Image Processing Reference
In-Depth Information
in terms of computing power in the nodes and available data rate and thus bandwidth in the network
were very limited at the time most fieldbus systems were invented. Hence, developers wanted to get
the most out of what was at hand, and they found that a straightforward application of the medium
access strategies used in computer networks was not efficient enough. So, they started designing new,
more efficient methods tailored to individual application areas and optimized for the particular type
of traffic and communication models they had in mind. In fact, the efficiency or overhead in the sense
of how many bits must be transmitted over the network to convey a given amount of user data was
a very common parameter to compare (and market) fieldbus systems—even though the significance
of such a parameter alone was and is more than questionable.
An important classification for the way the data transfer is controlled is the distinction between
single and multi-master systems. The single-master (or master-slave) approach reflects the tradi-
tion of centralized, PLC-based automation systems and is typically used for fieldbus systems in
thelowestlevelsoftheautomationpyramidwheretherolesofthenodesinthenetworkcanbe
clearly distributed. Such networks typically have a limited size and a simple, mostly single-segment
structure. he master either retrieves data from the slaves in direct request-response communication
or explicitly synchronizes time slots that are then used by the slaves to send their data.
The alternative strategy is the multi-master approach, where all nodes are basically equal and
must share the communication medium in a democratic and fair way. Such networks can have
more complex structures, including hierarchical ones with large numbers of nodes, and are found
mostly on the middle level of the automation pyramid (the cell level) and in building automation.
Historically, multi-master fieldbus systems originated from the wish to have truly distributed systems
for process control, and in many cases such fieldbus systems were more than pure communication
networks—they were embedded in comprehensive operating-system-like software frameworks.
As far as the actual MAC strategies are concerned, all fieldbus systems use a time division mul-
tipleaccess(TDMA)strategyinthesensethatthebandwidthissharedinthetimedomain,i.e.,
the network nodes use the communication line sequentially. Other multiplexing methods known in
particular from telecommunications, such as frequency division multiple access, code division mul-
tiple access, or space division multiple access play no or only subordinate roles in the fieldbus world.
More precisely, medium access is managed either in a centralized fashion by polling or time-slot-
based techniques, or it is done in a decentralized way by token passing or random access methods.
Within these basic methods, which will be briefly reviewed below, various variants and blends
areinuse.
20.5.4.1 Polling
Pollingisamaster-slaveaccessschemeandforeseesthataslavenodeisonlyallowedtosenddata
when explicitly told so by a central master (sometime also called arbiter or arbitrator). On the
network, there is a constant alternation between poll messages from the master to each slave and
response messages from the slaves back. In its purest form, polling is strictly cyclic, which means
that the master polls the entire list of slaves one by one and then restarts the cycle. he polling rate
might be adapted (in theory even dynamically) if individual nodes have more data to transmit than
others [], but this is rarely done in practice.
Evidently, polling is perfectly suited for periodic traffic where all process variables need equidis-
tant sampling (Figure .). As the traffic is strictly controlled by one node, the cycle time can be
kept constant provided that slaves implement anti-jabber mechanisms so that they cannot block the
mediuminthecaseoffailure.Likewise,jittercanbekeptlow.Simplepollingisemployedintyp-
ical sensor-actuator bus systems like AS-I, HART, MODBUS, or lower-level fieldbus systems like
BITBUS, PROFIBUS-DP, PROFIBUS-PA. Polling can also be used as an underlying bus access mech-
anism in a multi-master system, such that the permission to access the slaves is rotated between
several master nodes. his strategy is used, e.g., by PROFIBUS-FMS or P-NET.
