LIVE: Making sensors network-ready: shopfloor data directly into IT

Hello and welcome, dear listeners, to the IoT Use Case Podcast. Today we’re at Hannover Messe in the Podcast Bus, a converted VW T5 van. Joining me today are Karsten Walther, Managing Director of Perinet, and Georg Bassenge, Chief Sales Officer at Perinet. Today we’re talking about the availability of shopfloor data and seamless transmission all the way into IT — a major challenge. Since we’re here at Hannover Messe: what have been your highlights?

Karsten

We’re now on our second day at Hannover Messe. What has stood out to me in a positive way is that the topic of IoT really is very widespread. Data acquisition from the field is being approached in very different ways by different players. But the presence is there, and I like the fact that people recognize that they need data in IT. Up to now, it was often looked at more from the automation perspective, or in a very automation-centered way. In my view, that has changed, and the focus is shifting more toward the question: how do I get the data into IT?

And Georg, what has been your impression of the trade fair so far?

Georg

It’s only one day old so far — or one day in, depending on how you look at it. What I found is that the topic of artificial intelligence — which was also picked up by the Chancellor, where automation using AI was demonstrated — is moving more and more into the foreground. That’s exciting for us because we create the foundation for delivering data and feeding artificial intelligence. In that respect, I’d say industry and the trends are moving in the right direction.

I personally find that super exciting as well, because from my own background I also come from IoT, from my time at Siemens, and now my focus is on AI. I think it’s really exciting to see these worlds coming together more and more.
And that coming together is supposed to be our main topic today. Karsten, this issue: data is generated at sensors and machines, and in principle you want to bring it all the way into IT. But there are a few layers in between. From the perspective of your customers and your implementation work, can you explain what the complications are? Why is it so difficult to get data from the field into IT?

Karsten

I’d like to use an analogy for how data gets from the field into IT today. You can imagine it like taking a train. I have a lot of local interactions, which would typically be the automation level, at the lowest layer. I talk to the conductor, I have to show my ticket, and I have quite a lot of interactions there on the train.
But then I also have another communication partner at home: my wife. She wants to know when I’ll be home. In a traditional communication architecture, the data passes through several layers. That means I would say to the conductor: please tell the train driver, who should tell the control center, and they should call my wife to let her know I’ll be home on time. That’s cumbersome, there are several translation steps involved. And that makes the path of the data difficult and time-consuming.
On the other hand, today we have the right technology so that sensors can become intelligent, and they can open several communication channels at the same time. That means I can talk locally with the conductor, but I can also talk directly to my wife. And I talk to both of them exactly about what interests them in that specific interaction.
That’s also an issue in traditional automation. In that case, all data has to be passed upward, including data the individual layers may not even care about. By being able to communicate directly with my respective counterparts and exchange only the data or information that is actually needed, communication becomes much easier.

That’s a great image — the conductor essentially calling your wife. It would be interesting to test whether that would actually work.
Can you make that clear with an example from a factory? Which data would be sent directly, perhaps bypassing the MES system, bypassing the controller — depending on the case? Which layers do you see there, and what would the path of specific data look like?

Karsten

Right now, we have quite a lot of customers and partners in the packaging sector that we work with. There, I have many work steps that are controlled locally by a PLC. Cartons are folded, bags are inserted. There’s a lot to coordinate in a small area — how robot arms move, or how mechanical parts move. They have to run in sync with each other. That’s what matters locally to the machine: that the packaging works and that the right products come out.
At the higher-level layer, the more important questions are: how fast is my machine running? Is it running at all? Do I have problems because material needs to be replenished, or because too much output is being generated? That’s also a typical problem: one machine is running, the next one has just stopped, and suddenly I have an issue with the intermediate storage of semi-finished products.
That’s the level I’m interested in from an operations perspective. In other words, I need to see: are my machines running in sync? This is not hard real-time in the sense that I have to shut something down within half a second. But I do need to see: this one machine is going into a fault condition. Then I also need to see that I shouldn’t produce too much material on the machine upstream.
Routing this data through the machine controller is a problem on the one hand because the controller has its own very tight cycle and doesn’t really handle occasional upward data delivery well. And on the other hand, operations would be completely flooded with data if all raw data were passed upward in every cycle, and then people would first have to laboriously extract the actual information from it.

Interesting that the discussion is always about data being the new oil, data being gold — but in the end it depends on having the right data. How would you look at that? What are the right data, and how can that be described?

Karsten

A term from IT helps us here. IT stands for information technology. The difference is between data and information. Data is, first of all, just the factual capture of everything that is there, and information is the novelty contained in that data that I’m actually interested in. In other words: what is the information in it? What is the new insight that enables me to make good decisions?
Seen that way, you could say data is actually the hay, and information is the needle. If I only increase the amount of data, I’m making my problem more complex than it really is.

Interesting — finding the needle in the haystack. That’s a bigger challenge than if everything were oil. From your perspective, can you describe what you changed here? We have these challenges, we have a lot of data, we need to transmit the right data — and in different ways to different recipients. What has Perinet contributed here?

Karsten

We made sensors network-capable — even the smallest things. We looked at how IT has evolved: at the beginning, only large servers were connected. Then it moved to individual computers, and eventually in industry very successfully to edge computers. And the last layer still missing for this kind of IT communication is the boundary to the physical world — individual sensors and actuators.
We enable sensors and actuators to be connected directly to IT systems so that I can read the relevant information from those sensors. At the same time, I enable other forms of communication. In a traditional fieldbus communication setup, I always have one central unit polling all sensors cyclically. In a network communication setup, the sensor can report on its own when something happens. That is more efficient.
We were talking about hay earlier: I can drastically reduce the data volume directly at the sensor and send only relevant information upward — and above all, already in the language of IT. Those are IP protocols, in other words IP communication.

Georg, how do you present this to customers? What’s the storyline that makes customers understand it quickly and intuitively?

Georg

There are different storylines. Here at Hannover Messe, we’re using the storyline of IoT retrofit in the brownfield space. We tell the customer: if they have an existing machine, an old machine that still runs largely in analog mode, we enable them to make that machine IoT-capable and read out data at low cost and with minimal integration effort. That way, they can continue to leverage the asset and keep generating revenue with it, instead of having to replace the machine or even shut down the site.

So especially retrofit, brownfield, existing systems: they’re improved again and their lifecycle is extended.

Georg

Yes, that’s one of the core messages we’re putting forward here at Hannover Messe.

Karsten, from your perspective: what priorities are you setting in this area as Managing Director? Where are you investing in development? What are the latest trends you see there?

Karsten

In the end, it’s a philosophy you have to understand, based on a phrase: we see IT first. IT has taken over many areas of life, and we see the same thing happening in this environment. That changes the world both for automation engineers and for IT. Bringing those worlds together is what we do.
Technically, we’ve made the electronics so small that we can create a network path. For automation engineers, the mode of communication changes: I’m no longer operating cyclically in that clock and exchanging my process image; instead, I now have IT communication where I can transmit information.
But something also changes for IT. Up to now, IT had centralized servers on the internet, and now suddenly every sensor on a machine becomes a server. Their worldview changes as well. We’re working on making that change as easy as possible for the customer, so that ideally they don’t even notice it. The goal is to make this technology possible and as user-friendly as we can.
We’ve had the fundamentals in place since 2021, when we entered the market. Now the main development focus is on making it even easier and smoother to use in practice.

Georg

Long term, we’re also trying to win field device manufacturers as partners, so they can use our technology as an in-design solution. That enables companies to set up new business models such as condition monitoring or subscription models that they previously couldn’t run with their existing analog technology.
On top of that, we also enable companies to design their devices securely, or to let them communicate securely with the internet, which is a requirement under the EU CRA regulation.

Do you also encounter resistance there? Sometimes regulations are helpful in breaking through resistance. But often OT and IT are almost in ideological opposition. More and more requirements around update cycles are coming from IT into OT, where systems have traditionally been used very stably and over long periods. When you say “IT first,” are there people with an OT heart who say, “We don’t want that at all”? How do customers perceive this?

Karsten

It really varies a lot. It depends on the individual people involved. What matters is that both sides can benefit from it. On the OT side, I’m not going to achieve security — especially if we’re talking about the CRA — unless I plan for certain things that have to be done, like updates and key exchange.
On the other hand, IT might also be able to learn from OT about how to build stable systems, because that’s what it’s all about. It doesn’t help me to have a machine that can’t be attacked if it ends up standing still because key exchange didn’t work.
The hardest part is getting both sides to see the benefit, because at first they’re looking into a world that feels foreign to them. And in fact, we don’t just have OT — beneath that we also have AT, meaning automation technology.
I find that extremely exciting. I studied computer science and electrical engineering at the same time.
Among the electrical engineers I was the computer scientist, and among the computer scientists I was the electrical engineer. That exact kind of mediation is exciting. When people come together, they create more value for both sides.

Was that also the trigger for why you founded your own company? How did that come about? What gap in the market did you see?

Karsten

We had the edge computer and we had the data analyst. They could have analyzed all the data. The problem was that the projects failed because the accessibility of the data wasn’t there.
We had several projects. One example: we had a machine park at a company with 30 injection molding machines. That machine park had grown over decades, and every machine had a different interface. The engineering effort required to access the data — even though it was in the machine — was extremely high, and that caused the project to fail.
At the same time, we had a new technology, Single Pair Ethernet, which makes it possible to bring network communication into the smallest devices. We seized that opportunity. And that’s where my profile also comes into play, because I’m both hardware-oriented and software-oriented, and I was able to make that integration work and make the system so small.
Ultimately, the goal is to improve access to data and thus make these digitization projects possible in the first place.

Two questions on that. Single Pair Ethernet: what are the advantages, and what does “making sensors small” actually mean? Do you have dimensions or orders of magnitude? And where does the data go? Are you going both ways? Are you going into MES systems, but also directly into IT, bypassing other systems, out to the internet? Do you use both paths, or are you focused on one?

Karsten

Our current approach is to deliver the data upward into the IT system in parallel to the existing automation. That means we go directly to the sensor. The sensor continues to be queried by the PLC or the MES system, and we send the relevant data or information directly over the network into the IT layers.
This is where the advantage of Single Pair Ethernet comes in. Single Pair Ethernet is another Ethernet physical layer. The packets sent over it don’t care whether it’s Single Pair Ethernet, classic Ethernet, Wi-Fi, or fiber optic. They remain the same packets. That’s a major advantage, because I can now use the network for data transmission and connect any points I want.
The big added value is that the electronics have become smaller. We all know it: CPUs are getting faster and smaller all the time. Today, on the size of a fingernail, we have complete servers. But the network interface hasn’t shrunk at the same rate over the last 30 or 40 years. We know the RJ45 jack with the magnetics behind it — that’s huge.
The available computing power per unit of space has continued to increase, but the physical interface has not shrunk to the same extent. With Single Pair Ethernet, the interfaces in the electronics have also become smaller. That means I can integrate these systems extremely well, compress them, and make them very compact. From the communication point of view, the data is just normal network communication.

And now in very concrete terms about implementation: let’s say someone says, this sounds cool, we need information directly from machines. What’s the best way to go about it? What mistakes should be avoided?

Karsten

First, you should have a plan: which data do I actually need? For customers, selecting the data points they want to tap into is usually the key point. Typically, there aren’t that many data points, but first I need to identify which ones matter. Once that’s done, connecting them is relatively simple.
Projects are more likely to fail when someone says: I don’t yet know what I want to do, I just want everything first, and later I’ll figure out what information I can gain from it. That’s a story that was sold for a long time: just give us the data, and then AI will solve all problems. That’s not how it works.
Even AI can work better when the data input — or rather the information input — is better.

Semantics are relevant here — you need to understand what the data actually means.

Karsten

Yes, I sat opposite a data analyst for five years. I got a very clear view of where he spent his working time. One part of it was data normalization. That means: a measuring range on a sensor had been changed. Because that wasn’t supplied in parallel, the same numbers coming from the sensor suddenly had a different meaning. And then you constantly have to identify those cases: why did this happen? Did the sensor break? Was this an intentional change to the measuring range? That was investigative work — figuring out those things. A large part of his work was normalizing data. With our solution, that disappears completely.

Can you give a few examples of where else you use this? We’ve talked about KPIs and a few facts about the size of the sensors. Are there any concrete figures you can share?

Karsten

As for size, our electronics fit into the form factor of a typical M12 connector — in other words, they’re no larger than an M12 plug. That makes them very easy to use right at the sensor. I don’t need a new control cabinet to retrofit the machine. And that’s exactly the point: we’re working in the retrofit space, and the control cabinet is already full. So where do I put my new electronics now? Do I need a new control cabinet? No. In our case, you can place it right on the machine, right at the sensor, because the electronics are so small.
That reduces installation costs enormously, because for example I can wire ad hoc and don’t have to plan everything in advance. I can go to the sensor points I want and run my cables there. That reduces the integration effort. Commissioning and installation costs in particular drop dramatically.
What that means for each individual application, and how valuable that leverage point is, depends on the application. But with our technology, retrofitting becomes drastically easier and drastically cheaper.

I have to admit, I don’t know how big an M12 connector is. Can you describe it more concretely?

Karsten

It’s used a lot in industrial environments. The “12” stands for millimeters, so 12 millimeters in diameter. So it’s basically finger-thick. That means I can route it through the cable duct, and sensors are typically in that size range as well.

I think it’s really good that we’ve talked through this whole chain and how data can be used directly. At one point, when talking about the small sensor, you mentioned there’s something like a founding myth behind it. Can you tell that briefly?

Karsten

In many projects, we had problems implementing things economically because access to the data was so difficult. On the other hand, we had always drawn a picture on the board: what if every element were IT-capable?
That idea was planted for me in 1998, when the IPv6 standard was introduced. The professor is standing at the front and says: we now have so many IP addresses that we can give every atom in the universe several million addresses. That gave us a uniform addressing scheme across all layers. That always stayed in the back of my mind.
Then on the one hand there was that knowledge, and on the other hand the problem that we couldn’t get to the data. And then came the turning point: now we can bring it together with the right technology so we can use it, give every sensor and actuator an IP address, bring it into the world of IT, and thereby make accessibility much easier.

You’re not doing this alone, but within an ecosystem. Who do you work with, and how does that collaboration work?

Karsten

You could say that we provide the core technology: the basic computer with the operating system. A sensor is a server, and that’s what we provide. Then we work with partners: system integrators, solution providers, and OEMs who build their own adaptations on top of it.
One example: a system integrator in Spain built a complete OEE solution on top of it in order to equip legacy systems generically with these indicators and extract data from the equipment. The system integrator brings in the specific application know-how. We bring the technical know-how: how do you make something network-capable? What about security? The specific application knowledge — meaning applying the technology to a concrete problem — comes from solution providers, system integrators, or OEMs who integrate our periCORE into their products.
Right here at Hannover Messe, for example, there are display component manufacturers who are integrating periCORE into their products.

What’s your plan for the rest of Hannover Messe? Do you still have meetings coming up, and what are your next steps?

Karsten

Very clearly: this is an industry meeting point. My calendar is full, there are quite a lot of meetings here. I said earlier that many companies here are showing the topic of IT or IoT and how they solve it. What sets us apart is that we radically take the IT perspective, based on the belief that IT has taken over many areas of life. And this will be the next area. That’s a different standpoint compared to other companies. Our goal is to make that understandable.

Georg, same question to you: what does the rest of your time at Hannover Messe look like?

Georg

We’re trying to schedule meetings. We also have meetings with Brazil as the partner country, and we’re building sales partners worldwide — in North America, South America, and the Far East. Accordingly, we’re meeting with companies and evaluating them as resellers and distributors during Hannover Messe. That’s what my calendar looks like.
We’re disruptive in the sense that we make the entire data treasure fully transparent to our customers or end customers, without subscription tiers in between or pay-as-you-go or recurring revenue models, but instead by making it continuously transparent for our customers. In that respect, we are challenging existing structures, because value creation in the OT space is in some cases locked away behind very high levels of specialized expertise. In an environment where Germany is also facing a growing shortage of skilled workers, we are the enabler and the equalizer that makes the data directly available to customers in IT. In my opinion, that’s the future.

That’s certainly not always easy when there are established structures in place. Another exciting point is Brazil as partner country. I had an experience on the train yesterday — it was completely overcrowded because there’s a strike going on here. I was standing next to Brazilians, and I once spent time in Brazil. So I was able to dust off my slightly rusty Portuguese. Those are exactly the kinds of encounters that also make a trade fair special. So I wish you both continued success at the fair and with your conversations. Thank you very much for this interesting discussion about how to get data directly into IT — from sensors, from machines. I’m looking forward to more. Until next time.