Hold that thought.... I am interested enough to do the digging, but I am shorter on time lately than I have ever been. I have a paying project that I can't even get to.

Regards
cww

 

I agree that it's an implementation issue not a technology one - that's been the case for years. I think the problem is that nobody (in a wide sense) seems to trust the idea because of failed crappy implementations in the past - and all the half baked open source projects that you wouldn't want to run your plant.

I, for one, would like to see such a technology. IMO it would take a monster push, like a Google project to make that happen overnight.

That "nobody" comment was about a piece of software that would make a PC into a "better" general purpose PLC. Highly possible, but a monster task. Do you know of any software that does this?

----
Nathan Boeger
http://www.inductiveautomation.com
"Total SCADA Freedom"

 

In reply to Nathan Boeger:

NB: "PCs do crash routinely compared to PLCs"

MG: If you use poor hardware and software you will have a problem. If you use
reliable hardware and software you won't. A "PC" doesn't necessarily mean the same sort of computer you use to play video games with.

NB: "They are not designed to handle unanticipated shutdowns,"

MG: A journaling file system will automatically recover. A read-only file
system will not have a problem to begin with.

NB: "They are not designed to handle ... mechanical vibration, ... more moving parts ... fan in your power supply ... will eventually seize"

MG: Use a fanless, diskless, PC. There are no moving parts. Some PLCs by the way, have fans which seize and cause the PLC to overheat.

NB: "PCs are also most often connected to your industrial I/O via some sort of network that can go down."

MG: They use networks that were designed for PLCs. If this is a problem for PCs, then this is a problem for PLCs as well. Many PLCs these days use networked I/O. The local rack is disappearing for all but the smallest installations.

NB: "Mission critical corporate asset or not, Windows boxes lose their reliability over the years."

MG: Then don't use Windows, or is that too obvious an answer?

The WinPLC by the way seems to be taking a lot of flack because people see "Win" and immediately think "Windows - shoddy, unreliable, crash-prone, etc.). The company behind it might want to think about a re-branding effort where they drop all association with "Windows" and just call it a "PLC with flowcharts".

 

What he said!  They are converging and pretty soon you will need a score card to tell the difference. Check out the $200 Linux computer at
WalMart. 20 watts working, 2 watts snoozing. Throw away the drives and use Flash and you  have a  pretty good candidate. And made in the hundreds of thousands, the reliability should be _better_ , better than the Micrologix I've
been using in any case.

Regards
cww

 

Hi Nathan

Nathan Boeger wrote: "I agree that it's an implementation issue not a technology one - that's been the case for years. I think the problem is that nobody (in a wide sense) seems to trust the idea because of failed crappy implementations in the past - and all the half baked open source projects that you wouldn't want to run your plant."

Actually, most automation projects are half baked, the FDTTW thing (First Damn Thing That Works) and you usually get every part of the solution that can be modeled with relays. :^)

Systems that last a few years often gain refinement, but if it works, people tend to leave it alone. But, you should think through what you are saying.

A few k of machine code on a minimal computer tied to some fairly bush league applications on a PC does not constitute a gargantuan task.  Many single hobbyists working in isolation have done more on a lark. And the parts are all there. The amount of free code available is staggering and you have the parsers and compilers (and parser makers and compiler compilers) and screen builders and all kinds of examples of pretty much every part.

On reflection, with the rare perspective of a few years around such a project, there are a few issues.

Everybody has a different idea of what it should look and work like. The desire to produce a be all, end all,  product. More is better.

This has lead me to a conclusion that the commercial offerings succeed most often because of _lack_of resources. They have a specification, a time and a budget. And to the extent possible, they stick to them.

Also, we have to separate the PLC from the rest.  Automation people are almost always talking about the rest. When you are an AB fan or a Siemens fan it has almost nothing to do with the PLC. In fact, quite often in the past and perhaps even now, they would argue about the merits of vendors who were running on the _same_ PLC.  Example: Koyo. 

All PLCs do the same things, in much the same way, and the rest of a product could probably be configured to run on a different PLC if they wanted to do that.

SoftPLC has succeeded to some extent , IMHO, by picking A single target and emulating that faithfully enough to demonstrate that it was nothing special and could easily be done by more powerful computers.  They have progressed beyond that, but that success produced the foundation, the basis, and believers.

A PLC is so much less complex than everyday objects, that a greeting card is probably running more firmware.

The software and hardware built around the PLC need to be bounded by a PLC design and not a moving target.  <br> The interface between them has to be finite and well planned.

The tools are where free and open can really shine. These are, after all pretty, much standard PC applications and there are a lot of people who are pretty good at that.

These are not technical problems and are not insurmountable.  But to do so will require a different structure than what has been tried.  And commitment to that structure. Perhaps  two groups.  A small disciplined group for the core PLC and the rest of the community. Or perhaps a benevolent dictator. Or perhaps a sponsor with 49% control. I am absolutely certain it can be done with the right organization.  Just my thoughts on the subject.

Regards
cww

 

We have had Windows 95 PC's in operation up until this year when the final one was replaced with an XP Pro PC. Simply because they were reliable and the customers had no desire to change something that worked. We never experienced a problem with monthly crashes as suggested.

My use of "crash" in my earlier post was in reference to the software, not the hardware. If you are going to use exposed electromechanical devices such as fan motors in harsh environments, then protect them appropriately (climate controlled control room) or use solid state devices. Whether PC or PLC, why are you filling it with dirt? Either will become less reliable. Come to think of it, how are you treating employees? They will become less reliable too in such environments.

The idea that it "becomes less reliable with age" applies to most manufactured and living things on this planet. However, unless you are dealing with a virus, software is exactly the same set of bits you installed five years ago.

I don't understand the suggestion that an Ethernet network would be more or less reliable based on which devices (PC's or PLC's) are connected to it. Perhaps you could elaborate. Its a design given that you don't connect your corporate business network to your control network.

Regards,
Victor

 

In addition to Michael Griffin comments: Another application where PC are used 24x7 is model predictive control. It is a computationally intensive task, especially if it is nonlinear MPC with first principle based process models, and that is why it usually resides on a PC. An exception to this is parametric MPC in which the solution is precomputed off-line.

Regards,
Andrey Romanenko // CIengis

 

In reply to Victor Robert: I was referring to the infamous 49.7 day bug. Microsoft Windows used a 32 bit counter to track the number of milli-seconds since start up. After 49.7 days (4294967296 / (1000 * 60 * 60 * 24) = 49.7) the counter would wrap around. This in turn would cause a crash.

The following is a news reference from time when this was discovered (approximately 4 years after Windows 95 was released). Windows 98 was also affected. If you do a search for other reports on this issue, you will find that few people considered this to be a major issue, as they expected Windows to crash much more frequently than that anyway.

http://www.news.com/2100-1040-222391.html

 

I'm trying to avoid anecdotal stories of PCs lasting for several years - sure they can, but statistically, many will fail. Consider server rooms that are clean, air conditioned environments. You're constantly replacing power supplies and hard drives - that's why you put in redundant setups. You nailed that one on the head - electromechanical devices.

As far as software, you're living in the same electromechanical world. The bits are NOT exactly the same after several years of operation. Think hard drive fragmentation, defrags, Windows opening and retaining a file lock on the same .dlls, etc. Can you run a controls environment on the same Windows PC for years? Sure. Is it the same? No. Should you expect to have to rebuild the machine in a 5 year lifecycle - I hope you plan on it. This is aggravated by a few points. One, a hard power cycle or brownout can cause serious problems. Keep your UPS batteries maintained because they die over time (chemistry). Two, maintaining security patches and updates has a history of breaking industrial software. This is a much smaller consideration for PLCs than PCs.

Why the network makes PCs more susceptible to problems than the PLC. This is due to the fact that your important things are plugged directly into I/O modules in the PLC. No network required (besides the wire). This is done by design because the PLC is more reliable for 24/7 operation than the PC. This allows production in a typical setup to continue without the Ethernet network - minimally, part of the process could work. That is changing as everything becomes IP based. For the PC, presumably nothing useful will work if the Ethernet network is down. Ie, you could have a modem plugged in to send a page/text message, but all the PC knows is that its network is down.

Make sense?

----
Nathan Boeger
Inductive Automation
"Total SCADA Freedom"

 

Some thoughts on selecting PC vs PLC...

Many if not most applications today use a display with touch screen or keyboard/mouse for the user interface. The most cost effective solution for this hardware is the PC. Although most of us probably shudder at the idea of what we might get for $200 at WalMart, the cost of name brand equipment is not much more. The alternative is to use hardened devices from the PLC vendors.

The biggest advantages to the PC are the cost and the fact that I am confident that in ten years I will be able to find a replacement at almost any electronics retailer on Main Street and be able to load my CD and be up and running in a few hours. I would fully expect the manufacturer of a low volume, specialized device to tell me it is out of production.

Comparing the average PLC to a PC with 3GHz Pentium, 2GB RAM, 300 GB disk, database engine, network connectivity, etc. is like comparing apples to watermelons. They are two very different products.

If you already have a PC in the application, do you need to add a PLC? It depends entirely on the application. Networked I/O modules can easily be added to the PC with common Cat5/6 Ethernet cables. If your application is farming, and it really doesn't matter if the animals get fed 5 seconds later on Mondays and Fridays, then you probably don't need a PLC. If you are manufacturing many parts per second, then you do need a PLC.

The down side to adding the PLC is that you add a level of complexity to your application. Now you need additional skills to program and to maintain the system, adding initial cost as well as ongoing cost of ownership. Programming skills for a specific PLC, particularly an older one are far more difficult to find than general VB/C skills.

The up side of the PLC is that you are free to improve your PC hardware and software (user interface, faster processing, reporting, charting, data storage and analysis) knowing you will not adversely affect the underlying real-time process.

There is no "best" solution that will cover all automation applications.

 

Just as a note:

You can buy the mainboard used in the $200 WalMart PC for $60. And I don't think you are accepting any more risk than the $100 PLC you get from AB. The AB boards are Korean and the PC2500 is by VIA group partner Idot in Taiwan.

VIA makes millions of boards a year.  I'm fairly sure they know how.  The  only hangup with the PC2500 is that it uses a fan on the processor, but a little package engineering to transfer the heat to the skin would take care of that.

The boards are sold out at least till after the holidays or I'd have one. At $60 it's an incredible bargain if you've priced SBCs.  Intel also has a board out in the same ballpark for price and power.

No, these aren't the supercomputer class needed to run Vista, but they have vastly more compute resource than most PLCs.

Regards
cww

 

Sorry, Curt. I didn't mean to put down the product; I don't know anything about it. Its just that the terms "$200 PC", "WalMart" and "Industrial Automation" in the same sentence will take some getting used to, just as "Made in Japan" has had vastly different connotations over the past 50 years.

Nathan, your point "but statistically, many will fail" might be more accurately stated as "statistically, everything will fail". Light bulbs and automobiles fail, but we seem to keep on buying and using them regardless. If its a mission critical application, we always install two or more redundant PC's. Its slightly more involved than changing a light bulb, but takes about the same number of engineers . They can and will fail, but that is just part of doing business. We will replace PC's every 3 to 5 years of 24/365 use whether they need it or not, usually in conjunction with a software upgrade. Using name brand products, our experienced failure rate within that period has been extremely low. An advantage is that we have a "new" system every few years and can re-start the obsolescence clock (excluding I/O hardware).

Regards, Victor

 

...More thoughts on product selection.

We choose what we are familiar with. Any of us could get into any car on a rental lot, figure out the seat and mirror adjustments, and be driving away in 60 seconds. Now put a full size John Deere tractor, an 18 wheeler and a bicycle on the lot and apply the same 60 second requirement. It is clear which of the vehicles the farmer, the trucker and the cyclist will choose. Did any of them choose the right or wrong product? No, given the objective, but their selections were based entirely on personal comfort level, not technology evaluation.

If you ask granpa, who has owned and driven only full-size Fords for the past 60 years, what kind of car you should buy, are you limiting your possible outcomes?

If the objective is to get the product programmed and running by the end of the week, is there any doubt as to which method the C programmer, the VB programmer or the ladder logic programmer will choose?

If you are asked to recommend the best widget for a job, will your selection be one of the 5 products you have worked with, or could it possibly be one of the 10 you have never seen before?

Another consideration is that the objective of any project is to get it done on time, within budget and make everyone involved look good. If you tell the J programmer that he must program in the K language because it is the "best" choice, will he buy in to the project, or will there be some subliminal desire to prove that this was the wrong decision? The team has a better chance of getting to their goal if they are all facing in the same direction.

No easy solutions, but that's we why get the big bucks, right?

VR

 

I disagree with the point behind your version of my statement. I was pointing out that the MTBF on an industrial PLC is SIGNIFICANTLY greater than your off the shelf PC due largely to electro-mechanical parts and design specifications. That's important - you can't adequately summarize it with "everything breaks".

As to redundant PCs - we're discussing a specific implementation, not vague theoretical applications. Specifically, using cheap Automation Direct WinPLCs to control one of the most complex fish hatcheries in North America.

To set the record straight - on the theoretical approach, I think PCs could do much better for controls than PLCs - PERIOD. However, the software/implementation needs a lot of work. We're certainly not there yet with random non-programmers quick C++/VBA jobs. A heck of a lot of complexity comes with redundancy and clustering.

----
Nathan Boeger
Inductive Automation
"Total SCADA Freedom"

 

In reply to Curt Wuollet: VIA hardware is used in quite a few name brand "industrial" rated systems because they have versions with very low power consumption, fanless operation, and small size. They make boards themselves and other companies also base their designs on VIA chips (including VIA x86 compatible CPUs). The board you are referring to has a fan because it uses a faster CPU (fanless boards use lower clock frequencies to cut heat output).

Quite a few people reading this probably have VIA boards or VIA based boards in their factories with the name of some industrial automation company on the case (e.g. in many panel mount PCs). VIA is probably better known for their chipsets used in many mass market PCs, but they are also the leader in the low power x86 market. Intel however is also coming out with a new lower power x86 CPU and chipset design targeted at the same market (embedded PCs, which is much bigger than just industry).

A typical low power fanless x86 PC today runs at about 800MHz, and is probably equivalent in processing power to a higher power CPU running at about 600MHz. That isn't very fast, but it's more than adequate for a lot of embedded tasks. The motherboard you referred to has a fan and runs at about twice this speed. Typical power consumption for an entire low power motherboard is about 14 watts, as compared to close to ten times that for a high power Pentium CPU alone (which is why the Pentium CPUs need fans).

The PC you mention with gOS (Linux) is $199. The equivalent PC is available from the same source with MS Windows Vista Home Basic for $278 (40% more). That PC isn't offered with MS Windows Vista Business, so it's hard to get a clear idea of what that would cost. However, Dell charges $90 to upgrade a new order from MS Windows XP Home to Professional so it's probably reasonable to assume a similar price differential for Vista. So a price of $368 (85% more) for a PC with MS Windows Vista Business would be a better comparison. If you look at it that way, then the $200 price doesn't look so anomalous. This by the way wasn't supposed to be an accurate price comparison of Linux versus Windows. I am just trying to point out that the $200 price tag doesn't mean they used hardware that was any worse than any other low end PC.

 

I feel that both the PC and the PLC scada will have their palces in the Induatrail automation. However our experience is that in PC based systems Hardware and sofwtare maintanace / upgrades is still a issue. In every place PC is used it has been replaced after a maximum of five years as either the hardware or the software OS is not maintanable. This compares not so well with PLC hardware.

However we are getting machine control applications in the recent past where PC is a necdcessity because of visualisation / database requirements and forms part of the control system. Here the problem is that although we use the PLC for critical areas in the machine the total system reliability is the reliability of the PC.

I am yet to get a solution for this problem amd would be glad if somebody can suggest.

Real time PLC development software is still expensive on the PC. Not to speak of the motion control on PC.

 

Well put - Good points!

----
Nathan Boeger
"Design Simplicity Cures Engineered Complexity"
Inductive Automation

 

Actually, I think this is part of a trend, If they can build a green PC for less, the only limitation is bloatware. It's just better and easier all around if your PC doesn't dim the lights and double as a space heater. I think this is another case where MS is doing us a disservice with their incredibly inefficient software. If it won't run Vista, it won't sell, even if it's more computer than it takes to run anything within reason. The glimmer of hope is that with Linux on board, Walmart can't seem to keep them in stock.

Regards
cww

 

The really interesting part of the mini mb push is
that they hear this and are specifically setting up a longer life-cycle with a 5 year minimum. Much
better than a new generation every 6 weeks. I'm glad someone wants to stop the madness.

Regards

cww

 

I have read that Microsoft plans to release a "lite" version of XP next year. (What is taking them so long?) It is supposed to be used on "low end" PC's and have a relatively lower cost. They are presumably trying to compete in this low cost, energy wise market. This is a good thing for dedicated industrial systems that must run Windows.