analysis

The performance of the system is quantified by using the concepts of (1) Mean Square Error

(MSE) of the retrieved constellation points, and (2) Bit Error Rate (BER).

40


Figure 4.15: Block Diagram of Partial FFT.

4.3.1 MSE

The MSE is evaluated as the distance from the detected symbols (before a decision is made)

to the transmitted complex points in the constellation. Two types of MSE are evaluated,

one that indicates the performance depending on the frequency subcarriers (MSE-frequency),

and another that indicates the performance over the transmitted OFDM blocks (MSE-time)

[3].

The MSE-time calculation is evaluated over the K subcarriers, for each OFDM block, as

MSEt(n) =1

K

K−1∑

k=0

|dk(n)− dk(n)|2 (4.25)

And the MSE-frequency gives the mean square error over N received OFDM blocks, as

MSEf (k) =1

N

N∑

n=1

|dk(n)− dk(n)|2 (4.26)

The system performance can be then evaluated over time, while the subcarriers’ perfor-

mance can also be studied (normally those subcarriers that are received with more energy

will have lower MSE).

4.3.2 Bit and Symbol Error Rates

The Bit Error Rate (BER) is calculated as the total number of wrong bits divided by the

amount of transmitted bits. It is assessed for both the coded sequence (before the FEC

41


decoding is performed) and the decoded sequence of bits. This last result gives a direct

indication of whether the transmitted sequence is correctly retrieved or not.

42

Chapter 5

In-Air Acoustic Testing

As mentioned in Chapter 1, there is no widely accepted statistical model for an underwater

acoustic channel. The results obtained by software simulations thus do no provide full

justice to the actual channel. That leaves one with underwater testing as the only option for

testing the performance of various underwater communication techniques. Such underwater

experiments are not only time consuming but also expensive.

A mid way solution to this problem should provide better results than software simulation,

but should be less time consuming and expensive. An in-air acoustic test bed has been

proposed as a part of this thesis to provide for a more realistic, scaled version of the actual

underwater communication channel. The advantages of such a test bed would be

• Introduction of channel impairments such as motion-induced Doppler in a controlled

manner.

• Ability to reproduce the experiments easily.

• Build multi-node networks to test network protocols.

• Demonstration of proof-of-concept for various communication technologies.

• Online processing and adaptive feedback as opposed to offline processing of pre-recorded

experimental data

5.1 Hardware Used

EDIROL FA-101, an external sound card with built-in amplifier was found to be a perfect

match for this set-up.

43


5.1.1 EDIROL FA-101

Edirol FA-101 is a feature-packed 10x10 audio interface that can handle a full 10 channels of

input/output at 24-bit/96kHz. It can record and monitor all 10 channels simultaneously in

full duplex.In addition, the FA-101 offers 6-channel recording and playback at 24-bit/192kHz

for performance at DVD-A quality. Some of the key features of the Edirol FA-101 are

• 10x10 24-bit/96kHz audio performance

• 6x6 24-bit/192kHz audio performance

• Firewire bus-powered

• Direct monitoring

• Low-latency driver support

• 2x phantom-powered microphone preamplifier

• 8x8 balanced analog I/O

• S/PDIF optical I/O

• MIDI I/O

The Edirol-FA101 soundcard connects to the computer using a high speed IEEE 1394

port. Once correctly set-up, this device can be made to act as the primary soundcard for

a computer which gives the ease to use MATLAB for signal generation and processing.

Transmitter and receivers are discussed in the next section.

5.1.2 Transmitter

A conventional multimedia speaker (Figure 5.5)with a cut-off frequency of 16kHz (Figure

5.2) shows as the transmitter. The speaker is connected to the EDIROL FA-101 card using

a stereo-to-mono (1/8thto1/4thinch) connector. The speakers are in general not operated at

full power to avoid clipping by the speaker circuitry.

5.1.3 Receiver

A capacitive microphone that requires a 40V phantom power supply has been used as the

receiver. The microphone (Figure 5.3) connects to the EDIROL FA-101 sound card using

the XLR port that provides the 40V phantom power required.

44


Figure 5.1: Transmitter: conventional multimedia speaker

0 5 10 15 20

−60

−50

−40

−30

−20

−10

0

Frequency (kHz)

Mag

nitu

de (

dB)

Magnitude Response (dB)

Figure 5.2: Transmitter frequency response.

Figure 5.3: Capacitive microphone.

45


Figure 5.4: Deployment in lab.

Figure 5.5: Transmitter set-up.

5.1.4 Laptop

Dell Latitude E6400 Laptop computer running MATLAB has been used to drive the EDIROL

FA-101. All signal processing operations takes place on these laptop computers. The advan-

tage of such a system is that the laptop computers are used as all purpose machines with

fast processors.

5.2 Deployment

The entire hardware set-up has been deployed in the the laboratory as shown in Figure 5.4.

The transmitter set-up is shown in Figure 5.5

The receiver set-up is shown in Figure 5.6

46


Figure 5.6: Receiver set-up.

5.2.1 OFDM Parameters

The usable signal bandwidth is limited by the sampling frequency of the external sound card

and the cut off frequency of the transmitter and receiver. Typically a sampling frequency of

44.1 kHz has been used with an oversampling rate of 4. The OFDM parameters that have

been used in performing the tests have been tabulated below.

No of subcarriers FFT size Carrier spacing OFDM symbol duration

K Ns = 4K ∆f = B/K(kHz) T = 1/∆f(ms)

512 2048 21.55 46

1024 4096 10.67 92.9

2048 8192 5.383 182.8

Table 5.1: OFDM signal parameters

5.2.2 Results

As mentioned earlier in this chapter, one of the advantages of having an in-air testbed is the

ability to reproduce tests precisely for easy comparison. To study the performance of the

receiver algorithms, a number of tests were performed.

Static Test

This test was mainly carried out to test the performance of the algorithm and to check the

channel estimation algorithm. The channel estimates are shown in Figure 5.7 corresponding

to the deployment of the system in the laboratory shown in Figure 5.4.

The different multipath arrivals can be seen in the channel estimates. The main arrival is

observed at 0ms followed by a second arrival at 2.18ms and a third arrival at 5.71ms. These

times have been converted to distances to check if the arrivals correspond with actual real

time channel and shown in Table. 5.2.

47


Figure 5.7: Channel time domain estimates.

Arrival Time Path Distance Correspondence to actual channel

2.18 0.74 Reflection off wall

5.71 1.94 Reflection off ceiling

Table 5.2: Actual channel calculations.

The speed of sound in water is approximately 5 times the speed of sound in air. This

means the multipath arrivals reach faster in water than in air. This would correspondingly

mean a guard time that is 5 times shorter than in-air, would give the same performance in

water.

Linear Motion

In this test, the receiver microphone was slowly moved in various directions. This test induces

Doppler distortion in the received signal. Figure 5.8 shows the received constellation before

and after detection. It also has the plot for the phase estimates corresponding to k = 0,

k = 511 and k = 1023. The Doppler factor estimates a is also plotted. The relative

speed between the transmitter and receiver is 10cm/s, i.e the receiver is moved towards the

transmitter and a steady speed of 10m/s The motion of the receiver can be clearly seen in

the phase estimates. For each sub-carrier is different because they suffer from different phase

offsets due to the wideband of the signal.

The channel estimates calculated during the first block of transmission is shown in Figure

5.10

The bit-error rate plot for each block is shown in Figure 5.9

To-and-Fro Motion

In this test a moderated motion involving the movement of the receiver all the way close to the

receiver and bringing it to a complete stop followed by moving it back to its original position.

The phase estimates of this test are shown in Figure 5.11. It can be clearly ascertained

48


−1 −0.5 0 0.5 1−1

−0.5

0

0.5

1Constellation before detection (until block 2)

Real value

Imag

inar

y va

lue

−1 −0.5 0 0.5 1−1

−0.5

0

0.5

1Constellation after detection (until block 2)

Real value

Imag

inar

y va

lue

0 20 40 60 80 100 120 140 160−3

−2.5

−2

−1.5

−1

−0.5

0

0.5Phase estimate

Block index

Val

ue

0 20 40 60 80 100 120 140 160−8

−6

−4

−2

0

2

4

6

8x 10

−5 a−estimate

Block index

Val

ue

Figure 5.8: Plots of received constellation after detection, phase estimates and Doppler factor

estimates.

Figure 5.9: Bit error rate with coherent detection K = 1024QPSK.

49


0 2 4 6 8 10 12 140

0.2

0.4

0.6

0.8

1H(f) estimates for first block (abs.)

Frequency (kHz)

Val

ue

−5 0 5 10 15 20 25 30 35 40 450

0.2

0.4

0.6

0.8

1h(t) estimates for first block (abs.)

Time (msec)

Val

ue

Figure 5.10: Channel estimates for first block, frequency and time

50


Figure 5.11: Phase estimates for to-and-fro motion.

Figure 5.12: Bit error rate of received sequence.

from the figure that the phase tracking algorithm has tracked the phase perfectly. As the

microphone is moved towards the speaker a clear increase in the phase is seen followed by a

clear decrease in phase due to the backwards motion.

The BER plot for this test by using the differential detection algorithm is shown in Figure

5.12 and Figure 5.13. An increase in error is seen at points with high phase gradients or in

other words at points where the acceleration is high.

Fast Motion

In this test, the receiver was kept static initially. This was followed by a quick motion from

left to right and back to the original location followed by a static period. The importance of

51


Figure 5.13: BER of decoded sequence with errors reduced.

Figure 5.14: Phase estimates for fast motion.

this test is that it provides a means for testing the algorithm in a fast varying channel. The

phase estimates for this test are shown in Figure 5.14

As expected, the performance of the coherent detection algorithm which relies on accurate

channel estimation is found to be sub-optimal. The errors start to propagate and hence leads

to a very high BER (Figure 5.15 and Figure 5.16 )

The differential detection algorithm outperforms the coherent detection algorithm in this

case as it does not rely on channel estimation (Figure 5.17. The differential detection

algorithm does not have any block-by-block dependency and hence we can see that there is

no error propagation.

52


Figure 5.15: BER with differential detection for fast to-and-fro motion.

Figure 5.16: BER using coherent detection for fast motion.

53


Figure 5.17: BER using differential detection for fast motion.

Results of Differentially Coherent Detection with Partial FFT

In this test, the ICI reduction capabilities of the partial FFT algorithm has been tested.

Doppler distortion has been induced in the signal by moving the transmitter towards the

receiver. Figure 5.18 shows the mean square error and the bit error rate plot for this exper-

iment.

Performance of the partial FFT withM = 2 andM = 4 has been compared to differential

detection with no partial FFT (i.e M = 1). As expected, when the signal suffered from

Doppler Distortion, the MSE using partial FFT demodulator is at least 3dB lower than

conventional demodulator output. The overall BER improvement by the using of partial

FFT can also be observed from Figure 5.18

Receiver Diversity

The last set of in-air experiments aim to demonstrate the performance improvement with the

use of multiple receiving elements, i.e by receiver diversity. In the practical implementation,

another microphone is used to receive the same OFDM signal, and an MMSE combining of

the signal is performed in the detector.

Two sets of results are evaluated. The first is for static conditions (Figure 5.19), where an

improvement of around 50% in the coded BER is observed. The second set of results (Figure

5.20) corresponds to the fast motion case, and a BER reduction for the multiple receiving

elements case is also observed even with extreme detection conditions (the high BER peaks

are due to the high acceleration). The MSE curves are also lower when spatial diversity is

used, representing a more accurate retrieval of the constellation. It is worth mentioning that

54


(a) MSE

1 1.5 2 2.5 3 3.5 410

−3

10−2

10−1

M

BE

R

(b) BER

Figure 5.18: Differentially coherent results using Partial FFT.

55


performance improvement would be observed only when the distance between the receiving

elements is atleast a integer multiple of the wavelenght of the signal

These tests give us a feel of what to expect in the underwater channel. These tests have

been conducted in a controlled manner so as to facilitate very easy reproduction of these

test at a future time.

56


(a) MSE

(b) BER

Figure 5.19: Receiver diversity using two receivers. No motion.

57


(a) MSE

(b) BER

Figure 5.20: Receiver diversity using two receivers. Fast Motion.

58

Chapter 6

At Sea Testing

In this chapter, the layout and deployment of the Mobile Acoustic Communication Experi-

ment (MACE) conducted in July 2010, is discussed.

6.1 Deployment

The experiment was carried out in the Atlantic Ocean about 100 miles south of Cape Cod.

The focus of this experiment was on mobile multi input multi output (MIMO) systems.

Receivers were deployed in the sea using anchors and floats while the transmitter was being

towed by a ship in a pre-determined track. The system deployment is discussed in the fol-

lowing sections.

Figure 6.1 shows a representation of the actual system deployment. The ship was moving

continuously around the loop at speeds of 1m/s and 2m/s. B1,B2 show the location of the

receiver buoys and M1,M2 show the location of the receiver moorings. These transmitter

and receiver arrangements are further explain in the following sections

6.1.1 Transmitter

The transmitter used in this experiment was a 4 element ITC−1007 transducer array. This

array was mounted on a V-fin which was towed by the Oceanus (Research Vessel owned and

operated by the Woods Hole Oceanographic Institute). During the course of the experiment,

two different orientations of the transducer array were used. During the first tow, array were

oriented horizontally, and in the second tow, it was oriented vertically (Figure 6.2). The

experiment consisted of transmitting different sets of signals continuously for the duration

59

At Sea Testing

Figure 6.1: System deployment during MACE 2010.

of a tow. During the duration of a tow, the ship moved from the minimum range out to the

maximum range and back to the minimum range.

Figure 6.2: Left: Horizontal transducer orientation. Right: Right vertical transducer orien-

tation.

6.1.2 Receiver

Two different types of receivers were used in this experiment.

1. The two receiver buoys (marked by B1 and B2 in Figure 6.1) had a 4 element 1m

array. The receiver elements were mounted on a casing and deployed using a float

60

At Sea Testing

to make sure it stayed vertical. The receiver buoys had a processor on board which

could be remotely accessed by a laptop computer on-board the ship using an RF

connection. The receiver buoys used a Advantech single board computer connected to

National Instruments USB-6259 DAQ. DAQ was connected to output preamplifier of

WHOI Micromodmem Multi-Channel Analog Interface. The DAQ was operated using

the Data Acquisition toolbox of MATLAB. Since the batteries powering the receiver

buoys did not have sufficient power to last the entire duration of the experiment, they

had to be switched after successful completion of each orientation of the transducer

array.

2. The two receiver moorings (marked by B1 and B2 in Figure 6.1) had a 12 element

hydrophone array. The moorings were anchored to the ocean floor using 500lbs anchors

which made sure the set-up stayed vertical. The moorings set-up had sufficient battery

life to last the entire duration of the experiment and hence did not need switching.

6.2 Results

The experiment was carried out for a period of three days. The transmitted signal of interest

was an OFDM signal with the parameters shown in Table 6.1.

No of subcarriers , FFT size Sampling Frequency fs Lowest carrier frequency f0

K = 128, 256, 512, 1024, 2048 Ns = 8K 39063Hz 10580Hz

Table 6.1: OFDM signal parameters for MACE 2010.

Figure 6.3 shows a plot of the original transmitted OFDM signal and its frequency

spectrum.

The signal received at the receiver buoy B1 has been used for testing the performance

of differentially coherent detection. Figure 6.4 shows a plot of the received signal and its

frequency spectrum.

The signal received at the buoy is scaled due to the relative motion between the trans-

mitter and the receiver. Received signals were resampled manually before applying to the

algorithm so that the time scaling due to Doppler is compensated. The time synchronization

using the probe PN sequence as described in Chapter 3 is shown in Figure 6.5

The plot of the received constellation for K = 256 and K = 1024 is shown in Figure 6.6

The plot of the constellation after differential detection for K = 256 and K = 1024 is

shown in Figure 6.7

61

At Sea Testing

Figure 6.3: Transmitted signal and frequency spectrum.

62

At Sea Testing

Figure 6.4: Received signal at Buoy and its frequency spectrum.

63

At Sea Testing

0 200 400 600 800 1000 1200−1

−0.5

0

0.5

1

Samples

Mag

nitu

de

Transmitter Probe

0 200 400 600 800 1000 1200−0.2

−0.1

0

0.1

0.2

Samples

Mag

nitu

de

Received Probe

0 500 1000 1500 2000 2500−10

−5

0

5

10

Samples

Mag

nitu

de

Cross Correlation

Figure 6.5: Synchronization of preamble.

−3 −2 −1 0 1 2

−3

−2

−1

0

1

2

3

Qua

drat

ure

In−Phase

Scatter plot

(a) K = 256

−3 −2 −1 0 1 2 3

−3

−2

−1

0

1

2

Qua

drat

ure

In−Phase

Scatter plot

(b) K = 1024

Figure 6.6: Received constellation.

64

At Sea Testing

−1.5 −1 −0.5 0 0.5 1 1.5

−1.5

−1

−0.5

0

0.5

1

1.5

Qua

drat

ure

In−Phase

Scatter plot

(a) K = 256

−1 −0.5 0 0.5 1 1.5

−1

−0.5

0

0.5

1

1.5

Qua

drat

ure

In−Phase

Scatter plot

(b) K = 1024

Figure 6.7: Constellation after Differential Detection.

5 10 15 20 25 3010

−2

10−1

100

Block Index

BE

R

(a) K = 256

1 2 3 4 5 6 7 810

−2

10−1

100

Block Index

BE

R

(b) K = 1024

Figure 6.8: BER for K=256 and K=1024.

65

At Sea Testing

The plot of the BER for K = 256 and K = 1024 is shown in Figure 6.8

The plot of the MSE in time for K = 256 and K = 1024 is shown in Figure 6.9

5 10 15 20 25 30−10

−8

−6

−4

−2

0

2

4

6

8

10

Block Index

MS

E(d

B)

(a) K = 256

1 2 3 4 5 6 7 8−10

−8

−6

−4

−2

0

2

4

6

8

10

Block Index

MS

E(d

B)

(b) K = 1024

Figure 6.9: MSE for K=256 and K=1024.

66

Chapter 7

Online Processing

In this chapter, two main extensions are considered for the testbed. The first application is

an adaptive power control for OFDM and the second application is a text messaging system

to demonstrate the networking capabilities of the testbed.

7.1 Online Processing

All the experiments that have been conducted previously were based on offline processing.

Offline processing means the signals were transmitter, recorded at the receiver and stored on

the processing computer. These signals were then processed at a later time using MATLAB.

Although offline processing was beneficial for testing of various communication algo-

rithms, it cannot be used for actual implementations. The signals have to be processed as

and when they are received to make necessary alterations. Hence online processing has been

explored and tested with power control applications. This would also require some minor

changes to the transmitter receiver set-up previously described.

7.2 Transceivers

The transmitter/receiver set-up described in Chapter 5 has been modified so that each can be

used as a transceiver. A capacitive microphone has been added to the transmitter a speaker

has been added to the receiver to make each of then a transceiver as shown in Figure 7.1

The transceiver set-up has been used in performing adaptive power control and text

messaging.

67

Online Processing

Figure 7.1: Transceiver set-up.

68

Online Processing

7.3 Power Control

OFDM signals suffer from frequency selective fading and channel variation. To partially

offset for these distortions, adaptive power control based on receiver feedback has been

explored.

7.3.1 System Model

The transmitter has been modified to transmit signals at four discrete power levels based on

receiver feedback. At the receiver, once the signal has been recorded and demodulated, the

SNR is calculated using the eqn 7.1.

SNRk = H2k/Nk (7.1)

Figure 7.2: Block diagram of power control algorithm.

The receiver algorithm is shown in Figure 7.2. Once the SNR is calculated, it is modu-

lated using OFDM and sent back to the transmitter on the same channel. After obtaining

the receiver SNR, the transmitter compares the SNR value with a pre-defined threshold.

If the receiver SNR is lower than the threshold, the signal power is increased to the next

discrete power level. If the receiver SNR is 2dB higher than the threshold, the signal power

is decreased to the next lower power lever. Signal power is maintained at the same level, if

the SNR is within 2dB of the threshold.

69

Online Processing

7.3.2 Results

Two different tests were performed to to test the adaptive power control algorithm; static

and mobile. To begin with, a study of the received SNR at various distances is performed.

The plot of receiver SNR vs distance is shown in Figure 7.3.

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.816

17

18

19

20

21

22

23

24

Distance(m)

SN

R (

dB)

Figure 7.3: Measured SNR at receives plotted against distance.

Static Test

Figure 7.3.2 shows a comparison between the power level chosen and the corresponding

received SNR. The algorithm always starts with the lowest power lever, 0.3× fullpower in

this case. OFDM signals with QPSK mapping and 1024 subcarriers were used at a sampling

frequency of 44.1kHz. After the first transmission, the received SNR that was fed back to the

transmitter was 15.3 dB. The transmitter runs the power control algorithm and increases the

power in subsequent transmission until the required SNR of 20dB is achieved. The required

SNR was achieved at a power lever of 0.75× fullpower.

70

Online Processing

1 2 3 4 5 6 7 8 9 1014

16

18

20

22

Rec

eive

d S

NR

Transmission Number

1 2 3 4 5 6 7 8 9 100.2

0.4

0.6

0.8

1

Transmission Number

Pow

er L

evel

Figure 7.4: Case: static. Top: plot of power level for each transmission. Bottom: received

SNR for the transmission for the given power level.

71

Online Processing

Mobile Test

Figure 7.5 shows a comparison between the power level and the corresponding receiver

SNR. In this case, the receiver was static initially and then moved towards the transmitter.

It is shown in the first part of the plot when the SNR is increasing. During transmission of

packet 4, the receiver was moved towards the transmitter which resulted in a higher SNR. In

transmission 6 the receiver was moved away from the transmitter which resulted in a large

drop in the SNR. This drop was compensated by the transmitter in the next transmission

by increasing the power level to the next available power level.

1 2 3 4 5 6 7 8 9 100.2

0.4

0.6

0.8

1

Transmission Number

Pow

er L

evel

1 2 3 4 5 6 7 8 9 1015

20

25

Transmission Number

Rec

eive

r S

NR

Figure 7.5: Case: Motion. Top: plot of power level for each transmission. Bottom: Received

SNR for the transmission for the given power level.

These preliminary tests show the operation of what is called the ‘delayed feedback’ adap-

tive power control. These test form the basis for performing better adaptive power control

by using estimated values of receiver SNR, at some later stage.

72

Online Processing

7.4 Text Messaging

Text messaging has been used as an application to demonstrate the working of the testbed.

Two laptop computers communicate with each other using the testbed, employing OFDM

with differentially coherent detection. The OFDM parameters used for this system are shown

in Table 7.1

No of subcarriers FFT size Sampling frequency Bandwidth Symbol Mapping

K Ns = 4K fs[kHz] B[kHz] QPSK/8PSK

4096 16384 44.1 11.025 QPSK

Table 7.1: OFDM signal parameters for text messaging.

7.4.1 Network Protocol

A token ring network protocol has been used to implement the text messaging system. In

this network protocol, any node that has a special packet called the ’token’ is allowed to

transmit while all the other nodes are in the listening mode. Figure 7.6 shows a typical

token ring network with 5 nodes.

Figure 7.6: Token ring network protocol.

Token ring local area network (LAN) technology is a local area network protocol which

resides at the data link layer (DLL) of the OSI model. It uses a special frame called a

token that travels around the ring. Token possession grants the possessor the permission

to transmit on the medium. Token ring frames travel completely around the loop.Stations

73

Online Processing

on a token ring LAN are logically organized in a ring topology with data being transmitted

sequentially from each station with a control token circulating around the ring controlling

access. Each station passes or repeats the special token frame around the ring to its nearest

downstream neighbor. This token passing process is used to arbitrate access to the shared

ring medium. Stations that have data frames to transmit must first acquire the token before

they can transmit.

It has to be emphasized that we are dealing with acoustics which have a long propagation

delay. This means that the network should be timed in such a way that the system not go out

of synchronization. For this purpose, the clocks on the all the nodes have been synchronized

to the second. By doing this, every node is forced to either listen or transmit every fifth

second.

7.4.2 Results

Two computers were used in the experiment. They were able to communicate successfully

with each other using the network protocol. Text messaging system based on instant mes-

senger was implemented. The exchange of messages between the two computers is shown in

Figure 7.7

Figure 7.7: Text messaging.

74

Chapter 8

Future Work and Conclusion

OFDM has been researched for application to high speed wireless underwater communica-

tion channel, which is characterized by its frequency selectivity and motion-induced Doppler

distortion.

OFDM implementation using IFFT/FFT pair has been studied and implemented in

MATLAB. Two different detection techniques; one based on coherent detection and another

based on differentially coherent detection has been researched and tested experimentally.

The intention of building a testbed to test the performance of various communication

technologies has been achieved. The testbed not only saved time and cost, but also proved to

be the best system to easily induce various channel distortions like motion-induced Doppler

distortion.

The effects of motion and distance on system performance was tested using the testbed.

As expected, coherent detection relied heavily on accurate channel estimation and was not

ideal in cases where channel variation was very fast (like the fast motion). Differentially

coherent detection, which takes advantage of the differentially encoded data was found to

out perform coherent detection in such condition. Differential detection applied to real data

from MACE experiment shows performance similar to the testbed set-up.

Future work will address:

• Adaptive Power control can be extended to run an estimation algorithm at the trans-

mitter to predict the receiver SNR to improve performance. The disadvantage of a

delayed feedback system would be channel variation before the SNR is received.

75

Future Work and Conclusion

• Adaptive bit loading and adaptive modulation can be used at the transmitter to offset

the channel variations.

• The knowledge of the SNR at the receiver can be used by the receiver to work in

conjugation with control system to move the SNR if it is in a ’black hole’. The control

system can make use of the SNR to physically move the receiver in the direction of

increasing SNR.

• Multiple nodes can be introduced to test network protocols. The text messaging ap-

plication will be extended for iDiver, a multi-node communication system for divers.

76

Bibliography

[1] W Konrad and A. Quazi, “Underwater acoustic communications,” IEEE Com-

munication Magazine, 1982.

[2] B. Muquet, Z. Wang, G.B. Giannakis, M. de Courville, and P. Duhamel, “Cyclic

prefixing or zero padding for wireless multicarrier transmissions,” Communica-

tions, IEEE Transactions on, vol. 50, no. 12, pp. 2136 – 2148, Dec. 2002.

[3] M. Stojanovic, “Low complexity OFDM detector for underwater acoustic chan-

nels,” in OCEANS 2006, 2006, pp. 1 –6.

[4] M. Stojanovic, “A method for differentially coherent detection of OFDM signals on

doppler-distorted channels,” in Sensor Array and Multichannel Signal Processing

Workshop (SAM), 2010 IEEE, 2010, pp. 85 –88.

[5] L. Berkhovskikh and Y. Lysanov, Fundamentals of ocean acoustics, Springer, New

York, 1982.

[6] R. Coates, Underwater acoustic systems, Wiley, New York, 1989.

[7] .T Wang, J.G. Proakis, and J.R. Zeidler, “Techniques for suppression of intercar-

rier interference in OFDM systems,” in Wireless Communications and Networking

Conference, 2005 IEEE, 2005, vol. 1, pp. 39 – 44 Vol. 1.

[8] D.E. Lucani, M. Stojanovic, and M. Medard, “On the relationship between trans-

mission power and capacity of an underwater acoustic communication channel,”

in OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean, 2008, pp. 1 –6.

[9] M. Stojanovic, “Capacity of a relay acoustic channel,” in OCEANS 2007, 29 2007.

[10] J.G. Proakis, Digital Communications, McGraw-Hill Series in Electrical and

Computer Engineering, New York, 2001.

77

BIBLIOGRAPHY

[11] B. Li, S. Zhou, M. Stojanovic, L. Freitag, and P. Willett, “Multicarrier com-

munication over underwater acoustic channels with nonuniform doppler shifts,”

Oceanic Engineering, IEEE Journal of, vol. 33, no. 2, pp. 198 –209, 2008.

[12] M. Stojanovic, J.A. Catipovic, and J.G. Proakis, “Phase-coherent digital commu-

nications for underwater acoustic channels,” Oceanic Engineering, IEEE Journal

of, vol. 19, no. 1, pp. 100 –111, Jan. 1994.

[13] H. Wallace, “Error detection and correction using BCH codes,” .

[14] M. Stojanovic, “OFDM for underwater acoustic communications: Adaptive syn-

chronization and sparse channel estimation,” in Acoustics, Speech and Signal

Processing, 2008. ICASSP 2008. IEEE International Conference on, 31 2008.

[15] M. Stojanovic, “MIMO OFDM over underwater acoustic channels,” in Signals,

Systems and Computers, 2009 Conference Record of the Forty-Third Asilomar

Conference on, 2009, pp. 605 –609.

[16] G. Palou and M. Stojanovic, “Underwater acoustic MIMO OFDM: An experi-

mental analysis,” in OCEANS 2009, MTS/IEEE Biloxi - Marine Technology for

Our Future: Global and Local Challenges, 2009, pp. 1 –8.

[17] .S Lu and N. Al-Dhahir, “Coherent and differential ICI cancellation for mobile

ofdm with application to dvb-h,” Wireless Communications, IEEE Transactions

on, vol. 7, no. 11, pp. 4110 –4116, 2008.

[18] S. Yerramalli, M. Stojanovic, and U. Mitra, “Partial fft demodulation: A detection

method for doppler distorted OFDM systems,” June, International Workshop on

Signal Processing Advances in Wireless Communications (SPAWC).

[19] S. Yerramalli, M. Stojanovic, and U. Mitra, “Data detection techniques for OFDM

signals over doppler-distorted channels,” in Fifth ACM International Workshop

on Underwater Networks (WUWNeT’10), woods Hole, MA, 2010.

[20] S. Haykin, Adaptive Signal Processing, Prentice-Hall, Inc., New Jersey, 1996.

[21] X. Li and Jr. Cimini, L.J., “Effects of clipping and filtering on the performance of

OFDM,” in Vehicular Technology Conference, 1997 IEEE 47th, May 1997, vol. 3,

pp. 1634 –1638 vol.3.

[22] B. Li, S. Zhou, M. Stojanovic, L. Freitag, and P. Willett, “Multicarrier com-

munication over underwater acoustic channels with nonuniform doppler shifts,”

Oceanic Engineering, IEEE Journal of, vol. 33, no. 2, pp. 198 –209, 2008.

78

BIBLIOGRAPHY

[23] M. Stojanovic, “An adaptive algorithm for differentially coherent detection in the

presence of intersymbol interference,” Selected Areas in Communications, IEEE

Journal on, vol. 23, no. 9, pp. 1884 – 1890, 2005.

[24] K. Sathananthan, R.M.A.P. Rajatheva, and S.B. Slimane, “Cancellation technique

to reduce intercarrier interference in OFDM,” Electronics Letters, vol. 36, no. 25,

pp. 2078 –2079, Dec. 2000.

[25] M. Speth, S. Fechtel, G. Fock, and H. Meyr, “Optimum receiver design for OFDM-

based broadband transmission .II. a case study,” Communications, IEEE Trans-

actions on, vol. 49, no. 4, pp. 571 –578, 2001.

[26] A.-B. Salberg and A. Swami, “Doppler and frequency-offset synchronization in

wideband OFDM,” Wireless Communications, IEEE Transactions on, vol. 4, no.

6, pp. 2870 – 2881, 2005.

[27] M. Suzuki and T. Sasaki, “Digital acoustic image transmission system for deep

sea research submersible,” in Proceedings IEEE 92 Conference, 1992.

79

An experimental study of OFDM in a software defined ...1322/fulltext.pdf · The second technique is...

Documents

Transcript of An experimental study of OFDM in a software defined ...1322/fulltext.pdf · The second technique is...