Programmable Logic Controllers: From Relay Panels to Industrial Control

The wires were the program

A relay is a switch pulled in by a coil. Put enough of them with pushbuttons, limit switches, timers, and motor starters, and you can make a machine refuse a motion until the last one has finished. That is real automatic sequencing. Plants had it before anybody said PLC. A transfer line in an auto factory could be a wall of panels, each wire a decision, each decision documented on a drawing that drifted from the panel the first time somebody fixed a fault at two in the morning.

The pain was the change, not the concept. A new model year moved a clamp or added a station. Someone had to lift wires, ring them out, and hope the print got updated. I will not hand you a number of days that every plant lost. The complaint, from the people who ran those lines, was that the rewire was slow, the drawings lied, and the relays themselves were a maintenance load. Solid-state electronics were already in the building. What they wanted was a box that would survive the floor and take a new sequence without a new forest of wire.

1968 was a crowd, not a spark

Morley’s later account is the one the trade likes. On January 1, 1968, he drafted a short memo for a controller. Bedford Associates, the firm he was in, had a prototype up that spring. They nicknamed it Stupid. It was the company’s eighty-fourth project, numbered that way by co-founder George Schwenk, so it became the 084. Early demonstrations were at machine-tool shops, Bryant Grinders and Landis among the names in that account, not a ribbon-cutting at GM. Modicon — the name usually read as modular digital controller — was incorporated in October 1968 to build it. A patent in Morley’s name, U.S. 3,761,893, is one the trade points at. A patent is not the morning of the memo, and it is not proof no one else was in the room.

General Motors was in its own room. Bill Stone, at Hydra-Matic, the transmission division, set out the plant’s trouble with relay reliability and documentation in a 1968 paper, and GM asked vendors for a solid-state controller that could replace hard-wired panels. Morley’s group later said they had not seen that paper while they were building the 084. Believe both clocks. A specification and a prototype can grow toward each other without one man handing the other a finished idea. The automotive plants were the customer who made the box a product. They were not the only hands on it.

Allen-Bradley was building programmable controllers in the same years. Odo Struger’s name sits on a lot of that work, and he has been called the father of the PLC on that side of the trade, just as Morley has been called it on his. Morley, in a later interview, refused “inventor,” accepted a loose “father,” and said it had been fifty people. A competitor started a fight over the title and then dropped it. Walker’s dissertation is where you go when you want the prehistory instead of the title. I will not pick a winner. The machine showed up because the panels had become the problem, and because electronics could finally live on a factory floor.

You still wire the world. You no longer wire every decision

On a relay panel the decision is a contact and a wire. On a PLC the pushbutton and the valve still land on terminals. What changed is the relationship between them. The program says which input, in which state, is allowed to drive which output. Change the sequence and, a lot of the time, you change the program, download it, and test it. You do not pull half the panel.

A lot of the time is not every time. A new photoeye, a larger drive, another rack of inputs, a guard that was not in the original risk review: those are still metal, wire, and a signature. Anyone who tells a young technician that “it’s all software now” is about to get someone hurt or a machine that will not start. The program is the sequence. The field is still the plant.

Ladder logic looks like the drawing the electrician already knew

The early sell was that a man who could read a relay schematic could read the program. Ladder diagram is drawn as rungs. Contacts on the left, a coil on the right, power imagined as flowing across. It worked as a bridge, and it is still the language a huge number of plants maintain in. It is a picture of the logic. It is not a picture of the wires in the cabinet. A contact on the screen may be a real switch, or it may be a bit the program invented and no screwdriver will ever find.

The trade did not stop at ladder. IEC 61131-3, whose current edition is the 2025 issue, names a set of languages for these controllers: ladder, function-block diagram, structured text, and sequential function chart for the structure of a program. Older editions and older controllers also carried an instruction list that many shops still have to read. The right language is the one the next person on shift can debug at two in the morning, inside the rules your plant actually follows. A standard is not a law that every brand obeyed on the day it was published.

The scan is the heartbeat. It is not every controller’s whole life

The picture to start with is a loop. Read the inputs into a table. Execute the program against that table. Write the outputs. Do it again, as fast as this box and this program allow. Between one read and the next, the controller is deciding on a snapshot. A switch that flickers in the gap can be missed or, if you are careless, counted twice. That is why noisy sensors, filters, and a programmer who understands the scan are not decorations.

Real controllers outgrew the one-loop cartoon. Periodic tasks, events, interrupts, motion, a network that delivers a number late: the timing is a design, not a slogan. “Read, decide, act, repeat” will keep a beginner from getting lost. It will not tell you why a drive faulted on a machine whose scan time you never measured. William Bolton’s textbook and E. A. Parr’s guide are how the trade teaches this machine. They are not the history of 1968, and they are not a promise that your brand scans the way the introduction draws it.

Timers and interlocks moved into the box. The hazard did not

A timer, a counter, a step that waits for a permissive: those used to be devices on the panel. A program can hold them now, and often should, so the sequence is in one place a person can read. Physical relays did not leave. A motor starter is still a starter. An emergency stop that depends only on a line of code, in a controller that was never rated for safety, is a mistake with a body attached.

A standard PLC is not a safety controller because it is reliable on Tuesdays. Safety-rated logic is a different design, with its own hardware, its own checks, and a validation somebody signed. Barry Brusso’s short IEEE history is useful as an industry sketch of the fifty years after the first boxes. It is not a license to treat every processor as a guard. An wrong program, or a right program reading a failed sensor, will run the hazard with great consistency. That is the nasty gift of a stored sequence. It repeats.

The robot does not replace this box. They split the job

A robot controller owns the arm: the path, the joints, the grip. A PLC, in a great many cells, owns the room: conveyor, clamp, gate, the word from the last station, the word to the next. They talk across a handshake somebody had to define. Part present, area safe, robot clear, task done. Get the handshake wrong and both boxes are “working” while the cell is not.

There is no law that says the PLC must be the boss. Some designs put more of the cell in the robot controller, or in a drive, or in a safety relay that neither of them is allowed to override. The split should be written down. A plant that discovers the split during a crash has already paid for the documentation it skipped.

The network reports. It does not excuse the local logic

Later controllers grew ports. Remote racks, drives, screens, other PLCs, a data system upstairs. A screen in an office can show you a station that is down. That is a real gain, and it is a new way to be wrong: a bad address, a switch left unmanaged, a number that arrived late and got treated as fresh. The time-critical permissive still belongs next to the machine, in the controller or the safety device that was designed to hold it. Upstairs can coordinate a schedule. Upstairs should not be the only thing standing between a press and a hand.

Where this sits in the history of automation

PLCs are still the ordinary brain of discrete machines: packaging, machining, handling, a lot of building equipment, pieces of process plants. Newer units have more memory, more motion, more ways to talk. The job did not become a different job. Hold the sequence. Do not start the hazard. Tell a person what faulted.

A model that spots a drift in the data can advise. It does not replace a permissive that has to be true every scan, and it does not inherit a safety rating by being clever. I am describing tools a plant can buy. I am not describing every line, and I am not moving the interlock into a cloud because the cloud drew a graph.

The long use, without the straight line

Compare a PLC to a cam, or to the Jacquard cards on the textile page, and you can teach a true thing: a sequence stored outside the operator’s hands. Do not teach it as a bloodline. Those mechanisms do not descend into this box. The PLC comes out of relay panels, solid-state electronics, and a factory that changed models faster than the wires could. Walker’s word is prehistory. Use it.

The controller earns its place because a plant could finally keep a sequence in a program, test a change, and leave the power wiring alone. It does not earn the claim that the machine became intelligent, or that the person left. Somebody still writes the logic, still proves the sensor, still gets called when the snapshot and the machine disagree. The box does not care what we call it. It only energizes what the program, on this scan, believes the inputs allowed.