DIGITAL COMMUNICATIONS LEARNING TOOLS

gtTAL: Graphical Tool for Testbed-assisted Learning

Jos

e A. Garc

ıa-Naya, H

ector J. P

erez-Iglesias, Adriana Dapena and Miguel Gonz

alez-L

opez

Department of Electronics and Systems, University of A Coru

na, Spain

Keywords:

Digital communications, Multiple-antenna testbed, Testbed-assisted learning.

Abstract:

We introduce gtTAL, a graphical tool implemented on top of a distributed multilayer architecture which is

speciﬁcally suitable for multiple-antenna hardware testbeds. gtTAL helps in teaching digital communications

by allowing interaction with the hardware testbed at an abstraction level suitable for undergraduate students.

Instead of using the low-level interfaces provided by hardware manufacturers, the multilayer software archi-

tecture supplies a high level interface access for testbeds, releasing students from the necessity of knowing

low-level details of the hardware to start to practice with it. Therefore, they can easily test algorithms without

developing a new program from scratch, speeding up the time needed for both the implementation and the

debugging tasks. Indeed, the multilayer software architecture allows learning how to deal with real-world dig-

ital communication systems at different abstraction levels, varying from the lowest level software running in

real-time in DSPs or FPGAs, to the highest level software like gtTAL. These three elements: hardware testbed,

multilayer software architecture and graphical tool (gtTAL), constitutes what we termed testbed-assisted learn-

ing.

1 MOTIVATION

In wireless digital communication courses is common

to use computer simulations to illustrate the theoret-

ical concepts presented to the students. Unless very

complex simulations are carried out, the most typ-

ical simulations in graduate courses show the cor-

responding results under controlled and ideal condi-

tions. Thus, very important effects introduced by

hardware elements are often ignored, like for exam-

ple those caused by the antennas, by the D/A and

A/D converters or by the radio-frequency (RF) am-

pliﬁers. Some of these effects can only be well under-

stood if the students experience the problem by them-

selves. For this reason, computer simulations are only

useful as an starting point in understanding the key

concepts of modern wireless digital communications.

However, computer simulations do not allow to study

and understand very important implementation issues

present in real-world transceivers.

There is a variety of multi-antenna testbeds that

provides many educational opportunities (Rao et al.,

2004), allowing the students to learn, step by step,

all signal processing stages involved in the generation

of the signals to be transmitted through the antennas.

This transmit signal processing chain is very similar

to that usually found in computer simulations. How-

ever, when the students have to implement the corre-

sponding signal processing blocks of a real-world re-

ceiver, they ﬁnd several important differences. Some

of them are caused by implementation impairments,

i.e. frequency and phase noise or non-linear distor-

tions caused by RF power ampliﬁers. Other differ-

ences are inherent to the fact that transmitter and re-

ceiver are situated at different physical locations.

Consequently, both time and frequency synchro-

nization steps have to be carried out prior to any other

operation. Also, some operations like ﬁltering, dec-

imation and channel estimation (instead of perfect

channel knowledge) are mandatory. In contrast to

computer simulations, the sources of error are now

out of control, which on one hand makes analyzing

the results more difﬁcult but, on the other hand, the

students have to deal with a more realistic problem.

It is important to stress that some effects can also be

seen, and later understood, if the whole system im-

plementation is considered. For example, most of

the theoretical multiple-antenna models consider the

same average gain for all individual single-antenna

links. However, testbed measurements show that this

is not true and the average gain depends on the spe-

ciﬁc propagation seen by each of the transmit-receive

419

A. García-Naya J., J. Pérez-Iglesias H., Dapena A. and González-López M. (2010).

DIGITAL COMMUNICATIONS LEARNING TOOLS gtTAL: Graphical Tool for Testbed-assisted Learning.

In Proceedings of the 2nd International Conference on Computer Supported Education, pages 419-426

DOI: 10.5220/0002784904190426

 SciTePress

Tx PC

• Middleware

• Signal

Processing Layer

Network

Rx PC

• Middleware

• Signal

Processing Layer

Student PC

• End-User Layer

• gtTAL

Figure 1: General organization of the testbed and the stu-

dent PC running gtTAL.

antenna pairs.

In addition, being able to implement and test al-

gorithms in a testbed requires more multi-discipline

skills than the theoretical approach. For instance,

implementing a time synchronization algorithm in a

testbed requires to deal with low-level details of the

testbed hardware components, as for example the type

and format of the A/D converter outputs. Also, it re-

quires skills in instrumentation equipment usage (e.g.

oscilloscopes, spectrum analyzers, network analyz-

ers, etc.). In summary, testbed-assisted learning pro-

vides a much wider perspective of communications

engineering to the student.

The rest of this paper is organized as follows.

2 shows the main contributions provided by the

so-called testbed-assisted learning. 3 presents the

distributed multilayer software architecture of the

testbed, 4 describes the testbed hardware components,

and 5 details the architecture layers. 6 illustrates how

to use gtTAL in a digital communications lesson. Fi-

nally, 7 is devoted to the conclusions.

2 BENEFITS OF

TESTBED-ASSISTED

LEARNING

Some of the most relevant researchers in the ﬁeld

of digital wireless communications (e.g. (Rao et al.,

2004)) claim that students involved in testbed im-

plementations experience a work that is more time-

consuming, but more rewarding than single-discipline

tasks. For instance, rarely will a communication the-

ory student need to spend time understanding the im-

pact of I/Q imbalances, while a student working on

a testbed has to take into account such effects. Also,

students working on system implementations notice

that the testbed approach is more detailed and com-

prehensive than only theory and computer simula-

tions. Therefore, the testbed approach leads to a

greater satisfaction in seeing a fully functional testbed

transmitting and acquiring real-world signals.

Being involved in testbed development tasks

forces the student to work in a multidisciplinary envi-

ronment (e.g. computer engineers and electrical engi-

neers use to work together in the development of such

testbeds). The student learns how to work in a team

and also how to discuss with senior members about

different issues found during the implementation.

During the last years, different general-purpose

multiple-antenna testbeds have been constructed to

evaluate the performance of diverse signal process-

ing techniques and/or standards (e.g. (Caban et al.,

2006; Borkowski et al., 2006)). At a ﬁrst glance, one

may think that undergraduate students can participate

in the development of testbeds but, in fact, only post-

graduate students with high expertise are involved in

such teams. Unfortunately, setting up and later de-

velop software allowing to transmit, acquire and prop-

erly process signals involves cumbersome low-level

programming to access the hardware, making difﬁ-

cult to test new methods which allows the students

to start to interact with testbeds (Rupp et al., 2007).

Due to this reason, it is convenient to add a mecha-

nism to the testbed that allows to access it at differ-

ent levels of abstraction. This means that a student

starting to implement his ﬁrst algorithms should ac-

cess the testbed at a higher level than another student

who is prepared to deal with more low-level details of

the testbed. Consequently, such a mechanism allows

the students to focus exclusively on the development

and proof techniques, releasing them from the task of

low-level programming.

Our aim is, thus, to provide a mechanism that

allows even undergraduate students to access the

testbed at an abstraction level similar to that of com-

puter simulations. Once the student acquires skills

enough to deal with lower level details, this mecha-

nism should also permit using the testbed but still hid-

ing even lower level details. The understanding of a

multi-layer scheme, where each layer deals with spe-

ciﬁc problems at a different abstraction level, clearly

leads to enforce the knowledge of the students.

To the knowledge of the authors, a multitude of

universities as well as public and private research cen-

ters have been investing a lot of efforts in setting up

testbeds with research purposes. However, very few

works (i.e. (Rao et al., 2004)) consider the possibil-

ity of taking advantage of the educational possibilities

offered by testbeds. Therefore, we introduce the term

“testbed-assisted learning” that consists in involving

CSEDU 2010 - 2nd International Conference on Computer Supported Education

420

Figure 2: Testbed picture showing the Tx PC, the Rx PC,

and gtTAL running on a laptop.

testbeds in the learning process. This approach is not

restricted to the wireless communications ﬁeld but our

testbed and most of our research knowledge comes

from this area. Consequently, we will restrict our dis-

cussion to this ﬁeld.

3 TESTBED OVERVIEW

In Figure 1, the general organization of our testbed

is shown. The testbed is hosted by two ordinary PCs

(see Figure 2), one for the transmitter (referred to as

Tx PC), and another one for the receiver (called Rx

PC). Figure Figure 3 shows the block diagram of the

entire system. Three main parts can be distinguished

(from bottom to top): the testbed hardware that al-

lows us to transmit discrete-time signals over multi-

ple antennas at 2.4 GHz; the multilayer software ar-

chitecture that makes the hardware accessible to end

users (in the following, we consider the students also

as end users of the testbed); and ﬁnally, the graphi-

cal tool gtTAL, which is implemented on top of the

multilayer software architecture.

The lowest level layer (i.e. the middleware) is re-

quired to be installed in the same computers as the

testbed hardware is hosted, but the other two layers

can be installed in any other available PCs. However,

the conﬁguration we present here simpliﬁes the set-

up because all software needed is installed in the Tx

PC and in the Rx PC. Only the user layer is installed

together with gtTAL in the student PCs (see Figure 1).

The multilayer software architecture presented

above provides a high abstraction level, which allows

to implement user applications without knowing the

testbed hardware details. For instance, in this paper

we present gtTAL, a graphical tool designed to ex-

plain the so-called Alamouti code (Alamouti, 1998),

constituting one of the simplest and most known

space-time block codes frequently used to introduce

multiple-antenna systems to the students. Such graph-

ical application (detailed below) constitutes the fun-

damental tool to enable the teacher to show the main

effects caused by real-world transmissions through

the testbed. In this work we will restrict our approach

to describe the beneﬁts provided by the use of such

tool in an academic environment, but the testbed plus

the software architecture also enables many teaching

possibilities for more advanced students (e.g. access-

ing the testbed at speciﬁc layers).

4 TESTBED HARDWARE

DESCRIPTION

A picture of the Testbed PCs (Tx PC and Rx PC) is

shown in Figure 2. The hardware of the testbed is

entirely based on Sundance Multiprocessor Ltd (see

the bottom of Figure 3). The transmitter is based on a

PCI carrier board SMT310Q and the SMT365, a basic

processing module equipped with an FPGA, a DSP,

memory buffers and two buses capable to sustain a

transfer rate of 400 MB/s. The basic processing mod-

ule is directly connected to the data acquisition mod-

ule (DAQ module), the SMT370. It contains a dual

D/A converter with dedicated memory accessible at

the same speed of the D/A converter. The DAQ mod-

ule also has two A/D converters. Finally, the DAC

module is connected to the RF front-end module, the

SMT349, which performs up and down conversion

operations from 70 MHz to 2.45 GHz with 16 MHz

of maximum bandwidth. The receiver employs the

same conﬁguration as the transmitter (see the bottom

of Figure 3) but incorporating a buffer memory mod-

ule, the SMT351, allowing to store in real-time the

data acquired by the A/D converters to be later trans-

ferred to the Rx PC.

In order to show the advantages derived from the

use of a multilayer software architecture, we will ex-

plain a frame transmission step by step (see Figure 4).

• Once the symbols to be transmitted have been

generated at the graphical tool, a function is called

passing to it the corresponding symbol vectors

(one vector per transmit antenna).

• These symbols are sent to the signal processing

layer (TXPROC) through the user layer, where

they are converted to pass band signals that are

subsequently sent to the middleware.

• When both the Tx PC and the Rx PC are ready to

complete a transmission, the signals are passed to

the testbed hardware to be transmitted through the

antennas.

DIGITAL COMMUNICATIONS LEARNING TOOLS gtTAL: Graphical Tool for Testbed-assisted Learning

421

Sundance SMT310Q

SMT370

1 dual DAC

4 MB memory

2 x ADC

FPGA Virtex II

Middleware – TxDSP

PCI Bus

Tx PC

Rx PC

Middleware – RxHost

Middleware – RxDSP

Middleware – TxHost

Socket connection

Frame synchronization

Signal Processing Layer – TxProc

Signal Processing Layer – RxProc

Student PC

gtTAL

User Layer

Socket connection

SMT365

DSP@600 MHz

FPGA Virtex-II

16 MB memory

SHB buses

SMT365

DSP@600 MHz

FPGA Virtex-II

16 MB memory

SHB buses

SMT351

1 GB FIFO

memory for the

SHB buses

SMT349

IF ↔ RF

16 MHz BW

IF 70 MHz

RF 2.45 GHz

SMT370

1 dual DAC

4 MB memory

2 x ADC

FPGA Virtex II

SMT349

IF ↔ RF

16 MHz BW

IF 70 MHz

RF 2.45 GHz

Figure 3: Platform Scheme.

• At the receiver side, the middleware stores the sig-

nals into the hardware buffers and then they are

forwarded to the receiver signal processing layer

(RXPROC).

• Finally, the acquired signals are forwarded to the

user application through the user layer, complet-

ing the entire process.

function call

symbol/sample

sequences

gtTAL

User

Layer

TxProc

Middleware

RxProc

socket

connection

discrete

symbol

sequences

socket

connection

IF signals

transmission

socket

connection

acquired IF

signals

socket

connection

acquired

symbol/sample

sequences

function

return

acquired

symbol/sample

sequences

Figure 4: Example of a transmission using gtTAL, the dis-

tributed multilayer software architecture, and the testbed

hardware.

The user layer allows conﬁguring several param-

eters. For example, the signal processing layer can

perform time and frequency synchronization or ac-

quire non-synchronized signals, also raw data or al-

ready demodulated discrete symbols can be acquired.

5 ARCHITECTURE LAYERS

This section describes the three layers developed by

us to provide high-level access to the testbed (see Fig-

ure 3): the user layer, the signal processing layer and

the middleware layer.

5.1 User Layer

The user layer interacts with the user application (the

graphical tool in this case) by using a simple func-

tion implemented in MATLAB

(any other software

implementing socket connections is also valid). Its

main task consists in sending to the signal processing

layer the symbols to be transmitted plus the necessary

parameters, at the transmitter side. In the same way,

the user layer receives the acquired symbols, being

noticed if any error occurs.

The main target of the user layer is making the rest

of the layers accessible to the high level applications,

taking into account the type of development environ-

ment they use. For this reason, the user layer is jointly

executed with that application (see Figure 3).

5.2 Signal Processing Layer

The signal processing layer is network-connected

with both the user and the middleware layers (see Fig-

CSEDU 2010 - 2nd International Conference on Computer Supported Education

422

ure 3). It provides remote access and makes them in-

dependent with respect to the other layers. This layer

consists of two different processes that carry out the

signal processing operations needed to link the user

and middleware layers. The ﬁrst process (TXPROC)

receives the symbol vectors from the user layer and

performs the up sampling, pulse-shape ﬁltering, I/Q

modulation and frame assembling operations in or-

der to generate the IF signals that will be sent to the

middleware. Similarly, the second process (RXPROC)

waits for the acquired signals from the middleware

and performs the time and frequency synchronization

operations followed by the I/Q demodulation, ﬁltering

and down sampling. The resulting vectors are sent to

the user layer.

5.3 MIDDLEWARE LAYER

The middleware concept constitutes a great leap for-

ward in multiple-antenna testbed technology, making

the testbed hardware accessible through ordinary net-

work connections. This layer ﬁlls the gap between the

testbed hardware and the signal processing layer, al-

lowing discrete-time signals to be transferred through

the PCI bus and making possible the synchronization

between the Tx PC and the Rx PC using a network

connection.

The middleware architecture is split into two dif-

ferent sub-layers (see Figure 3). The top sub-layer is

responsible of establishing the network connections

between the transmitter and the receiver, and with the

higher layer (the signal processing layer). The bottom

sub-layer corresponds to the testbed hardware conﬁg-

uration and control software.

The middleware is constituted by four different

processes. The ﬁrst two (TxHost and RxHost) im-

plement the so-called top sub-layer and run, respec-

tively, on the Tx PC and the Rx PC. They are im-

plemented in standard C++ language and use sockets

to establish the necessary network connections: one

between TxHost and RxHost processes (used to syn-

chronize the transmitter and the receiver, thus the re-

ceiver knows when the signal acquisition process has

to start); another one, established between the TX-

HOST process and the Tx signal processing layer;

and, ﬁnally, another one between the RXHOST pro-

cess and the Rx signal processing layer. The re-

maining two processes are the transmitter and the

receiver processes that run on their respective Digi-

tal Signal Processors (DSPs) available in the testbed

hardware. They implement the so-called bottom sub-

layer. The transmitter DSP process (TXDSP) per-

forms data transfers through the PCI bus jointly with

the TXHOST process and conﬁgures and controls

the hardware components at the Tx PC. In the same

way, the RXHOST process and the DSP receiver pro-

cess (RxDSP) are responsible of transferring the data

through the PCI bus and, from the DSP side, con-

trolling and conﬁguring the testbed hardware compo-

nents at the Rx PC.

6 A LESSON IN DIGITAL

COMMUNICATIONS

Thanks to the abstraction level of our distributed ar-

chitecture, it is very easy to devise a graphical tool for

testbed-assisted learning (gtTAL). This tool helps in

explaining basic concepts about wireless digital com-

munication transceivers, including multiple-antenna

systems. For instance, we have developed a ﬁrst re-

lease of gtTAL oriented to explain Orthogonal Space-

Time Codes (OSTBC), including the popular 2×1

Alamouti OSTBC (Alamouti, 1998).

The students should be familiar with some ba-

sic concepts in the ﬁeld of wireless digital commu-

nications, such as modulation types (e.g. PAM, PSK,

QAM), symbol rate and bandwidth, the concept of

SNR (Signal-to-Noise Ratio), matched ﬁltering, etc.

6.1 The 2×1 Alamouti OSTBC

Signal Model

Let us start with the explanation describing the base

band model of the 2×1 Alamouti OSTBC. Two anten-

nas are used at the transmitter side and only one an-

tenna is employed at the receiver side (see Figure 5).

As shown in Figure 5, the input binary data stream b

is ﬁrst mapped to the corresponding symbols, which

are then split into two sub-streams s

and s

. Each

pair of modulated symbols {s

, s

} is then transmit-

ted during two consecutive time slots using the fol-

lowing strategy: during the ﬁrst time slot, s

and s

are respectively transmitted through the ﬁrst and the

second antenna. During the second time slot, −s

∗

transmitted through the ﬁrst antenna while s

∗

is trans-

mitted through the second one

Since the source symbols are sent through the an-

tennas during two consecutive time slots, they experi-

ence different fading realizations h

and h

(see Fig-

ure 5), but the fading value is assumed to be the same

during two time slots (i.e. a block fading channel with

a duration of at least two channel uses). Hence, the

The operator (·)

∗

denotes complex conjugation.

DIGITAL COMMUNICATIONS LEARNING TOOLS gtTAL: Graphical Tool for Testbed-assisted Learning

423

Alamouti

STBC

Compute

− 1

( )

ˆs

= z

−s

Modulator

Figure 5: Alamouti OSTBC scheme.

signal received during the ﬁrst time slot has the fol-

lowing form:

= s

+ s

+ n

. (1)

Given that the channel remains constant during two

time slots, the observation in the second time slot is

given by

= s

∗

− s

∗

+ n

. (2)

In the expressions presented above, n

denotes the ad-

ditive white Gaussian noise (AWGN). By deﬁning the

vector of observations as:

x = [x

, x

]

= [z

, z

∗

]

(3)

where the operator (·)

stands for the transposition.

The relationship between the observation vector x and

the source vector s = [s

, s

]

is given by

x = Hs + n (4)

where n = [n

, n

∗

]

is the noise vector and H repre-

sents the (2 × 2) matrix obtained following the Alam-

outi coding scheme from the two channel coefﬁcients

, and h

(mixing matrix):

H =



| h





∗

−h

∗



(5)

It is interesting to note that H is unitary up to a scalar

factor:

= H

H = khk

(6)

where khk

= |h

+ |h

is the squared Euclidean

norm of the channel vector, I

is the 2×2 identity ma-

trix and (·)

denotes the Hermitian operator. Conse-

quently, the transmitted symbols can be recovered up

to scalar factor:

s =

where

H is a suitable estimate of the mixing matrix.

As a result, this scheme supports maximum likelihood

(ML) detection based only on linear processing at the

receiver.

Channel Estimation Strategies

The performance of communication systems based

on the Alamouti coding scheme strongly depends on

the accurate estimation of the channel matrix H. For

this reason, this lesson is focused on supervised and

unsupervised algorithms to estimate the mixing ma-

trix. The standard way to estimate this matrix con-

sists in utilizing pilot symbols (Budianu and Tong,

2001) known by both the transmitter and the receiver.

Among the supervised methods, the Least Squares

(LS) criterion (Haykin, 2001) constitutes a frequent

starting point, given the simplicity of the resulting

technique.

The main inconvenient caused by the use of pilot

symbols is the energy spent during their transmission.

Because pilot symbols do not convey data, the result

is a loss in terms of spectral efﬁciency. This drawback

can be avoided by using unsupervised approaches

(also known as blind channel estimation methods).

Blind Source Separation (BSS) algorithms can esti-

mate the mixing matrix, H, and therefore the realiza-

tions of the source vector s, from the corresponding

observations x. The lack of a priori knowledge may

limit the achievable performance, but makes blind ap-

proaches more robust to calibration errors (i.e. devia-

tions from the assumed theoretical model) than con-

ventional array processing techniques (Cardoso and

Souloumiac, 1993). The best known BSS algorithms

are the so-called joint approximate diagonalization of

eigenmatrices (JADE), and FastICA (Bingham and

Hyv

arinen, 2000). More recently, speciﬁc algorithms

for OSTBCs have been designed taking advantage of

the speciﬁc structure of these codes (e.g. (Beres and

Adve, 2007)).

6.2 Exercise

Our graphical software tool, gtTAL, allows testing ei-

ther with complex-valued discrete symbol sequences

CSEDU 2010 - 2nd International Conference on Computer Supported Education

424

Figure 6: Screen shot of gtTAL main window.

Figure 7: Result window showing the transmitted symbols

(black) and signals (red), the acquired signals after the I/Q

demodulation (green), and after the matched ﬁlter (blue).

or pulse-shaped signals, both base band or pass band.

In all cases, the signals are transmitted using the

testbed or through randomly generated channels (i.e.

like in a conventional computer simulation). Sev-

eral parameters can be modiﬁed such as the num-

ber of bits to be transmitted, or how they are gen-

erated (equiprobable source or using a function pro-

vided by the user); the modulation type (PAM, PSK

or QAM), and the number of levels of the modula-

tion; the number of samples for each symbol or the

pulse-shape form to be used (square, root raise cosine

or user-deﬁned). Figure 6 shows the main window of

gtTAL, used to introduce these parameters. Students

can measure the performance of several channel es-

Figure 8: Symbol constellation corresponding to the ac-

quired signals.

timation algorithms (Least Squares, Fast-ICA, JADE,

etc.) and compare it with the performance obtained

when perfect channel state information is available at

the receiver side. In the latter case, it is required that

the mixing matrix H be randomly generated using dif-

ferent distribution models (i.e. not using the testbed).

For example, the parameters introduced in the

main window of gtTAL shown in Figure 6 allows

to generate k = 2000 equiprobable bits mapped us-

ing the Gray code. The symbols are 4-QAM modu-

lated, ﬁltered using a root raised cosine pulse shape,

and ﬁnally, the symbols are space-time coded using

the Alamouti scheme. gtTAL displays the modulation

constellation and the pulse shape.

After pressing the “Transmit” button, the se-

quence is transmitted over the testbed as it is shown

in Figure 4, and the tool plots Figure 7 and Figure 8.

Figure 7 shows the transmitted signal s

(another ﬁg-

ure is plotted for s

), the transmitted symbol values,

the transmitted signal after the pulse-shape ﬁlter, and

the acquired signals after the matched ﬁlter. Figure 8

shows the received symbol constellation. At the bot-

tom of the ﬁgure, the estimated values for the symbol

error ratio (SER), the bit error ratio (BER) and the

signal-to-noise ratio (SNR) are displayed.

7 CONCLUSIONS

We have introduced gtTAL, a graphical tool to help

in the teaching process of wireless digital communi-

cations. In its current release, gtTAL results very ad-

equate to explain the basic concepts about space-time

block codes for multiple-antenna transceivers. In par-

DIGITAL COMMUNICATIONS LEARNING TOOLS gtTAL: Graphical Tool for Testbed-assisted Learning

425

ticular, it includes the well known Alamouti space-

time code. Also, it offers the possibility of using dif-

ferent channel estimation strategies, including super-

vised and unsupervised methods.

gtTAL can be used as a conventional graphical

user interface to perform computer simulations, i.e.

without needing any additional hardware component.

However, its main potential resides on its ability to

transmit and acquire signals by using a hardware

testbed. To do so, gtTAL perfectly integrates with a

distributed multilayer software architecture, that en-

ables to access the hardware testbed avoiding low-

level programming. In this sense, gtTAL represents a

high-level tool to operate a hardware testbed. Thus,

even undergraduate students can easily experiment

with real data transmissions in realistic environments.

This clearly contributes to increase their motivation as

well as their personal reward in learning digital com-

munications.

The hardware testbed, the multilayer software ar-

chitecture, and gtTAL, constitute what we termed

testbed-assisted learning.

ACKNOWLEDGEMENTS

This work has been funded by the Xunta de Gali-

cia, the Ministerio de Ciencia e Innovaci

on of Spain,

and the FEDER funds of the European Union under

the grants with numbers 09TIC008105PR, TEC2007-

68020-C04-01, and CSD2008-00010.

REFERENCES

Alamouti, S. (1998). A simple transmit diversity technique

for wireless communications. IEEE Journal on Se-

lected Areas in Communications, 16(8):1451–1458.

Beres, E. and Adve, R. (2007). Blind channel estimation for

orthogonal stbc in miso systems. IEEE Transactions

on Vehicular Technology, 56(4):2042–2050.

Bingham, E. and Hyv

arinen, A. (2000). A fast ﬁxed-

point algorithm for independent component analysis

of complex valued signals. International Journal of

Neural Systems, 10:1–8.

Borkowski, D., Br

uhl, L., Degen, C., Keusgen, W.,

Alirezaei, G., Geschewski, F., Oikonomopoulos, C.,

and Rembold, B. (2006). SABA: A testbed for real-

time MIMO system. EURASIP Journal on Applied

Signal Processing, 2006.

Budianu, C. and Tong, L. (2001). Channel estimation for

space-time orthogonal block codes. In IEEE Inter-

national Conference on Communications, ICC 2001,

volume 4, pages 1127–1131.

Caban, S., Mehlf

uhrer, C., Langwieser, R., Scholtz, A. L.,

and Rupp, M. (2006). Vienna MIMO Testbed.

EURASIP Journal on Applied Signal Processing,

2006, Article ID 54868.

Cardoso, J. and Souloumiac, A. (1993). Blind beamforming

for non-gaussian signals. IEE Proceedings F Radar

and Signal Processing, 140(6):362–370.

Haykin, S. (2001). Adaptive Filter Theory. Prentice Hall

Information and System Sciences Series, 4th edition.

Rao, R., Zhu, W., Lang, S., Oberli, C., Browne, D., Bhatia,

J., Frigon, J.-F., Wang, J., Gupta, P., Lee, H., Liu, D.,

Wong, S., Fitz, M., Daneshrad, B., and Takeshita, O.

(2004). Multi-antenna testbeds for research and edu-

cation in wireless communications. IEEE Communi-

cations Magazine, 42(12):72–81.

Rupp, M., Caban, S., and Mehlf

uhrer, C. (2007). Chal-

lenges in building MIMO testbeds. In Proc. of the 13th

European Signal Processing Conference (EUSIPCO

2007), Poznan, Poland.

CSEDU 2010 - 2nd International Conference on Computer Supported Education

426