What’s missing from Profinet

I spent a couple of days last week at a Profinet seminar in Troy, Michigan last week. Lots and lots to say about it. Probably some folks that won’t like what I have to say.

The first day of the seminar is mostly for end users. It is incredibly well done. Carl Henning is the best in the business that I have seen at running these things. You get really great information. It’s entertaining. The slides are first rate. The technical expert (Hunter Harrington) is wonderful. If only I could be that young, good looking and competent again (again? I’d take once.). All in all a really enjoyable way to spend a day if you like spending 7 or 8 hours with a 100 or so other guys.

There’s only two things wrong with what they do. One, they bash EtherNet/IP quite a bit. Not necessary in my opinion. EtherNet/IP has its place. It’s adequate. It works. In fact, it’s small (RTA’s solution is only 20K), exceptionally easy and cheap to integrate and quick to implement (2 days for our solution). A great developer solution.

Neither Profinet IO or EtherNet/IP are perfect. People use the technologies they have to use because of the architectures they already have or the market their in. Nobody evaluates and picks. Second, Carl and Hunter don’t really talk about the downside of Profinet. And there is a downside to it. And I’ll get to that in a second.

As I understand it, the seminar was done because of some heavy arm twisting by Chrysler. They pushed PTO to schedule it and they told their heavy duty suppliers to be there. Chrysler has made a huge commitment to Profinet. Not sure what groups but lots of different suppliers where there.

The mood was pretty grim if you talked to anyone in depth. Travel budgets have been slashed. Layoffs among everyone. One guy told me how it was going to be touch and go if they survive. Knowing their product pretty intimately I think they are going to go belly up. They have a single product that relies heavily on product development and shipping of lots of devices. Very hard for them to make a go of it if the recession lasts 18 to 24 months.

There were a number of Profinet solutions present at the development seminar the next day. Truthfully, I don’t like any of them, even mine! Here’s the round up if you are in the market to develop a Profinet device.

RTA – Our company. We are a reseller of the Softing Source Code stack (see Softing below) and a provider of a little module that does EtherNet/IP, Modbus TCP and Profinet IO. You add it to your board, talk to it over serial (Modbus RTU) and you’re in business without bothering with any of the details. But it’s a gateway with all the inherent limitations of a gateway; response time, customization, bom cost…etc. It’s not really inexpensive either.

SOFTING – Softing has two solutions. Their source code solution and an Alterra FPGA. The source code solution is difficult to integrate, excessively large and costly. When I tell prospects what the cost is going to be they faint. The hardware solution is OK but again it’s not as customizable as I would like. My guys don’t want hardware.

HILSCHERER – If I was to give out an Engineering excellence award it would go to their NetX solution. What an incredible feat of engineering. There are like 5 processors in there and they can all run different protocols and communicate amongst themselves. Wow! But it’s way over engineered and pricey. My biggest complaint is that I can’t customize it. I’d like to add value by tailoring the application side of Profinet but it’s not possible.

SIEMENS – Even larger code than Softing. Same issues with code size, difficulty of integration and price.

ANYBUS – If you have a small number of units to do this is a perfectly acceptable solution. The user interface is a little odd but it supports lots and lots of different protocols so I’ll cut them some slack there. I wouldn’t call this a good solution if you were going to ship 10,000 valve blocks where footprint, cost and time to market matter.

In fact, there is no good footprint, cost and time to market solution for the 10,000 valve block application. If I was building that device I would want to use a real cheap Ethernet enabled processor like a Freescale Coldfire or some Arm processor.

I certainly like the incredible amount of functionality that Profinet brings to the table but as a developer I run away screaming into the night when faced with the complexity.

If I was to do an app like a valve block I’d want to do it with the Flash and RAM I have on board. I wouldn’t want to add additional components. Those extra parts cost me a lot; bom cost, bigger pcb, more assembly labor. And I certainly wouldn’t want to add megabytes and megabytes of Flash to do a source code solution. That’s the one place where Modbus TCP and EtherNet/IP are heads and shoulders better than Profinet.

You see, there just isn’t a good developer solution for Profinet! Devices with Profinet are complex and incredibly expensive. And that’s what’s missing from the Profinet presentation that Carl and Hunter did last week.