Image Processing Reference
In-Depth Information
of bus access requires proper error containment mechanisms to avoid faulty nodes (babbling idiots)
from blocking the medium and jeopardizing real-time behavior.
The real-time properties of time-slot mechanisms also led to the introduction of such approaches
in fieldbus systems that originally use different access control methods. As prominent example, CAN
wasenhancedinthiswaybysuperimposingTDMAstructureslikeintime-triggeredCAN(TT-CAN)
or flexible time-triggered CAN (FTT-CAN). Last not least, it should be noted that the boundaries
between TDMA and other MAC techniques are sometimes blurred. For instance, the access control
method of ControlNet is called concurrent time domain multiple access, although it is rather an
implicit token passing strategy. he flexible time domain multiple access (FTDMA) of byteflight also
has some similarity to token passing.
20.5.4.4 Random Access
Random access basically means that a network node tries to access the communication medium
whenever it wants to—without limitations imposed for instance by any precomputed access schedule.
This principle is called carrier sense multiple access (CSMA), it was first implemented in the ALOHA
network in and since then has been modiied in various ways. Obviously, it is perfectly suited for
spontaneous, peer-to-peer communication. Its conceptual simplicity comes at the expense of a severe
drawback: collisions inevitably occur when several nodes try to access the network at the same time,
even if they first check if the line is idle. he variants of CSMA therefore have one common goal—to
deal with these collisions in an effective way without wasting too much communication bandwidth
and thereby avoiding excessive communication delays.
The best-known CSMA variant is CSMA-CD (Collision Detection) used in Ethernet, more pre-
cisely in its original version with shared medium. Here, collisions are immediately detected by the
sending nodes which monitor the bus while sending. After CD, the nodes abort the data transfer
and wait for a random time before trying again. In fieldbus systems, this variant is not very com-
mon because despite its stochastic component, there is a high probability that collisions remain if
thesourcedataratesaretoohigh.Inpractice,theefectivelyattainablethroughputislimitedtoval-
ues well below % of the maximum data rate, which was always deemed insufficient for automation
networks. It should be noted, though, that things changed with the introduction of switched Ethernet.
A variant where the backoff time (the time a node waits before a retry) is not just random, but
adaptable is called predictive
p
-persistent CSMA and used in LonWorks. Here,
p
denotes the prob-
ability that the node will try again in a certain time interval after the collision. his probability can
be adapted by each node based on an estimation of its backlog and the monitored network load. In
high-load conditions, the nodes reduce their probability for starting a medium access, which in turn
reduces the probability for collisions. LonWorks additionally has a number of high-priority “time
slots” with differing probabilities for medium access, so that messages with a higher priority still
have a better chance to be sent without long delays.
he most widely used CSMA strategy in the fieldbus world, however, uses asymmetric symbols for
coding the bits on the bus line, so that when two different bits are sent at a time, one of them wins
over the other. In this case, there is no actual collision, bus access is nondestructive, and this is why
this strategy is called CSMA-CA (Collision Avoidance), sometimes also called CSMA-BA (Bitwise
Arbitration).heirstieldbustousethismethodwasCAN,andthisexamplewillbeusedinthe
sequel to explain the idea in more detail.
The bus line is designed as an open collector bus so that the low level is “dominant” and the high
level remains “recessive.” his means that a “” sent from an end device can be overwritten by a “.”
CAN uses message-based addressing, and after sending a synchronization bit, the nodes write the
identifier of their message to the bus bit by bit and at the same time observe the current bus level.
If the bits sent and read back differ, arbitration is lost, and the node stops sending. In the end, the
