Sie sind auf Seite 1von 8

Wireless Link using OFDM Modulation:

Performance Prediction, Modeling


and Implementation
Adrian Bohdanowicz, Chris van den Bos, Maarten Ditzel,
Wouter A. Serdijn, Gerard J.M. Janssen, Ed F.A. Deprettere
Delft University of Technology
Faculty of Information Technology and Systems
Mekelweg 4, 2628 CD Delft, The Netherlands
E-mail: adrian@merijn.et.tudelft.nl
Abstract The Ubiquitous Communications program
(UbiCom) at Delft University of Technology aims at the realization of a wireless communication system for the transfer
of audiovisual data, meant to provide augmented reality to
mobile users. In this paper, the design and simulation of
a high-speed wireless data link for the UbiCom program is
presented. The design (strawman prototype) uses an infrared link for data transmission, whereas the simulation model
simulates a 17 GHz radio link.
The main difficulty in the design is the fading multipath
communication channel, which makes high data-rate communication hard to achieve with conventional single-carrier
modulation schemes. By using orthogonal frequency division modulation (OFDM) as modulation scheme, multipath
effects have less influence. In OFDM, which is a multicarrier modulation scheme, many low bit-rate channels are
multiplexed in the frequency domain. The data rate per subcarrier is chosen such that inter symbol interference (ISI) is
no longer a problem.
The OFDM modulation is used in the strawman prototype
as well as in the simulation. For the design of the strawman
prototype, a custom infrared transmitter and receiver have
been designed. For the simulation, a model of the physical
channel is used.
It is expected that OFDM is capable of using a infra-red
channel efficiently. In that case, a high bit-rate link is feasible. It is argued that a Rician channel model is appropriate
for the simulation of the 17 GHz radio link, however, further
measurements are necessary. The results are a suitable simulation of the radio link and a design for an infrared link,
which will be implemented in hardware and demonstrated.
Keywords orthogonal frequency division multiplexing,
OFDM, multi carrier modulation, radio channel modeling, multipath fading, frequency selective fading, infra-red
transceiver, infra-red link, capacity estimation

The authors are members of the Ubiquitous Communications program


ISBN:

43

I. I NTRODUCTION
The Ubiquitous Communications program (UbiCom) is
a multi-disciplinary research program at Delft University
of Technology. The goal is to arrive at a system for wireless visual communications, meant for augmented reality. The Ubiquitous Communications program consists of
three projects:
Base station and personal transceiver (P1),
Visual information processing and application (P2), and
System and application specification, emulation and
evaluation (P3).
Within the program many challenging research topics
are covered. Key issues of the program include:
low power design at all levels of the system i.e. at the
device, circuit, system and application level
wearable and distributed computing
quality of service management in a rapidly changing environment
high performance computing due to perceptual constraints of augmented reality
multi-level design space exploration and performance
analysis
high quality imaging by means of a retinal scanning
display combined with an liquid crystal display for its
blocking capabilities
high performance image processing for identification of
real world objects
Wireless visual communication demands a mobile high
bit-rate communication system. This topic of the Ubiquitous Communications program is covered by the P1
project.
The aim of the P1 project is to design a fourth generation communication system, suitable for wireless visual
communications. The eventual bit-rate to be attained is
155 Mbit/s. A radio link at 17 GHz is envisioned for outc
STW,
1998

44

Adrian Bohdanowicz et al.

door applications. This link may also be used for indoor


applications or alternatively, be replaced by an infra-red
link. The focus of the P1 project lies at:

Characterization and modeling of the wireless broadband multipath fading channel


Suitable modulation and multiple access techniques for
high data-rate wireless communications
Design techniques for high-frequency radio communication systems

As the UbiCom system is targeting the 17 GHz wireless


indoor and outdoor environment, the 17 GHz channel has
to be characterized and modeled thoroughly. The outdoor
channel is a hostile, multipath environment where due to
refraction, reflection and scattering the transmitted signal
reaches the receiver via many different paths. Each path
signal is received with a different amplitude attenuation,
phase shift and time delay. Multipath reception causes amplitude fluctuations in the frequency domain at the receiver
antenna. This phenomenon is known as a frequency selective fading channel.
The difference in the arrival times of the paths also cause
inter symbol interference, which seriously degrades the
performance of wireless conventional single carrier communication systems at high bit-rates. The changes in the
propagation conditions due to moving objects (e.g. vehicles) in the vicinity of the receiver or the transmitter also
degrades the performance. Furthermore, random effects
such as noise and co-channel interference cause an even
worse reception of the transmitted signal.
To conquer the effects of frequency selective fading in
high data-rate wireless communications OFDM is chosen
to be the modulation technique for the UbiCom system.
OFDM is a promising technique for the transmission of
high bit-rates in parallel over a number of sub-carriers,
thus effectively increasing the symbol period. In this way,
it can tolerate much larger delay spreads in the channel.
Hence, a sophisticated equalization unit as needed in single carrier systems becomes obsolete.
In section II, the strawman prototype and the simulation
model are discussed. First, the principle of OFDM is explained. Then the infra-red link capacity is evaluated to
determine the maximum bit-rate that is theoretically possible. Finally, the radio channel modeling is discussed.
In section III, the implementation of the strawman prototype is discussed. It is explained how the OFDM modulator and demodulator are implemented. Then the implementation of the infra-red receiver and transmitter is described.
Finally in section IV, conclusions are presented.

11111
00000
00000
11111
00000
11111
00000
11111
00000
11111
00000
11111
cyclic prefix

11111
00000
00000
11111
00000
11111
00000
11111
00000
11111
00000
11111

time

OFDM symbol

Fig. 1. OFDM signal with cyclically extended guard interval

II. S TRAWMAN

PROTOTYPE AND SIMULATION MODEL

The Ubiquitous Communications program plans the realization of a strawman prototype of the whole system in
December 1998. Parallel to the realization a simulation
model is implemented as well.
The purpose of the strawman prototype is twofold: to
supply a test environment, which can easily be extended
or modified and to serve as a demonstrator towards people
who are not directly involved in the program. From the
test environment valuable information can be obtained for
the design of the final UbiCom system.
The purpose of the simulation model is to do performance analysis at each level of the system and of the system as a whole.
As the modulation scheme of the final UbiCom system
is chosen to be OFDM, the prototype will also apply this
scheme. Initially infra-red light is chosen as the information carrier. Already at an early stage this choice will
give valuable information for the design of the final indoor
communication system, in which infra-red is an option.
A. Orthogonal frequency division multiplexing
In a conventional single carrier system, data symbols
are transmitted sequentially. In high data-rate communications the symbol period becomes smaller than the delay
spread of the channel and inter symbol interference occurs. In multi-carrier systems a number of data symbols
are transmitted at different sub-carriers in parallel thus increasing the symbol length.
Another advantage of transmitting the data symbols in
parallel is that the complete frequency band available is
divided into many narrow sub-bands. To increase the
bandwidth efficiency, an orthogonal multi-carrier scheme
is used, in which the sub-bands are overlapping. Every
sub-band only covers a small part of the total available
frequency band and as a consequence channel equalization becomes much simpler than in a single carrier system.
Also burst errors caused by fading do not distort several
adjacent symbols severely, but only distort many symbols
slightly.
To obtain orthogonality between the sub-carriers the
data-symbols are mapped on the sub-carriers using an inSTW SAFE98

Wireless Link using OFDM Modulation: Performance Prediction, Modeling and Implementation
add cyclic prefix

45
cos(x)

1.4

1.2

inverse
discrete
fourier
transform

parallel
to
serial

Fig. 2. OFDM modulation

1
relative intensity

symbol
mapping

serial
to
parallel

0.8

0.6

0.4

0.2

timing
recovery

0
-1.5

-1

remove cyclic prefix

serial
to
parallel

discrete
fourier
transform

0
angle in radians

0.5

1.5

Fig. 4. Lambertian pattern

parallel
to
serial

inverse
symbol
mapping

In case of a Lambertian radiation pattern (see figure 4),


the signal light power at the receiver is given by [3]
Prec =

Fig. 3. OFDM demodulation

verse discrete Fourier transform (IDFT). To reduce ISI


the last part of the OFDM symbol is copied and put as
a preamble, which serves as a cyclically extended guard
interval, which is called a cyclic prefix (see figure 1).
The use of a cyclic prefix instead of a plain guard interval, simplifies the channel equalization in the demodulator.
Also it is advantageous to maintain carrier synchronization
in the receiver [1].
OFDM demodulation consists of three steps:
locate the starting point of an OFDM symbol,
separate all sub-carriers by applying the discrete Fourier
transform (DFT), and
map the symbols into bits.
The complete modulation and demodulation schemes are
depicted in figures 2 and 3 respectively.
For evaluation purposes, both schemes are simulated using the Ptolemy simulation environment [2]. A simplified
model of the multipath channel is incorporated in the simulations.
B. Infra-red link capacity estimation
The infra-red link capacity (the maximum error-less
bit-rate) is given by Shannons formula:


S
C = B log2 1 +
N
where S is the total signal power, B is the bandwidth, and
N is the noise power. The ultimate link will exhibit multipath fading which necessitates the lowering of the signal
bandwidth with respect to the single-path link. Hence, the
performance of a single path LOS channel gives an upper
bound.
ProRISC/IEEE CSSP98

-0.5

Ps Arec
cos()
2 (r)2

where Ps is the transmitted power, r is the distance between receiver and transmitter, is the angle between the
normal to the transmitter and the beam, and Arec is the effective infra-red detector area. If the detectors responsiveness is given by R (in A/W ), the resulting signal current
from the receiver photodiode is
is = R

Ps Arec
cos()
2 (r)2

and the received signal power is proportional to:



i2s

Ps2 R2

2
Arec
cos()
2(r)2

Now a noise current is generated by incoming ambient


light:
i2n = 2qRPamb B = 2qREamb Arec B
where q = 1.60 1019 C and Eamb is the irradiance of
the ambient light on the receiver photodiode. Thus, the
maximally attainable signal to noise ratio (SNR) is given
by
SNR =

Ps2 R Arec cos2 ()


8qEamb B
(r)4

in case of a receiver amplifier noise figure of 0dB, which


corresponds to a noiseless amplifier.
The capacity of an additive, white Gaussian noise
(AWGN) communication channel is given by [4]


S
S
B
C = B log2 1 +

bits/s
N0 B
N0 ln(2)

46

Adrian Bohdanowicz et al.

where S is the signal power and N0 is the noise spectral


density. The capacity of the channel approaches a maximum limit value when for a given NS0 the bandwidth goes
to infinity. Thus, if the available channel bandwidth is unlimited, the necessary SNR to attain a given capacity approaches a lower limit if B . For an infra-red channel
the available bandwidth is not restricted by governmental
regulations, so that the necessary signal level can be minimized by maximizing the bandwidth.

components. The major difficulty of modeling the radio


channel impulse response is that in event of motion each
parameter becomes random and time variant. If the channel is assumed to be constant for a short period, this equation can be applied for each point in a three dimensional
space (for example see figure 5).

Example calculation
For infinite bandwidth, the capacity is given by
40

Ps2 R
Arec
8qEamb ln(2)
(r)4

Now, assume the infra-red communication link has the


following parameters:
Ps = 100 mW
R = 0.5 A/W (typical value)
Eamb = 10 W/m2 (no optical filtering, indoor environment)
= 45o
r =3m
Arec = 100 mm2
then C 3.6 Mbit/s for B . For r = 2 m, Cmax
is already 18 Mbit/s. These figures are in accordance with
the experimentally found channel capacitance in [5].
If the bandwidth is restricted to for example 10 MHz,
as is the target for the first strawman prototype, the capacities for r = 3 m and r = 2 m reduce to 3.2 Mbit/s
and 16 Mbit/s , respectively. It is seen that the actual reduction is moderate, only 11 percent. It should be noted,
though, that the actual maximally attainable bit-rates are
lower, as the capacity is only attainable by using the ideal
transmission scheme. OFDM will introduce an additional
performance loss, but it is expected to be small.
C. Modeling of the radio channel
For the purpose of simulations, characterization and
modeling of the radio channel is needed. The linear filter approach based on the discrete multipath model is used
to characterize the behavior of the radio channel. In this
approach, its impulse response function gives the full characterization of the radio channel. The complex baseband
impulse response h(t) is given by:
h(t) =

L
X

k (t k )ejk

k=0

where k , k and k are the amplitude, phase and time


delay of the kth path and L is the number of multipath

30
Power [dB]

CB =

cos2 ()

20

10

5
4

0
0

3
2
2

4
6

1
8
10

Distance [cm]

Time delay [ns]

Fig. 5. Power delay profile of adjacent impulse responses

To simulate the variation of the channel model, the characteristic parameters are obtained from measurements and
implemented in the model. In this case the channel is assumed to have a dominant, line-of-sight path. Then the
instantaneously received signal power can be described by
the Rice distribution. The one-value characterization of
this property of the channel is the Rice K-factor defined as
the ratio of the power of the dominant path to the average
power of the scattered paths:
K=

Pdominant path
Preflected paths

The small-scale fading effect is implemented in the channel model by variations of the Rician K-factor, which,
based on the measurements, is assumed to follow a Gaussian [6] distribution as depicted in figure 6.
The time dispersive nature of the channel is represented
by a root mean square (rms) of the delay spread, given
by the square root of the second central moment of the
power delay profile (PDP), where the PDP is given by
|h(t)|2 . Caused by the multipath environment, the rms delay spread specifies the maximum obtainable rate of symbols that can be transmitted without serious distortion over
the channel. As far as obstacles are located with complete randomness, a standard Poisson process describes the
time sequence of the components. The envelope of the impulse response is assumed to be exponential [7], which is
in agreement with measurements.
STW SAFE98

Wireless Link using OFDM Modulation: Performance Prediction, Modeling and Implementation
1

- FFT

0.9

0.8

0.7
Prob( K<abscissa )






f~

0.6

47

f~eq

-@
- M
,
,
@
6

- ~
y

~g

A - @
,  A
,
@
A
A

M1

~s

Fig. 7. Channel estimation and equalization

0.5

0.4

0.3

0.2

simulation
theory

0.1

0
10

10

15

Kfactor [dB]

Fig. 6. Cumulative distribution function of Rician K-factor

III. S TRAWMAN

OFDM symbol. A simplified least square error algorithm


is applied to estimate the start of an OFDM frame [9].
The algorithm calculates the power of the differences
between the input samples r(n) spaced N samples apart,
where N is the number of sub-carriers. The results are
then summed over an interval of length S. The minimum
of this squared error gives an estimate of the starting time
of the frame:

PROTOTYPE IMPLEMENTATION

()

ISSUES

In the strawman prototype, low power consumption is


not an issue. There are two reasons for this. First, the
prototype must be available soon, so time is limited. Second, the strawman prototype must provide an environment
for experiments. For the P1 project, the focus is on the
evaluation of the OFDM system performance. Without a
constraint on the power consumption, an indication can be
obtained on how good OFDM can be. By applying the
power constraint on later prototypes, it becomes possible
to see how far away that prototype performance is from the
optimum.

Frame synchronization
The frame synchronization is based on the principle that
the cyclic prefix contains a copy of the last part of the
ProRISC/IEEE CSSP98

|r( + i) r( + i + N )|2

i=0

min {()}

=0..m

in which m = N + L S with L the number of samples


in the cyclic prefix.
Channel estimation and equalization
The channel estimation algorithm uses a single training
frame to estimate the channel characteristics. This estimate ~g(n) is continuously updated using the received data
(see figure 7,  specifies a unit delay):
f~eq (n) = f~(n) ~g(n)



~g(n) = ~s(n 1) I f~(n 1)

A. OFDM (de)modulation
As the available time is limited for the implementation
of the strawman prototype, a straightforward implementation of OFDM is chosen. The selected scheme uses the
properties of the cyclic prefix, which effectively reduces
ISI. The main problems of this scheme are related to synchronization, i.e.:
frame synchronization: recovery of the timing of a single
OFDM frame, i.e. an OFDM symbol preceded by the
cyclic prefix;
carrier synchronization: loss of carrier synchronization
causes the loss of orthogonality of the sub-carriers and
thus degrades the performance of the system [8].
Additional problems are related to channel estimation and
equalization.

S
X

~s(n) = M (~y (n))




~y(n) = M1 f~eq (n)
with

x0
x1
..
.
xn

1
x0
1
x1

= .
..
1
xn

in which f~(n) is the result of the fast Fourier transform


(FFT), f~eq (n) is the equalized FFT result, ~g(n) is the inverse channel estimate, ~s(n) is an estimation of the received symbols, ~y (n) is the decoded data stream and M
and M1 stand for the symbol and inverse symbol mapping respectively.

48

Adrian Bohdanowicz et al.

The initial estimation of the channel ~g(0) is the all ones


vector. The initial values of the estimated symbols ~s(0) are
~training :
the training symbols from the training frame S

~g (0) =

1
1
..
.

'B72B$B,

,1

,1B,

287

75$160,7B'5,9(5

2873

6,1
0(*
&26

2871

'B72B$B4

,1

,1B4

/('

,1

287

1
~training
~s(0) = S

$B72B'B,

,1

Implementation
The Fourier transforms are the most computational intensive parts of an OFDM modulator and demodulator.
For implementing the Fourier transforms required by the
OFDM link several options are available:
design and implement a dedicated IC:
Since the demonstrator is to be completed in December
1998 there is too little time to design and implement a
dedicated IC for the FFTs as the development of a complete chip takes much longer. The resulting hardware
cannot be modified at all.
buy a dedicated IC (off the shelf):
Several dedicated ICs are available for FFTs and even
complete OFDM modulators. The advantage of this approach is that it is possible to implement a complete
OFDM system before December 1998.
However, using an existing design will not give much
insight in the problems encountered when designing and
implementing an OFDM system.
use programmable devices as complex programmable
logic devices (CPLDs) or field programmable gate arrays (FPGAs):
It is possible to use programmable devices like a CPLD
or an FPGA for the modulator and the demodulator.
Then the hardware has to be designed using a hardware description language such as VHDL, which is a
time consuming extra step in the trajectory. The resulting hardware is less flexible and more difficult to modify
than for instance the software for a digital signal processor.
use a general purpose processor:
If the processor has enough processing power it is possible to implement the FFTs and also the entire OFDM
(de)modulator completely in software.
use a dedicated digital signal processor (DSP):
Several digital signal processors are available to implement the Fourier transforms and the symbol mapping.
Complete boards and development tools are available,
including a C-compiler with additional optimized runtime libraries (including FFTs).

287

287B,

5(&(,9(5
,13
3+272',2'(

6,1
,1

,11

&26
$B72B'B4

,1

287

287B4

Fig. 8. Overview of the infra-red link system

As the first two options do not fit the requirements


of a demonstrator and the third option has some drawbacks, only the last two are suited for the demonstrator.
In both options the OFDM (de)modulation is implemented
by writing software for a processor. If a programming language is chosen that is supported by both processors (for
instance C) the complexity of the two options is practically
the same.
Implementing the modulator and demodulator in software has several advantages:
flexible (software is easily modified in contrast with
hardware);
simple hence robust;
low learning curve (no new techniques have to be
learned).
Therefore the OFDM modulation and demodulation is implemented in software on a dedicated digital signal processor. This processor is mounted on a standard PC-card
(ISA/PCI). For the prototype the evaluation boards of the
Texas Instruments C6000 fixed point processor are used.
B. Infra-red transmitter and receiver
For the generation of infra-red light, a choice must be
made from two devices. These are the laser diode and the
Light Emitting Diode (LED).
The laser diode is a device whose light output power
can be modulated very fast, so high data-rate transmission
is possible. Further, the power efficiency of a laser diode
is higher than that of a LED. However, a laser diode produces a very narrow beam of light with a high intensity. If
this beam strikes the human eye directly, the retina may be
STW SAFE98

Wireless Link using OFDM Modulation: Performance Prediction, Modeling and Implementation
ceiling

111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000

Fig. 9. Laser diode and diffuser

damaged quite quickly. Therefore, the beam width should


be enlarged by some means. Some options are diffraction
(lenses or roughly-surfaced transparent materials), and reflection (curved mirrors or roughly-surfaced white materials). A reflector of rough, white material is simple, cheap
and safe [5]; a layer of plaster would do the job [10] (see
figure 9 for a possible setup). Other disadvantages of laser
diodes are the high price and the poor availability.
Light Emitting Diodes cannot be modulated as fast as
laser diodes. The power efficiency is also lower, typically
10% or less [3]. However, the light is spread over a wide
angle, so the danger of permanently damaging the retina
is small. They are cheap and can be bought within a week
typically. Therefore, IR LEDs will be used as light emitters in the first UbiCom demonstrator.
For the detection of infra-red light, there are two devices
to be considered. One is the PIN diode, the other one is the
Avalanche Photo Diode (APD). Although the latter may
be useful in case of very weak received signals, it has two
major disadvantages: it is expensive and it usually requires
very high voltages (up to 200V) for proper operation. This
makes the use of PIN diodes unavoidable.
To reduce the influence of ambient light, optical filtering may be used. The noise current, generated by the PIN
photodiode, is caused by the ambient light. As the incident ambient light has a broad frequency range, its influence can be reduced by the application of filters which only
transmit light with the same wavelength as the light emitting diode.
There are two main classes of filters: absorption filters
and interference filters [3].
Absorption filters are made of materials that absorb
light with a wavelength shorter than some specific value.
These filters have a broad bandwidth (several hundreds of
nanometers). They can only moderately attenuate light
with an undesired wavelength. However, they are cheap
and transmit light no matter what the angle of incidence is.
Interference filters are built of many thin layers of transparent material (dielectric slabs). In and between these
ProRISC/IEEE CSSP98

49

layers, reflection and transmission of light occurs. Only


at very particular frequencies, constructive interference of
light occurs in all layers, so that light is transmitted. These
filters are very narrow-band (tens of nanometers) and can
therefore greatly reduce the influence of ambient light.
However, they are directionally sensitive, so that a set of
photodiodes would be necessary, each pointing in a different direction, to cover a wide range. Further, they are
very expensive and hard to obtain. Therefore only absorption filters will be used in the first demonstrator. It should
be noted however that the use of interference based filters
could greatly enhance the overall system performance.
The infra-red transmitter consists of an infra-red emitter
(a light emitting diode) and an amplifier to drive the light
emitting diode. The diode transfer is very linear, so the
main source of distortion in the system is the driver amplifier.
The infra-red receiver consists of a PIN photodiode and
a current-to-voltage converter. As the signal photo current is very small (typically <1 A) and the output voltage
must be large (>1 V) the overall gain is large. Therefore,
the infra-red receiver amplifier consists of two stages: a
low-noise preamplifier, and a second amplifier to boost the
signal.
IV. C ONCLUSIONS
In this paper, the designs of a wireless infra-red link,
and a simulation framework are presented, for evaluation
of the OFDM transmission link in the Ubiquitous Communication system.
To ensure good modulation performance, proper choice
of OFDM parameters (e.g. cyclic prefix) is needed.
It was found that the wireless infra-red link was suitable for high bit-rate transmission (up to 18Mbit/s). The
the maximum channel capacity, given the bandwidth of
10MHz, was approximated within 11 percent (3.2 Mbit/s
vs. 3.6 Mbit/s). This means that the OFDM modulation
scheme may be capable of attaining a signaling speed that
is close to the theoretical maximum. The ambient light
was shown to be the major limiting factor on the maximum bit-rate. By using interference-based optical filtering
and using multiple photodetectors, the bit-rate could be increased further.
To come up with useful simulation results, a channel
model is needed. In this paper, a Rician channel model is
used. The important characteristic parameters are incorporated in the presented model. As the channel model behavior depends on the characteristic parameters, the proper
value of each must be determined. In order to obtain these
values, further measurements on the target 17 GHz outdoor
environment should be conducted.

50

Adrian Bohdanowicz et al.


R EFERENCES

[1]

William Y. Zou, Yiyan Wu, COFDM: an overview, IEEE transactions on broadcasting, vol. 41, no. 1, pp.18, March 1995.
[2] J.T. Buck et al, The Almagest, Vol. 1 - Ptolemy 0.7 Users Manual,
University of California at Berkeley, 1997.
[3] R. Otte, Low-power wireless optical transmission, Ph.D. thesis,
Delft University of Technology, The Netherlands, 1998.
[4] J.G. Proakis, Digital Communications, 3rd ed. New York:
McGraw-Hill 1995
[5] D.C. Lee, J.M. Kahn, Experimental 25Mb/s Infrared Link Using 4-PPM with Scalar Decision-Feedback Equalization, ICC98
conference record, 1998.
[6] G.J.M. Janssen, Robust receiver techniques for interferencelimited radio channels, PhD Thesis, Delft University Press, 1998.
[7] H. Hashemi, Impulse Response Modeling of Indoor Radio Propagation Channels, IEEE J. Selected Areas on Communication, vol.
11, no. 7, pp. 967978, September 1993.
[8] Ferdinand Classen, Heinrich Meyr. Frequency synchronization
algorithms for OFDM systems suitable for communication over
frequency selective fading channels, IEEE transactions on vehicular technology, vol. 3, pp. 16551659, 1994.
[9] Michael Speth, Ferdinand Classen, Heinrich Meyr. Frame synchronization algorithms of OFDM systems in frequency selective fading channels, IEEE transactions on vehicular technology,
vol. 3, pp. 18071811, 1997.
[10] C.R. Lomba a.o., Experimental characterization and modelling
of the reflection of infrared signals on indoor surfaces, IEE proc.
optoelectronics, vol. 145, pp.191197, June 1998.

STW SAFE98

Das könnte Ihnen auch gefallen