Programmable logic controllers: the deterministic layer the rest of the floor depends on

Determinism is the whole point

The value is the scan: inputs sampled, logic solved, outputs written, in a bounded and repeatable interval. An office operating system will happily pause for something more interesting, which is harmless for a spreadsheet and unacceptable for a press interlock or a filling valve. The hardware follows from the same logic — conformal coating, wide temperature range, screw terminals rather than plugs, and a power supply that rides through a dip. Engineers coming from software backgrounds often find the programming environment primitive and want to replace it; the primitive parts are mostly deliberate constraints that keep behaviour predictable and the fault visible to an electrician with a meter.

Whoever holds the program controls the machine

Many machines arrive with the logic locked, or with a compiled download and no source. That is a commercial position, not a technical necessity, and it determines who can modify a sequence, add an interlock or diagnose an intermittent fault at midnight. Settle it at purchase order stage: what is supplied, in what form, under what licence, and whether the plant may change it without voiding support. Where source cannot be obtained, price the consequence honestly, because you are buying a dependency on one supplier for the life of the asset. Getting this concession afterwards, when the machine is installed and paid for, almost never works.

Backups nobody has ever tried to restore

Ask a plant for its controller backups and you will usually be shown a folder. Ask when a restore was last tested and the room goes quiet. A usable backup set is more than the program file: it includes firmware version, hardware configuration, drive and network device parameters, human-machine interface projects, and any recipes or set-points held in retentive memory. It must be stored somewhere that survives the loss of the engineer's laptop, and it must be re-taken after every change. The test that matters is loading it into a spare processor and watching the machine run, ideally during a planned shutdown rather than during a breakdown.

Spares, firmware and parts you cannot buy quickly

Processors, communication cards and specialist input modules are the items with long procurement times exactly when you need them fastest. Decide holdings by consequence, not by unit price: a cheap analogue card on the constraint may deserve a shelf spare while an expensive one on a redundant utility does not. Firmware brings its own trap. A replacement module often ships with newer firmware than the original, which can refuse the stored configuration or subtly change behaviour, so record versions and test the swap rather than assuming compatibility. Machine builders sometimes use a controller family the plant standardises against, which multiplies the spares burden across the site.

Change control and the access question

Controllers were designed for a world where physical presence was the security model. Once they sit on a network reachable from the office, or from a supplier's remote support connection, that assumption no longer holds, and the international framework covering industrial control security is IEC 62443. Practical measures are unglamorous: keep the programming port key switch in run, log who connected and why, segment the control network from general business traffic, and treat a support connection as something opened deliberately and closed afterwards. Alongside that, insist on a written change record. Undocumented shop-floor edits are the reason two supposedly identical machines behave differently.

Frequently asked questions

Should we insist on the source program when buying a machine?

Ask for it, and be prepared to negotiate. Builders protect logic because it embodies their process knowledge, so a common compromise gives the customer readable source with defined areas the builder still owns, plus an escrow arrangement if the builder ceases trading. What you should not accept without pricing it is a locked machine you cannot modify or diagnose, because every future interlock change, line integration and fault investigation then runs through one supplier's schedule and rates.

Can an industrial PC replace a controller for machine sequencing?

It can, and soft control on industrial hardware is used in real production, but the trade is real. You gain computing power, easier integration with data systems and a richer programming environment. You take on operating system patching, storage that wears out, boot times, and a fault-finding model your electricians may not share. For safety-related functions the requirements are stricter still. Mixed approaches are common: sequencing on a controller, heavy computation such as image processing or analytics alongside it.

How should changes to controller logic be managed on a live line?

Treat a logic edit like a physical modification. Record what changed and why, have a second competent person review anything touching interlocks or motion, take a backup before and after, and verify behaviour against a written test before releasing the machine. Online editing while production runs is powerful and occasionally necessary, but it is also how a plant ends up with running logic that matches no stored file. If you edit online, save immediately and update the controlled copy the same shift.