Image Processing Reference
In-Depth Information
hardwired system with a “Packet Sniffer” node capability providing for real-time monitoring and
time-stamping of information for performance evaluation. The experiment consists of DeviceNet
requests sent out over the Bluetooth network to change the value (i.e., “flip”) of a bit on the I/O block.
When a packet is returned indicating that the bit value has been changed (i.e., a “change-of-state” is
observed), a request is then placed to flip it back, etc. he packet interval of DeviceNet transmissions
and poll rate of the system are set to be representative of typical DeviceNet operations and to ensure
that the packet interval is significantly shorter than the response time of the Bluetooth system. hus, a
practical performance metric of “number of DeviceNet packets between observed COS occurrences”
can be used. Figure .b shows a baseline result with minimal noise interference and ideal distance
between the two Bluetooth nodes; note the tight and bounded distribution around these packets
indicating relatively high determinism. Figure .c shows the same system, but with the distance
increased to m and the antennae of the two nodes oriented to be perpendicular to each other. he
inset reveals that, while average performance does not degrade significantly, there is a significant loss
of determinism due to the expanded tail of the distribution. Figure .d illustrates that the same
problem can occur when noise interference is introduced (note that the noise levels were purposely
set to be below the threshold where the Bluetooth system would invoke frequency hopping to avoid
noisy areas of the spectrum).
The results summarized in Figure . do not represent an exhaustive evaluation of Bluetooth,
nor do they illustrate performance issues that are exclusive to Bluetooth. Rather the results illustrate
one of the general problems with wireless that must be addressed if wireless is to be utilized in time-
critical networked control and safety systems, namely, lack of determinism of the protocol in the
face of external factors and the lack of focus on determinism in the design of the protocol as well as
products and systems utilizing the protocol. he need for determinism in NCSs is explored further
in Section ..
23.4 NCS Characterization
Determination of an optimal NCS (which includes solutions for control, diagnostics, and/or safety
applications) design for a particular manufacturing application requires not only knowing the theory
and trade-offs of the NCS protocols being considered (i.e., a theoretical perspective), but also exper-
imental data to augment the theoretical understanding of the NCS capabilities (i.e., experimental
perspective), and an analytical study that allows for gauging the relative importance of performance
metrics (e.g., end-to-end speed and jitter) to the specific application environment (i.e., analytical
perspective). In this section, we explore these three perspectives in detail and show how they can be
utilized collectively to provide a methodology for NCS design decision making [].
23.4.1 Theoretical Perspective
The theoretical information perspective involves a general overview of network operation and the
metrics that should be evaluated when assessing the best NCS solution for a particular system. he
comparative evaluation of network technologies and implementations is primarily a study of trade-
offs. Indeed, the move from point-to-point to networked systems itself is a study of trade-offs. One
important aspect of the trade-off study is illustrated with the Lian curve, discussed in Section ...
This curve illustrates the impact of network congestion on NCS performance in a digital control
environment. Finding and maintaining NCS operation in the sweet spot for optimized performance
isastudyoftrade-ofsthatisatopicofongoingresearch.Itisimportanttonotethattheterm“net-
work” can be used very loosely in this analysis. For example, oftentimes the network congestion is
actually observed in the devices themselves as they parse and encode data for end-to-end application
communication []. The Lian curve phenomena can still be observed in these environments with
the network definition extended to include the node processing.
