Fieldbus technology in industrial automation

Jean-Pierre Thomesse

To cite this version:

Jean-Pierre Thomesse. Fieldbus technology in industrial automation. Proceedings of the IEEE, 2005,
93 (6), pp.1073-1101. �10.1109/JPROC.2005.849724�. �inria-00000807�

HAL Id: inria-00000807

https://inria.hal.science/inria-00000807
Submitted on 21 Nov 2007

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

Fieldbus Technology in Industrial Automation

Jean-Pierre Thomesse

solutions in a given sector. These companies realized the

Abstract— Fieldbus technology in industrial automation strategic importance of the fieldbus in the industrial
is not only relatively complex because of the number of automation systems. Once their products were developed,
solutions possible, but also, and above all, because of the
they pushed to have them standardized, and hence we have
variety of applications. Ironically, these in turn are
responsible for the multitude of solutions available. If the the current vast range of standards, de facto and de jure.
analysis of the basic needs is relatively standard, as they will This paper will try to establish the origins and current
always involve connecting sensors, actuators, and field status of the fieldbus technology, including technical and
controllers with each other, the options in architecture are scientific analysis, and standardization. Taking into account
numerous and can impose the need for certain services. The
the number of systems and the number of contenders, the
required performances themselves and the quality of service
expected fundamentally depend on the applications. history of the fieldbus is long and the different episodes are
This article traces this technology from its beginnings, numerous. Because of this, a lot of details will not be
which go back to the first industrial networks in the 1970’s. included, and we will simply focus on the essential steps in
The principal stages of development are recounted, from the its evolution.
initial requirement specifications to the current state of There are two main parts to this paper. The first focuses
international standardization. The diverse technical
solutions are then analyzed and classified. In particular, we on the origin of the concept of the fieldbus and on the
study the temporal aspects, the medium access control requirements which led to the beginning of fieldbuses,
protocols and application relationships. ending with the current state of standardization. The second
part is dedicated to the technical aspects, the services, the
protocols, and then the quality of services, ending with
Index Terms—Application relationships, Architecture, some communication architecture considerations. The
Client – server, Cooperation models, Fieldbus history,
Medium Access Control, Protocol classification, Publisher – conclusion will focus on perspectives and future possible
Consumer, Real-time, Standardization. evolutions.

I. INTRODUCTION II. FIELDBUS CONCEPTS: ORIGIN AND REQUIREMENTS

F OR about 20 years now, the word "fieldbus" has been

very widely used. Its common meaning is a network
for connecting field devices such as sensors, actuators,
A. How the fieldbus concept began
The fieldbus concept has several origins, but they can be
classified into two main groups: the end-user needs and the
field controllers such as PLC’s, regulators, drive
technological capabilities.
controllers, etc., and man-machine interfaces. But this is
1) Functional end-users needs
only an informal definition and needs more in-depth
analysis. Fieldbus technology represents a wide domain of a) Needs for standardization
problems which are similar, but not exactly identical in
It is appropriate to start with the history of fieldbuses, in
nature. Fieldbus technology involves a variety of solutions
order to understand the differences in approaches. As with
and techniques which, although frequently seen as closely
most histories, this one has a prehistory. The fieldbus
related, are different from each other. Fieldbus technology is
ancestors which stood out the most in industrial automation
a kind of technical, political and human adventure, which
were Modbus from Modicon (Modicon Bus) [62], [63] and
for more than 20 years has led to a lot of articles in
WDPF from Westinghouse (Westinghouse Distributed
journals, a lot of announcements, a lot of so-called scoops,
a lot of conferences and workshops, and a lot of products Processing Family) [140] [112] because of their seniority,
and standards. their functionalities, and their worldwide acceptance. Other
There are many standards. Why? networks, were already in existence, but did not go beyond
• The need for a fieldbus technology identified by a few specific applications or domains. For example, the
a number of different end-user companies in a network ARCNET (Alliance Research Centre Network)
number of different sectors. started in 1977 and primarily covered office communication
• The variety of possible hosts to be connected needs [114], before being used in data acquisition [19].
(the variety of sensors, of actuators, of Another network which was highly used in avionics and
controllers). aerospace applications was the Military Standard 1553,
Initially, there was no existing standard so each IT [56], [133]. In the nuclear instrumentation domain, the
(Information Technology) provider developed their own CAMAC [28] network, created in the 60s, is considered as
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

the first instrumentation network. Several proprietary

networks were in use at the end of the 70s to connect PLCs Fig. 2 Operational Architecture
(Programmable Logic Controllers) (Allen Bradley Data
HighWay and Tiway – Texas Instrument Way), [126] as In Fig. 2, a TOP network is situated between the
well as in the process industry [158], [136]. enterprise and the factory control levels, a MAP network is
between the latter and the cell control level, Mini-MAP or
b) MAP and TOP projects sometimes Proway (Process Data Highway) [52], [81] just
The integration of heterogeneous systems was difficult below that, and then finally the fieldbus network between
due to the lack of standards, and was expensive on account the machine control and the sensor-actuator levels, leading
of necessary gateways, adaptors, and protocol converters. It to the operational architecture. Mini MAP [58] or
was at this moment that two US companies started two MAP/EPA (MAP - Enhanced Performances Architecture)
projects, aim of which was “the definition of a standard [108], [125] was added, based on the FAIS (Factory
communication profile”. Boeing Company launched the Automation Interconnect System) specification [115]
TOP project (Technical and Office Protocol) [26], [36], and developed in Japan.
General Motors the MAP project (Manufacturing This notion of hierarchical architecture was also
Automation Protocol) [41], [145], [48], [79]. The main developed in process industries, [159], [11] but with a
objective of the MAP project was the definition of a difference in functions. Indeed, the following layers were
standard communication profile suited for communication most often considered: a first layer for reflex automation, a
between the design offices and factories, and inside the second for the supervision and a third for optimization.
factory, between workshops and machine tools or robots.
d) Wiring simplification needs
General Motors wanted a communication profile for
manufacturing applications that could allow all devices to At the lowest level of communication, before the
communicate without having to develop specific hardware fieldbus era, a lot of standards reigned, for example, the 4-
and/or software. Boeing’s idea with the TOP project was 20 mA standard for analog sensors or the 0 – 24 V for
similar but concerned another issue, namely communication digital inputs, etc. These standards led to a cabling of 2
between business and technical offices. wires for each analog point and for each Boolean point
(true, false), or each binary digit in a number. The result
c) The concept of CIM and hierarchical was the need for a great number of cables in the factories.
architecture The design and installation of the wiring were expensive
Computer Integrated Manufacturing refers to automated operations, and maintenance or evolution was difficult. This
manufacturing, automated transport of pieces and materials, was one of the reasons why end-users requested a solution
using computer technologies at all the stages of a product for simplifying these operations: the fieldbus was an answer
from its design to the manufacture and the quality control. to this request. This need had already been stated in 1971
The idea of structuring applications in a hierarchical way by [80].
abstracting levels has been used for decades to simplify
their design. Fig. 1 shows the five levels model of 2) Technological capabilities (enabling technologies)
application architecture defined by the NBS (National
Bureau of Standard) in the USA [13], [79]. These models a) OSI-ISO model
were initially functional, meaning that the main interest lay In 1978, work on the communication reference
in the function organization, but not in how it was architecture model started in ISO (International Organization
implemented. They were based on the structure of discrete for Standardization), and was to become the OSI model
part manufacturing factories. It was later that they were also (Open System Interconnection) that we all know today
used as implementation models (or operational models) (as [168], [86]. This model, originally conceived for computer
in Fig. 2). interconnection, brought the right concepts for the
Each network governs the functions of the layer below understanding of data communication, for the design and
and serves as an interface for the layer above. This is how a standardization of new communication protocols. The 80s
MAP network in a factory works. All the controllers of were to be a very rich period in creativity and innovation in
cells or of a workshop are connected to the MAP backbone. the field of services as well as in protocols.
But each of these controllers is connected to a Mini-MAP The OSI model was sometimes poorly understood, we
network which interconnects the machines in the cell. And will come back to this point later in the second part of the
each machine can use one or several fieldbuses which paper (in the first section dedicated to the OSI model).
interconnects the instrumentation to the machine
controllers. Notice that the fieldbus and the Mini-MAP b) Local Area Networks and MAC protocols
locations and roles are more or less similar. In Local Area Networks, the stations shared the same
transmission support. Logically, without some kind of
intervention, all stations would transmit simultaneously.
Fig. 1: CIM Architecture issued from [13] For one station to send at a time, it was necessary to
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

develop medium access control (MAC) protocols. Some of the most important reason for fieldbus development was the
these protocols were called deterministic, i.e. a transmission awareness that it could become the backbone of the future
could occur within a bounded delay. The other protocols, distributed and real time systems for automation. (And then
which did not have this property, were called non- the bone of contention for the competition between
deterministic. A deterministic MAC protocol based on the automation companies.) Thus, the specification and the
token mechanism [88] was chosen by the MAP project. The design of numerous fieldbuses began. An initial experiment
TOP project chose a non-deterministic protocol called of a digital fieldbus (1981) was carried out by Brown
Ethernet (CSMA-CD) (Carrier Sense Multiple Access – Bovery Company and Electricité de France with the KSU
Collision Detection) [87]. network at the Thémis [53] solar power plant in the south
Because Ethernet lacked the ability to guarantee latency of France.
delay, research for making Ethernet deterministic led to the In parallel with this innovative design work was the real
protocols known as CSMA-BA (Carrier Sense Multiple start of protocol engineering activity, formalization of
Access – Bitwise Arbitration), CSMA-CA (CSMA – protocols in terms of automata, Petri Nets, etc., and proofs
Collision Avoidance), [16], [91], CSMA-DCR (CSMA – of property, development of languages for specification
Deterministic Collision Resolution) [101]. (ESTELLE, LOTOS, SDL), conformance testing methods,
With all these varieties of MAC protocols, Local Area conformance testing procedures and institutions,
Networks exploded. It was attractive to specify one’s own arrangements and recognition between national
protocol, well suited to one’s need. The trend was organizations.
facilitated by the progress made in microelectronics, and
B. Development of fieldbus
design automation.
From another point of view, the LAN technology gave In the beginning of the 1980s, several projects started in
an opportunity to a lot of users to experiment with the Europe after the MAP project had began in the US. In
distribution of applications. It was a great temptation to France, the FIP fieldbus project saw light in 1982 under the
experiment with distributing functions on microcomputers, aegis of the French Ministry for Research and Industry. It is
and testing their cooperation through a network. It was the a similar process which led to the PROFIBUS fieldbus
moment in the evolution of industrial applications that project in Germany in 1984, and to the P-Net [40] project
Digital Control System (DCS) or Direct Digital Control in Denmark in 1983. At the same time, in 1983, the Bosch
(DDC) migrated to Distributed Control System (DCS) Company developed the specifications of CAN for cars
[110], [131], ultimately leading to the systems used today. manufactured in Germany [16], [91], [123]. FIP [151]
stands for Factory Instrumentation Protocol and is now
c) Microelectronics and Integrated Circuits known as WorldFIP. PROFIBUS [9], [37] stands for
Process Field Bus, CAN stands for Controller Area
The 1970s and the 1980s saw the development of
Network.
microelectronics, of semi-custom and full-custom integrated
The standardization process began at this time in these
circuits, the development of microcontrollers, and of DSP
different countries and at an international level, with IECTC
(Digital Signal Processing). These were the state of the art
65/SC65C/WG6 [51], simultaneously with ISA in the US
technologies that made it possible to design new
(in the ISA SP50 (Instrumentation Society of America -
communication controllers. The first I2C network (Inter
Standard Practice)).
Integrated Circuits) was created in 1982 by Philips for the
This beginning shows, with the number of fieldbuses
interconnection of ICs in television sets [124]. However,
now, that ideas, old or new, were not and are not lacking.
the perspective was not only the integration of protocols
The contenders for the IEC international standard at the
into silicon, but also the capability to put “intelligence”
early beginning were classified into two sub-groups: the
inside the smallest device, inside any sensor, or actuator.
first group included solutions based on existing protocols;
This digital treatment capability found in each sensor and
the second group included only new paper-proposals
actuator necessitated new communication means [59]. This
without experiments. Some details on these proposals can
was another reason for the development of the fieldbus, and
be found in several publications, [15], [164]-[167]. Two
was stated in a report from Professor Soutif of Grenoble
fieldbus types were to be considered, the H1 fieldbus at a
University [143] and during a dedicated colloquium in the
low data rate for the connection of some sensors essentially
UK [65].
in process control, and the H2 fieldbus at high data rate for
manufacturing or for interconnection of several H1
3) Conclusion
All the elements were in place for entering into the networks.
fieldbus saga. The discussion at that time centered on 1) 1st group
sensor and actuator networks, or instrumentation networks.
a) ERA Technology
The term “fieldbus” had not yet been coined. It would
appear only in 1985 at an IEC meeting. The UK company ERA Technology proposed a fieldbus
The needs were many and the provider companies based on the existing Mil Std 1553B. The proposal
recognized great potential in this emerging market. Perhaps extended the current standard for physical performances:
• 1900 m at a data rate of 62,5 kbit/s, 750 m at
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4

the data rate of 250 kbit/s, and 350 m at the a) FIP

data rate of 500 kbit/s; The FIP requirements, published in 1984, were
• changes to specifications for spur isolation developed by a group of end users and labs. The
resistors; requirements focused on the application needs, periodic
• optional addition of power; updates of data, independence of addresses and locations,
• 32 nodes possible with power and active coherence and consistency of data. The FIP Club was
repeaters. created in 1986 for the promotion of these specifications.

b) IEEE P1118 b) PROFIBUS

A US group proposed a fieldbus based on the P1118 The PROFIBUS-FMS (Fieldbus Message Specification)
project (based on Bitbus [10] from Intel) dedicated to fieldbus was already a technical solution: a character-based
exchanges between microcontrollers for all types of transmission, according to the RS 485 standard, with a
applications. The specifications covered the physical, the token passing method distinguishing master and slave
data link, and the application layers. stations. Only the master stations were included in the
Physical layer: virtual ring. FMS was a subset of MMS (Manufacturing
The covered distance was from 2000 m to 5000 m with Message Specification) [89].
data rates between 50 and 500 kbit/s, with 250V isolation,
optional intrinsic safety, and power with signal. The 3) Organization positions
proposed medium was a twisted pair with possible
redundancy. a) IEC TC65C/WG6
Data link layer The technical committee IEC TC65C/WG6, after having
A master/slave protocol with an optional back-up master defined Proway (PROcess Data Highway) (IEC 955) [52]
was required. In case of failure with initial master, the back- (Gellie, 1980) “Inter sub-system communications for
up master assured availability. industrial process”, was in charge of the fieldbus
Application layer standardization after a meeting in Montreal (Canada) in May
Different types of messages and services were specified 1985.
(broadcast and multicast, datagram, acknowledged
datagram, connection oriented) with a response time b) ISA SP50
between 10 ms and 50 ms, and a minimum of 1 ms, to The American position was given by ISA SP50. ANSI
ensure the physical procedures. The fieldbus had to be entrusted ISA for the definition of the American Fieldbus
optimized for small frames (128 bytes), with downloading Standard. The position was that it was not necessary to
and task control, management tools for device status. The develop a specific American Standard, as they had to
P1118 proposal scope was underlined for distributed cooperate directly with the IEC committee for a unique
intelligent devices in all industries. international standard. After a "call for proposals", the
diversity of the protocols proposed made any convergence
c) Foxboro proposal difficult, if not impossible. The group SP50 of ISA, then
The Foxboro company presented two complementary defined their requirements, which were not very different
solutions for the H1 applications, using enhanced HDLC from those of IEC, and tried to find a common solution
with Manchester encoding and baseband communication among the proposals (Rosemount, PROFIBUS, FIP, ERA,
and HDLC with NRZI encoding, and RS 485 for H2 Foxboro).
networks. The enhancements of HDLC were related to the
error detection mechanism. c) NEMA
The number of undetected errors were to be less than one In the US, the NEMA (National Electrical Manufacturers
such error in 40 years [6] (Armitage et al, 1988) at a data Association) created a task force (SC21) which worked with
rate of one Mbit/s. ISA SP50 and IEC to determine a single American and
international standard.
d) Rosemount proposal
Rosemount Inc. presented two solutions, one for the H1 4) Conclusion
bus using IEEE 802.4 with FSK phase coherent, and one The proposals were very different in terms of
for H2 using IEEE 802.4 with FSK phase continuous. requirements and solutions. Many concerned the physical
Rosemount started the development of the Hart system in layer, and related aspects such as connectivity, topology,
1985. distances, without any deep investigation into the
functional application aspects. All this was something new.
2) 2nd group For the first time ISA and IEC had to consider:
Two European proposals, FIP and PROFIBUS were only -on the one hand, existing products,
paper proposals at this time. -on the other, paper proposals, largely based on
different views of what a fieldbus should be.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

The decision was to write requirements in order to define • More advanced control strategies can be
a fieldbus. The contenders tried to push for the implemented because of improved field data.
requirement(s) corresponding to their solutions. These The description of the devices consisted essentially of
requirements are presented in the following section. (for each type of sensor):
• the maximum message response time (time
C. Fieldbus requirements
between request and delivery of information),
The establishment of fieldbus requirements started the • the message frequency (in average).
standardization, by both the IEC and ISA committees. The “information flows” part dealt with
Before examining the different proposals, it was decided to • the design philosophies (grouping of devices on
first express the requirements before choosing or defining a a bus based on functional analysis, on
standard solution. ISA SP50 gave a questionnaire to all the geography, etc.),
members to try to state the real needs of the user. ISA and • the bus control and the exchanges (master/slave;
IEC committees started writing the requirements in 1986. peer-to-peer, etc.),
Without going into details, a very deductive approach • the address allocation,
was advocated and described in an ISA document entitled • the fieldbus topology (with distinction of
“Field Instrument Bus Standard Specification” [82], [83]. lengths between master and junction box and
But it was difficult to follow the stages described in this between junction box and slaves),
document strictly because of the members’s various levels • the fieldbus size in number of stations,
of progress. Some were working on the needs analysis, • the redundancy possibility.
others on the protocol specifications and still others on the The application environment analyzed the power
first implementations. However, it was only in February requirements, the type of wires, the insulation requirements
1987 that the first version of the final requirements was and the capability to support flammable atmospheres.
drafted. For one year, new needs or requirements were As can be seen, the questions were very end-user oriented
proposed at each meeting. But work on the definition of a at this early stage; the environment and management were
solution started in Spring 1987 [164]-[165]. Therefore, we the two key points of the questionnaire. Technical
can see an evolution in the requirements from end users’ communication aspects were not dealt with, except on a few
needs at the beginning moving to more and more technical points such as the notion of masters and slaves and
aspects in the later versions. bidirectional communications. It was implicit that the
This next section presents, first, the questionnaire by fieldbus had to provide two services READ and WRITE.
ISA SP50, then a table summarizing the requirements Other services were not considered.
issued from IEC and ISA. Some requirements from the FIP The committees were very optimistic. At the end of
proposal are included [50] because of their specificity in this 1986, it was expected that the functional guidelines would
arena; an example of operational architecture issued from be available in January 1987, and a standard set in June
[166] is also given. 1989 [84], [85]. The tables of contents of two future
documents (Architecture and Overview, Messaging Service)
1) ISA questionnaire were published in the working groups on December 11,
ISA published a 15 page “Discussion draft and 1986.
questionnaire for functional requirements” [82]. This
document discussed the requirements for a “low level” 2) Requirements summary
industrial fieldbus that connected field devices to higher- The following table (Table I) summarizes the
level monitoring and control systems. Some of the requirements from IEC, ISA and WorldFIP.
following features were used to distinguish a “low level”
field bus from a “high level” bus system such as Proway or
MAP. It was structured in four chapters entitled
respectively: Benefits of fieldbuses, Describing field
TABLE 1. PRINCIPAL REQUIREMENTS
devices, Information flows, and Application environment.
Seven benefits were identified, and each was to be
N.B. The response time is defined as follows:
qualified according to its importance from greatest to least.
for IEC: time delay between event occurrence and
The benefits were:
signaling
• Lowering the installation costs,
for ISA: the time elapsed between a request and the
• Ease of adding field devices,
delivery of information
• Providing two-way communication with field
devices,
• Improving the accuracy of information delivered
3) Architecture and functional aspects
at control room, Fig. 3 explains the position of the fieldbus and the types of
• Enhancing the maintainability of field devices, equipment to be connected. This presentation of distributed
• Providing remote access to measurement data architecture was issued from the IEC functional
through handheld interface,
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6

requirements dated July 1986, which can be found in [164] manufacturing.

. This standard provides for serial digital communication to Regarding the mechanisms relevant to MAC and LLC,
and from field devices, it also provides for attaching more the question seemed to be eluded; they were to become the
than one addressable field device on one bus”.. The general major point of discussion and the stumbling block
requirements were introduced as follows. throughout the following years.
“The fieldbus will be a serial digital communication The fieldbus was not yet considered as a real time
standards which can replace present signaling techniques network but as relative to MAP and Mini-MAP [6]
such as 20 mA and 24 VDC, so that more information can (Armitage, 1988). This was reinforced by the position of
flow in both directions between intelligent field devices and PROFIBUS FMS which appeared at this moment more as a
the higher level control system over shared communication mini-MAP network than a fieldbus.
medium.” It was after the appearance of fieldbus in other
applications that the concept of a real time network would
Fig. 3. System with MAP/PROWAY and fieldbus (issued second draft – be considered at the international level for standardization.
Field Bus standard for use in industrial control systems. From [166].
It was after the publication of real time communication
Just one fieldbus is needed to allow multi-point needs by the European MAP’s Users’s group [45], [127],
attachments for a number of addressable devices. The most that the work group ISO TC 184, SC5, WG6 TCCA was
common justifications for this design are: created to study real time communication independently of
Better quality and quantity of information flow the network’s position in the system’s architecture.
Save cable and installation cost D. Application domains
Ease of adding or deleting field devices in a system At the beginning of the fieldbus era, only two main
Fewer connections to devices mounted on moving domains of application (process control and discrete
equipment manufacturing) were considered by the standardization
Fewer penetrations through process containment walls bodies (IEC, and national organizations). We have seen that
Save cable and installation weight the requirements were quite similar in both cases. And now,
Reduce installation errors as fieldbus technology has penetrated all application
Reduce terminal and junction boxes”. domains, it is interesting to observe their similar needs,
even a posteriori. This section analyzes the different
4) Conclusion application domains in order to show that the variety of
This conclusion resumes the official requirements from
fieldbuses existing today are also a result of these different
IEC-ISA and adds some comments related to real time
requirements.
networking. The requirements stated that two fieldbuses
The applications can be seen as a set of criteria for
were needed, H1 and H2. Even if they presented some
classification of industrial local area networks and, more
similar functionalities, they differed in speed, distances
especially, fieldbuses. The criteria are related to:
between stations, number of stations, and services to the
• the types of traffic, which influence the services
user. We can see the needs expressed on the one hand for
and the required quality of service (real time
the low speed process control applications (H1), and on the
constraints, synchronization needs, etc.),
other hand for high-speed process applications (H2) (in
• the environment characteristics (EMC
discrete part manufacturing or in certain process control).
[ElectroMagnetic Compatibility], intrinsic
The former required a robust physical layer, with a powered
safety, power),
bus, with intrinsic safety, and possible reuse of the existing
• the dependability constraints (availability,
wiring, but in terms of services, it needed READ and
reliability, safety, security, etc.).
WRITE services, without particular synchronization needs.
We will not analyze the environment and dependability
The latter expressed more requirements in terms of
characteristics because they are too dependent upon the
synchronization and distributed control.
application itself, and dependent even on the location of the
Both concluded that fieldbus traffic was either periodic or
application. Also, the presence of perturbations is not the
aperiodic, that it was composed of data and of messages.
property of a given domain; for example, a flammable
The data was coming from or going to the final elements in
atmosphere may be encountered in several contexts.
the devices; it was transmitted with status. The messages
The question of real time is common to all domains as
contained other information.
well. It is not expressed in the same “units”, but the
Regarding the services provided to the user, the
constraints are potentially the same. This question will be
requirements cited the services for the exchange of values
addressed in the next part.
(READ, WRITE, Information Report or Notification) and
Now, as traffic is always periodic or aperiodic, we will
for synchronization. Time stamping was required, but the
briefly analyze its characteristics for the following types of
concepts of consistency or of coherence were not really
applications: discrete manufacturing applications, process
recognized as necessary. The concept of response time, even
control industry including energy production, building
if cited and quantified, was not really studied. A rapid
automation, control of utilities networks, transportation
classification distinguished process control and
systems, and embedded systems.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7

1) Discrete Manufacturing applications In terms of environment, continuous processes,

A discrete manufacturing application is characterized by especially in the chemical industry, are the domain of
the fact that, between two operations, a product is in a intrinsic safety for devices powered by the network. It is
stable state, i.e., it is not damaged if stocked between these also the domain for protection against electromagnetic
operations. This criterion allows for decomposing the perturbations (industries with electrical motors).
application into sub-applications relatively independent Redundancy is often desired, sometimes necessary, for
from a time point of view, with each sub-application being dependability and safety (people, environment, production
attached to an operation or to a machine. tools).
In such applications, it is then natural to distinguish the 3) Building automation
communications within a machine from those between This type of application concerns the surveillance of
machines. In the former, the stations are the sensors, the houses or buildings, access control, heating and air
actuators, the axis controllers, the regulators, and other conditioning, management of utilities and electric domestic
PLC’s. Traffic is essentially periodic and is relevant to the appliances. The applications are more relevant to data
fieldbus. There are needs for broadcasting, for distribution acquisition and supervision, than to control. The control
of control algorithms and for synchronization between functions are often very simple.
application processes. The range of sensors is very broad. A lot of sensors are
In the latter, the exchanges are more asynchronous On/Off (open/close, enable/disable). Others measure the
(production orders, report of activity, downloading of usual physical input variables (temperatures, levels, speeds,
programs (or of “domains” in MMS terminology)). The etc.). Finally, some are camera-based and need image
synchronization between the sub-applications is more analysis for remote monitoring. The real time constraints
relevant to production management and productivity criteria are not numerous. Only some cases, such as burglar alarms
than to real time constraints and process dependability or of or control of elevators and access, are constrained.
product quality criteria. Indeed, such a synchronization As far as the wiring is concerned, it usually represents a
failure does not affect the process nor the product. Even if significant part of the cost. Therefore, wireless or power line
this inter-machine traffic may be supported by some communications are being used more and more in this kind
fieldbuses, it is more the role of cell networks, or factory of application. The environmental conditions are not really
networks. too demanding, but the great number of devices and
The environment depends essentially on the factory controllers lead to very complex systems. All possible
domain and may lead to the use of special media adapted to topologies must be available for adapting the architecture of
the EMC. the system to any type of building or group of buildings.
The reliability criterion applies to the fieldbus inside a The dependability is not specified as in industry but it is
machine, for the quality of the products and for human also very important:
(operators) safety. Other criteria, such as availability, are not • reliability, an elevator or a heating system cannot
usually critical but are important for productivity. fail on Christmas evening,
2) Process Control industries • availability of communication resources in case of
The continuous processes are characterized by the fact an emergency in a remote health care monitoring,
• safety and security for protection against
that the products are continuously produced through a
vandalism, or unauthorized people.
sequence of operations (assumed by different machines)
.
with no stable state between two successive operations.
4) Control of utilities networks
Iron-steel industry processes, many chemical and biological
These applications consist of the remote monitoring and
processes, the paper mill industry, and energy production
the control of very large networks for the distribution of
are considered in this category.
water, gas, hot steam, or electricity. They are no longer
Inside a given machine, the traffic is very similar to that
located in a small area such as a factory or a building, and
described in the previous section. The fieldbus must assume
the networks are no longer really local area networks, and
real time traffic between sensors, controllers and actuators.
yet, these applications really are of the same nature as the
But considering the need for synchronization between
previous. The functions are the remote monitoring and
successive machines, some real time traffic introduced by
control of stations (pumping, stocking, transformer
the distributed control between the concerned controllers is
stations), pipes and lines. Operators are only in central
also supported by the fieldbus.
control rooms for exploitation and maintenance
The characteristics of these processes differ essentially in
organization.
their time constants, e.g., the speed control in a steel mill
Traffic consists of status variables and events as well as the
and the temperature control in a blast furnace. The lack of a transfer of information between the intermediate stations of
stable state leads to very strict time constraints for the fluid/current transport.
synchronization of operations, e.g., the controller The synchronization of data acquisition is often
coordination of the sequential elements of a rolling mill. important for establishing the order of events. The data rate
The real time quality of service depends on the criticality of depends on the complexity of the system considered.
the application. The networks that convey the data for monitoring and
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8

control have the same dependability roles as a fieldbus in a also for maintenance and management. We will see that if
factory. The only difference is in the distances covered; the the requirement is relatively simple, the solutions are
medium and the physical layer protocol must be adapted to numerous (in terms of protocols). Other functions are
the distance. Power line protocols are used in electrical required but as options or “nice” functions. Synchronization
networks. Optic networks are used in transformer stations. is one of these functions. It is, nevertheless, necessary for
Radio waves are often used to connect very remote stations. the management of distributed systems. The fact that this
It is also now a preferred domain for Internet use. function was not considered twenty years ago shows that a
5) Transportation systems lot of people did not think that the fieldbus would change
A transport system is an infrastructure for the transport of the application architecture and design. They thought it
people and/or freight. The applications in these systems would only be a simplification of the wiring.
cover the management of a railway network, the remote Consequently, the only cited time constraints were the
control of urban traffic, the monitoring of highways, etc. response time and the frequency of the exchanges, which
Traffic is composed of status variables, events, and device allow for a very simple calculation of the load on a fieldbus
command and control. The topology of the fieldbus and then its proportioning.
depends on the geography of the system considered. The The maximum values given for each fieldbus (maximum
safety constraints are often very important. Dependability is length, maximum number of stations, maximum data rate,
also very crucial, especially availability, even during maximum frequency of data update, etc.) are limitations for
maintenance or updating operations. each application design. Because of these limitations it is
It is possible to include in this category of applications, sometimes necessary to use several fieldbuses (and other
the control and management of telecommunications networks) on which the architecture of the application will
networks (telephone networks, mobile networks, etc.) then depend.
6) Embedded Systems This notion of architecture is not as simple as is usually
These systems are now in many products, from cars to understood. The word “architecture” is sometimes used for
buses to trains, but also in major electrical domestic topology, it may be used with the same sense as in the title
appliances (refrigerators, etc.). In vehicles, the application of the OSI model, and in this case, it then represents an
consists of various functions: organization of services and protocols. Here, the word
Control of the motor(s), of the braking system, of the “architecture” represents the organization of the automatic
stability, of the gearbox. control application implemented around a fieldbus and other
Assistance to the driver, or to automatically pilot the industrial networks. Architecture is typically defined by
vehicle (as in some trains) diagrams as seen in Fig. 3. The question of architecture is
Other functions are related to the energy consumption, inherent in the requirements for setting up a fieldbus. This
such as optimization, was not the question before.
Management of lights, glass-cleaner, Fig. 3 shows an operational architecture, because it
Management of passenger access in trains, ticketing, indicates the devices and the fieldbuses actually in
Maintenance operation. But this kind of diagram leaves much to be
The distances are short; the environment possibly very desired. Indeed, nothing is said about the functions
demanding (in cars, for example); safety is a major implemented in each station represented by a box, and
constraint. nothing is said about the cooperation between the functions
These applications present time constraints depending on nor the exchanges supported by the fieldbuses and other
the functions considered. The motor is controlled every 10 networks. Before designing such an architecture, it is
ms; the response time of a braking request must be as short necessary to carry out a functional analysis which must
as possible. specify the application, the functions and sub-functions,
The term “embedded system” is also used for different their interactions, and their communications. The result is a
equipment such as refrigerators, coffee-machines, washing functional architecture which ideally may relegate the
machines. Each time a piece of equipment is built with components and the networks to second place, focusing on
“intelligence” and communication capabilities, it can be the functions. A functional architecture being specified, the
considered as an embedded system. We also speak of designer has to choose the networks, the components, the
“ambient intelligence” [1] with the expectation of a fully devices, and the distribution of the functions in the devices.
communicating environment and many autonomous This is the real design stage of the solution, taking into
intelligent devices in the near future. New problems will account the constraints, defining which station is connected
occur such as connectivity, safety, confidentiality, and to which fieldbus, and distributing the functions in the
integrity, etc. devices.
7) Synthesis It is only after this stage that the choice of the
This brief study of the application domains shows that architecture takes place, along with the choice of the
the basic fieldbus specifications in terms of functions and of fieldbus. Now, it is the choice of the fieldbus which has an
services are very similar in each of the applications. impact on, and sometimes imposes, the choice of the
The exchange of data (values, status and events) is the architecture, depending on the services provided by the
main function of the fieldbuses for automatic control, but
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 9

communications system. It is clear that the existence of incompatible fieldbus families.

certain services determines the distribution facilities. 2) IEC Standardization
These architectures will be applied to all domains of After a lot of episodes and developments, the IEC 61158
application, with a hierarchy from the first level fieldbus, standard, including a large set of services and protocols, is
up to the highest level. Let us take some examples. defined as follows. It is structured by layer, according to the
In a train, the usual architecture has two levels; each OSI model architecture reduced as mini-MAP, or MAP –
wagon has its own fieldbus, and another fieldbus EPA, to the Physical Layer, the Data Link Layer including
interconnects them all and has a gateway to the external the Medium Access Control and the Application Layer (cf.
world. The first section of the second part).
In a building, an architecture with three levels can be
considered; here, each apartment has its own fieldbus; they, a) IEC 61158 standard
in turn, are interconnected by a floor fieldbus, themselves The main standard is IEC 61158. The first standard (IEC
interconnected by a building fieldbus. Other fieldbuses may 61158-2), published in 1993 [67], defined the physical
be associated with the control of the elevators. layer.
In the control of a pipe (be it for gas or water, or some The other parts are:
other medium), we could have a fieldbus along the pipe, 61158-3: Data Link Layer service specification
structured in “segments” of the maximum length for the 61158-4: Data Link Layer protocol specification
chosen fieldbus. Each fieldbus of a segment connects all the 61158-5: Application Layer service specification
devices of the segment, and a special site serves as 61158-6: Application Layer protocol specification.
concentrator. The concentrators may be interconnected by These specifications are a collection of different national
another upper level fieldbus or network. standards or specifications.
The data link parts (IEC 61158-3 and 61158-4) cover 8
E. Current Standardization
types listed below:
1) Introduction Type 1 is the TR1158, the compromise standard
Fieldbus standardization is a subject which has led to a proposal refused by a minority of members in 1999, which
great number of publications for the past twenty years. led to a publication of indignation by Patricio Leviti. [104].
Regularly, during the 90’s, journals published the progress Type 2 is the ControlNet specification,
of the standardization process (Control and Instrumentation, Type 3 is the PROFIBUS specification,
Control Engineering, Measurement and Control). Words Type 4 is the P-Net specification,
such as “war”, “battle”, “winner”, “peace” appear in the Type 5 is the FOUNDATION Fieldbus specification,
titles [25], [122], [47]. Optimistic opinions (“a standard Type 6 is the SwiftNet specification,
will be obtained at the end of the 80’s”) expressed in 1987 Type 7 is the WorldFIP specification,
all the way up to 1996 [54] decrease after this period. Type 8 is the INTERBUS specification.
Skepticism and interrogations start to appear [135], [119], The application layer specifications (IEC 61158-5 and
[64], [163], [117]. 61158-6) covered 10 different types. The first eight are
Several updates on the situation have been regularly associated with the data link layer. The two others, Type 9
published, some are listed here in chronological order: [32], and Type 10, define the FOUNDATION Fieldbus H1
[60], [153], [46], [128], [154], [35], [33], [138], [104]. network and PROFInet, respectively.
The international standardization concerned essentially Two other parts were planned, 61158-7 for network
the IEC, but also the ISO (TC 184 SC5 WG6), which was management and 61158-8 for conformance testing
in charge of the TCCA (Time Critical Communication procedures. They were cancelled, because of the existence of
Architecture) specification. In Europe, in the middle of the proprietary tools for configuration and network
1990’s, the CENELEC decided to define European norms management, and for conformance tests of the different
while waiting for an international solution from IEC. types. The maintenance of this standard is now entrusted to
Indeed, at the beginning of the 90’s, different lobbying the SC65C/MT1, (MT stands for Maintenance Team).
groups appeared only to disappear, OFC (Open Fieldbus
Consortium), IFG (International Fieldbus Group), ISP, b) IEC 61784
(Interoperable System Project), etc. There was no
A project for a new standard has also been started for the
compromise and no possible consensus between any two
definition of the CPFs (Communication Profile Family)
opposing blocks. Even though a group of experts wrote a
inside the IEC 61784 standard. Its objective is to clarify the
complete specification for the Data Link Layer [70], [71],
situation created by the number of variants and options in
including the different concepts and mechanisms, the
the IEC 61158 standard. While it defines services and
opposition continued. Because this specification was
protocols by layer, according to the OSI model reference
refused by the minimum minority in 1998, and caused a
architecture, the 61784 standard proposes a specification for
great problem at the highest IEC level, the Committee of
a complete stack of protocols based on the previous
Action of the IEC issued some kind of ultimatum for the
options. These communication stacks are called “profiles”.
working group. The result today is the current content of
This standard is composed of two parts; the first, the
the IEC 61158 with eight completely heterogeneous and
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 10

IEC 61784-1 standard, is composed of 18 profiles; and the Fieldbus was added to EN 50170 as an addendum in 2000,
second part, 61784-2 on Real-Time Ethernet, in project jointly with ControlNet [27] and PROFIBUS-PA [39].
(work started mid-2003), is composed of 9 proposals, all
based on Ethernet. b) EN 50 254 [22], [23]
This new project has been entrusted to the new working The EN 50 254 was also published to include fieldbuses
group IEC SC65C WG11. It is structured as follows. with higher performances for the transmission of short
Structure of IEC 61784-1 frames: INTERBUS, PROFIBUS DP, and Device
The current CPF are defined in the first part. WorldFIP [3], [22].
CPF 1 FOUNDATION fieldbus CPF 1/1 H1, CPF 1/2
HSE (High Speed Ethernet) c) EN 50 325
CPF 2 ControlNet CPF 2/1 ControlNet, CPF 2/2 The EN 50 325 standard covers the profiles derived from
EtherNet/IP the CAN protocol (and of the ISO 11898 standard), as
CPF 3 PROFIBUS, CPF 3/1 PROFIBUS-DP, CPF 3/2 DeviceNet, SDS, CANopen, which are also parts of the IEC
PROFIBUS-PA, CPF 3/3 PROFInet 62026.
CPF 4 P-Net, CPF 4/1 P-Net RS 485, CPF 4/2 P-Net
RS 232 d) EN 50 295
CPF 5 WorldFIP, CPF 5/1 WorldFIP, CPF 5/2 The EN 50 295 standard is a standard defining AS-i
WorldFIP Device WFIP (Actuator and Sensor Interface) protocol [7].
CPF 6 INTERBUS, CPF 6/1 INTERBUS, CPF 6/2
INTERBUS TCP/IP, CPF 6/3 subset F. Conclusion
CPF 7 SwiftNet, CPF 7/1 SwiftNet transport (without This first part presented the history of fieldbus and its
application layer), CPF 7/2 Full stack. requirements. They were written between 1984 and 1987,
after which, choosing a standard was possible. But it was
Structure of 61784-2 in project not to be so simple, particularly with the development of
CPF 2 ControlNet other standards by other committees, especially in the ISO,
CPF 3 PROFIBUS, PROFInet and with the start of other fieldbuses for car automation,
CPF 6 INTERBUS building automation, trains, etc. Obviously, the entire
CPF 10 VNET/IP (Virtual Network Protocol) story was not to be played out in a single scene.
CPF 11 TCnet The concept of architecture must not be forgotten because
CPF 12 EtherCAT (Ethernet for Control Automation the ultimate desire of the end user is really not a fieldbus,
Technology) but an operational architecture, which meets his needs in
CPF 13 EPL (Ethernet PowerLink) terms of dependability, in terms of performance, and in
CPF 14 EPA (Ethernet for Plant Automation) terms of cost.
CPF 15 MODBUS-RTPS (Real-Time Publish –
Subscribe) III. FIELDBUS TECHNICAL ASPECTS: SERVICES AND
3) ISO standardization PROTOCOLS
In 1990, a new work item for the ISO TC184 SC5 WG6 This second part is dedicated to the technical analysis of
TCCA group (Time Critical Communication Architecture) fieldbuses.
was started, following the analysis of the MAP experiments What services are provided by a fieldbus? According to
to define real-time communication requirements and what protocols? According to what communication stack?
recommendations [127]. The European MAP user group Looking at the requirements, we can see that some are
published a list of requirements for real-time already structured in terms of OSI layers: more generally,
communication. At the same time, the fieldbus appeared as some of them address the services required by the end-users
a real-time network [130]. The study of a communication (i.e., the application layer in OSI terms), others are related
architecture was published as a technical report [57], [90]. to the physical transmission and coding, and still others
Following this work, the network management of TCCS express properties or performances.
(Time Critical Communication Systems) was also studied For twenty years, all papers and contenders (except some
[93]. such as LonWorks [105]), agreed with the fact that a
4) CENELEC Standardization fieldbus is designed according to a reduced OSI model.
Four European standards have been published and But what is the reality? The first section will analyze this
updated several times in order to provide international model and the particularities of fieldbuses regarding the OSI
standards where the IEC lacked. concepts.
The OSI model gives the structure for analyzing the
a) EN 50 170 [21] different technical aspects, from the topology and the
The EN 50170 was published in 1996 with three national cabling to the application services provided to the users. We
standards: P-Net from Denmark, PROFIBUS-FMS from will follow this structure for presenting the different
Germany, and WorldFIP from France. FOUNDATION choices, the different solutions to fulfill the requirements.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 11

A normal approach would lead to choosing the services is included in the data link layer, because it is thus in all
from the requirements, and then the protocols from the fieldbus standards. The MAC is obviously necessary and all
chosen services and required quality of service. But as was existing protocols can be used.
already said, the diversity of the applications, the
approaches and the competitors did not allow for simple c) Network layer
deductive and objective reasoning. The network layer is not a part of the usual fieldbus
The requirements being determined, the choices for architecture model. It was introduced in the OSI model to
services were not too broad, but the same cannot be said for integrate the routing function in the topologies allowing for
the protocols… especially at the Medium Access Control several paths. The network layer is not necessary if only a
layer. single path is possible between stations. In most cases,
Before going into detail, we will present the architecture even if the general application architecture is complex,
recognized for the fieldbus and analyze the necessary OSI bridges may interconnect the different fieldbuses, and no
concepts in the first section. Traffic will be analyzed in the network protocol is needed.
second section, before studying the main relationships (or
cooperation models) at the application level in the third d) Transport layer
section. The fourth section is dedicated to the study of The transport layer was introduced in the OSI model for
Medium Access Control Protocols, and the fifth to the providing end-to-end control of the exchanges between two
communication architectures. end-stations, without considering the underlying
A. OSI model and fieldbus mechanisms (routing, data link protocol, physical wiring,
etc.).
The first version of the OSI model was published [86],
To do this, the emitting transport layer cuts the messages
[168] when the work on fieldbuses started. Most fieldbuses
into small packets which are transmitted separately from
were presented according to a three layer architecture. It was
one point to another until they reach the receiving transport
not new; other architectures had also been considered, the
layer. They are then reassembled to reconstitute the initial
Mini MAP, FAIS (Factory Automation Interconnection
message and delivered to the application. There is a
System), Collapsed Architectures, Enhanced Performances
mechanism to control the proper reception and possible
Architecture. This reduced architecture came from the MAP
retransmission.
Task Force, which claimed that a real time network must
These protocol mechanisms are carried on the messages
have only the physical, the MAC and the LLC (Logical
and are similar to those in the data link layer which are
Link Control), and the application layers. This concept was
applied to a frame (DL-PDU, Data Link Protocol Data
introduced to reduce the delays observed in the first
Unit).
implementations of the MAP seven layer profile. It was a
Regarding the lack of a transport layer in a fieldbus, the
mistake which led to concluding that to improve
end-to-end control is then done at the data link layer. And it
performance (to boost communication), it was necessary to
may only apply to a frame and not to another PDU. The
reduce the OSI model to a more simplified one.
Application Protocol Data Unit (A-PDU) must then be
Let us analyze this point and the particularities of
shorter than the longest Data Link PDU (DL-PDU). Even if
fieldbuses regarding the general-purpose mechanisms and
most of the fieldbus A-PDUs are short, operations such as
concepts defined by the OSI model.
program downloading and uploading are then made more
1) Fieldbus Architecture Model
complicated, and even impossible. Therefore, functions
It is common to say “a fieldbus has three layers:
such as fragmentation and reassembling are sometimes
the physical layer,
included in the application layer implementation.
the data link layer, including implicitly the MAC layer,
the application layer”.
e) Session layer
What happened with the other layers of the original OSI
model? What about the network, transport, session and Regarding the session layer, it was introduced in the OSI
presentation layers, as well as another layer that was added, model to facilitate managing the exchange of very large
the user or 8th layer (to be discussed later)? messages. It does not have a role in most fieldbuses even if
some synchronization functions could be considered as
a) Physical layer and topologies relevant to an OSI session concept.
The physical layer is always necessary. All the
f) Presentation layer
topologies in fieldbuses are found here; bus, star, ring, tree,
and other topologies supporting store and forward Regarding the presentation layer, its role is not only
transmissions. necessary, but also fundamental, in order to provide a
common language of exchange between stations with
b) Data link layer different internal and local syntaxes. It is often included in
the application layer. For a reciprocal understanding of the
The data link layer is also necessary, but we will see that
exchanged information, a comprehensible coding of the A-
the problem of transmission errors is not treated the same as
PDUs from both parties is necessary. With OSI, the ISO
in OSI networks. This paper considers that the MAC layer
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 12

standards ASN1 (Abstract Syntax Notation One) and BER Furthermore, considering that some protocols could be
(Basic Encoding Rules) are used but are not efficient for implemented in different stacks, it was necessary to define
fieldbus. Therefore, other kinds of coding are used [129], some kind of “interface layer” (sometimes called a “glue”)
often associated with the name of the data exchanged. to satisfy the implementation constraints. These “glue
layers” may also implement some intermediate layer
g) Application layer functionalities. LLI (Lower Layer Interface) in the
The application layer is obviously necessary. It may be PROFIBUS fieldbus, MCS (Message Control Services) in
defined according to different models. This layer normally the WorldFIP fieldbus are such examples.
includes the “user layer” presented below, because, as stated But the OSI model is not just a layered architecture, it is
in the OSI standard, “the Application layer has no upper also the definition of several concepts, services and
interface”. protocols, addressing, Service Access Point, multiplexing,
grouping, point-to-point or not, broadcasting, flow control,
h) “User layer” acknowledgement, etc. Considering these concepts,
The application layer defines elementary types of objects fieldbuses differ from general-purpose networks. The
such as integers or chains of characters. But applications differences are studied in the next section.
manipulate several types of objects such as speed,
temperature, pressure, etc. The need to define these types of 2) Basic OSI mechanisms and fieldbuses
objects, in addition to those that exist in application This section analyzes some OSI concepts highlighted by
protocols, was felt early on in the first networks. This is communication needs.
how the “Companion Standards” were defined. These
standards proposed specific objects for each application a) Point-to-point, multipoint, broadcasting
domain, like robotics, numeric commands, process All fieldbuses provide point-to-point communication,
controls, etc. With fieldbus, these functions are integrated and some provide broadcasting capabilities. When provided,
in what is called the “user layer”. We find here, obviously, physical broadcasting is obtained thanks to fieldbus
definitions for types and objects, but also for standardized topology through the diffusion of signals. But the data link
functions that are called “function blocks” [72], [75]-[77], layer protocols are all point-to-point and do not take into
which correspond to particular treatments of the objects, account the fact that a given frame may have several
such as conversion between units, filtering, linearization, simultaneous receivers. The problem of reliable
etc. broadcasting [24] has never been dealt with in existing
It is called “user layer” in order to express the idea that it fieldbuses. A single protocol (WorldFIP) addresses this
is the way by which the user “sees” the fieldbus and problem at the application layer and proposes a mechanism
communication [106]. This is directly issued from the for verifying space consistency. Space consistency is a
necessity for the end user to “ignore” the communication property, which defines copies of data. It is verified when
techniques. This approach, which came from the MMS the copies of data on different stations are equal. This
works [89], was already recognized by Pimentel [129], and mechanism provides a kind of global acknowledgement
is now the base for defining EDDL (Electronic Device (acknowledgement of a set of frames in a single A-PDU)
Description Language) [77] coming from the DDL defined [137].
in the HART Fieldbus (Highway Addressable Remote
Transducer) [109]. b) Connection or connectionless protocols
“Profile” is also used to described the concept of The connection mechanism was introduced to
possible options in the protocols of the stack and in the dynamically manage the resources necessary for
companion standards. For example, we see the “pressure communication between two entities. In the case of the
sensor profile”, different “actuator profiles”, etc. This word fieldbus, as a lot of operations are statically defined, it may
“profile” has, then, two meanings: one for designating the be considered that the connections are permanently
choices of protocols and protocol options in a real OSI established at the configuration or commissioning stage. In
stack implementation, and the other for the integration of fieldbuses, operations with or without connections should
the dedicated functions of given devices. be possible. But it is important to define multipoint
connections for multi-peer communications.
i) Conclusion
In conclusion, to say that a fieldbus is always based on a c) Buffer vs. queues
reduced model is a gross misunderstanding. Let us recall This item analyzes how the PDUs to be sent and received
that the OSI model is a conceptual, not an implementation, are stored in the communication stacks at the sender and
model. A fieldbus usually presents all the functionalities receiver sites. Usually, the PDUs to be transmitted are
provided by the seven layers of the OSI model. But, in stored in queues at the sender site; they are also stored in
terms of implementations, other choices are possible. For queues at the receiver site. They are generally managed
example, the transport functionalities may be implemented according to the First In, First Out strategy (FIFO) but
within the application and presentation layers. another schedule may be obtained through priorities or
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 13

deadlines. The idea in these classical communication control is useful for random traffic management.
systems is that all the PDUs must be processed.
B. Traffic classification and characteristics
In fieldbus-based applications, due to periodic traffic, we
can drop old PDUs in favor of the most recent. The strategy 1) Typical exchanges
The requirements have specified the types of traffic and
of storing all the PDUs in queues is not suited to this
their main functional characteristics. They are mainly
behavior. Therefore, this data is not stored in queues but in
constituted of input and output variables, what we call state
buffers, which always contain the last value produced or
variables, and events. Input-output variables, internal
received. The reader can find a very good analysis of these
variables and states (as in state – transition models or state
mechanisms in [103].
control) are considered as status. Changes in the status are
d) Control of errors or status vs. event considered as events.
But traffic may also include some files for downloading
The detection of errors or the control of the exchanges is
device domains, and service requests and responses,
either done by the sender, or by the receiver.
especially for the management of application processes and
When does the sender control the exchanges? It must
stations.
control the exchanges when they are randomly initiated, or
For some of the transfers, the temporal characteristics are
when the message has the semantics of an event. The
frequency, jitter, lifetime, response time, simultaneity and
sender decides the transmission; the receiver is not informed
the temporal and space coherences or consistencies.
and will only be so at the reception of the message. The
Frequency indicates the rate at which the data is updated,
sender controls the transmission by waiting for an
and jitter is a variation in the periods; lifetime indicates the
acknowledgement of the receiver.
duration that the data values are significant, and response
When does the receiver control the exchanges? It controls
time is the delay between a demand and the result.
the exchanges when they are regularly initiated; when the
Simultaneity indicates that several operations or events
message has the semantics of a status, independent of the
occur at the same time, i.e., in a predefined time interval or
time-triggered or event-triggered paradigm (see latter
time window. When several operations occur in a given
subsections for description). The receiver waits for a
time window, they are called time coherent.
periodic reception in time-triggered systems, or waits for
For other transfers, no such constraints exist, but their
the response to a request if the exchange has been so
required quality of service is more related to the absence of
initiated. It is the receiver, who is in charge of transmission
errors, to the delivery order and/or to the recovery
control.
mechanisms. In other words, the required quality of service
In fieldbuses, both of these situations are encountered
depends on the traffic considered. Safe and secure
and so, fieldbuses normally have to provide both of
transmission is required for file transfers, and respect of
communication mechanisms.
time constraints is required in the case of exchanges of
status.
e) Acknowledgement or not
Following these requirements, traffic may be considered
Acknowledgements were introduced in protocols so that as composed of two types of information exchanges:
a receiver informs the transmitter of a message whether or identified data and usual messages, as in all ISO
not it has been well received or not. In fieldbus communication systems (Fig. 4). Identified data is all the
applications, aperiodic exchanges must be correctly data known by the control system such as the input issued
received, then acknowledged and possibly repeated. On the from the sensors, the commands to the actuators, and so on.
other hand, periodic exchanges do not need to be They are essentially real time and periodic data. Identified
acknowledged. So, if there is an error in periodic traffic, data has only one producer, but one or more consumers.
the receiver can ignore it and wait for correct data to follow. Rather than producer – consumer, we may say publisher -
But, it is not sufficient that a message be received without subscriber (see section on cooperation models and quality of
error, it must be received at the right moment. The temporal service).
aspect is important. The management of errors, the recovery These so-called messages are issued from any application
strategy, must be placed under the control of the user, i.e., process which needs to send something to another one.
the application processes [45].
Fig. 4. Traffic classification
f) Flow control
In general purpose networks, flow control is necessary for Considering the different fieldbuses, this classification is
preventing congestion, for satisfying previous requests, for not always so clear. Some provide only exchanges of
keeping one’s engagements; flow control starts with an identified data; some provide only exchanges of messages.
admission strategy and test. Flow control is important This distinction is useful for two different reasons:
when traffic changes very quickly. • firstly, because considering the identified data,
In the case of fieldbuses, the flow control may be seen as only the successive values are of interest, and
a function of the configuration stage, it is essentially a they can often be immediately accessible
feasibility study, a test of schedulability. At run time, flow through the name of the object without having
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 14

to treat a message in the different layers. 3) Messages

• secondly, the values of this data can be stored in We call messages all the exchanges that are not relevant
buffers and not in queues as with messages. to the previous exchanges of identified data. Messages are
This point will be examined later. exchanged during configuration and maintenance stages.
2) Typical traffic of identified data They are used for downloading and uploading. In
A fieldbus has to transmit essentially the values of data fieldbuses which do not consider the traffic of identified
between sensors and controllers, between controllers and data, everything is considered a message. When only this
actuators, and between controllers themselves. These traffic is considered, it is then necessary to distinguish the
exchanges, called "identified data traffic", are known once real time and non real time messages.
the application is specified. They may be managed in the 4) Conclusion
client-server as well as in the producer-consumer model or All this traffic may be managed in very different ways,
their extensions [148], [29], [44], [162]. with priorities to one type or another, with more or less
predictability, and so on. These mechanisms are relevant to
a) Periodic traffic the MAC layer and will be studied later (see Section “Data
Periodic traffic is induced by the sampled systems Link and MAC”). The service and protocol characteristics
theory, which is the basis for automatic control and will be summarized according to the following points:
detection of events. Most identified data is the input or peer-to-peer vs. multi-peer or multicast, confirmed (or not)
output of control algorithms. They must often be services, acknowledgement or not, connection or
transmitted periodically. This traffic is deduced from the connectionless protocols, flow control.
periodic polling of input in normal centralized systems. The Time constraints are statically or dynamically specified
periods of exchanges may be different for each kind of data. and managed [17]. The specification of the time constraints
A jitter may or may not be accepted. It is clear that the may be determined at the connection opening, at the
protocols will play a major role in the respect for configuration stage, or dynamically at the service request.
periodicity without jitter. These systems are based on state They are then managed differently by static or dynamic
traffic and are sometimes called "Time-triggered systems". scheduling [14] [134]. Jointly scheduling tasks and
Fig. 5 shows a general example of periodic traffic. It messages is still an open problem [18].
shows the updating of A at each elementary period, of C C. Cooperation models, quality of service
and D every two elementary periods, B and E every three
Cooperation models represent how two or more
elementary periods, and F every six elementary periods. The
application entities cooperate to obtain a given objective.
macro-cycle is the period equal to the LCM (Lowest
Two distinct transaction families can be distinguished: the
Common Multiple) of the periods. And the micro-cycle is a
client–server family and the publisher–subscriber, also
time interval equal to the HCD (Highest Common
known under the name producer–consumer.
Denominator).

Fig. 5. Example of periodic traffic 1) Basic models

b) Aperiodic traffic a) Client – server

All data may be transmitted cyclically, as is the case in In the client – server model, two entities cooperate (Fig.
some fieldbuses. But the obtained global traffic may be too 6). The server is an entity, which provides a service, i.e.,
great for its nominal data rate. In this case, aperiodic which executes an action on the account of a requester,
exchanges of some data is more advantageous. Indeed, some which is called a client. This model is more useful for
state values do not change at a predefined period and may transmitting state data than event data. Event data detected
be transmitted only on a change basis. by a server is only transmitted if the client requests the
The random or “on demand” traffic takes place in the free transfer through a READ service.
time slots left by the periodic traffic.
Fig. 6. Client – server model
The schedulability of traffic may be analyzed at a
configuration stage and on line [4], [116], [134], [161]. Normal client - server
The client – server interactions are broken down into four
c) Time-triggered or event-triggered systems steps, request, indication, response and confirmation (cf.
Distinguishing periodic and aperiodic traffic is relevant Fig. 7). An indication is an event by the server that
from two points of view of the application. Kopetz indicates the reception of a request. And the confirmation is
compared these approaches [94]. The main comparison the counterpart regarding a response by the client. This
criterion is the capability to meet the application time model is used by all the application protocols, which are
constraints. Most fieldbuses favor a kind of time-triggered more or less derived from MMS. The client – server model
system. But some of them combine both approaches, events is used in INTERBUS, PROFIBUS-FMS and DP, in AS-i,
being managed by a periodic server. in P-Net and in WorldFIP.

Fig. 7. Usual client – server interactions

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 15

it from the other types of interaction.

The semantics of the response may vary from one service First, a typical confirmed request/response exchange is
to another. For example, the answer may be significant to performed between the publishing manager and the
the request acceptation; it may be significant to the service publisher. However, the underlying conveyance mechanism
execution beginning, or to the result of the service returns the response not only to the publishing manager,
execution. but also to all subscribers wishing to receive the published
In the case of a READ service, the value of the read information. This may be accomplished by having a
objects is carried by the response. The request contains the protocol mechanism in an underlying layer, which transmits
name of the object, and depending on the local addressing the response to a group address, rather than to the
mechanism, the means to access the object. The response individual address of the publishing manager. Therefore, the
contains the value or the reason for failure, and when response sent by the publisher contains the published data
provided, the timeliness attributes. The object may be, a and is multicast to the publishing manager and to all
priori, a simple variable or a complex structure. subscribers.
All the application layers which provide services The second difference occurs in the behavior of the
according to the client – server model are more or less built subscribers. Pull model subscribers, referred to as pull
on the MMS model. They propose the management of subscribers, are capable of accepting published data in
objects such as tasks (create, kill, start, resume, and stop), confirmed service responses without having issued the
variables (read, write), domains (downloading and corresponding request. Fig. 9 illustrates these concepts.
uploading), and so on. Only one subset of services is
generally provided. Fig. 9. Pull Publisher – Subscriber Model
In terms of timeliness, the duration of such an operation
• Push model
can be subdivided into three terms: request transfer, action In the "push" model, two services may be used, one
execution, response transfer. The duration may vary confirmed (1 and 2) and one unconfirmed (3). A confirmed
according to the latency time of transfers depending on service is used by the subscriber to request a binding to the
medium access control, and on latency time on the server publisher. The response to this request is returned to the
depending on its current load. subscriber, following the client-server model of interaction.
Unusual client - server The unconfirmed service (3) in the “push” model is used
An “unusual client – server” model is, in fact, composed by the publisher to distribute the information to
of two sequences of unconfirmed services. This model (Fig. subscribers. In this case, the publisher is responsible for
8) is also called client – server, even if the client is in invoking the correct unconfirmed service at the appropriate
charge of the association of the indication to the previous time and for supplying the appropriate information. In this
request. The same services as in the previous section may fashion, it is configured to "push" its data onto the
be defined, but with only the Request and Indication network. Subscribers for the Push Model receive the
primitives. The response to a READ Indication is a WRITE published unconfirmed services distributed by publishers.
Request. The response to a WRITE Indication is also a Fig. 10 illustrates the concept of the Push Model. In Fig.
WRITE Request. This model is used, for example, in the 10, the sequence “request – confirmation” (noted 1 and 2)
BatiBus network [8]. represents a subscribing phase. The publishing operation (3)
is triggered by the Push Publisher itself, each time
Fig. 8. Unusual client – server interactions
necessary.
b) Publisher – Subscriber
Fig. 10. Push Publisher – Subscriber Model
The Publisher-Subscriber interactions, involve a single According to the Pull model, it is possible to define a
publisher Application Process (AP), and a group of one or READ service initiated by the Pull Publishing Manager,
more subscriber APs. This type of interaction has been similar to the READ service of the Client - Server model.
defined to support variations of two models of interaction, The difference is that, in the Pull model, all the subscribers
the "pull" model and the "push" model. receive the READ confirmation under the form of an
Pull model indication, because there is no prior request.
In the "pull" model, the publisher receives a request to According to the Push model, it is more a service
publish from a remote publishing manager, and broadcasts resembling an Information Report initiated by the Push
(or multicasts) its response across the network. The publisher which may be considered as a server..
publishing manager is responsible only for initiating The Push Publisher – Subscriber model is well suited for
publishing by sending a request to the publisher. transmitting event data. It may be used for services as
Subscribers wishing to receive the published data listen for “Event Notification” request and indication, or “Information
responses transmitted by the publisher. In this fashion, data Report” request and indication defined in MMS.
is "pulled" from the publisher by requests from the The Publisher – Subscriber models are used for
publishing manager. A confirmed service is used to support exchanges (read and write services) between buffers. The
this type of interaction. following fieldbuses use this model, WorldFIP, CAN,
Two characteristics of this type of interaction differentiate
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 16

LonWorks, EIBus, ControlNet, SwiftNet, FOUNDATION of the response. This model has been studied in different
Fieldbus. works [29]. It can be implemented above MMS protocols,
In terms of timeliness, the duration of the exchanges but is not recognized in standards.
done by these models depends only on the MAC protocol,
coming from the latency of the sending operation by the Fig. 12 Client – Multi-server
Push Publisher.
c) Third-part model
c) Manager - agent This model is a particular case of the previous one. A
The manager-agent model is similar to the client – server client requests a service from a server which is unable to
one. It is the model used by the SNMP protocol (Simple provide the service but which knows the appropriate server
Network Management Protocol), in conjunction with a MIB able to do it (Fig. 13). Several scenarios may be considered
(Management Information Base) based on a tree structure. and possible failures must be detected and corrected.
This protocol provides a Push Publisher – Subscriber
Fig. 13. Third Part Model
service, the so-called TRAP request and indication.
2) Other models d) Multi - Publisher Multi – Subscriber

a) Client – server multi confirmations This model is of interest for synchronizing the activities
of publishers. If, for example, several pieces of data must be
This service model defines several responses (and then
produced at the same time (i.e., in a given time window), it
confirmations) for a single request (Fig. 11). It is of interest
is easy to synchronize the producers and then to apply one
in the case of long service execution. The semantics of the
of the two publisher – subscriber models to provide
responses and of the confirmations may be the following:
information to the subscribers.
the first response indicates that the request is possible and
The time constraints, which may be specified, are the
taken into account by the server. The second response
global response time (less than the period, obviously)
indicates that the service starts its execution. The last one
between request and confirmation, or the server response
delivers the results of the service.
time between the indication and the response.
Fig. 11. Multi-confirmation Client – Server This model of cooperation is essential when properties
such as the time coherence of data production, of data
publishing, or of data consumption are required.
This model of cooperation is well suited for long
This model is used in WorldFIP networks and in all the
duration services, when a server is overloaded, and when the
IEC 61158 Type 1 compatible networks.
execution of a service may take a long time. It allows the
client to know the status of its request, and it is possible e) CS mono request multi response
for the client to establish time constraints (delays or
This model provides the following timed behavior. It is
deadlines) for each of the responses. Cancellation of a
similar to the Push Publisher – Subscriber model. The
service request may then occur when the constraints are not
request “Read-Rq” may be compared to the request to
met. Such a model may be used for any service. This model
become a subscriber.
is not implemented in standardized protocols.
Fig. 14. Multi-responses
b) Client multi - server
The client multi-server model (Fig. 12) is a particular This model (Fig. 14) is of interest in fieldbuses for the
case of the client – server model. A given request, which managing of periodic exchanges, with a single request.
can not be processed by a single server, can have several This model may be extended with the means to define a
servers that may answer it. In this case, there is a function starting event, for example, a date or a condition and,
to break down the request into sub-requests adapted to the similarly, an ending event, for example, duration or a
capabilities of the different partial servers. condition or a date.
The client does not know all the partial servers. The This model can be compared to a periodic client - server
decomposition of the initial request into several is not consisting of periodically requesting the service provided by
known to the client. Some synchronization between the the server. The difference is on the temporal quality of
partial server actions can be requested and verified by the service. A periodic client – server is periodically triggered
principal server. on the client site, but with transmission and server delays,
The problem is that, upon the definition of the result in the period may not be met at the server site. In the multi-
the case of all partial servers not being able to provide a response model, the period is managed at the server site and
correct response, the global response to the client then can then be more strictly respected.
becomes partial. If the response can only be complete or
negative, the problem does not exist. f) Conclusion
If a response is partial, the client must know the The different cooperation models represent the rules,
composition, in order to correctly identify the lacking parts which are furnished by the application and presentation OSI
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 17

layers. They make up the ASE (Application Service 1) the moment when the buffer is written and
Element), which is called “industrial messaging” or 2) the moment when the buffer is read
“Fieldbus Application Layer” (FAL) [74]. A lot of Given ∆T the length of the residence time interval, the
messaging services and protocols have been defined in Data Link residence is defined a s follows:
different domains, which may be used in fieldbus 0 ≤ (RT – WT) < ∆T, where RT and WT are the READ and
applications: MMS [89], SNMP, MPS [2], [149], [69], WRITE instants.
[150], [152], IEC 870-5, [66]. All the application layers in This type of timeliness was called Asynchronous in
existing fieldbuses provide a subset of these application previous French and European standards.
service elements. Applying this principle to the communication between a
publisher and a subscriber, it can be seen that three
3) Timeliness operations in three intervals can be controlled. A global
The word “Timeliness” means all the temporal aspects of analysis of these attributes may be found in [107].
operations, of data, and more generally of dynamic system The publishing operation is controlled at the publisher
components. Timeliness is expressed through dates or time site by an attribute which is transmitted along with the
stamps, through durations and through Boolean attributes, value to the subscribers. They, in turn, can then know if the
which determine if a temporal property has been met or not. publisher has met its own constraint or not. For example,
Timeliness has been studied (and is still being studied) such an attribute can indicate if the period of a periodic
for a long time in different communities, with the first publishing operation has been respected. Or, such an
work on real time languages and on formal methods for attribute can indicate if the publishing occurred before a
time and dynamic system modeling. A lot of papers have given deadline after a request, Action 1 in Fig. 16.
been published on these subjects. Historically speaking, the A similar control can be placed at the receiving site. For
following are of interest because they introduced the example, a given value must arrive periodically or before a
concepts now available with the truly “real-time” fieldbuses. given deadline after a request. An attribute may be
For the topics related to real time systems and languages computed at the receiving instant (Action 3 in Fig. 16), in
the reader may consult different papers: [12] for the PEARL order to be transmitted along with the value to the
language, Gertler in [55] proposed the first synthesis on real application entity afterwards. Such an attribute can also be
time languages, Deschizeaux [34] and Kronental, [97] defined at the Data Link layer to control if the sending
proposed elements for the standardization of real time instant occurs in a given time window (see Action 2 in Fig.
operating systems; Thomesse, in [147], introduced timing 16).
considerations and mechanisms in real time distributed A subscriber then receives, not only a value of data, but
applications. Le Lann [100] and Lamport [98] were among also attributes which indicate if the successive operations
the first to formally introduce the problems of time in have occurred in the right time window, or on time. These
distributed systems. For time concepts and modeling, the attributes represent the quality of service from a time point
reader can consult: [30] which explains the different types of of view. The subscriber may then decide what to do
time constraints, [113] for an overview of time concepts in according to the quality of the data. The reader may find the
real time applications, [5] and [121] for a formal basis of the quality of service in [92].
presentation of the introduction of time in logic and state –
Fig. 16. Application layer Residence mechanism (issued from [74])
transitions systems. [160] proposes extensions to UML for
time consideration and modeling. D. Data link and MAC
The objective of this section is to give the particularities 1) MAC Classification
of timeliness in fieldbus services and protocol. We will not The usual MAC protocols are based on one of three
present the generalities such as time stamping, or clock following classes: controlled access, TDMA or contention
synchronization. This section will focus on the definition of (see Fig. 17). If a control is used, it can be centralized or
timeliness attributes in order to verify if time constraints decentralized. In the case of TDMA, the classification is not
have been met or not. so easy, the access is always decentralized, because the
The first such attributes were proposed in the WorldFIP decision to send is taken individually by each station, but
application layer. And they are now redefined in some the clock synchronization function itself may or may not be
profiles of the IEC 61158 standard. The idea was to verify centralized. In applying such a classification to fieldbus
if a given operation had occurred in the right time interval, MAC protocols, it is necessary to distinguish the
i.e., in a given time window. Here we will only give an management of periodic and of random traffic, as shown in
example in Fig. 15, issued from the WorldFIP standard and Figure 4 (see also [102]).
from the IEC 61158-3 Data Link Layer standard [73].
Fig. 17. MAC protocol classification
Fig. 15. Residence Attribute
Regarding how periodic traffic is managed, it is either
The “Residence” timeliness is an assessment based on centralized, or decentralized. Fig. 18 and 19 show two
the length of time that a datum has been resident in a classifications of the fieldbuses according to their MAC
buffer, which is the time interval between: protocol, regarding the periodic and the aperiodic traffic. In
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 18

the case of decentralized management each station must former, a same station may then have several rights in a
decide, at its allotted time to send, which traffic it should same round, possibly to send more often than others.
prioritize. In the case of centralized management, a periodic The length of slots: all the slots are of the same length
server deals with this problem. A station can ask for an (as in digital phone systems, because all the traffic is the
additional right to send when it is periodically polled same), or of different lengths in order to take into account
(special frame on demand) or the server systematically and the needs of each station.
periodically allots a time slot for aperiodic traffic (time slot The clock synchronization: the clock synchronization is
in each frame). the basis for defining the starting instant of transmission for
each node. This synchronization may be done in a
centralized way as in TTP-A (Time-Triggered Protocol) [96]
Fig. 18: Periodic traffic in fieldbuses
or by a distributed algorithm as in TTP-C. It is also
important to note TT-CAN (Time Triggered CAN) and
FTT-CAN (Flexible Time-Triggered Protocol), which,
Fig. 19. Aperiodic traffic
being based on CAN, introduce a time-triggered
mechanism.
The contention protocols cover all CSMA variants. The
controlled access protocols are used the most in large c) Examples
fieldbuses, with thousands of stations. All protocols use The TDMA principle is used for periodic traffic by TTP
MAC addresses which are either a station address or a [94], [95], ARINC protocol family, SERCOS [68], and
logical address (source address), which is more efficient for ControlNet [27]. INTERBUS on a ring topology is similar
cyclic traffic. Addressing by the name of the identifier is to TDMA; a single frame is divided into as many fields as
used by WorldFIP, CAN, BatiBus, EIBus and the number of stations. Each station has the right to send in
FOUNDATION Fieldbus. Otherwise, a classical addressing its own field.
mode is used.
2) TDMA class d) Quality of service
This class represents the protocols, which give the right The temporal quality of service is generally good, the
to send n the medium according to a rule, such as the Time frequencies are met, and no jitter occurs when the clocks are
Division Multiple Access. well synchronized. It is supposed that each station respects
its sending time. The periodic transmission of the time-
a) General principle constrained data may be guaranteed under certain hypotheses
TDMA is based on dividing the access time of the [142]. The clock synchronization is not a topic of this
medium into slots, which are allotted to the stations paper, but the reader may consult [78] for an example of
according to a given strategy. The slots may or may not be clock synchronization algorithm.
equal in duration. Each station may send a frame of a given 3) Polling class
length at a definedmoment. In synchronous TDMA, the
access is periodically allotted, as indicated in Fig. 20. In a) General principle
ATDMA, (asynchronous TDMA Fig. 21), the slots are The polling class represents the protocols that allow the
allotted to the stations according to their needs. This means right to send by sending an explicit message (the Poll
that a station without generated traffic does not use its message) to the station, enabling it to send. The Poll
slots, such as the Sites 2 and 4 in Figure 21. While in message is always sent by a special station, called Master,
synchronous TDMA the address of the sender is implicitly Arbitrator, or Manager, etc.
given by the relative position of the slot, in asynchronous
TDMA, each slot must contain its address or its b) Variants
identification. In STDMA, the nominal data rate of the The variants are related to the addressing method, and to
network is equal to the sum of the stations’ loads; in aperiodic traffic management. Some are static, others
ATDMA, the total load of the stations may be greater than dynamic [134].
the nominal data rate of the network. Addressing methods: there are two main sub-classes: the
Fig. 20. Synchronous TDMA first designates each station by its address, the second by
the identification of the data to be sent. The former
Fig. 21. Asynchronous TDMA indicates the station explicitly and, in the latter, it is
implicitly designated, as in the producer - consumer
cooperation models.
b) Variants
Aperiodic traffic: different techniques are used to manage
The variants concern the following points: aperiodic traffic. For example, WorldFIP uses a dynamic
The content of a slot: the content of a slot may be the scheduling of requests for aperiodic traffic taking place in
value of the data, or a frame issued from a station the free time slots of the periodic traffic; INTERBUS uses,
containing (possibly) the values of different data. In the in each cycle, a two byte field in the periodic frame to
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 19

transmit information on demand. ControlNet uses a Round passed between stations with successive addresses.
Robin algorithm for managing aperiodic traffic.
c) Examples
c) Examples The first was PROFIBUS - FMS, which defined a token
Centralized MAC fieldbus representatives are P-Net, passing mechanism between the master stations, and a
WorldFIP, AS-i [7], PROFIBUS-DP, PROFIBUS-PA. A polling mechanism between a master station and the slaves.
station is in charge of the distribution of the access control. P-Net provides a similar mechanism but with an implicit
INTERBUS [38] may also be considered as a polling token, as ControlNet for the aperiodic traffic management.
protocol, because each station periodically receives the right
to send from a central master. It may also be considered as a d) Quality of service
kind of TDMA on a ring topology, analogous to the The temporal quality of service guarantees that bounded
Cambridge ring. It could also be modified in a multi-master transmissions (with bounded jitter) are respected due to
protocol [20]. dependability hypotheses. The respect for periods is less
strict than with TDMA or polling because of token
d) Quality of service management. If the token holding time of each station is
A polling MAC can guarantee the periods without jitter strictly constant, and if no errors occur, the periodicity is
if some mechanisms (anti-jabber) are developed to avoid respected. Jitters may appear in the case where the previous
overly long frames from being transmitted. The polling hypothesis is false [31]. Two successive polling operations
technique favors periodic traffic and time-triggered systems. of a same slave by two masters may lead to temporary
The time coherence constraints are easier to manage if a inconsistencies between the state information. From this
multicast is allowed and a consensus mechanism used in point of view, PROFIBUS FMS was more a Mini-MAP-
order to ensure the distributed copies are identical. like profile than a fieldbus.
WorldFIP is typically such a fieldbus [137]. 5) Link Active Scheduler (LAS)
One may raise the objection that a centralized system is
not robust. Some fieldbuses allow a redundancy of the bus a) General principle
controller, or of the bus control function, which may be The general principle consists of giving the
implemented on several stations (PLC, regulators, sensors, responsibility of traffic scheduling to a specific station (the
and so on). LAS). But it has the capacity to delegate responsibility to
To introduce dynamic behavior in statically defined another station with token passing or with an order to
systems, different operating modes may be defined as distribute data for a given duration. It is based on a mixed
according to the Fohler proposal [49]. mechanism of PROFIBUS Token Passing, with the
The bus arbitrator of WorldFIP may be duplicated; a WorldFIP Bus Arbitrator. It comes from the IEC TC65C
token-like mechanism allows a bus arbitrator to give the WG6 - Data Link layer working group committee (IEC
bus control to another arbitrator, as with the master stations 61158-3 Type1), which tried to find a common solution for
in PROFIBUS. the much wanted international standard. No variant is
The draft proposal IEC 1158-3 included services coming known at this moment.
from WorldFIP and PROFIBUS standards, allowing
centralized access control as well as decentralized. b) Example
4) Token class The FOUNDATION Fieldbus has implemented this
mechanism.
a) General principle
This class represents the protocols which provide a c) Quality of service
control access similar to the polling class, which can be The temporal quality of service is similar to the one
used with a bus or a ring topology, but is decentralized. obtained with the polling technique.
6) Contention or CSMA class
b) Variants
The variants are related to the role of the stations in the a) General principle
fieldbus, be they masters or slaves, to the form of the The CSMA (Carrier Sense Multiple Access) class
token, and to the passing method. The role of the stations: represents all the protocols, which are based on any variant
master and slave stations may be distinguished from each of the Ethernet principle. The principle is to wait for the
other, such as in PROFIBUS. Master stations constitute a channel to be free to send a frame. Collisions may occur,
virtual ring over a bus topology. They poll slave stations and the variants propose different recovery mechanisms.
when they hold the token.
The form of the token: it may be an explicit message but b) Variants
it may also be implicit; for example, when a Round Robin The variants are CSMA-CD, CSMA-CA, CSMA DCR,
scheduling is used (ControlNet for the aperiodic traffic and predictive p-persistent CSMA as in LonWorks.
management), the token is automatically and implicitly The most known variant is CSMA-CD (Collision
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 20

Detection), which is not very common in fieldbuses except E. Communication Architectures

when the maximum load is relatively low in relation to the After having analyzed the cooperation models at the
nominal data rate. An example is the PCCN (Poste de application layer, and the Medium Access Control, let us
Contrôle-Commande Numérique) network for electrical now further investigate the communication architectures.
transformers.
CSMA-CA is used in building automation networks 1) Two stack architectures
such as BatiBus [8], EIBus [43], EHS [42], in car networks The communication architectures have been veritably and
such as CAN [91]. It is often a CSMA with forcing explicitly developed according to, and thanks to, the OSI
capabilities, often called CSMA-CA for Collision reference model. The first modifications (or extensions) of
Avoidance; it means that even in case of a collision, a this model were introduced with the IEEE 802 model and
single frame may be transmitted, the one with the most with the MAP-Enhanced Performances Architecture reduced
priority. model [99]. Both of these extensions have their own
Other CSMA variants (CSMA-DCR) have been defined reasons for introduction, but arguments for real-time needs
to guarantee an upper limit to transmit all collided frames were put forward to promote MAP-EPA. It is a known fact
[101]. These protocols are based on a partitioning of the that speed alone is not enough to meet the real-time
stations’ ability to transmit. The protocol is robust in the constraints [144] and that, when a minimum amount of
sense that if a station fails, the others are not concerned at resources are available, the scheduling of the tasks and
the MAC level. The problem is that a more urgent frame messages within resource allocation is the only solution.
must wait until the end of the transmission of all the That is the main reason for the definition of most of the
previous collided frames. fieldbus “real-time” MAC protocols including, more or less
Another variant is found in LonTalk protocol [105]. It is explicitly, a scheduling of the messages. In parallel, for
based on predictive p-persistent CSMA. This method configuration or maintenance operations (downloading),
consists of estimating the backlog to adjust the medium normal (not real-time) protocols are necessary. In brief, the
access delay according to the network’s current load. internal communications can only be ensured with a two
• Examples stack architecture (see Fig. 22). Another reason for a two
The examples of CSMA variants are CAN, SDS [141], stack architecture is that the fieldbus needs to communicate
[118], DeviceNet [120], LonWorks, BatiBus, and EIBus. with the outside world.
CAN is used in DeviceNet and in SDS as a The use of Web technologies and the emergence of
"subfieldbus" in a machine. Ethernet at the field level contributed to the study of
• Quality of service architectures. Even if some papers in this issue deal with
The temporal quality of service may be predictable in the these questions, let us introduce the problem before going
case of CAN based networks. This mechanism may back to the future.
guarantee the periodicity under some hypotheses [156],
[157].(Tindell et al, 1995a; Tindell et al, 1995b)). The data Fig. 22. PROFIBUS and WorldFIP architectures
rate is limited by the length of the medium, typically 1 2) Introduction of Web technology
Mbps for a length of 40 meters. Considering a fieldbus with dedicated Time Critical Data
The industrial Ethernet solutions and switched Ethernet Link Layer protocol (TC-DLL), a solution for compatibility
are not studied here; other papers in this issue are dedicated with Web technology is the tunneling of IP (Internet
to these solutions. Protocol) datagrams inside TC-DLL PDU (Fig. 23). Station
1 is a gateway to the outside world. The HTTP frames
7) Logical Link Control which are fragmented into IP datagrams can be transported
The LLC is not distinguishable from the MAC in between fieldbus stations after being encapsulated into TC-
fieldbuses. However, the LLC services may be identified in DLL PDUs. Szymanski in [146] analyzes the solutions for
the data link layer specifications. The usual LLC services introducing Web technology into process control.
are known under the names LLC type 1, 2 or 3. The major
Fig. 23. Encapsulation
fieldbuses provide LLC type 3-like services, without
connection in the OSI sense, with immediate 3) Introduction of Ethernet as Fieldbus MAC protocol
acknowledgement for the real time traffic. LLC Type 1 is A trend was started some years ago to use Ethernet as the
also used. Transmission without acknowledgement may be Time Critical Data Link Layer [61] in automation [111].
of interest for periodic traffic. The failure or error detection For that, some mechanisms were added to Ethernet to
is then made by the receiver(s), when it is made by the obtain two channels, one for real time traffic, one for the
sender with LLC Type 2 services and protocols. The latter rest. This is the principle used by all the data link layers,
are used for messaging traffic, in order to provide the right which support different types of traffic.
transmission safety. Even if from a service point of view, Since the end of the 90’s, Ethernet has been proposed as
fieldbuses are very close to IEEE 802.2 specifications, from a standard for the Medium Access Control. And upon
the protocol point of view, they are all different and Ethernet, naturally, the promoters thought to use the
incompatible. TCP/IP stack and the Internet application layer protocols.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 21

This was the reason for the standard project “Real-Time” IV. CONCLUSION
Ethernet (RTE) of the TC65 SC65C WG11, standard IEC The standardization of protocols is far from being
61 784 part 2. The drawbacks of Internet are, moreover, the finished. The needs of the end-users expressed in the
non-predictability and the connector technology for the European MAP user group are still more or less valid.
industrial environment. The latter problem was resolved, Maybe the new working group of IEC TC65 on Real-Time
but as for the former, some mechanisms need to be added to Ethernet will take into account the TCCA
make Ethernet predictable, if this expression can be used. recommendations, and contribute to the design of a
Both of the solutions shown in Fig. 24 can be used. It is common architecture, which could improve the
also important to notice that the same application layer can interoperability of heterogeneous components. Another
be used over the different stacks as has already been done in challenge could be the definition of a common Time
the CIP solutions family [139]. Critical Data Link Service and Protocol, with the right
parameters to dynamically tune the protocol to the
Fig. 24. Ethernet-based architectures application needs (quality of service required vs. possible
quality of service).
4) Towards a common stack? The fieldbus technology covers a very large spectrum of
In a 1991 paper, a three stack architecture was proposed techniques and applications. The fieldbus is present
by Tom Phinney [127], as indicated in Fig. 25. The idea of everywhere. This phenomenon may explain the diversity
this architecture was to provide a common data link layer and the lack of a real standard, but it is not the only reason.
with different qualities of service for typical fieldbus traffic, One could write a paper, making a parody of Louis
as analyzed above in this paper as well as for file transfers Pouzin’s well known paper [132], entitled “virtual circuits
with all the necessary security and dependability. Such a vs. datagrams: technical and political issues”, written when
data link layer provides real-time features, associated with IP and X25 (in the mid 70’s) were fighting as network
connection mechanisms, acknowledgements, bridging protocol standard candidates in the standardization bodies.
capabilities, etc. It was obviously possible to adapt this Such a paper could now be entitled “client – server vs.
layer to any kind of physical layer, and to build upon publisher – subscriber: technical and political issues” or
different stacks starting from a full OSI stack, for covering “token bus vs. bus arbitration: technical and political
general purpose networking needs, going all the way down issues” to explain the importance of political or economical
to a reduced stack for very specific real time fieldbus-based and strategic aspects in choosing a standard or not.
applications. A full stack is used for normal communication This paper has tried to explain the different approaches
and can be implemented with TCP/UDP and IP protocols. and solutions in order to give the reader the most complete
A medium stack can be used when neither routing nor overview on the history of the fieldbus and on its current
fragmentation/reassembling are necessary. The stack on the situation.
right is the normal Time Critical stack. Not all the aspects have been treated, and that, for
different reasons:
Fig. 25. A Time Critical Communications Architecture (from [127])
- different physical layers, the powering of devices
by the network, the intrinsic safety; because the
The Time Critical Data Link Layer should be the solutions are numerous and, if important
proposal for IEC 61158 [70], [71]. concerning the applications, these points are not
This proposal should be re-examined for two reasons: really strategic,
firstly, in light of a future Real-Time Ethernet protocol, and - network management, which is out of the scope
secondly, in light of the real-time mechanisms introduced of standardization and covered by proprietary
in the Internet stack needed to implement such applications solutions,
as phone over IP, video transmission, videoconferencing, - conformance testing, which is very closely
etc. With the capability for the user to control protocol attached to each solution,
behavior, according to TCCA recommendations, we could - the problem of interoperability and
hope for a common, general purpose and real-time interchangeability, which was (and is) an open
communication architecture. problem until now,
Two methods are possible for solving this problem: one - the problem of scheduling policies, which are the
is based on the encapsulation of IP datagrams in the time basic element of solution for the real-time
critical data link layer frames, the other is based on the constraints management.
modification of Ethernet frame scheduling to meet real-time And to conclude, going back to the title of this paper, is
constraints. The former was chosen for years by several fieldbus a technology? Fieldbus may be considered as a
fieldbus vendors, and the latter was supported by the technology for the design of automation systems, like any
defenders of Ethernet (or Ethernet variants) as the data link other component or artifact. It is an essential component of
layer for fieldbuses. any automation system, and a major component of a lot of
systems. Several solutions have been promoted,
implemented and tested in real industrial applications. All
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 22

the recent power plants, new factories, trains, cars, new REFERENCES
buildings, etc. include fieldbuses, even if they are invisible [1] E. H. Aarts, “Ambient Intelligence: calming, enriching
to the user. The technology may then be considered as and empowering our lives”, Password, issue 8, July
mature. 2001.
But it is also more than a technology. [2] AFNOR, French Standards NF C46601 to C46607.
Fieldbuses in industrial automation represent more than a FIP bus for exchange of information between
technology because they are, today, true real-time transmitters, actuators a n d programmable
communication networks. And as such, they are also controllers. Published between 1989 and 1992.
relevant to time modeling, time management, and to the [ 3 ] AFNOR, French Standard C46-638, Système de
Communication haute performance pour petits
sciences that have time as an object of study.
modules de données (WorldFIP Profil 1), 1996.
Fieldbuses represent more than a technology because they [4] L. Almeida, “Flexibility and timeliness in fieldbus
are the basis for the emergence of new paradigms for based real time system”, Ph.D. thesis, 1999,
communication and cooperation between agents. The University of Aveiro, Portugal.
difference with the normal OSI world comes from the [5] R. Alur and T. A. Henzinger, “Logics and models of
different expressions of qualities of service when real time: a survey”. In REX Workshop, Springer
considering the applications. New communication Verlag, LNCS 600.1991, pp 74-106.
paradigms [1], [155] have been created with the [6] B. Armitage, G. Dunlop, D. Hutchinson, and S. Yu.
development of sensor networks, of ad-hoc networks, of “Fieldbus: an emerging communications standard”.
ambient intelligence. They are: an extreme mobility, a Microprocessors and Microsystems, Vol 12, N°10,
variable connectivity, a great number of stations, an December 1988, pp 555, 562.
[7] AS-i. Actuator and Sensor Interface, Low voltage
opportunity to discover new stations, the autonomy of
switchgear and controlgear, CENELEC TC17B
agents, and the list goes on. A convergence could perhaps (secretariat)146. 1996.
be found with the definition of a real Time Critical Data [8] BatiBus. French Standards AFNOR, NFC 46 621 to
Link service and protocol. 623 and 629. Un réseau pour la gestion technique et
Fieldbuses represent more than a technology because they administrative des bâtiments. 1991.
have provided an opportunity for extensive research, [9] K. Bender, “PROFIBUS, the fieldbus for industrial
although we did not consider this point, this research automation”, Prentice Hall International, 1993, 240
concerns: pages.
- the protocol verification, [ 1 0 ]BITBUS, IEEE 1118, Standard Microcontroller
- the performance evaluation, System Serial Control Bus. 1991.
- the distributed application design methods, [11] L. Borsi, and E. Pavlik, “The concepts and structures
of distributed process automation systems”. Process-
- the scheduling, and especially the joint
Automation. 1980. (2), 63-70.
scheduling of tasks and of messages, [12] J. Brandes et al. “A real time programming language
- the joint modeling of the system and of the and its application for measuring processes”. In IFAC
network, in order, for example, to analyze the 5th World Congress, Paris. 1972.
impact of network behavior on the system itself, [13] P. F. Brown and C. R. Mac Lean, “The architecture of
- the modeling of devices with different objectives the NBS factory automation”. IFAC Congress,
(proof of interoperability, documentation, Münich, Germany. 1987.
configuration, maintenance, etc.). [14]A Burns, “Scheduling hard real time systems: a
The road was long to arrive at the Internet solution as a review”. Software Engineering Journal. 1991, pp
common communication stack for a large spectrum of 116-126.
applications. This choice was accompanied by a drastic [ 1 5 ]P. Burton, “FieldBus : an overview of current
proposals”. IEE Colloquium on « Industrial LANs :
reduction in the number of operating systems. Will we see
the real issues ». London, UK, 1987. 24p, pp 5/1-3.
the same evolution for fieldbuses and for automation [16] CAN, Bosch CAN Specification-Version 2.0 Part A,
operating systems in the near future? R. Bosch Gmbh, Germany, 1991.
[17] C. Cardeira and Z. Mammeri, “Scheduling in fieldbus
ACKNOWLEDGMENT based real-time systems”. Real-time computing,
The author would like to thank all his colleagues and the Editors (W. A. Halang and A. D. Stoyenko), Springer
Ph D. students who have contributed to the fieldbus Verlag, 1994, pp. 568-573.
concept development and its story. He would like to thank [18]C. Cardeira and Z. Mammeri, “A schedulability
analysis of tasks and network traffic in distributed
also the reviewers whose remarks have contributed to
real-time systems”, Measurement, 15, 1995, pp. 71-
improve the first version, and Mrs T. Wagner for the 83.
linguistic aspects. [19] R. P. Carson, “Distributed data analysis in computer
Special thanks are addressed to R. Berthoumieux and M. networks”. Industrial-Research/Development. May
Desjardins without whose confidence this adventure would 1981; 23(5): 130-5.
never have happened. [20]S. Cavalieri S., Consoli A. and O. Mirabella,
“Adding multi-master capabilities to Interbus-S”. In
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 23

Fieldbus Technology, Systems Integration, IEEE-Computer-Society-International-Conference. 23-

Networking and Engineering, Proceedings of the 27 Feb. 1987 San Francisco, CA, USA, pp. 80-81.
fieldbus Conference FeT’99, Springer, pp. 30-37. [ 3 7 DIN,
] German Standards 19245-1 to 19245-3,
[21] CENELEC European standard EN 50170. Fieldbus. PROFIBUS, Process fieldbus. 1990.
Volume 1 : P-Net, Volume 2 : PROFIBUS, Volume 3 : [38] DIN, German Draft Standard 19 258-1. Interbus-S,
WorldFIP. 1996. sensor and actuator network for industrial control
[ 2 2 CENELEC,
] High efficiency communications systems for for CENELEC (1996b).
subsystems for small data packages, CLC TC/65CX, [39] DIN, German Draft Standard 19245-4 PROFIBUS-PA,
EN 50254 Project. 1996. Profibus for Process Automation.
[ 2 3CENELEC,
] Criteria f o r "High efficiency [40] DS, Danish Standard, DS 21906. P-Net, Multi-master,
communications subsystems for small data packages" multi-net fieldbus for sensor, actuator and controller
CLC/TC65CX(SEC)040. 1996. communications.
[24]J. M. Chang and N. F. Maxemchuck. “Reliable [ 4 1 ]J. Dwyer, “Why General Motors' Manufacturing
broadcast protocols”. ACM Trans on Computer Automation Protocol is here to stay” ? Automation.
systems. 1984. Vol 2, N°3, pp. 251-272. May-June 1985; 21(5), pp. 19-21.
[25] A. Chatha and C. Polsonetti. “Fieldbus standard: we [ 4 2 ]EHS, European Conference on Integrated Home
need a winner”. I&CS, 1992, vol 65, N° 10 pp. 29- Applications, 13-15 January 1991, Amsterdam,
31. Netherlands.
[26] N. Collins, “Boeing architecture and TOP (technical [43]EIBus, French Standard NF C46 624 à 628, Un
and office protocol)”. Networking:-A-Large- réseau pour la gestion technique et administrative des
Organization-Perspective. April 1986, Melbourne, bâtiments.1991.
FL, USA, pp. 49-54. [44] Elloy J.P., P. Molinaro and R. Ricordel, “A temporal
[27]ControlNet, ControlNet Specification, release 2.0. extension to the MMS protocol with the Kernext
ControlNet International. 1997. tool”. In Proc. IEEE-IES Workshop on Factory
[ 2 8 ]L. Costrell, “CAMAC instrumentation system - Communication Systems. 1995. (JD Decotignie Ed).
introduction and general description”. IEEE- EPFL, Lausanne, Switzerland.
Transactions-on-Nuclear-Science. April 1971; ns- [45] EMUG, European Map Users Group. User requirements
18(2), pp. 3-8. for communications in time critical applications. Feb
[29] Y. Dakroury and J.P. Elloy, “A new multi-server 1989, réf ISO/TC 184/SC5/WG2-N180.
concept for the MMS Environnement”. In Proc of 9th [46]D. Fantoni, “A never-ending story: the Fieldbus
IFAC Workshop on DCCS. 1989. standardization”. Automazione-e-Strumentazione.
[30] B. Dasarathy, “Timing constraints of real time systems vol.47, no.11, p.69-75.
: constructs for expressing them, methods for [47] M. Felser and T. Sauter, “The fieldbus war: history or
validating them”. IEEE Trans on Software short break between battles”. Proc. IEEE-IES
Engineering, 1985. 11(1) pp. 80-86. Workshop on Factory Communication Systems, 2002,
[31] J. D.Decotignie, P. Raja, G. Noubir, L. Ruiz and J. Sweden.
Hernandez, “Analysis of polling protocols for fieldbus [48] R.-D Floyd. “Manufacturing Automation Protocol”. In
networks”. Computer communication review, ACM IEEE International Conference on Communications,
Sigcomm, vol 23, 1993, N° 3, pp. 69-90. 1985, 23, 26 June. Chicago, IL, USA. pp 620-624.
[32] J. D. Decotignie and R. Prasad, “Die Feldbusse-der [49] G. Fohler, “Realizing changes of operational modes
grosse Bazar”. Bulletin SEV/VSE, Teil 1, Die with a pre-run time scheduled hard real time system”,
Probleme der industriellen Kommunication, 21/93 pp. Dependable computing and fault tolerant systems,
11-17. Teil 2, Das Pflichtenheft, 25/93, pp. 26-29, Springer Verlag, 1993, pp287-300.
1993, Switzerland. [50]D. Galara and J.P. Thomesse, Groupe de réflexion
[33] J. D. Decotignie, “Some future directions in fieldbus FIP. Proposition d'un système de transmission série
research and development”. In Fieldbus Technology, multiplexée pour les échanges d'informations entre des
Systems Integration, Networking and Engineering, capteurs, des actionneurs et des automates réflexes.
Proceedings of the fieldbus Conference FeT’99, Ministère de l'Industrie et de la Recherche. mai 1984,
Springer, 1999, pp. 308-312. 56 pages. Published by Editions KIRK, Paris, France,
[34] P. Deschizeaux, R. Griesner and P. Ladet, “A real time 1991.
operating system for microcomputer network”. [51]M. Gault and J..P. Lobert, Contribution for the
SOCOCO’1979. 2 nd IFAC/IFIP Symposium on fieldbus standard, to IEC/TC65/SC65C/WG6, 1985.
Software for Computer Control. Praha. 1979, pp. [52] R. W. Gellie, “`PROWAY'-a standard for distributed
MIII-1, MIII-4. control systems”. Proceedings of the Specialists'-
[35] D. Dietrich and T. Sauter, “Evolution potentials for Meeting on Distributed Systems for Nuclear Power
fieldbus systems”. WFCS2000. IEEE International Plants. 14-16 May 1980 Chalk River, Ont., Canada,
Workshop on Factory Communication Systems. Sept pp. 65-100.
6-8 2000. Porto Portugal. pp. 343, 350. [53] R. Genier, “The THEMIS solar power plant.” Epure,
[36] S. R. Dillon, “Manufacturing automation protocol and France. Jan. 1985; (5): 3-17.
technical and office protocols - success through the [54] W. George, “A good year for fieldbuses”, August 1996,
OSI model”. COMPCON-Spring-'87.-Thirty-Second- C&I , vol 28, N°8, pp. 26-28.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 24

[55] J. Gertler and J. Sedlak, “Software for process control – [72] IEC IEC 61804-1: Function Blocks (FB) for process
A survey.” Automatica, vol 11, 1975, pp. 613-625. control – Part 1: General requirements. 65C 269
[ 5 6 ]Gifford,-C.-A., (1974). “A military standard for CDV. 2002.
multiplex data bus”. Proceedings of the IEEE-1974, [ 7 3 ]IEC, IEC 61158-3. TC65/SC65C. Digital Data
National Aerospace and Electronics Conference. 13- Communications for measurement and control-
15 May 1974 Dayton, OH, USA, pp. 85-8. Fieldbus for use in industrial control systems. Part 3
[57]K. Grant, Users Requirements on Time Critical Data Link Service Specification.2003.
Communications Architectures. Technical Report [ 7 4 ]IEC, IEC 61158-5. TC65/SC65C. Digital Data
1992, ISO TC184/SC5/WG2/TCCA. Communications for measurement and control-
[58] M. Graube, “The carrier band network and Mini-MAP: Fieldbus for use in industrial control systems. Part 3
low-cost solutions”. Control Engineering. Oct. 1986; Application Service Specification. 2003.
33(11): 30-1. [75]IEC, IEC 61499 - 1Function block for industrial
[59]P. Griem and J. W. Bernard, “Considerations for process measurement and control; Part 1 Architecture
distributed industrial control systems”, International CDV 65 – 338. 2004.
telemetering conference. 28-30 Sept. 1976 Los [76] IEC, IEC 61499 –2 Function Block - Part 2 Software
Angeles, CA, USA. . pp. 686-91. tool requirements – CDV 65 – 339. 2004.
[60] G. H. Gürtler, “Fieldbus standardization, the European [77] IEC, IEC 61804-2: Function Blocks (FB) for process
approach and experiences”. In Feldbustechnik in control – Part 2: Specification of FB concept and
Forschung, Entwicklung und Anwendung. Springer, Electronic Device Description Language (EDDL),
1997, pp. 211-216. 65C 324 FDIS. 2004.
[ 6 1D. ] Harrold, “Ethernet Everywhere”, C o n t r o l [78]IEEE, IEEE 1588 Standard for a Precision Clock
Engineering, vol 46, 1999, N°6, pp. 46-52. Synchronization Protocol f o r Networked
[ 6 2E. ] Hassler, “Industrial master/slave system”. Measurement and Control Systems. 2002.
Technische-Rundschau. 8 April 1980; 72(14), pp.17- [79]INI, The MAP Book: An introduction to Industrial
18. Networking. Industrial Networking Incorporated, 3990
[63]E. Hassler, “An industrial master-slave system. Freedom Circle, Santa Clara, CA 95052 8030. 1987.
Organisation of communications for decentralized [80]F. Inose, K. Takasugi, M. Hiroshima,-M, “A data
freely-programmable controls”. T e c h n i s c h e - highway system”, Instrumentation-Technology. Jan.
Rundschau. 29 April 1980; 72(17), pp. 25-6. 1971; 18(1). pp 63-67.
[64] W. Hodson, “Will fieldbus kill the DCS?” Control [81] ISA, Proway LAN Industrial Data Highway, ISA –
systems, vol 15.1988, N°2, pp. 21-23. S72.01, Instrument Society of America, 1985.
[65]IEE, Colloquium on `Distributed Process Control; [82] ISA, SP50 “Field bus standard for use in industrial
Today and Tomorrow', IEE, 1 March 1982 London, control systems”, “discussion draft and
UK questionnaire for functional requirements”. 1986.
[66] IEC, IEC 870-5. Telecontrol equipment and systems- [83] ISA, SP50 “Field bus standard for use in industrial
Part 5: Transmission protocol- Section 1: control systems”, “Functional Guidelines”. Various
Transmission frame formats, Section 2: Link versions in 1986.
transmission procedures, Section 3 : General [ 8 4I]S A , . ISA-SP50-1987, Fieldbus « functional
structure of application data, Section 4 : Definition guidelines », September 1987.
and coding of application information elements. [85]ISA, ISA-SP50-1987, Fieldbus – draft standard,
1990. September 1987.
[67] IEC, IEC Standard 1158-2, Fieldbus standard for use [86] ISO, ISO 7498, Data Communications, Open System
in industrial control systems- Part 2 Physical layer Interconnection – Basic Reference Model.
specification and service definition + AMD1 (1995). [87]ISO, Information processing systems, Local area
[ 6 8 ]IEC. I E C . TC44 (Sec)148 Draft Standard for networks-Part3: Carrier Sense Multiple Access-
Electrical equipment of industrial machines-Serial Collision Detection. 1990.
data link for real time communications between [88]ISO, Information processing systems, Local area
controls and drives, (SERCOS). networks-Part4: Token Bus access method. 1990.
[69] IEC, IEC/TC57. TR 870-1-4. Telecontrol equipment [ 8 9ISO,] ISO/IEC IS 9506 Manufacturing Message
and systems-Part 1: General considerations: Basic Specification. 1990.
aspects of telecontrol data transmission and [90] ISO, TR 12178, Industrial automation, Time critical
organization of standards IEC 870-5 and 870-6. communication architectures-User requirements.
1994. TC184/SC5/WG2. 1994.
[70] IEC, TC65/SC65C. Digital Data Communications for [91]ISO, IS 11898 Road Vehicle-Interchange of digital
measurement and control-Fieldbus for use in Information- Controller Area Network for high speed
industrial control systems. Part 3 Data Link Service communication.. 1995.
Specification, IEC 1158-3, IEC 65C/160/CDV.1996. [92]ISO, DIS 13236 Information technology-Quality of
[71] IEC, TC65/SC65C. Digital Data Communications for Service-Framework. 1996.
measurement and control-Fieldbus for use in [93] ISO, WD 13283, User requirements for TCCS and
industrial control systems. Part 4 Data Link Protocol Network Management requirements,
Specification, IEC 1158-4, IEC 65C/161/CDV. 1996. TC184/SC5/WG2-N582, 1996.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 25

[94] H. Kopetz, “Event triggered vs time triggered real time [ 1 1 3L.] Motus L. “Time concepts in real-time
systems”. LNCS Vol 563 Springer Verlag, 1990, pp. programming.” IFAC/IFIP Workshop on real time
87-101. programming, 24, 26 June 1992, Brussels, pp. 1-10.
[95] H. Kopetz and G. Gründsteidl. “TTP, A time triggered [114]J. A. Murphy, “Token-passing protocol boosts
protocol for fault tolerant real time systems”, throughput in local networks”. Electronics. 8 Sept.
Computer, vol 27, 1994, N°1, pp. 14-23. 1982; 55(18), pp. 158-163.
[ 9 6 ]H. Kopetz, W. Elmenreich and C. Mack, “A [115] N. Nakano, “Time critical communication architecture
comparison of LIN and TTP/A”. WFCS2000. IEEE in Factory automation”. IFIP Information
International Workshop on Factory Communication Infrastructure Systems for Manufacturing (B-14). Ed
Systems. Sept 6-8 2000. Porto Portugal, pp. 99, 107. H. Yoshikawa and J. Goosenaerts. Elsevier Science
[97]M. Kronental, “Towards the standardization of real B.V. 1993, pp. 363-374.
time operating system kernels”. SOCOCO’1979. 2nd [116] N. Navet, Evaluation de performances temporelles et
IFAC/IFIP Symposium on Software for Computer optimisation de l’ordonnancement de tâches et de
Control. Praha.. messages. Ph. D. Thesis, Nov 1999, INPL, Nancy,
[98] L. Lamport, “Time, clocks and the ordering of events France.
in a distributed system”. Communications of the [117]P. Neumann, “Locally distributed automation-but
ACM, July 1978, Vol 21, N°7, pp. 558-565. with which fieldbus system?” Assembly-Automation.
[99]N. Laurance, “MMS over the three layer stack”. vol.19, 1999, no.4, pp. 308-12.
Contribution to ISO TC184/SC5/WG2-TCCA Oct [118]D. Norris, “Smart Distributed System : distributed
1992 Réf TC184/SC5/WG2/N340. control for factory floor automation.” FieldComms’95
[100] G. Le Lann, “Distributed systems, towards a formal Making the most of the fieldbus Conference
approach”. IFIP World Congress, North Holland, Proceedings GGH Marketing Commun. Titchfield,
1977, pp. 155-160. UK Vol 1, 14 pages.
[ 1 0 1 G.
] Le Lann and N. Rivierre, “Real time [119] M. Ochsuer and M. Schrier. (1997) “To fieldbus or
communications over broadcast networks: the CSMA- not to fieldbus”. InTech, vol 44, 1997, N°10, pp. 44-
DCR and the DOD/CSMA-CD protocols”, RTS'94, 48.
pp. 67-84, Teknéa, Toulouse, France. [ 1 2 0ODVA
] Open DeviceNet Vendor Association,
[102] M. Leon Chavez M, “Fieldbus and real time MAC DeviceNet Specification, release 2.0. (2001).
protocols”, SICICA 2000, IFAC Conference, Buenos [121] J. S. Ostroff, “Formal methods for the specification
Aires. and design of real time safety critical systems”.
[103] P. Leviti, IEC 65C/166/INF, June 1966, part A and Journal of Systems and Software, 18 1992 (1).
B, 80 pages. Tutorial on the Data Link Service [122]D. Pancucci, “Peace breaking out on fieldbus”,
Specification, Digital Data Communications for Manufacturing Computer Solutions, vol 5, N°7,
measurement and control-Fieldbus for use in 1999, pp. 48-51.
industrial control systems, (Part 3 of the IEC 1158-3, [123] D. Paret, Le réseau CAN, Controller Area Network,
IEC 65C/160/CDV), IEC TC65/SC65C. Dunod, 1996, France.
[104]P. Leviti, “IEC 61158: an offence to technicians”. [124] D. Paret and C. Fenger, The I2C bus from theory to
IFAC Int. Conf. on Fieldbus Systems and their practice (book and disk). Publ. Wiley, 1997.
Applications, FET’2001. Nov 15-16 2001. Nancy, [ 1 2 5 ]M. Pearson, “Implementing MAP/EPA in the
France, Ed by Dietrich, Neumann and Thomesse, manufacturing cell”. FMS-Magazine. Jan. 1987; 5(1):
Pergamon, pp 9-16. 15-18.
[ 1 0 5 ]LonWorks, Documentation Echelon Corporation, [126]J. F. Peyrucat, “Interautomaton communication via
4015 Miranda Avenue, Palo Alto, CA 94304, 1995. networks”. Mesures,-Regulation,-Automatisme. 24
[106] G. Loose, “A user view of the user layer”. C&I, Vol Jan. 1983; 48(1), pp. 35-37, 39, 41.
28, N°5, May 1996, pp. 60-61. [127]T. P. Phinney, D. Brett, D. McGowan and Y.
[107] Z. Mammeri and P. Lorenz, “Integration of temporal Kumeda, “FieldBus-Real-Time comes to OSI”. 10th
mechanisms in communication protocols for time Annual International Phoenix Conference on
critical distributed systems”. 12th IFAC Workshop on Computers and Communications, IEEE Comput Soc.
Distributed Computer Control Systems. 1994, Press, Los Alamitos, CA, USA, 1991, pp. 594-599.
Toledo, Spain. [128]R. Piggin, K. Young and R. McLaughlin, “The
[ 1 0 8 ]J. Marx, “MAP and EPA: on the road to current fieldbus standards situation – a European
connectivity”. I&CS-. Dec. 1987; 60(12). pp. 28-30. view”. Assembly automation, Vol 19, N°4, 1999, pp.
[109]S. McClelland, “A Hart to Hart with Rosemount”. 286-289.
Sensor-Review. April 1989; 9(2): 71-74. [ 1 2 9J.] Pimentel,. “Fieldbus application layer:
[110]M. J. McGowan, “Process bus protocol orchestrates functionality and models”. Proc. Of the 8th IFAC
distributed or centralized control”. Control- Distributed Computer Control Systems. Pergamon,
Engineering. Sept. 1980; 27(9): 129-132. Oxford, 1988, pp. 15-20.
[111]G. A. Mitchell, “Ethernet’s in control”. Control [130]P. Pleinevaux and J. D. Decotignie, “Time critical
Engineering, May 2000, vol 47, N°5, pp. 46-54. Communication Networks: Field Busses”, IEEE
[112] H. M. Morris, “Distributed system makes wide use of network magazine, vol 2, 1988, pp. 55-63.
bubble memories”. Control-Engineering. Jan. 1982;
29(1), pp. 68-70.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 26

[131]K. Pluhar, “Introducing four more new integrated control”. IFAC/IFIP Workshop Real time
distributed control systems”. Control-Engineering. programming. Pergamon Press, 1977, pp. 47-54.
Aug. 1980; 27(8), pp. 45-57. [148] J.P. Thomesse and P. Noury, Communication models
[132] L. Pouzin, “Virtual circuits vs. datagrams – Technical Client-server vs Producer - Distributor-Consumer.
and political problems”. National Computer Contribution à ISO TC 184/SC5/WG2-TCCA. 1989.
Conference, 1976, pp. 483-494. [149]J.P. Thomesse and Lainé T. “The field bus
[133]L. R. Qualls, “Advantages of a time division application Services”. Proceedings IECON'89 15th
multiplex data bus for remotely piloted vehicle built- Conference IEEE-IES Factory Automation,
in test”. Proceedings of the IEEE-1976 National Philadelphie, USA, pp. 526-530.
Aerospace and Electronics Conference. 18-20 May [150] J.P. Thomesse, P. Lorenz, J.P. Bardinet, P Leterrier
1976 Dayton, OH, USA. pp. 203-207. and T. Valentin, “Factory Instrumentation Protocol:
[134] N. Raja, “Static and dynamic polling mechanisms for model, products and tools”, Control E n g . 38,
fieldbus networks”. Operating systems review 27(3), 12,1991, pp. 65-67.
1993, pp. 34-45. [151]J.P. Thomesse, “Le réseau de terrain FIP”, Revue
[135]A. Reeve A, “Which fieldbus will you use –and Réseaux et Informatique Répartie, Hermès, Vol 3,
when?” Control & Instrumentation, vol 25, 1993, N°3, 1993, pp. 287-321.
N°5, pp. 67-70. [152]J.P. Thomesse, “Time and industrial local area
[136] C. W. Rose and J. D. Schoeffler, Microcomputers in networks”, Proc COMPEURO’93 Paris. May 24-27
instrumentation and data acquisition systems, 1993. pp. 365-375.
Analysis-instrumentation,-vol.12. 1974, pp. 157-163. [153] J.P. Thomesse, “A review of the fieldbuses”. Annual
[137] G. Saba, J. P. Thomesse and Y. Q.Song, “Space and reviews in Control 22, 1998, pp. 35-45.
time consistency qualification in a distributed [154]J.P. Thomesse, “Fieldbuses and interoperability”.
communication system”. Proc of IMACS/IFAC Int Control Engineering Practice 7, Pergamon, 1999, pp.
Symposium on Mathematical and Intelligent Models 81-94.
in System Simulation, Brussels, Belgium, April 12-16 [ 1 5 5 ]J.P. Thomesse and M. Leon Chavez, “Main
1993, Vol 1, pp. 383-391. paradigms for current fieldbus concepts”, In
[138]T. Sauter and M Wollschlaeger, “Feldbussysteme- Proceedings of the fieldbus Conference FeT’99,
Historie, Eigenschaften und entwicklungstrends.” Springer, pp. 2-15.
Informationstechnik und Technische Informatik, vol [156] K. Tindell, A Burns and J. Welling, “Analysis of real
42, 2000, pp. 7-16. time communication”. The Journal of real time
[139]V. Schiffer,. “The CIP family of fieldbus protocols systems, 9, 1995, pp. 147-171.
and its newest member – Ethernet/IP”. Conference on [157]K. Tindell, A Burns and J. Welling, “Calculating
emerging technologies and factory automation, ETFA controller area network message response times”,
2001, pp. 377-384 Control Engineering Practice, Vol 3, (8), 1995, pp.
[140]T. H. Schwalenstocker, “A process control system 1163-1168.
using multibus”. Proceedings of the First Annual [158] P. H. Troutman, “A digital link for controllers and
Control Engineering Conference. 18-20 May 1982 valves”. Instrumentation-Technology. July 1978;
Rosemont, IL, USA 1982, pp. 133-137. 25(7), pp. 55-7.
[141]SDS, “Smart Data Systems”, CENELEC Working [159]Y. Tsukada, “Trends in process control systems”.
Draft 0.2, Low voltage switchgear and controlgear- JEE-Japan-Electronic-Engineering. Feb. 1976; (110):
Part 5, control circuits and switching elements- EN 24-7.
60947-5.X., 1997. [ 1 6 0UML-TR,
] UML™ Profile for Schedulability,
[142]Y. Q. Song, F Simonot and J.P. Thomesse, Performance, and Time Specification, January 2002
“Message sojourn time for TDM schemes with any draft, OMG Headquarters, 250 First Avenue,
buffer capacity”. IEEE Trans on Communications 43, Needham, MA 02494, USA.
1995, pp. 1013-1021. [ 1 6 1 ]F. Vasquez and G. Juanole, “Pre run time
[143] M. Soutif, Rapport sur l'Industrie des Instruments de schedulability analysis in fieldbus networks,” Proc of
Mesure, Ministère de la recherche, 1982, Paris, IECON’94, IEEE Conf on Industrial Electronics, pp.
France. 1200-1204.
[144]J. A. Stankovic, “Misconceptions about real time [162]L. Vega-Saenz and J.P. Thomesse, “Time in
systems”. Computer, Vol 21, N°10, October 1988, distributed systems-Cooperation models and
pp. 10-19. communication types”, 5th workshop on Future
[145]K. M. Sturgis, “GMs manufacturing automation trends of distributed computing systems, IEEE
protocol”.-Proceedings of the Conference Local-Net- Computer Society Press, 1995, pp. 41-49.
84. 10-12 Oct. 1984 San Diego, CA, USA, 1984: 13- [163] M. Welburn, “Take the bus… but don’t get on the
22. wrong one”. Control and Instrumentation, vol 30, N°
[146] J. W. Szymanski, “Embedded internet technology in 8, 1998, pp. 41-42.
process control devices”. WFCS2000. IEEE [164] G. G. Wood, “Fieldbus services under MAP”. ISATA
International Workshop on Factory Communication 17th International Symposium on Automotive
Systems. Sept 6-8 2000. Porto Portugal, pp. 301-308. Technology and Automation Proceedings. Allied
[147]J. P. Thomesse, “A new set of software tools for Autom, Croydon, UK 1987. Vol 1. pp. 87134/1-9.
designing and realizing distributed systems in process
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 27

[165]G. G. Wood, “Survey of LANs and standards”.

Computer standards and interfaces. Vol 6, N°1 ;
1987, pp. 27-36.
[166] G. G. Wood, “Current fieldbus activities”. Computer
Communications. Vol 11, N°3, pp 118-123.
[167]G. G. Wood, “International standards emerging for
fieldbus”. Control Engineering. Vol 35, N° 10, pt2,
pp. 22-25.
[168]H. Zimmermann, “OSI reference model. The ISO
model of architecture for open system
interconnection”, IEEE Trans. COM-28 n°4, April
80, pp. 425-432.