The $6 million question

The $6 million question

Finally we come to the six million dollar question 'How do I decide on the communication for my plant'?
Some of the points mentioned before are pretty clear.
Cost is a very big factor, both the cable and the installation. It is not always wise to skimp; however,
engineering is becoming by far the most expensive part of any automation system and an undersized
communication network will eventually cost much more in engineering to achieve the required
performance. On the other hand, a lot of money can be wasted on an expensive network which is badly
engineered. The solution is to get some advice, but know enough to make sure you are given what you want.
Redundancy of networks is becoming very popular. This is not a standard product, and if it is, should it be?
Each customer has his own requirements from a redundant network. Should all data be sent on a master
network and the second a stand-by; should data be shared on both networks if they are healthy; should a
complete network be failed because of one fault; or should only that data be re-routed, the options are endless.
Remember though the cost of engineering these solutions. Add this to the cost of the extra interfaces
required and it gets expensive. Most people at the end of the day are worried about cable break, a better
(and cheaper) solution then is dual cable not dual networks.
Speed can be misleading. A baud rate of 9600 bits/second is exactly that. This is a bit rate of the interface
protocol. It does not reflect the amount of data which can be transmitted on the network in one second.
Remember that data transfer is only one of the sections of a transmission protocol. Tests by Modicon for
instance show that the Modbus Plus communication network, operating at 1 Mbaud, has a guaranteed
minimum throughput of 20,000 registers (16 bit) per second, per network.
The most expensive is not always the best. The use of ethernet for inter-PLC/PC communications has long
been debated. At 10 Mbaud it is surely the 'fastest' PLC network available, but is it always? A critical
feature of any control network is determinism. Simply - can you the user guarantee that data in one PLC
when sent will reach its destination within a specified time?
The answer is no. By the very nature of ethernet, CSMA/CD, as more data is put onto the network, so more
collisions occur, more random 'back-off' times and no guarantee of performance. Token passing, although
slower, generally 1 to 2 Mbaud has guaranteed performance, and is generally cheaper. The maximum
transmission time for any network configuration can be calculated.
A classic rule for any PLC application is that repeatability is more important than speed. It is better that data
is received within 1 second all the time than within 100 ms most of the time. Don't write off ethernet
though. It definitely has a role at a higher level (Figure 6) where file transfer is occurring between computers.
Bridges are very useful for isolating similar networks. Separate areas of plant can run on dedicated
networks providing very high speed data transfer between users. The bridges then provide a link between
the networks for the occasional data transfer required between plant areas.
Fibre-optics is coming more and more into its own. The cost of cable is dropping and will eventually rival
that of twisted pair, certainly cheaper than co-ax. 3M has now released hand installation kits with relatively
low dB losses, doing away with the need for expensive installation equipment. The immunity to lightning
(very useful in SA) and HV interference is a big plus point. The cost of the fibre-optic modems, however,
is still relatively expensive.