Interview with the Indu-Sol Germany managing directors Karl-Heinz Richter and René Heidl about how reliable industrial communication works.
Mr. Richter, the core business of Indu-Sol has been focused on reliable communication infrastructures for the production level for 18 years. How has this task changed in recent years?
Karl-Heinz Richter: In the beginning we had a lot of educational work to do. PROFIBUS was celebrated everywhere, and nobody wanted to believe that it could also have weak points. We had to explain a lot: That there is something like physical wear and tear, for example, which can lead to communication problems in the medium term. However, practical experience has confirmed our statement. With the switch to PROFINET, not only has the communication landscape changed, but also the problems that can go hand in hand with it. In PROFINET we mainly struggle with sporadic, non-reproducible events that can initially be easily acknowledged and actually appear unproblematic.
But are these sporadic events still problematic?
Richter: Yes, because data is also lost in the process. This becomes visible in consistent network monitoring. And what's more: These sporadic disruptions are always the beginning of an impending network problem.
What has changed with PROFINET in relation to the communication networks?
René Heidl: As an argument in favor of switching to TCP / IP-based communication, we were touted that very different applications can be connected. While with PROFIBUS networks were homogeneous, with PROFINET you suddenly had to deal with heterogeneous networks. In the meantime, many are rowing back again and laying several homogeneous networks in parallel. However, most of them are not really aware of this development.
Why are many people going back to homogeneous networks?
Heidl: In short: Because homogeneous networks are more stable, safer and easier to monitor. But you also have to monitor each network individually and create interfaces to each of these networks.
And that is much more complex?
Heidl: Of course. Just imagine, four lanes are being built in your city center: one for motor vehicles, one for buses and trucks, one for ambulances and one for the fire department. Who wants to pay for that, both for construction and maintenance?
No one. But what does that have to do with the communication networks? Can't all protocols simply drive on one lane?
Richter: Yes, but the traffic lights are the core problem. While six cars are speeding through the traffic lights during a traffic light cycle, one truck has just started, and it turns red again. Sooner or later this leads to a traffic jam. In the network world, the traffic light is the switch that manages the communication protocols. If short protocols, e.g., PROFINET, and long protocols, e.g., visualization data, run via the same switch, there is no jam, but rather data loss if the memory is not large enough. It is precisely because of this that the networks become unstable and that sooner or later will lead to a communication failure and, in the worst case, to a system downtime.
So is the switch the real problem?
Heidl: Exactly. Coming from the homogeneous PROFIBUS network, selection criteria for the right switch were often the housing size, the price, the protection class and the manufacturer. Anyone who used a PLC from Siemens, VIPA, Rockwell or Phoenix usually always used a switch from this provider. This also made perfect sense in homogeneous networks. Because the PLC and switch are very well coordinated with one another.
But should machine and system builders apply different criteria in heterogeneous networks? Why?
Richter: Now you have to deal with a multitude of applications and communication protocols. Some are short and are often sent, others are long and come less often. A classic PROFINET switch is designed for short PROFINET protocols, for example. However, if you now run a visualization application via the same switch, problems may arise. We have already talked about data loss and the like.
What's the alternative?
Richter: Well, either you decide to lay a separate line for each application. However, this becomes expensive both in terms of installation and maintenance. Or you use other criteria when choosing the switch.
Which criteria are important when choosing a switch today?
Heidl: The range. You should think about running the backbone of a machine in gigabit, right from the start, i.e., using backplane capacity, data throughput, buffer size and buffer page size as the main arguments for choosing the switch instead of the name of the manufacturer. Incidentally, this has long been the case in the IT sector. Why automation behaves completely differently here is incomprehensible. Then you would have enough leeway to reliably transfer the data from all applications in heterogeneous networks.
Let's go back to our introductory question: Does your company's field of activity look different than it was 18 years ago?
Richter: Yes. For years we have been changing more and more from a diagnosis provider to an OT network supplier. However, one company branch will probably never completely replace the other. Nevertheless, we have consistently expanded the hardware area. A new development in this area is our Managed Industrial Ethernet Switch PROmesh P10.
What makes it different from other industrial switches?
Richter: For example, it does justice to the new PROFINET 2.4 standard via TSN. And then there is a lot of diagnostic data in the infrastructure of the switch, which actually only needs to be made visible. Therefore, in addition to the shield current measurement, which was already implemented in the previous model, we now also offer line diagnostics that are carried out so intelligently that the switch signals a warning of a total communication breakdown. This enables preventive maintenance of the network.
Why are these diagnostics necessary?
Heidl: If the quality value CRC errors accumulate, this indicates line problems and the question of the cause of the error arises. Either EMC coupling, defective lines or a weakening end device can cause these errors. If the P10 detects errors in connection with high leakage current values, then the cause of the error is in the area of EMC. If the P10 detects errors in connection with poor line quality values, either the connection is defective, or the connected device is defective.
How do you implement the line diagnosis?
Heidl: During the line diagnosis, the form of the real signals is evaluated and processed further via a data matrix. Clear conclusions can be drawn about the line quality based on the transmission quality of the signals. Since this solution was preceded by a large number of tests and also AI-like valuations, we see an absolute market leader here.
Final question: What is the performance data of your switch?
Heidl: We currently offer a backplane capacity of 51.2Gbit / s, a throughput of 19.3Mpps. The buffer size is 2Mbit and the buffer page size is 256Byte.
Technical terms of communication and network technology
Backplane capacity: The backplane capacity is a measure of the amount of data a switch can process per unit of time. It is given in bit/s.
Data throughput: The data throughput describes how many telegrams a switch can process per time unit. Its unit of measurement is pps (packets per second).
Buffer size: The buffer size provides information on the amount of data that a switch can temporarily store (buffer). It is given in bits.
Buffer Pagesize: The total memory is divided into individual memory pages. A maximum of one telegram can be saved per page. The buffer page size specifies the fixed memory size of the memory page and is specified in bytes.
Note: This interview was originally published by SPS-MAGAZIN on February 12, 2021 and has been translated from German to English. To read the original article, please click here: https://www.sps-magazin.de/kommunikation/wie-industrielle-kommunikation-zuverlaessig-gelingt/