Since the discussion began about the extent to which Ethernet can be applied to automation technology and perform the tasks of the classical fieldbus in that environment, manufacturers and users alike have been paying a great deal of attention to the question of speed. Talk of 100 Mbs or even of gigabits suggests almost unlimited possibilities for transferring data at increasing speeds.
The considerable difficulties involved in both operation and installation are however, often not mentioned. The approach of designing more efficient protocols, and thus achieving high net data rates at the lowest possible transfer rates, seems to have fallen into neglect. This article provides an overview of the efficiency of common fieldbus protocols, and shows that the performance of a system cannot usefully be measured by its transmission rate.
Topology of common fieldbus systems
The following observations and calculations are related to application within an automation environment. Market studies such as ARC indicate that as far as Germany and Europe are concerned, CAN, Interbus and Profibus are the systems most relevant to the market. The number of devices networked in each application is, according to information provided by the corresponding fieldbus organisations, generally in the region of 10 or 20 - with a tendency to increase. This has also been taken into account in our analyses. No reliable data is yet available about the sizes of networks and the data quantity involved in Ethernet applications in automation. For the purposes of comparison, therefore, the same system parameters have been assumed for Ethernet.
Consideration of the efficiency of various
protocol types
In order to calculate efficiency, the ratio of user data to the total data transmitted is determined, where user data refers to the information that a device has to transmit. Efficiency is calculated for two different network topologies. Topology 1 contains 50 devices, each of which has 8 bits of information. This represents an example of a highly distributed structure. The second topology assumes 20 devices, each of which has 32 bits of information, as an example of a typical network today.
CAN - controller area network
Figure 1 shows the structure of the CAN protocol. The data width can be between 0 and 64 bits. This information data is compared with an overhead of 47 bits, assuming that the interframe space is 3 bits. This yields the following efficiency figures:
Topology 1 (50 devices, each with 8 bits
of information)
Total user data: 50 x 8 bits = 400 bits
Total data transmitted: 50 x 55 (47 + 8) bits
= 2750 bits
Efficiency: 400/2750 = 14%
Topology 2 (20 devices, each with 32 bits
of information)
Total user data: 20 x 32 bits = 640 bits
Total data transmitted: 20 x 79 (47 + 32) bits
= 1580
Efficiency: 640/1580 = 40%
Ethernet/TCP/IP
The messages exchanged over the transmission medium are based on Ethernet packets. An Ethernet packet itself (Figure 2) transports a minimum of 46 and a maximum of 1500 data bytes. As Ethernet involves unchecked data exchange (there are no acknowledgements, nor is data repeated in the event of an error), reliable transmission can only be ensured if communication software is used. TCP/IP is suitable for use here, and its structure is also illustrated in Figure 2. The following calculations can be used to evaluate efficiency.
Topology 1 (50 devices, each with 8 bits of information)
Total user data: 50 x 8 bits = 400 bits
Total data transmitted: 50 x 576 bits = 28 800 bits
Efficiency: 400/28 800 = 1%
Topology 2 (20 devices, each with 32 bits of information)
Total user data: 20 x 32 bits = 640 bits
Total data transmitted: 20 x 576 bits = 11 520 bits
Efficiency: 640/11 520 = 5%
These calculations have assumed minimum header size, so that the messages are 72 bytes in length.
Interbus
Figure 3 illustrates the structure of the Interbus protocol, whose most important components consist of the loop-back word (LB), the process data, the frame-check sequence (FCS) and the end marker. The Interbus ring structure means that the message is only sent once, regardless of the number of devices. Overhead in the form of addresses is only present once, independently of how many devices there may be, each byte having an additional 5 bits. This yields the following efficiency figures:
Topology 1 (50 devices, each with 8 bits of information)
Total user data: 50 x 8 bits = 400 bits
Total data transmitted: [50 x 13 (8 + 5 bits)] + 78 bits = 728 bits
Efficiency: 400/728 = 55%
Topology 2 (20 devices, each with 32 bits of information)
Total user data: 20 x 32 bits = 640 bits
Total data transmitted: [20 x 4 (8 + 5 bits)] + 78 bits = 1118 bits
Efficiency: 640/1118 = 57%
Profibus DP
Figure 4 shows the structure of the Profibus DP protocol. This is based on the variable information length data format. The efficiency can be calculated as follows:
Topology 1 (50 devices, each with 8 bits
of information)
Total user data: 50 x 8 bits = 400 bits
Total data transmitted: 50 x 80 (72 + 8) bits
= 4000 bits
Efficiency: 400/4000 = 10%
Topology 2 (20 devices, each with 32 bits
of information)Total user data: 20 x 32 bits = 640 bits Total data transmitted: 20 x 104 (72+32) bits
= 2080 bits
Efficiency: 640/2080 = 31%
Effects on the cycle time
The cycle time of a system is mainly determined by the data transmission time and the running time of the software. In order to show the effect of the efficiency on the cycle time, the running time of the software must be ignored, even though, depending on the particular implementation, it can be quite significant (0,2–1,5 ms). Making that assumption, the following transmission times* are found for the various systems for topology 1:
CAN (1 Mbps): 2,62 ms
Ethernet/TCP/IP (10 Mbps): 2,75 ms
Interbus (0,5 Mbps): 1,39 ms
Interbus (2 Mbps): 0,35 ms
Profibus DP (1,5 Mbps): 2,54 ms
Profibus DP (12 Mbps): 0,32 ms
*Neither software running times nor delays caused by collisions were taken into account.
Summary
The calculations performed indicate considerable differences in the ratios of the total transmitted data to the user data. Systems with an efficiency between 1% and 30% should, in the topologies described, only be used when the time restrictions are very loose or when correspondingly high transmission rates are available. Depending on the access method and the topology, a limit has already been reached here. If the topologies of serial fieldbus systems at high transmission rates are compared, it is seen that, for data rates above 10 Mbps, a point-to-point topology is always preferable from an electrical point of view. Only Interbus and Ethernet offer this topology. Moreover, the user pays for a high transmission rate through restrictions on cable length and in the choice of transmission medium (eg infrared transmission, radio). Manufacturers of automation equipment must invest more in developing the circuitry, the layout and the connection methods of products if they are to achieve acceptable EMC performance. The effect of the efficiency on the cycle time is considerable. In particular, when medium data quantities per device are combined with a high device count, it can be seen that other systems, having transmission rates 20 times greater (Ethernet) or six times greater (Profibus), do not in fact offer any speed advantage over Interbus.
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version