Choosing a processor board

[Epy]

At my work, we currently control large piping skids with PLCs. The PLCs are total overkill in reality, with what we need them to do, we could get away with much much less. We want to move towards a much smaller, cost-effective solution, and so I'm looking at embedded boards (SBCs). There are waaaaaaaay too many to choose from, and I've got no idea how I'm going to pick one over the other. I know a few of you here are in the embedded programming world, I was wondering if one of you could comment.

I don't even really have much information to offer up. I only need probably 128 MB of RAM (if that), something in the 100s of MHz, an ethernet port, and a USB port. I already have a USB I/O board picked out which should work anywhere Debian Linux is supported.

I just don't even have the first clue. I could pick something now and it'd work I'm sure, but I'm not sure it'd be the best choice out of the sea of choices out there.

Edit: this may seem pointless, i.e. how can someone give me an answer without more information, so I guess at this point it'd be helpful if someone who uses embedded boards could come in and say "our company uses X because Y"

[Salem]

My first thought is how you're going to manage the safety critical aspects of this.

Do the PLCs have clearly defined failure paths to always stop the skid safely?
Do these have to be safety certified? Any kind of certification doesn't come cheap at all.

Making s/w to handle the normal operational case is probably not that hard.
Making the whole system tolerant of all sorts of hardware and software failures (whether mechanical, accidental or malicious) is a much harder proposition.

[cyberfish]

Like Salem said, if safety is a concern, it's almost impossible to use a SBC in place of a PLC.

There is also the hard real time issue. Linux is not a hard real time OS, so there's always a chance that something will be delayed much longer than it can be, in certain situations, so it's impossible to prove that the system is correct.

Depending on how much computation power you need, it MAY make more sense to go with a barebone microcontroller without an OS, like STM32, LPC, or something like that. That requires much more embedded development experience, though.

If safety is NOT a concern (eg. there is already another controller in charge of safety), Iv'e had good luck with Gumstix, but they all pretty much do the same thing. You'll just have to pick the cheapest one that has all the IO ports you need (and if you only need ethernet and USB, that's just about all of them).

[Epy]

Luckily, these piping skids are designed so that if everything fails, the valves fail to the open position and liquid just flows through with no issues. Safety isn't a huge issue for our skids, mainly because there's no real threats, at least as far as the controller is concerned. Approvals are going to be a concern yes, I guess whatever I find has to be UL listed at least for customers to be happy. Such a pain in the @#$%#^.

[megafiddle]

There are some small inexpensive PLC's available.

Are the existing PLC's just much too big for the application?
Or are they already of the smaller type, and still overkill?

If they are just too large, and using ladder logic, the logic should be easy to reprogram
into a smaller PLC.

-

[Epy]

Everything is too big. I'm using a MicroLogix which is fairly small, but the idea is that everything in the box could be put on a couple of small boards.