Image Processing Reference
In-Depth Information
temporal deviation of the local clock to the virtual reference tick is formed. Next, the local clock
is adjusted such that the local deviation is minimized. Since all nodes in the network perform this
operation, all local clocks move their ticks toward the tick of the virtual global time. he operation
of observation, computing, and correction is performed in every communication cycle.
In case of wrong transmission times of faulty controllers on the communication channel, things
get more complicated. Here, a special part of the fault-tolerant median value algorithm takes over.
This algorithm uses only the best of the measured deviation values. All other values are discarded.
This algorithm ensures that the maximum influence of a faulty controller to the virtual global time is
strictly bound. Additionally, the protocol requires the marking of particular synchronization frames
that can be used for deviation measurement. he reasoning behind this mechanism is twofold. he
first reason is that it is used to pick exactly one frame from a controller in order to avoid monop-
olization of the global time by one controller with many transmit frames. The second reason is
that particular controllers can be excluded from clock synchronization, either because the crystal
is not trustworthy or more likely because there are system configurations in which a controller is not
available.
In case of a faulty local clock, the faulty controller perceives only deviation values that exceed a
particular value. This value is given by the precision value. This condition is checked by the syn-
chronization service and an error is reported to the application. A specific extension of the clock
synchronization services handles the compensation of permanent deviations of one node. In case
such a permanent deviation is detected, a permanent correction is applied. The detection and cal-
culation of such permanent deviations is executed less frequently than the correction of temporal
deviations.
heerrorhandlingoftheprotocolfollowsastrategy,whichidentiieseveryproblemasfastas
possible, but keeps the controller alive as long as possible. Problem indicators are frames that are
received outside their expected arrival intervals or when the clock synchronization does not receive
a sufficient number of synchronization frames. he automatic reaction of the controller is to degrade
the operation from a sending mode to a passive mode where reception is still possible. At the same
time the problem is indicated to the application. he application can react in an application-specific
manner to the detected problem. his strategy gives the designer of a system a maximum of flexibility
for the design of the safety required by the application.
16.3.3 Start-Up
The preceding protocol description handled only the case of an already running system. To reach
this state the start-up service is part of the protocol. Its purpose is to establish a common view on the
global time and the position in the communication cycle.
Generally the start-up service has to handle two different cases. he coldstart case is a start-up of
all nodes in the network, while the reintegration case means to integrate a starting controller into an
already running set of controllers (see Figure .).
During coldstart, the algorithm has to ensure that really a coldstart situation is given. Otherwise
the starting controller might disturb an already running set of controllers. For this reason, the starting
controller has to listen for the so-called listen timeout for traffic on the communication channel. In
case no traffic is detected, the controller assumes a coldstart situation and starts to transmit frames
for a limited number of rounds. In case another controller responds with frames that fit to the slot
counter of the coldstart node, start-up was successful.
In case traffic is detected during the observation period, the controller changes into the reinte-
gration mode. In this mode, the controller has to synchronize the slot counter with the frames seen
on the channel. herefore, the controller receives frames from the channel and sets the slot counter
accordingly.Foracertainperiodoftime,thecontrollercheckstheplausibilityofthereceivedframes
