You are on page 1of 13

J Sign Process Syst (2011) 62:173–185

DOI 10.1007/s11265-009-0353-4

Promising Technique of Parameterization For Reconfigurable


Radio, the Common Operators Technique: Fundamentals
and Examples
L. Alaus & J. Palicot & C. Roland & Y. Louët & D. Noguet

Received: 13 October 2008 / Revised: 28 January 2009 / Accepted: 23 February 2009 / Published online: 14 March 2009
# 2009 Springer Science + Business Media, LLC. Manufactured in The United States

Abstract In the field of Software Radio (SWR), parame- Register (R-LFSR), derived from the classical Linear
terization studies have become a very important topic. This Feedback Shift Register (LFSR) structure.
is mainly because parameterization will probably decrease
the size of the software to be downloaded, and also because Keywords Software defined radio . Parameterization .
it will limit the reconfiguration time. In this paper, Reconfigurable . Common operator . Fast Fourier Transform
parameterization is considered as a digital radio design (FFT) . Linear Feedback Shift Register (LFSR)
methodology. Two different techniques, namely common
functions and common operators are considered. In this
paper, the second view is developed and illustrated by two 1 Introduction
examples: the well known Fast Fourier Transform (FFT)
and the proposed Reconfigurable Linear Feedback Shift The term Software Radio (SWR) was first coined around
1990, thanks to the pioneering works of J. Mitola [1] and
W. Tuttlebee [2]. SWR refers to a set of techniques that
L. Alaus (*) : D. Noguet enables the reconfiguration of a communication system
CEA-LETI, Minatec, without the need to change any hardware system element
17 rue des Martyrs,
[2]. One of the benefits of SWR is that it considerably
38054 Grenoble Cedex 9, France
e-mail: laurent.alaus@cea.fr enhances transceiver flexibility by, for instance, offering the
ability to support several air interfaces or configurations.
D. Noguet
e-mail: dominique.noguet@cea.fr Although the objective of a purely software based radio is
critical due to real-time processing and power consumption
J. Palicot : Y. Louët constraints, it can be observed that modern transceivers
SUPELEC,
available on the market are more and more Software Radio
Avenue de la Boulaie CS 47601,
35576 Cesson-Sévigné Cedex, France powered. This can be concluded by considering the
regularly updated surveys from the SDR Forum [3]. Any
J. Palicot
e-mail: yves.louet@supelec.fr reconfiguration of SWR simply corresponds to a change in
their software. The required software does not even need
Y. Louët
e-mail: jacques.palicot@supelec.fr to be stored in the device itself, since it can be down-
loaded, thereby bringing easy maintenance capability to
C. Roland the radio.
Université de Bretagne Sud, UEB,
Parameterization is an important aspect of the SWR
CNRS Lab-STICC,
56321 Lorient, France concept. It consists in identifying all the common aspects of
e-mail: christian.roland@univ-ubs.fr the mobile communication modes and standards which the
174 J Sign Process Syst (2011) 62:173–185

receiver is intended to handle. In this paper we address to bring in mind the basics of the Parameterization and to
heterogeneous wireless systems such as GSM1, IS952, lay the foundation of the CO technique. We specify clearly
PDC3, DECT4 and UMTS5. In addition we consider the keystone of the method, the complete definition and the
wireless LAN6 as well as broadcasting technology like T- rules of design and implementation. Moreover, on the top
DMB7 and DVB-T8. The main drawback of this profusion of the TA, we introduce the Pragmatic Approach. In the
of wireless communication standards is that communicating search of a CO, the Pragmatic Approach is twofold; the
devices enclose more and more radios. This explains the Existing Search (ES) and the Constructive Search (CS). In
growing interest in multi-mode terminals enabled thanks to the following the Common Operator Technique is illustrat-
SWR techniques. ed by the Existing Search approach applied to two COs; the
In this context, parameterization can be considered as an Fast Fourier Transform (FFT) and a specific Linear
optimization process which can help in the design of Feedback Shift Registers, the R-LFSR. The Constructive
reconfigurable equipments. According to the methodology Search will be instanced by another LFSR operator in [13].
of parameterization, the common aspects of the different This paper is organized as follow: in the first section, we
standards will become one common processing procedure, discuss the concept of parameterization techniques as a
which could be installed in the device. This common methodology to design Software Radios. First, we detail the
procedure could be executed by a simple “call”, thereby required definitions and present the state of the art. Next,
resulting in a gain of both size and time with regards the the CO technique is introduced and two approaches for
amount of code to be downloaded or read in order to modify designing CO are discussed, i.e., the Pragmatic and the
the radio behavior. From the design point of view, the Theoretical Approach. Finally, a comparison between CF
parameterization will optimize the co-design and will proba- and CO is proposed.
bly reduce the time to market by enabling incremental design. Then in a second section of this article, the common
To perform parameterization, a first technique, the operator technique is applied to two operators. The first
Common Function (CF) technique, was introduced in [4] subsection (3.1) describes the FFT operator and gives a few
and also as a chapter in [5]. One example of the CF examples of its application. In the second subsection (3.2)
technique, for UMTS, GSM and PMR9, concerning channel we propose a specific LFSR common operator called
coding block is given in [4]. In this paper the author Reconfigurable Linear Feedback Shift Register (R-LFSR)
highlights the fact that “parameterization allows communi- and we describe its application to a tri-standard terminal.
cation systems to be built with flexible components, under
the restrictive assumption that these components belong to
a predefined set of transmission modes”. From this point of 2 Fundamentals of the Techniques of Parameterization
view, the Common Function technique could be seen as
highly dependent of the standards considered. 2.1 Definition and Existing Solution
At the opposite, the proposed common operator tech-
nique, presented in this article, aims to be less standards 2.1.1 Definition
dependant. We have already published some studies,
specific to the CO technique. The Fast Fourier Transform The initial work of J. Mitola [1] and W. Tuttlebee [2]
(FFT) developed in [6] was the starting point of the method, defined the Software Radio as “a set of techniques that
illustrated by two additional CO in [7] and [8]. Further- permit the reconfiguration of a communication system
more, in [9], a theoretical approach (TA) and an optimiza- without the need to change any hardware system element”.
tion process are proposed to identify CO in a multi standard The first way of implementing a SWR was the so called
terminal, viewed as a graph. All this previous work needs “Velcro” technique, which consists in instantiating all
secure foundation and the present article comes up to that required standard implementations and enabling reconfigu-
expectation. As a consequence, the purpose of this paper is ration by activating the desired one. Although this approach
meet the flexibility criteria, and is still considered for
1
GSM: Global System for Mobile Communications analogue stages, more efficient approaches were introduced
2
IS95: Interim Standard 95 (CDMA) to reduce silicon area, design effort and cost, especially for
3
PDC: Personal Digital Cellular the digital section. Then, the function sharing among
4
DECT: Digital Enhanced Cordless Telecommunications different standards was proposed and can be considered as
5
UMTS: Universal Mobile Telecommunication System the beginning of the techniques of Parameterization. This
6
LAN: Local Area Network (WiFi, WiMax, …) led to the definition of the Common Function Technique. It
7
T-DMB: Terrestrial — Digital Multimedia Broadcasting was clearly explained by P. F. Jondral in the introduction of
8
DVB-T: Digital Video Broadcasting-Terrestrial [5]: “It seems natural to ask whether there are construction
9
PMR: Professional Mobile Radio principles that are common to all standards and, if so, to
J Sign Process Syst (2011) 62:173–185 175

Figure 1 Common function


technique.

describe the standards according to these principles”. The A detailed study of this specific CF displays the cumulative
CF Technique could be perfectly illustrated by Fig. 1. For aspect of the General Modulator. In practical terms, the CF
each Transmitter Data Processing, all the components consists in the juxtaposition of several modulators and some
dedicated to the same “Functionality” were aggregated in very specific procedures roughly common. Furthermore, the
the same Common Function. Fig. 1 clarifies three CF of parameterization considered is limited to a simple choice
Channel Coding, Interleaving and Modulation. Each Com- between several paths depending on the standard.
mon Function is depicted as the combination of one The second example of the parameterization study, for
Common and several dedicated slices. The Common part UMTS, GSM and PMR17, concerning channel coding
includes the components required by at least two functions aspects is given in [4]. As in the previous example, it is
and each dedicated part is characteristic to the non-related more a matter of an assembly of boxes than a Common
components of each individual function. The resource Function. All the boxes are not used at the same time; their
sharing inferred by the CF technique allows the non- use depends on the considered standard. As stated by the
duplication of redundant components and a possible save of author, when a box is not used, the corresponding software
complexity (memory size, downloading speed, reconfigu- is not used and therefore it is not time- or CPU-consuming.
ration speed, computation complexity…).. According to the two Common Functions presented
From this point and based on the previous definitions, before, we can conclude on the high standards dependency
we suggest that “a parameterization technique” should be a aspect of the method. The basic design of the CF consists in
method that searches for and finds all the commonalities the aggregation of the required components in only one
between several different standards in order to optimize the function with minor sharing. The evolution to new standards
resources during the equipment’s implementation and/or the must require the addition of the distinct components of each
execution phases. function in the associated CF. As a consequence, the CF
should be redefined and redesigned. Finally, we follow the
2.1.2 Examples conclusion of the author of [4] “parameterization (imple-
mented with Common Function) makes it possible to build
This subsection brings up to light two different examples communication systems with flexible components, under the
and audits briefly the contribution of the method. restrictive assumption that these components belong to a
The first typical example focuses on a: “Generalized predefined set of transmission modes”. In consequence of its
parameterizable modulator” presented in [5]. The standards dependency, we describe the Common Function
concerned Common Entity is a general modulator. It has technique as a Fixed Technique. This drawback is one of
been designed for the GMSK10, QPSK11, π/4QPSK and the initial motivations in the definition of the CO technique.
dual QPSK modulations for the air interfaces of the
following standards: GSM (FDMA12/TDMA13, GMSK), 2.2 Proposed Solutions: The Common Operator Technique
IS 136 (DAMPS14) (FDMA/TDMA, π/4QPSK) and UTRA
FDD15 (DS-CDMA16, QPSK). 2.2.1 Definition

10
The CO technique follows the principles of the techniques
GMSK : Gaussian Minimum Shift Keying of parameterization and consists in identifying common
11
QPSK : Quadrature Phase-Shift Keying elements based on structural aspects. In reference to the
12
FDMA : Frequency-division multiple access definition of the pragmatic approach given in 2.2.2.1, the
13
TDMA : Time-division multiple access intrinsic design of the CO is created independently of any
14
DAMPS : Digital American Mobile Phone System
15
FDD : Frequency-Division Duplex
16 17
DS-CDMA : Direct Sequence Code Division Multiple Access PMR : Professional Mobile Radio
176 J Sign Process Syst (2011) 62:173–185

Figure 2 An example of a
breaking down of several
standards.

standards. In essence, a CO is used to perform operations 2.2.2 Methodology


without knowing their application and must be defined by a
general equation. In practical terms, we want to design CO In the context of techniques of parameterization, we designed
independent of “calling” functions and as a consequence of and proposed two distinct but convergent approaches to
typical standards. In contrast to the CF, one CO is not specify the CO: Theoretical and Pragmatic. In this paper, the
specific to the implementation of a terminal; we describe CO technique is illustrated by the FFT and the LFSR
the Common Operator technique as an Open Technique. Operator, both designed with Pragmatic Approach.
Following all these requirements, it is only sensible that
a common operator could be called and re-called by distinct Pragmatic Approach The pragmatic approach is the initial
functions several times all along the terminal. This “reuse” approach developed to identify or create possible COs. The
is an important issue in the design and the implementation approach follows two stages; an Existing Search and a
of the operator. One of the objectives is to execute the Constructive Search.
common operator in a minimal clock cycle number The Existing Search consists in identifying in literature
(analogy with the well known MAC operator in DSP). and in existing works, some operators potentially common.
Figure 2 represents the Graphical breakdown of a multi- The main interest is to enlarge the classical operations
standard terminal proposed in a previous work [10]. From available with the operator considered to non-expected
top to bottom, we decrease the granularity of the considered operations and define a new CO.
components up to basic LUT or MAC. The CO consists in The Constructive Search is basically the next step in the
increasing the granularity of the basic elements to address design of the CO, following the previous Existing Search. It
in a SDR. This operator could be used by as many future consists in the “handmade” building of CO. The approach
standards as possible. tries to transpose like-looking architectures into a “drawn
This operator has to be completely and successfully up” general block and step by step, “merge” them into
checked, optimized and tested. It is expected that the higher Common Operators of Higher Levels.
the level of the operator is, the shorter the development time of
new functionalities will be. In this article, we settle on LFSRs Theoretical Approach The theoretical approach (TA) con-
and FFT, presented in 3 and encircled in the graph, Fig. 2, sists in moving the classical breaking down of Fig. 2 into a
J Sign Process Syst (2011) 62:173–185 177

graphical approach where Fig. 2 is viewed as a Graph 2.3 Comparison of the Two Techniques
problem. In this context, the TA applies an optimization
process in order to find the most relevant CO (or The Common Function technique was historically the
combination of CO) present in the graph derived from Fig. 2. first one proposed by several articles from Karlsruhe
The optimization process related to the Theoretical University (Germany) [5]. According to the previous
Approach (TA) is performed by a specific cost function to definition, both CF and CO techniques converge to the
minimize. The definition of this cost function is not same objective: a reconfigurable architecture, which
addressed in this paper (see [10] for further description). becomes optimal in the sense of the theoretical approach.
Most of the systems we considered here are standard- That means finding paths in the previous scheme, which
based. This means, their air interface as well as all their decrease some costs. These two technical methods are
protocols are completely known. Using this standard descrip- very close. The main difference between them consists in
tion of Fig. 2, we are able to draw some of the possible paths the common blocks considered. The CF focuses on
between the functions so as to perform the required standard. similarities between functions whereas the CO looks for
Using the scheme introduced by TA, with the definition architectural similarities inside the functions. Both CO
of Section 2.1, we give another vision of the parameteri- and CF techniques could be carried out in a hardware
zation: “the parameterization techniques are the means to manner (by instance on FPGA or DSP or even on ASIC)
find the optimal path in the tree derived of Fig. 2. Here we in order to optimize their occupancy, execution rate,
are considering a global optimality. This means that even if general use, etc. As presented in Fig. 3, depending on the
a path is not locally optimal for one function, if it CF or CO standpoint we address either the Common
contributes to the global optimality then we consider it. Function (CF) technique or the Common Operator (CO)
The criteria for measuring this optimality are issued from technique.
requirements given in [9] (memory size, downloading Finally a clear summary of CF and CO can be
speed, reconfiguration speed, computation complexity…)”. suggested:
From the point of view of this definition, three difficult
– The CF Technique consists in seeking an optimized
problems remain: the definition of the graph, the definition
generic function (the expected common function) like
of the cost function and the algorithm optimization. These
coding, mapping, etc. which can replace the initial task
issues are out of the scope of this paper.
present in a predefined set of standards (NB: a CF
Finally; both approaches have the same objectives in the
should be not considered in this paper as a mathemat-
context of Techniques of Parameterization: Find the best set
ical function). It is a fixed technique.
of CO. The TA selects the best set among several operators,
– The CO technique claims to be independent of the
already in the graph. As a consequence the approach is
standards by finding the smallest set of highest-level
circumscribed by the quantity and the relevance of the
operators like MAC, FFT, etc., which are used by
operator present in the graph. The PA allows us to identify
the maximum number of functions. It is an open
and create “possible” CO, without necessarily certifying
technique.
their relevance. Actually, the PA improves the graph of the
TA and the TA validates (or not) the CO created by the PA. A tradeoff could be proposed between CF and CO, with
As a result, once a CO obtained by the Pragmatic the following rule: (1) a CF can call (one or several)
Approach, it could be integrated in the Theoretical functions, (2) a CF can use (one or several) operators, (3) a
Approach to find out the best optimization. CO can not call a function.

Common Function Common Operator


Standard Layer Function Coarse Fine
UMTS Application
Filtering grain grain
...

GSM Source coding FFT MAC


IS95 Transport
Channel coding Cordic LUT
DECT Network Access FIR cellular Butterfly
Bluetooth Link
Synchro
...

Physical Mapping

Figure 3 Two techniques for parameterization.


178 J Sign Process Syst (2011) 62:173–185

X(k) H(k) using the FFT function. In this subsection, we only keep
three examples of [6].

Channel Estimation and Equalizer A great deal of litera-


ture exists on channel estimation and equalization algo-
rithms. It goes from the classical Frequency Domain (FD)
N Last point
Least Mean Squares (FLMS) [14], and unconstrained
Frequency Domain Least Mean Squares (UFLMS) [15]
Decision
(see Fig. 4) through Constant Modulus Algorithm (CMA)
μ(k) implementation in the Frequency Domain (FD) [16], Quasi
Add N zeros
Newton algorithm in the FD [17] and Decision Feedback
Equalizer (DFE) in the FD [18] towards new Single Input
Multiple Output (SIMO) semi-blind algorithms like those
-
presented in [19].
+
Whatever equalizer is used in the receiver (except the
MLSE), a Frequency Domain implementation will be
μ(k) Average possible.
Complex
power
conjugate
signal
Multicarrier (De)modulation Multicarrier modulation and
Figure 4 UFLMS equalizer. demodulation are based on the FFT [20, 21]. Indeed, it is a
particular situation of “Channelization” (see following
3 Examples of Common Operators section). The scheme is the same as in Fig. 6.

This section will propose and describe two different Channelization Many different “pragmatic” architectures
common operators: the Fast Fourier Transform (FFT) and have been proposed recently [22, 24]. In the literature, these
the Reconfigurable Linear Feedback Shift Register (R- architectures are called Software Defined Radio (SDR)
LFSR) uses of the Existing Search approach presented in architectures instead of SWR architectures. We can consid-
2.2.2.1. Actually, we first identified the well known FFT er that Fig. 5 proposes a general valid vision of SDR
operator as a fair common operator with a large range of architecture. The analogue part is divided into two parts.
practical applications. Next, we focused on the general Linear The first one comprises some analogue functions which
Feedback Shift Registers architecture and point out that a large could not be performed digitally called Analog Front End
diversity of operations could be executed by a LFSR [12]. (AFE) then we find the ADC and the second part called
There are many ways of designing a LFSR structure. In this Digital Front End (DFE).
paper, we decided to center on an existing architecture called As proposed and described by Tim Hentschel in many
R-LFSR [7]. Afterwards, an additional LFSR operator was papers and particularly in [22] and in [23], the DFE
designed following the requirement of the Constructive Search; comprises the former analogue functions performed in the
the Extended Reconfigurable LFSR presented in [13]. digital domain.
As presented in Fig. 5, one of the main functions
3.1 The FFT Operator implemented by the Digital Front End is the so-called
channelization function. Firstly, we should consider the
Using the scheme of the Pragmatic Approach and more main difference between the mobile terminal and the base
precisely of the Existing Search, we identified numerous station channelization function. In a mobile terminal
algorithms, which could be performed by FFT. We have currently only one channel of interest has to be selected,
already presented some work in [6, 8]. In this subsection, while a base station needs to process many channels in
we will show that the Fast Fourier Transform (FFT) could parallel.
be considered as a basic operator for most of the receiver
functions.
AFE ADC DFE

3.1.1 Three Examples of Functions Performed


in the Frequency Domain
I/Q down Sample rate
Channelization
conversion conversion
In [6], we have already described how it is possible to
perform many receiver functions in the frequency domain Figure 5 SDR architecture with DFE functions.
J Sign Process Syst (2011) 62:173–185 179

X(z) N H 0 (z) FFT operator is mandatory for multicarrier modulation, we


Y0 (z) globally optimize its use with this method.
z −1
After taking a more detailed look at all these functions, it
N H1 (z) becomes very clear that, if they are performed in the
Y1 (z)
z −1 Frequency Domain, the FFT used could be very different
N H 2 (z) from one function to another. For example, for a short FIR,
FFT Y2 (z) or for a GSM channelization function, we need a short FFT
(typically 128 points), whereas for DVB-T demodulation,
the length of the FFT should be 8 K points. Knowing that it
is possible to perform large FFT with a short FFT (see
[28]), thus, the optimal operator will be a short FFT or may
z −1
be the basic butterfly.
N H Ν −1 (z)
Y N-1 (z)
3.2 The R-LFSR Operator
Figure 6 Channelization realization with FFT filterbank.
Besides the FFT, several architectures called Shift Registers
Two approaches are possible. The first is the classical and Linear Feedback Shift Registers have been identified as
per channel approach [23, 25]. The second is the filterbank mainly used in literature, standards or existing transceivers.
channelizer, which could be performed, under some From this point, we implemented a specific LFSR: the
assumptions, with transform operators, particularly with Reconfigurable Linear Feedback Shift Register [7]
the FFT [26]. The reader is referred to [22], for a very The R-LFSR has been keyed as an interesting common
interesting comparison between these two approaches. The operator for transceiver functions at the bit level. The
filterbank channelizer is presented in Fig. 6 [27]. genesis of R-LFSR structures came from the observation
The main drawback is that the channels should have a that many operations of several ongoing standards derive
regular distance between the carriers. from Linear Feedback Shift Registers or Shift Registers.
These operations mainly concern Pseudo Random Sequen-
3.1.2 Parameterization Examples with the FFT Operator ces Generators, Scramblers, Convolutional Coder, Cyclic
Redundancy Check and block channel coding (Extended to
The implementation of the CO in the frequency domain Reed-Solomon codes). The so-considered R-LFSR was
using the FFT applied to Fig. 2 removes “unnecessary paths” designed to substitute these operations as a common
and leads to the new graph of Fig. 7. Figure 7 focuses on a operator with specific parameterizations. This operator has
sub-graph of Fig. 2 to clearly present the decrease of path been implemented for three standards. As a consequence, it
with a basic example. In this diagram, the use of the CO not should be called a large number of times. But in a real
only causes an outstanding decrease of the number of paths implementation it is not sensible to consider only one
but also eliminates the need for some functions (such as the instance of a R-LFSR cell that would be called by the
FIR function). Like in Fig. 2, only one part of the possible functions of the standards. Thus is due to data and control
paths are represented in the scheme of Fig. 7. dependency which exist between these functions and
Although the FD version is not locally optimal (like in would lead to inefficient buffering and synchronization
the short FIR case), taking into account the fact that the overhead. Following the examples of filters banks, the

Multichannel OFDM
Equalization Multichannel OFDM
Equalization

channel Filter
channel bank
FIR

FIR
FFT CIC
FFT
FFT

Butterfly Butterfly Butterfly

Cordic
MAC MAC
MAC

Z-1 LUT LUT Z-1 LUT

Figure 7 Optimal path for the channelization, equalization and OFDM demodulation functions.
180 J Sign Process Syst (2011) 62:173–185

Figure 8 R-LFSR structure.

problem has been solved by implementing a bank of The R-LFSR is the juxtaposition of a smaller complexity
Common Operators. operator called “E” Operator, whose equation is: S1 ¼
S0 þ a:A þ b:B
3.2.1 Three Examples of R-LFSR Using To fully understand the constitution of the R-LFSR, the
architecture could be seen as the extension of an Infinite
R-LFSR Architecture In the search of Common Operators, Impulse Response Filter developed in [29] or as the
we identify the Shift Registers (SR) as common basic concatenation of a Finite Impulse Response Filter and a
elements. Due to some functional limitations of the Galois LFSR [12], as seen in Fig. 8.
classical LFSR with regards to non-recursive operations
[12], we round up and theorize its architecture into the R-LFSR and Pseudo Random Sequences A pseudo-random
proposed R-LFSR. The equation and the transfer function number generator is an algorithm which generates a
of such a structure are (when N is the number of register): sequence of numbers that approximate the properties of
X
N X
N random numbers [12]. The Fibonacci LFSR [30] is the
yn ¼ bNk : xnk  aN k :ynk ð1Þ classical technique to generate pseudo-random variables.
0 1 Examples of scramblers are given in the ongoing standards
P
N
[31–34]. A Galois and a Fibonacci LFSR could generate
bN k : zk
H ð zÞ ¼ 0
ð2Þ the same sequences. A specific parameterization of coef-
P
N ficients and a fitted seed give the same output as explained
1þ aN k : zk
1 in [12].

Figure 9 R-LFSR structure and


PN sequences generator.
J Sign Process Syst (2011) 62:173–185 181

Figure 10 R-LFSR structure


and NSC coder.

The R-LFSR includes a Galois structure and with the R-LFSR and CRC Cyclic redundancy check coding is an
adequate seed, it can play the part of a pseudo random error-control coding technique for detecting errors that
sequence generator and in consequence of a scrambler occur when a message is transmitted.
(Fig. 9). The most critical point in the CRC process is the
calculus of the remainder (coding) and of the syndrome
R-LFSR and Convolutional Encoder In telecommunication, (decoding). Both are performed by a binary division.
a convolutional code is an error-correcting code in which This division can be calculated using a circuit similar to
each m-bit information symbol is transformed into an n-bit the one of the R-LFSR [36]. The parameters ai and bi have
symbol, where m/n is the code rate (n≥m) [35]. The just to be parameterized in accordance with the CRC
convolutional coder can be divided into two main catego- calculation (Fig. 11).
ries: the Non Systematic Coder (NSC) or the Recursive
Systematic Coder (RSC). 3.2.2 The R-LFSR Hardware Implementation: R-LFSR4/8
A NSC is represented by a single Linear Shift Operators
Register and each output code depends on a generator
polynomial whose equation could be seen as the equation Introduction The common operator technique is a promis-
of a Finite-Impulse Response (FIR) Filter. A NSC is ing practical method. Next we prove with a case in point,
not a LFSR based operator. However, using the FIR the viability of the method in a practical implementation. To
structure of the R-LFSR, it can be implemented with the test the relevance of the common operator technique, we
adequate parameterization of feed-forwards coefficients focus on ongoing OFDM based air interfaces and narrows
(Fig. 10). down the standards to address to the IEEE 802.11 g, IEEE
A RSC is a convolutional coder, which uses a Linear 802.16-2005 and 3GPP LTE [31–34]. These standards use
Feedback Shift Register and each output code is repre- functions that can be derived as Linear Feedback Shift
sented by the equation of an Infinite Impulse Response Register operations.
(IIR) Filter. As mentioned above, the R-LFSR “is” a This section claims to define a new common operator to
transpose form IIR and can easily carry out each output of parameterize a tri-standard transceiver. Previously, we
a RSC. presented the R-LFSR architecture but a common operator

Figure 11 CRC 8 calculation with R-LFSR (polynomial generator: 1+D2 +D3 +D4 +D8).
182 J Sign Process Syst (2011) 62:173–185

Figure 12 Resources allocation


of optimized R-FLSR structure.

should be specified with a finite number of registers. As a implementations are realized on an ALTERA/Cyclone II
result, we plan out two dissimilar operators of size 4 and 8, target using Quartus synthesis tools
in accordance with the minimal and the most restated The results of Fig. 13 are given with regards to the
structures to address. These operators named ‘R-LFSR4’ Velcro method. In these first implementations, whatever the
and ‘R-LFSR8’ incorporate 4 and 8 “E” Operators and are R-LFSR structure applied, we obtain a gain about 50% with
built such as they can be used independently or “juxta- regard to logic cells comparison.
posed” to form a more longer “United” R-LFSR structure. To give a full evaluation of R-LFSR architectures, we
must let the reader know that whatever the structure
R-LFSR Operators Implementation We identify 24 LFSR- applied, scheduling of function mapping is not raised in
based or SR-based structures present in the standard to this paper.
address: scramblers, convolutional coders, CRC or The specific LFSR implementation presented in the
assimilated. article replaces exactly all the structures to execute for each
We developed an “Optimized R-LFSR structure” based standard. Once, the parameterization applied, the R-LFSR
on the networking either of R-LFSR4 or of R-LFSR8. architecture becomes the exact replica of ALL the struc-
Figure 12 represents this parameterizable network and the tures to replace for the selected standard, without the need
possible allocation of operator with regards to the oper- of extra mappings and scheduling considerations.
ations required. This structure can execute in turn all the All sets of binary parameters dedicated to a specific
modes of the three standards to address. In this section, architecture to replace are pre-defined in the implementa-
evaluations of performance are defined by calculating the tion. The parameterization consists in ‘downloading’ these
number of Logic Cells required for each architecture. All coefficients, i.e. ‘switch on/off’ coefficients.

Figure 13 Optimized R-LFSR


structure.
J Sign Process Syst (2011) 62:173–185 183

The R-LFSR is consequently a good structure in a 7. Alaus, L., Noguet, D., & Palicot, J. (2008). A reconfigurable
linear feedback shift register for software defined radio terminal.
parameterization method. In the case study considered, R-
Santorin: ISWPC2008.
LFSR8 is the most relevant specific operator. The proposed 8. Ghouwayel, A., Louët, Y., & Palicot, J. (2006). A reconfigurable
operators give not only a benefit of space but also a good butterfly architecture for fourier and fermat transforms. Karls-
flexibility to get prepared for upcoming evolutions. ruhe: WSR’06.
9. Moy, C., Palicot, J., Rodriguez, V., & Giri, D. (2006). Optimal
determination of common operators for multi-standards software-
defined radio. Karlsruhe, Germany: WSR’06.
4 Conclusion 10. Rodriguez, V., Moy, C., & Palicot, J. (2007). Install or invoke?:
The optimal tradeoff between performance and cost in the design
of multi-standard reconfigurable radios. In Wiley InterScience.
This paper defines the techniques of parameterization, Wireless Communications and Mobile Computing Journal, 7(9),
analyzes their objectives and their requirements to propose 1143–1156. doi:10.1002/wcm.487.
the new Common Operator technique. We conclude (1) that 11. Tessier, R., & Burleston, W. (2001). Reconfigurable computing
the CO is an alternative method to overcome the high for digital signal processing: a survey. Journal of VLSI Signal
Processing, 28, 7–27. doi:10.1023/A:1008155020711.
standards dependency of the Common Function Technique,
12. Klapper, A., & Goresky, M. (2002). Fibonacci and Galois
(2) that the CO technique is, in essence, a global representations of feedback-with-carry shift registers. IEEE Trans-
optimization technique and accounts for a promising line actions on Information Theory, 48(11), 2826–2836.
of research in the factoring of a whole SDR. 13. Alaus, L., Palicot, J., & Noguet, D. (2008). Extended reconfig-
urable linear feedback shift register operators for SDR. Bologna:
The second part concerns the two operators identified:
ISSSTA’08.
FFT and R-LFSR. We show, in this paper, that FFT and 14. Ferrara, E., Cowan, C., Grant, P. (1995). Frequency domain
LFSR operators can replace a wide range of structures adaptive filtering. Prentice-Hall.
for many operations to execute. Different functionalities 15. Mansour, D., & Gray, A. (1982). Unconstrained frequency-
domain adaptive filter. IEEE Transactions on Acoustic, Speech
such as equalization and multi channel modulations come and Signal Processing, 30(5), 726–734.
down to a FFT process. Moreover, general PN sequen- 16. Schirtzinger, T., Li, X., & Jenkins, W. (1995). A comparison of
ces, scrambling, convolutional coding and CRC are three algorithms for blind equalization based on the constant
specific implementations of the general equation of the modulus error criterion. New York: ICASSP’95.
17. Berberidis, K., Palicot, J. (1995). A block Quasi-Newton Algo-
LFSR. In addition, the next generation of wireless
rithm implemented in the frequency domain. EUSIPCO’96,
systems and modulation (BFDM [37], MC-CDMA Trieste, September.
[38]), use all the schemes. As a consequence, FFT and 18. Berberidis, K., Palicot, J.: (1995). A frequency domain decision
LFSR appear as relevant high level operators for upcom- feedback equalizer for multipath echo cancellation. Globecom’95,
Singapore.
ing SDR terminals.
19. Berberidis, K., Marava, A., Karaivazoglou, P., & Palicot, J.
We conclude that FFT and LFSR are good candidates for (2001). Robust and Fast Converging decision feedback equalizer
becoming basic common operators for almost all the based on a new adaptive semi blind estimation algorithm. San
transceiver’s functions in the context of parameterized Antonio: Globecom’01.
20. Alard, M., Lassalle, R. (1987). Principles of modulation and
Software Radio. The next step is to confirm this conclusion
channel coding for digital broadcasting for mobile receivers. EBU
by using a SWR receiver implementation to quantify the Review, Technical N 224.
gain offered by these operators. Further studies will concern 21. Akansu, A., Duhamel, P., Lin, X., & Courville, M. (1998).
a possible merge of them. Orthogonal transmultiplexers in communications: a review. IEE
Transactions on Signal Processing, 46(4), 979–995.
22. Hentschel, T. (2002). Channelization for software defined
base-stations. Annals of Telecommunications, 57(5), 386–
References 420.
23. Hentschel, T., Fettweis, G., & Bronzel, M. (1998). Channelization
1. Mitola, J. (1995). The software radio architecture. IEEE Commu- and sample rate adaptation in software radio terminals. Rhodes:
nications Magazine, 33, 26–38. doi:10.1109/35.393001. ACTS Mobile Communications Summit.
2. Tuttlebee, W. (1995). Evolution of radio systems into the 21st 24. Garcia, J., Gubolivic, Z., Diaz, F., Alonso, J., Macleod, J., Beach, M.
century. Bath, United-Kingdom: IEEE Radio receivers and (2000). TRUST approach to software defined radio: RF consid-
associated systems. erations. Summit’00, Galway, Ireland.
3. Forum, S. D.R. Software Defined Radio Forum, http://www. 25. Tuttlbee, W. (2001). Software defined radio: Enabling technology.
sdrforum.org/. Wiley.
4. Rhiemeier, A. (2002). Benefits and limits of parameterized 26. Zangi, K., & Koilpillai, D. (1999). Software radio issues in
channel coding for software radio. 2nd Karlsruhe Workshop on cellular base stations. IEEE Journal On Selected Areas in
Software Radios. Germany. Communications, 17(4), 561–573. doi:10.1109/49.761036.
5. Jondral, F. (2002). Parameterization-a technique for SDR Imple- 27. Crochiere, R., Rabiner, L. (1983). Multirate digital signal
mentation. In W. Tuttlebee (ed.), Software Defined Radio processing. Prentice-Hall.
Enabling Technologies (pp. 233–256) Wiley 28. Bailly, B., Bidet, E., Cardin, J., Djoko Kouam, M., Joanblanq, C., &
6. Palicot, J., & Roland, C. (2003). FFT: A basic function for a Palicot, J. (1995). FDF a 512 FIR filter using a mixed temporal-
reconfigurable receiver. Papeete: ICT’ 2003. frequential approach. Santa Clara: CICC’95.
184 J Sign Process Syst (2011) 62:173–185

29. Tseng, B. (1989). Direct realization of the transpose structure for


FIR and IIR Filters. Signals, Systems and Computers Twenty-
Third Asilomar Conference.
30. Proakis, J. (2007). Digital communications, McGraw-Hill Higher
Education.
31. ARIB STD-T63-25.212 V4.5.0, Multiplexing and channel coding,
(FDD), (Release 4).
32. IEEE Std 802.11b-1999/Cor 1-2001
33. IEEE Std 802.11g-2003
34. IEEE Std 802.16 - Part 16
35. Leveiller, S. (2004). Quelques Algorithmes de Cryptanalyse du
Registre Filtré. Thesis Report, Ecole Nationale Supérieure des
Télécommunications.
36. Kitsos, P., Sklavos, N., Zewas, N., Koufopavlou, O. (2001). A
reconfigurable linear feedback shift register (LFSR) for the
bluethooth System. University of Patras.
37. Siclet, C., Siohan, P., & Pinchon, D. (2002). Oversampled
orthogonal and biorthogonal multicarrier modulations with Dr. Jacques Palicot received, in 1983, his PhD degree in Signal
perfect reconstruction. Santorini: DSP ’00. Processing from the University of Rennes. Since 1988, he has been
38. Helard, M., Le Gouable, R., Helard, J., & Baudais, J. (2001). involved in studies about equalization techniques applied to digital
Multicarrier CDMA techniques for future wideband wireless transmissions and new analog TV systems. Since 1991 he has been
networks. Annals of Telecommunications, 56(5), 260–274. involved mainly in studies concerning the digital communications
area and automatic measurements techniques. He has taken an active
part in various international bodies EBU, CCIR, and within RACE
and ACTS projects. He has published various scientific articles
notably on equalization techniques, echo cancellation, hierarchical
modulations and Software Radio techniques. He is currently
involved in adaptive Signal Processing and in new techniques as
Software Radio and Cognitive radio. From November 2001 to
September 2003 he had a temporary position with INRIA/IRISA in
Rennes. Since October 2003 he is with Supélec in Rennes where he
leads the SCEE department

Laurent Alaus was a student at the “Ecole Supérieure de Chimie,


Physique et Electronique de Lyon” (ESCPE Lyon), where from he
received the Diploma of Engineer in Telecommunications and
Electronical Engineering in 2006. The same years, he obtained a
Master’s Degree (Former Diploma of Advanced Studies) in Tele-
communications of the National Institute of Applied Sciences (INSA).
He dedicated his research of Master to forward-looking Signal
Detection Methods in Software Defined Radio (SDR) Architectures.
Since the first term of 2007, as a Phd student at CEA-LETI, Laurent
Alaus carries through a thesis on the Architecture of SDR Transceivers
related to Cognitive Radios in collaboration with the Signal
Communication Embedded Electronics (aSCEE), lab of the “École Dr. Christian Roland received his Ph.D. degree in physical systems
Supérieure d’Électricité” (SUPELEC). and metrology from the Conservatoire National des Arts et Métiers
J Sign Process Syst (2011) 62:173–185 185

of Paris in 2001. Over the last 10 years he has worked in the field of
radio communications. Since September 2002 he has been with
LESTER, Lorient, France.

Dr. Dominique Noguet received his Diploma of Engineer of the


National Institute of Applied Sciences (INSA) in electrical engi-
neering in 1992, an MSc in microelectronics of the University of
Strasbourg in 1994. He received his PhD of Polytechnic National
Institute Grenoble (INPG) in microelectronics with honors in 1998.
Dr. Yves Louët received his Ph.D. degree in Digital Communica- He joined the LETI ASIC design team and has worked in the field of
tions in 2000 from Rennes University, France. He has been with specific architecture for telecommunications from 1998 to 2001. He
SIRADEL from 2000 to 2002 where he was working as a Research then joined the telecommunication laboratory where he coordinated
Engineer on channel modelling for cell planning. He is now the RNRT PETRUS project from 2001 to 2003. From 2002 to 2004
Associate Professor in SUPELEC since 2002 and his research he was involved in the FP5 IST-STRIKE project that he also
activities concern PAPR analysis and PAPR mitigation methods in coordinated. He lead radio design activities in Magnet and Magnet
any contexts (OFDM, MIMO-OFDM, software radio, etc.). Yves Beyond (WP5). He was the coordinator of the ORACLE project on
LOUËT has already organized special sessions in ISSPIT 06 opportunistic radio and leads flexible radio activities (WPRC) within
(“Peak to Average Power Ratio of a Multiplex of Modulated the NoE NEWCOM++.
Carriers”, Vancouver, Canada), EW 07 (“PAPR mitigation in He received a best paper award and the best PhD award from
multicarrier systems” Paris, France) and URSI 08 (“Signal INPG. He authored or co-authored about 30 papers in peer reviewed
Processing advances for Software Radio”, Chicago, USA). Yves journals and conferences. He is currently the head of “digital
LOUËT is Member of the IEEE and Member of the French architectures and validation platforms” ASP laboratory at LETI,
Engineering Society (SEE). where he also leads Cognitive Radio Activities.

You might also like