Every automation project starts the same way. Somebody says “we need a PLC,” everyone nods, and then somebody goes and buys whatever the last guy used. That’s how most machines end up running on a controller chosen by habit, not by engineering.
Which is fine — until it isn’t. Until you need a part in three days, or the software license expires, or you find out the only person who can edit the program retired last spring and took the password with him.
So here’s the unfiltered version of how to choose a PLC for a machine or a panel. Not the brochure version.
You’re Not Buying a Controller, You’re Marrying an Ecosystem
The box on the DIN rail is the cheap part. The expensive part is everything welded to it: the programming environment, the licensing model, the communication stack, the spare parts chain, and the shrinking pool of people who already know how to work on it.
Vendors know this. It’s the entire business model. Once a plant standardizes on a platform, switching costs get brutal — retraining, rewriting code, redrawing panels, re-validating everything. So pick the ecosystem, not the spec sheet.
Ask three questions before you look at a single datasheet:
- Who can program this thing if the original programmer vanishes?
- How fast can I get a replacement if one dies at 2 a.m. on a Sunday?
- What does the software cost, and does it expire?
The Specs That Actually Matter
I/O count — and how much headroom you’re willing to pay for
Count your inputs and outputs, then add 20–30%. Not because you’ll use them, but because adding a card later usually means a bigger rack, a bigger cabinet, and a bigger power supply. Buying small and expanding is almost always more expensive than buying one size up today.
Scan time and processing power
For most machines, any modern controller is fast enough. Scan time matters when you’re doing motion control, high-speed counting, or tight interlocking. If you’re just starting motors and reading sensors, you’re paying for horsepower you’ll never touch.
Communications
This is where people get quietly trapped. Some platforms speak open protocols happily. Others are happiest talking to themselves. If your machine has to hand data to a supervisory system, a database, a robot, or another vendor’s drive, test that integration before you buy — not after the panel is wired and the deadline is tomorrow.
Environment and ratings
Washdown area? Dust? Vibration? Wide temperature swings? That decides your enclosure rating and whether you need extended-temperature hardware. A controller that’s perfect in a clean room is dead in six months on a foundry floor.
Memory and data handling
Recipe management, data logging, trend history, barcode tracking — these eat memory fast. If the machine needs to store anything, check how the platform handles retentive data. Losing your recipe table on a power blip is a bad afternoon you’ll remember for years.
The Software Tax Nobody Warns You About
Here’s the uncomfortable part. On several major platforms, the programming software is licensed — per seat, sometimes per year, sometimes tied to a hardware key that can be lost, stolen, or bricked by a Windows update nobody approved.
Things that surprise people, every single time:
- Editing a program on a machine you already own may require a paid license.
- Older projects may not open in newer software versions, and downgrading isn’t always supported.
- Some platforms charge separately for features like motion, safety, or advanced networking.
- Free “lite” versions often can’t save or download beyond a tiny I/O limit.
None of this is hidden exactly. It’s just never on page one of the brochure.
Lock-In Is Real — Here’s How People Quietly Work Around It
Total platform independence is basically a myth, but partial escape is very doable, and plenty of shops do it without announcing it:
- Keep the logic portable. Write the machine sequence in a structured, documented way that can be re-implemented elsewhere, and keep a plain-language description of every step and interlock. That document is worth more than the code.
- Use open protocols wherever the hardware allows it — even when the proprietary option is easier to set up.
- Buy one spare CPU. A single shelf spare has saved more production days than most redundancy schemes ever will.
- Keep licenses and install media together, offline. Vendor accounts get locked. People leave. Companies get acquired.
- Document the I/O map ruthlessly. A machine with a clean tag list can be rebuilt on another platform in days. A machine without one is a hostage situation.
When You Don’t Need a “Real” PLC At All
A depressing number of panels are running a full-blown controller to do what a smart relay, a micro unit, or a handful of timers and relays could do. There’s no shame in going small — there’s shame in paying for overkill.
- A few logic steps and a couple of sensors? A smart relay or micro unit is fine.
- Simple on/off sequencing, no comms, no data? Relay logic still works and always will.
- Motion, coordinated axes, high-speed registration? Now you need real hardware, and now scan time genuinely matters.
- Anything with a person’s hands in a danger zone? You need certified safety hardware, not clever programming.
Safety Is a Separate System, Not a Feature
This one gets people hurt, so it goes blunt: standard control logic is not a safety function. E-stops, gates, light curtains, and two-hand controls belong on certified safety hardware — safety relays, a dedicated safety controller, or a safety-rated network — independent of whatever runs the machine’s normal operation.
You can pull safety status into the main controller for monitoring. You cannot rely on the main controller to stop the machine when someone’s arm is in it. Every serious automation standard says this in different words, and the reason it gets repeated constantly is that people keep ignoring it.
The Panel Side: Physical Reality Bites Back
Choosing the controller is half the job. The other half is whether it fits, cools, and wires sanely:
- DIN rail space. Add up widths before committing. Terminal blocks and power supplies eat more room than anyone plans for.
- Heat. Controllers, drives, and power supplies in a sealed box get hot. Check temperature rise, not just the ambient rating.
- Power budget. Add up the current draw of every module, then size the supply with real headroom.
- Wiring style. Removable terminal blocks and pre-wired cables save hours at commissioning and days at troubleshooting.
- Expansion. Leave physical room. Every machine grows eventually.
The Gray Market Question
You will absolutely find used, refurbished, and surplus controllers online for a fraction of list price. Sometimes that’s the only way to get an obsolete module and keep a line running — and it works fine for years. Sometimes you get counterfeit hardware, mismatched firmware, or a unit still locked by a previous owner.
Practical rules: buy from sellers who accept returns, ask for the exact firmware revision, expect to flash it yourself, and never put gray-market hardware on a safety function or on a machine where an unplanned stop costs real money. Also worth knowing — some manufacturers will refuse support or warranty on units not bought through their channel. Not illegal, just policy. But you should know it going in.
A Practical Order of Operations
- Write down what the machine actually has to do, in plain language.
- Count I/O with 25% headroom.
- Decide what has to talk to what, and over which protocol.
- Check the environment: heat, dust, water, vibration.
- Separate safety functions completely.
- Price the whole ecosystem — hardware, software, licenses, cables, training, spares.
- Check availability and who locally can service it.
- Confirm panel space, power, and heat before ordering anything.
- Buy one spare of whatever you can’t live without for a week.
The One Question That Settles It
If you only remember one thing: the best PLC is the one you can fix at 3 a.m.
Not the fastest. Not the cheapest. Not the one with the prettiest software or the slickest demo. The one where a spare exists, someone nearby can program it, the documentation is current, and the licensing doesn’t hold your production hostage.
Everything else — brand loyalty, spec-sheet bragging, “we’ve always used this” — is noise. Pick the platform you can live with for ten years, document it like your job depends on it (because it might), and keep a spare on the shelf. That’s the whole game.