Just how PC is PC-CONTROL?

Writing the draft of this article at home on an ageing 486 IBM-clone (complete with severely fragmented hard disc), before transferring it to the wonderful world of the PowerMac for page processing, prompted the question: just how long is PC-based control expected to run process plants before rapidly falling behind both hardware and software developments?

Equipment obsolescence over the ten or more years' lifespan of a typical plant is, to coin a phrase, nothing new. Planned maintenance and scheduled shutdowns can change the pumps and valves, mixers and filters as often as needed, which is generally not often. And certainly nowhere near as often as control engineers will be enticed to upgrade their systems to keep pace with Microsoft's continually evolving systems.

Take the interest currently being shown in Windows CE. According to the independent market research company Intex Management Services, as the market for PC-based control software grows rapidly in Europe over the next few years, Windows NT and CE `will become established as the dominant operating systems'.

More about NT later, but IMS acknowledges that the relatively new CE operating system has not yet been widely adopted in the industrial automation market. While PC-based control software designed for other operating systems has been successfully ported to CE, the first dedicated CE-based packages are only now coming on to the market. IMS forecasts that the real CE breakthrough will not happen until the end of this year with the release of Windows CE Version 3, which Microsoft has said will have improved real-time capabilities.

This view was echoed last month in Orlando, Florida, at the annual automation strategies forum hosted by industry analyst Automation Research Corporation. According to ARC, a panel discussion revealed that many CE-based control products will not be available until CE V3.0 is released and tested. Nevertheless, speakers from Siemens and Rockwell presented plans for their respective companies to deploy CE and develop CE-based products.

Although IMS takes the view that Windows CE will primarily be used in small embedded control applications, it still expects Windows NT to dominate the larger `feature rich, workstation-based applications' where users want to combine control and front-end HMIs or Scada systems on the same platform.

NT may have become, as IMS maintains, the de facto standard for manufacturing applications such as Scada and MES (manufacturing execution systems), but there is still concern over its suitablity for process control. This concern centres, as with the CE products above, on its real-time or deterministic capabilities. According to Jeremy Shinton, Mitsubishi's business development manager for PC-based control products, Microsoft itself, in a 1995 paper `Real-time systems with Windows NT', defined determinism as `the ability to, without fail, provide a response to some kind of event within a specified time period. This response must be predictable and independent of other tasks.'

In Shinton's view `off-the-shelf' Windows NT cannot do this with any certainty. `It is incapable of real-time control,' he says, `as external I/O transfer routines and driver routines can interrupt the execution of the thread running the control logic. This creates an unpredictable system.'

Mitsubishi's answer is to run a robust `hard' real-time operating system on the PC, with Windows NT operating as a service to this real-time system so that all control and real-time functions have the first opportunity to complete before an NT application can become active. In the company's MelSoft Mx PC-based control system, this real-time system is Radisys InTime - the PC version of the iRMX operating system that has had over 20 years proven use in advanced process control, defence and avionics applications - and deterministic scan times are possible down to 1ms.

Although InTime is completely transparent to the user, Shinton says it provides the level of integrity required to control a machine or process safely by ensuring that control is given the highest priority. With Windows NT running as a protected task within the real-time operating system, the control functions will still operate even if Windows crashes or the hard disc fails. If unaltered NT is used as the operating system for control, then an NT crash - `the blue screen of death' says Shinton - will halt the control system and make a safe and orderly shutdown impossible.

Not all PC-based control systems are the same, of course, and the IMS report referred to earlier dismisses claims that Windows NT has to be combined with a real-time extension before it is suitable for industrial control. However, IMS's definition of `industrial' seems to suggest that `the vast majority of control applications can be considered to be soft real-time, where the non-deterministic performance of NT is not an issue.'

But for those applications, mainly process, where it most definitely is an issue, Shinton suggests five fundamental rules that should be obeyed by a PC-based control system. The system must:

* Provide determinstic real-time performance

* Survive a Windows NT crash

* Be protected from poorly behaved

Windows applications

* Survive a hard disc failure

* Use the same robust, reliable real-time kernel as a PLC.

In the panels on pages 28 and 30 are a selection of other PC-based systems currently on the market, but with nearly 50 PLC manufacturers around the world, together with untold numbers of PC software houses, it is only a snapshot of what is on offer.