The PLC adventure Part 1: Considerations that went into the design of the PLC

I came up with the idea for the PLC while frustrated and hung-over on New Year’s Day in 1968. I won’t go into that part of the adventure here. Instead, I will list some of the considerations in the fundamental design of the Modicon PLC.

When my team and I first started working on the programmable controller in the Fall of 1968, we had to consider the bandwidth of a process. In processes, nothing much happens in fewer than eight milliseconds. The lumps of energy being supplied by 60 Hz generally obliterate the need to do anything faster. On that basis, we suggested that 60 Hz would be the quantum time that we would need to control external process elements.

We also knew that the architecture should not generate cross-modulation products, which meant that we had to take a snapshot of the status of the process and deliver that process to the software architecture to ensure that there were no cross-modulation products in the sampling.

We made the assumption that there would be no difference in the database between the snapshot coming from the outside world and internal references across the language.

Input/output, memory and a specialized scan processor were the three hardware elements in the programmable controller. We would agree that the scan processing could be software, but we wanted it to be very high speed on the projection that we could have 10,000 to 100,000 inputs to gather in our eight millisecond quantization of the process.

All of our suppliers said that we could get higher speed memory, higher quantities and smaller sizes. They also suggested that the cost would be less, but I knew that the signal-to-noise ratio in information theory was important. What I wanted was energy per bit.

One of my engineers suggested that there should be enough energy in the core memory that we could hear the cores click. We did not want 100 electrons per bit in the memory, but rather large, real power in each residual bit so that transients – which are manifest in any process – could not materially affect the memory and software processes.

We knew that the lifetime of a computer or component degrades by a factor of two every eight degrees Celsius rise in temperature. We had to keep the PLC as close to room temperature as we could.

The solution
Between each circuit board we put a sheet of aluminum. That meant that the circuit board thermal radiation would go directly to the aluminum board, and that the aluminum board would then be thermally conducted to the external world.

Our original tests were to take a variable transformer and slowly reduce the power all the way down to zero, slowly bring it back up to over-voltage and have no failures of the software in that time. We had to make the resiliency to power upsets strong.

My target user was the electrician, not the computer scientist. We did not want the “windows” of process control. We wanted the relays of process control. We looked at the programmable controller as a box of relays.

We were selling up-time. It used to take six months to get a process up using computers. A process engineer could do the same job in six days using a programmable controller.

The language had to be synchronous and use a “come from” approach. There were no “go to” commands in the original ladder logic. Each line of control can only reference data, not demand that data be transmitted to another line of code.

The output of all the lines was placed back to the memory, and mixed with real-world data. Each of these lines of logic was an independent operation, and if one of them failed, it would not collapse the other commands.

Because of limitations to the hardware power at the time, I had to limit how many elements could be put on a single line. I also wanted to constrain the elements to reduce complexity. So we used four elements because four elements allow sufficient complexity to control most simple processes.

Our investor at the time was Digital Equipment. I remember meeting one of the company’s founders at the airport with the president of Modicon, Bill Fletcher. We made a presentation and got a cheque for $200,000 US after a 30-minute discussion. We were impressed. Two hundred thousand dollars at that time would probably be equivalent to a million dollars today.

We were off and running. There were many trials and tribulations in the PLC’s manufacturing and marketing, but the technology was strategically targeted to run forever as a tool, not as a computer.

Next month I’ll talk more about the PLC adventure, including specific memories I have about bringing the PLC to market.