Hans Beckhoff, owner and CEO of Beckhoff Automation in Verl, Germany, just gave us the most coherent explanation of his company’s PC-Based control technology that I’ve ever heard.
The key thought is totally different from the myriad of PC-based control suppliers that sprang up in the late 1990s in the US. It was the value proposition of the latter that used to drive me crazy (metaphorically speaking, I hope).
I have to say that every time the marketing and PR people from Beckhoff USA would pitch PC-based control, I would give them a rough time. All of the contemporary software companies are gone for all intents and purposes. Meanwhile Beckhoff grows stronger.
Listening to Beckhoff is much like listening to Dr. Truchard, CEO of National Instruments–and the two have much in common although they’ve never met. They should, by the way, but only if I could be there to listen to a couple of very smart entrepreneurs discuss technology and the market.
Beckhoff’s view is that by putting control in the software in a central PC you gain tremendous benefits of leveraging current computer technology, flexibility, ability to incorporate multiple disciplines into one application, and more. The vision of TwinCAT is not that different from LabView–although TwinCAT began in automation and is moving toward measurement, while LabView began in measurement and has grown into automation.
The difference from the other PC-based control people is that they all seemed to push the value proposition of cheap, commercial hardware. It was a drive to the bottom, so to speak. Beckhoff, as well as NI and Opto 22 for that matter, all touted the benefits of leveraging the benefits of software control along with their industrially hardened hardware computer and I/O platforms.
While the other guys argued publicly about whose real-time kernel was better, the companies that are still around and growing explained other benefits. That is the direction that won.
The latest version of TwinCAT (v. 3) leverages Microsoft Visual Studio and the power of C and C++ programming along with Simulink from Matlab. The potential benefits for machine builders when these are coupled with the roadmap for more powerful chips from Intel should leave engineers drooling with delight over possibilities.
This was a day well spent for my education–and hopefully I can write some more to make you think of possibilities of the future of computing for industrial automation. I’ve got to leave now. More later.
[Disclaimer: Beckhoff paid my way to Germany, although I gave up a weekend to be here (sound of small violins, I know). Interesting side note at least to me and some of my readers who live close to me in western Ohio is that two families from Verl left home in the 19th century and wound up founding two cities in northwestern Ohio near where I’m from–Delphos and Ottoville. No wonder most of the people here look just like back home, and the names are also similar or the same.]
Maybe PC-based softPLC's haven't been successful, but general PC's for automation have been quite successful, just maybe not running a real time OS or a soft PLC. And some soft PLC companies are very successful — just like at Codesys.
We use a white box PCs with micro PLCs for I/O and motion controllers and custom boards for the hard real time parts. Using hardened PC's would only add cost, without any benefits, and using something like LabView or TwinCAT would actually make the software development more difficult. I know several of our customers also use a lot of PCs, too.
I'm pretty sure that a substantial amount of Beckhoff K-bus and E-bus I/O systems are used with PC's running regular non-real time OS's – or with regular PLCs. That's partly based on my informal eBay surveys of Beckhoff and Wago 750 stuff for sale; various fieldbus interfaces (e.g. DeviceNet, CANOpen, Profibus, Ethernet, Modbus/TCP and EtherCAT) are much more common than Beckhoff (or Wago) embedded PCs.
For designs with a long life time, going PC based (or phone/tablet based if you want to follow the hype) has some big disadvantages (the big advantage is software — which only applies if you can plug into the Windows or Linux ecosystems, and is less of an advantage with the advent of Android). The TI DSP used in our custom board is still active, with chips in stock, 15 years after its introduction; can you still get original 500MHz Pentiums? (Intel's idea of a long life product is 5 years). And semiconductor technology advances also happen in the embedded world; for example, TI is currently selling 8-core DSPs that are software compatible with the original. So there is a definite advantage is "tagging along" with high volume markets, but the PC market isn't the only one. Other examples include automotive (source of the CAN bus, and cheap safety MCUs), telecom, networking, and mobile (which is replacing the PC market as the center of semiconductor attention).
Beckhoff uses the "Win-tel" platform, but that does not mean the short lifecycles of commercial off the shelf technology–at least of the consumer variety.
First, there is a real-time kernel for running control. Trying to run control applications strictly on Microsoft Windows consumer platforms is not a good thing.
Next, Beckhoff uses the Windows CE embedded platform. This has been around for quite some time and is supported according to the industry standards of embedded computing (which includes defense and telephony applications). When I was at Control Engineering, I wrote several articles explaining CE. I have not checked the new Website, but I'd be surprised if you could find them anymore. But it is a robust embedded platform for control.
On the Intel side, Beckhoff (and others, by the way) uses the Intel embedded platform. Once again, this is a specific technology that leverages from the commercial/consumer side, but follows the industry standards for support in the embedded world.
I think that Beckhoff (and others of its ilk) have succeeded where companies such as ASAP, Steeplechase and Think 'N Do failed is that it has build a solid line of industrialized hardware. It doesn't market as cheap. It markets as powerful, rugged and industrialized. As far as the typical North American user is concerned, it could be a PLC.
Obviously a better path to success.
It sounds as if your platform is PC for certain, probably supervisory, functions with with control distributed among small, special-purpose devices. I consider that to be "modern," but Beckhoff people say that that architecture is OK, but that there are many benefits of a more centralized control architecture. I'm betting that depending upon the application and the skill of the engineer, in most cases it won't matter which you do.