IMDHMIControls LibraryShop main store

Comparison

How an HMI and a PLC divide the work in a control system

PLC programming Published 7 min read

Short answer

The PLC reads field inputs, executes the control logic, and drives the outputs; the HMI only reads selected PLC data for display and writes operator requests back for that logic to accept or refuse. Every screen symptom traces through four hops — tag binding, communication link, PLC data, field device — and the hop where it stops tells you which side owns the fault.

On this page
  1. What is the difference between an HMI and a PLC?
  2. How do the HMI and the PLC exchange data?
  3. How fast does each side react to a signal?
  4. Can the HMI run the control logic instead of the PLC?
  5. Why does the screen show a wrong value when the PLC is fine?
  6. How do you trace a screen symptom to its source?
  7. Which side should own each task in a new project?
  8. Questions asked

The PLC reads field inputs, executes the control logic, and drives the outputs. The HMI reads selected PLC data, draws it for an operator, and writes that operator's requests back for the same logic to accept or refuse. Neither substitutes for the other, and every screen symptom you will troubleshoot sits somewhere on the four-hop path between them: tag binding, communication link, PLC data, field device. This article gives the role split, the timing that follows from it, and the tracing order that tells you which side owns a fault.

What is the difference between an HMI and a PLC? #

The PLC is the controller: it wires to the sensors and actuators and executes the program that decides machine behavior. The HMI is the operator's window: it displays chosen PLC data and submits operator values and commands, but it wires to no field devices.

The split is about wiring and authority. Every sensor and actuator lands on the PLC's input and output modules, and the program turns those signals into machine actions. The HMI's only connection is the network to the controller, and its only authority is what the controller's logic grants it.

What you compare HMI PLC
Main job Present process data and collect operator requests Read inputs, execute logic, drive outputs
Field wiring None — one network connection to the controller Input and output modules wired to sensors and actuators
Programming Screens, tag bindings, alarms, trends Ladder logic, function blocks, structured text
Owns Presentation and the operator's request Machine state and every control decision
Reaction to a signal Whenever the screen next updates it One to two scan cycles, input change to output change
When it fails Values go stale and buttons stop responding The machine stops or misbehaves identically on every display

Talking to the controller is not controlling the machine. The rest of this article is what that sentence costs you when something breaks.

How do the HMI and the PLC exchange data? #

The HMI polls or subscribes to controller data over a configured protocol — Modbus TCP, EtherNet/IP, PROFINET and OPC UA are the usual ones — reading the tags it displays and writing the operator's entries to tags the PLC exposes. Every screen object carries a tag binding, a pointer at one controller address or symbol, and that binding is the whole interface between the two worlds.

The poll rate you configure, not the PLC's speed, sets how current the screen is. On the command side, use the request pattern: the HMI writes a request tag, and PLC logic decides whether conditions allow it and drives the output. Safe Modbus writes to PLC output registers from SCADA shows that pattern in detail, and it carries over to every protocol.

Because the entire screen rides on one link, commission a SCADA PLC communication alarm for link loss on day one, so a dead connection announces itself instead of hiding as data that quietly froze.

How fast does each side react to a signal? #

The PLC answers an input change within one to two scan cycles — worst case just under twice the scan time, because a change landing right after the input scan waits a full pass to be read and another before its output lands. Average is about one scan, plus input filter delay.

Program size and I/O count move the scan time up; hardware interrupts move a specific response down below it. One published engineering brief reports a measured 275 µs response with ±50 µs of jitter on a hardware-interrupt configuration — that is the territory interrupts buy, while a plain cyclic program answers in milliseconds.

Worked example (illustrative inputs): a compact PLC with a 15 ms scan, an HMI polling at 500 ms.

  • PLC worst case, input change to output change: 2 × 15 ms = 30 ms.
  • Operator sees the result no sooner than the next poll: 30 ms + 500 ms = 530 ms.
  • The screen therefore trails the controller by 530 ÷ 30 ≈ 17.7 times.

Decision from those inputs: any reaction the machine needs in under roughly 50 ms belongs in the PLC alone, and any logic routed through the screen inherits the poll rate plus total blindness whenever the link drops.

Can the HMI run the control logic instead of the PLC? #

No — for any machine that can hurt itself, the product or a person, the logic belongs in the controller, because the HMI's view of the process arrives late, disappears with the link, and is not the thing wired to the outputs. Splitting logic between controller and panel looks cheaper on small jobs and becomes the maintenance problem nobody can trace as projects grow: conditions live in two tools, two places, and two people's heads.

What may live in the HMI:

  • Screen navigation and operator login
  • Display formatting, scaling and units for reading
  • Keypad entry, with a range check at entry

What must live in the PLC:

  • Sequencing and machine state
  • Interlocks and permissives
  • Timers that govern outputs
  • Alarm conditions that trip or stop anything

Safety functions — E-stops, guard interlocks, light-curtain reactions — belong in the safety controller or hardwired relays, never behind any screen. Prove the split at commissioning by unplugging the panel: a machine that must run unattended should run its cycle with no display attached.

Why does the screen show a wrong value when the PLC is fine? #

Because the tag binding, not the process, feeds the object: a screen element that references the wrong controller item renders perfectly while showing a plausible but unrelated number, and only comparing screen against controller separates the two. A quality fault and a binding fault are different failures.

Quality says the data cannot be read: the object shows question marks or holds its last value. Many objects failing at once points at the link, and why HMI values show question marks when PLC communication times out walks that diagnosis. A wrong binding is readable but wrong: the object points at some other item and displays a number nobody asked for. Watch the intended data in the controller's programming tool — PLC right, screen wrong means fix the binding; both agreeing on a wrong value means keep walking toward the field.

How do you trace a screen symptom to its source? #

Walk the four hops in order — binding, link, PLC data, field device — and stop at the first hop where the story breaks: that hop, and the side that owns it, is the fault. The order matters because each hop is cheaper to check than the one after it.

  1. Name the screen object and the process state it claims to represent — a value, an alarm, a trend or a button.
  2. Open its tag binding and confirm the address or symbol, and that its read/write role matches the object's job.
  3. Watch the same data live in the controller's programming tool.
  4. Compare the PLC input with the sensor itself, or the PLC output with the actuator's feedback.
Symptom on the screen First hop that stops What it tells you
One value stale or wrong, the rest live Binding Presentation fault; fix the tag on that object
The whole screen shows question marks Link Exchange fault; network or controller connection
Screen and PLC agree on a wrong value Field input Sensing fault; check device and wiring
Button flashes but the machine ignores it Write path or permissives The request never landed, or logic correctly blocked it

For the third row, detecting a failed analog input channel and reporting it to SCADA gives the channel-level checks that separate a dead channel from a dead sensor.

Which side should own each task in a new project? #

Give the PLC everything that must happen while nobody is watching — sequencing, interlocks, timers, alarm conditions that stop the machine — and give the HMI everything that exists only because a person is standing there: screens, formatting, keypads, navigation and login. The test is what happens with the panel switched off.

Task Owner Why
Start/stop permissives and sequencing PLC Must hold with no screen attached
PID loop execution PLC Needs deterministic timing, not screen updates
Setpoint storage PLC, retentive memory A failed panel must not take settings with it
Setpoint entry HMI The person types at the screen; the value lands in the controller
Alarm detection PLC The controller sees the raw condition first
Alarm formatting and acknowledgment HMI Presentation only
Trend history HMI or supervisory system Nobody actuates a machine from a trend

Hold the line at the wire: the PLC owns behavior, the HMI owns presentation, and the link between them carries requests and status. Nothing the machine depends on for sequence or safety crosses that wire as the deciding path.

Questions asked

Can one panel be both the HMI and the PLC?

Yes — combined HMI-plus-controller panels exist and suit small standalone machines where the display can sit at the machine. Treat the controller side as a real PLC anyway: version-controlled code, defined power-fail behavior, and a spare on the shelf, because when that one unit fails you lose the display and the control together. Above a handful of I/O, or with any interlocked motion, separate them.

Does a faster HMI make the machine react faster?

No. The machine's reaction is set by the PLC's scan and its I/O; the screen update only decides how quickly a person notices. Cutting an HMI poll from 1 s to 250 ms multiplies the traffic on the link by four (1000 ÷ 250 = 4) to speed an observation the operator still has to read and act on. Set the poll rate for what the operator needs to see, and leave the reaction to the PLC.

Where should recipes and setpoints be stored?

In retentive controller memory, with the HMI holding only a working copy for editing. If the settings live in the panel, a failed panel takes the process settings with it and the replacement must be reloaded before the machine runs. Stored in the PLC, a dead display costs you visibility for the minutes it takes to swap, and the machine restarts with its last recipe intact.

Who owns the alarm list, the HMI or the PLC?

The PLC owns alarm conditions; the HMI owns their presentation. The panel's built-in alarm system can only report what the controller publishes, so any condition that must stop the machine or start equipment is detected in PLC logic, not in a screen-side threshold. Format, sort and acknowledge on the HMI; detect and act in the PLC.

Do we need an HMI on every machine?

Only where a person must watch or adjust without a programming laptop. A machine that runs a fixed cycle unattended needs no screen; its states can surface as stack lights and a handful of codes. Any machine an operator sets up, tunes or clears jams on needs one, because those tasks otherwise cost a laptop, a cable and a controls engineer.

Keep reading

  • Selection guide

    How to size an S7-400 CPU against a PLC-5/40 scan time

    No S7-400 CPU maps one-to-one to a PLC-5/40 by instruction timing, and none needs to: every CPU from the 414 upward is faster on logic execution. Select on memory, I/O and communications capacity, then accept the choice only when the translated program's measured worst-case cycle lands at or below two-thirds of the measured PLC-5/40 scan.

    PLC programming

  • Comparison

    How ControlLogix and S7-400 PLC architectures differ

    A ControlLogix application maps onto an S7-400 only at the requirements level. ControlLogix runs tasks that contain programs and routines on symbolic tags; the S7-400 runs organization blocks that call FBs and FCs against data blocks, mixing symbolic and absolute addressing. Match execution behavior, data ownership and I/O timing — never translate instruction names.

    PLC programming

  • Troubleshooting

    PLC-5/80E loses scan list on remote I/O after power cycle

    Auto-negotiate selected on Ethernet Channel 2 of a Series F Revision A PLC-5 clears the chassis scan list on the remote I/O scanner channels whenever the processor power cycles. Disable auto-negotiate and hard-set 100 Mbps full duplex, or flash Revision A.1 or newer; Auto-Config only rebuilds the list after each event.

    PLC programming