PLC Simulator
PLC field notesfundamentals

The PLC Scan Cycle Explained (With Diagrams)

The PLC scan cycle is the continuous read-execute-write loop at the heart of every controller. Learn what happens in each phase and why scan time matters for your control programs.

PLC Simulation Software8 min read

If your PLC program behaves in a way that seems to miss events or update outputs a fraction too late, the scan cycle is almost certainly involved.

The PLC scan cycle is the continuous loop — typically completing in 1–20 ms — during which the processor reads all inputs into memory, executes your control program from top to bottom, then writes the results to all outputs. Everything your program does happens within this cycle. Understanding it is the difference between writing reliable control logic and writing code that occasionally misbehaves in production.

The PLC scan cycle explained — the read-execute-write loop at the heart of every PLC

The Four Phases of the Scan Cycle

Modern PLCs typically break the scan into four distinct phases:

PLC scan cycle flowchart showing read inputs, execute program, update outputs, housekeeping and repeat

Phase 1: Input Scan

The processor reads every digital and analogue input wired to the I/O modules and copies the values into a region of RAM called the Process Image Input Table (PIIT) or Input Image Register. From this point forward, your program works against these snapshot values — not the live wire.

This is a critical design decision: if a sensor changes state while the program is running, the program does not see the change until the next scan. This prevents a rung in the middle of your program from seeing a different sensor state than a rung at the top, which would make deterministic logic nearly impossible.

PLC process image architecture showing field inputs, input image table, CPU, output image table and field outputs

Phase 2: Program Execution

The CPU works through your ladder program from the first rung to the last, left to right on each rung. For each rung it:

  1. Evaluates the input conditions (contacts) using the PIIT values.
  2. Computes the rung result (true or false).
  3. Updates the affected output coils and memory bits in the Process Image Output Table (PIOT) in RAM.

Note: coil values written in Phase 2 are immediately available for contacts further down the program in the same scan. This is the source of several common sequencing bugs — a bit set late in the program is invisible to rungs that already ran above it.

Phase 3: Output Scan

The processor writes the Process Image Output Table to the physical output terminals on the I/O modules. Field devices — motor starters, solenoid valves, indicator lights — respond to the change.

PLC scan timing diagram showing inputs sampled once per scan and outputs updated at the end of the scan

Phase 4: Housekeeping

The CPU handles internal tasks: communication servicing (Ethernet/IP, Modbus, serial), updating the real-time clock, checking the watchdog timer, and self-diagnostics. On some PLCs this phase is interleaved with program execution; on others it is a discrete step.

Why Scan Time Matters

The Watchdog Timer

Every PLC has a watchdog timer — a hardware timer that the firmware resets at the end of every successful scan. If the scan takes longer than the configured maximum (e.g. 500 ms), the watchdog trips a fault and the PLC stops. This prevents a runaway program from leaving outputs energised indefinitely in an unsafe state.

If you add computationally expensive operations (nested loops over large arrays, heavy floating-point maths) and the scan time exceeds the watchdog threshold, you will get a fault. Keep scan time well below the watchdog limit.

Input Timing and Short Pulses

If an input signal is present for less than one full scan time, the PLC may not see it. Consider a proximity sensor whose output pulse lasts 500 µs and a PLC with a 10 ms scan: in most configurations the input scan will never sample that pulse.

Solutions:

  • Use high-speed interrupt inputs (HSC — High Speed Counter) for fast events.
  • Use latching hardware inputs that hold the signal until the next scan.
  • Re-architect the sensor or mechanical system to produce longer signals.

Output Latency

There is always at least one scan of latency between a condition becoming true and the corresponding output energising. For a 10 ms scan cycle this is imperceptible to humans, but it matters for precise motion control applications.

Comparison table of scan-based PLC versus continuous and interrupt-driven execution models

Scan Cycle Timing in Practice

(* Estimated scan-time contribution of a simple motor control program *)
(* Rung count: ~20 rungs, no function blocks *)
(* Typical execution time: 0.05–0.2 ms on a mid-range CPU *)
(* Total scan including I/O: 2–5 ms *)

The execution phase is usually the smallest contributor to total scan time. I/O scanning — especially if you have a large rack of analogue cards or remote I/O over a field bus — often dominates.

Illustrative bar chart of PLC scan-time contributions showing I/O scanning dominating program execution

Periodic vs Event-Driven Tasks

Most modern PLCs let you organise your program into tasks with different priorities:

Reference tableSwipe
Task typeTriggerTypical use
Cyclic (Free-running)Continuous scanMain control logic
PeriodicFixed interval (e.g. every 10 ms)Predictable timing for motion, PID
EventRising edge on a bitInterrupt-style fast response

A periodic task with a 10 ms period will execute every 10 ms regardless of how long lower-priority cyclic tasks take. Use periodic tasks for PID loops (see PID Control for PLCs: Practical Tuning Guide) where consistent sample intervals are critical.

How a task updates its I/O also varies. With synchronous I/O the process image is refreshed at fixed points in the scan, keeping inputs frozen while the program runs; with asynchronous I/O the data can update independently for lower latency.

Comparison of synchronous versus asynchronous PLC I/O update behaviour

The Scan Cycle and Ladder Logic Execution Order

Here is a critical gotcha for ladder logic beginners:

(* Rung 1 — sets BitA based on Input1 *)
IF Input1 THEN
    BitA := TRUE;
ELSE
    BitA := FALSE;
END_IF;

(* Rung 2 — reads BitA, which was just updated this scan *)
IF BitA THEN
    Output1 := TRUE;
END_IF;

Ladder rung example of a same-scan dependency where a coil reads a bit written earlier in the same scan

In this example, Output1 will energise in the same scan that BitA is set by Input1. No scan delay. But if the rungs were reversed — Rung 2 above Rung 1 — Output1 would not energise until the next scan after Input1 goes high. This matters when you are building sequences and timers.

Flowchart of ladder logic execution order within a single PLC scan

Debugging Scan-Cycle Timing Issues

Checklist of PLC scan-time gotchas including missed short pulses, one-scan latency and watchdog limits

  1. Monitor scan time — most PLCs expose current scan time in diagnostics. Watch it under load.
  2. Use one-shot contacts — if you are detecting a rising edge, use an OSR (one-shot rising) or R_TRIG function block to capture the event for exactly one scan.
  3. Trace the data table — step through the output image table after a scan to verify which bits were set and by which rung.
  4. Simulate first — use a browser-based PLC simulator to iterate quickly. In the simulator you can pause after each scan and inspect every register.

For a deeper dive into how ladder logic rungs execute, see How to Read Ladder Logic (Step by Step) and work through the Debugging and the Scan Cycle lesson.


Practice this yourself in the simulator — 3 scenarios free. No install. No credit card. Write real ladder logic against a live machine model in your browser.

Try the simulator free →

ShareX / TwitterLinkedIn

From reading to running logic

Practice this yourself in the simulator

Start with guided PLC practice in your browser. No install and no credit card required.

Start practising free

Continue learning

Related field notes

All articles
ladder logic
fundamentals

Seal-In Rungs in Ladder Logic: The Complete Guide

A seal-in rung keeps a coil latched after the momentary start signal drops. Learn how to build reliable seal-in circuits with stop, overload, and E-stop protection.

8 min read
fundamentals
beginner

What Is a PLC? Plain-English Guide for Beginners

A PLC (Programmable Logic Controller) is an industrial computer that reads sensor inputs, runs a control program, and switches outputs on or off. Learn how it works in plain English.

7 min read
ladder logic
structured text

Ladder Logic vs Structured Text: Which One to Learn First

Ladder logic is graphical and widely supported; structured text is textual and powerful for maths. Compare syntax, use cases, and job market demand to choose your first PLC language.

9 min read

Technical reference and worked-example guide

PLC scan cycle explained: implementation, evidence and troubleshooting

Direct answer

PLC scan cycle explained becomes useful when it connects input sampling, program or task execution, output update, communications and housekeeping in the selected controller context with field state to input image, rung or code order, output image, physical output and feedback, then proves a one-scan state transition observed repeatedly with timestamps or trace state under normal, boundary, fault and recovery conditions. The objective is a repeatable engineering or learning result, not merely activity inside a page or tool.

This guide is written for pLC learners and troubleshooters who need to predict when inputs are sampled, logic is evaluated and outputs become visible. The intended result is specific: the reader can trace a signal across input image, program execution and output update, then explain one-scan effects, order and task timing.

System map / 02

Six concepts that control the result

Treat these as connected checkpoints. Each checkpoint has an expected state, an observable state and a boundary to the next part of the system. That structure prevents a software indication from being mistaken for physical proof.

NODE 01observable

Define the operating contract

input sampling, program or task execution, output update, communications and housekeeping in the selected controller context. For PLC scan-cycle behavior, record the initial condition, actor, requested change, observable result and stopping condition before selecting a tool or implementation.

NODE 02observable

Map the evidence path

field state to input image, rung or code order, output image, physical output and feedback. Separate request, internal state, output or service, physical or user-visible result and independent feedback so each boundary can be inspected.

NODE 03observable

Prove normal operation

a one-scan state transition observed repeatedly with timestamps or trace state. Run more than one cycle from a known state and retain the values, timings or artifacts that demonstrate repeatability.

NODE 04observable

Exercise a boundary case

short pulses, immediate I/O, multiple tasks, asynchronous data, first scan and watchdog limits. Choose minimum, maximum, simultaneous, delayed or restart conditions that reveal assumptions hidden by the happy path.

NODE 05observable

Diagnose a controlled fault

a missed edge, overwritten output, race, stale input or task-order assumption. Preserve the first symptom, divide the system at a measurable boundary and change one condition only after predicting the result.

NODE 06observable

Transfer and hand over

the reduced timing case reproduced with target traces and official task documentation. Restore normal state, remove temporary changes, repeat affected checks and document which claims remain limited to the learning environment.

Procedure / 03

A six-step practice and commissioning workflow

Run the steps in order the first time. Later, the same structure becomes a diagnostic loop: define the expected condition, observe the boundary, interpret the difference and choose one proving action.

  1. 01

    Write the acceptance case

    Convert input sampling, program or task execution, output update, communications and housekeeping in the selected controller context into initial conditions, one stimulus and observable pass criteria.

    Evidence: Another person can repeat the case without guessing the intended result.

    Avoid: Using page completion or an animation as the acceptance criterion.

  2. 02

    Build the map

    Document field state to input image, rung or code order, output image, physical output and feedback and name who owns each state or decision.

    Evidence: Every request and result has a source, destination and useful inspection point.

    Avoid: Using the same value as command, status and independent feedback.

  3. 03

    Run the baseline

    Apply a one-scan state transition observed repeatedly with timestamps or trace state from a clean start and record the expected evidence.

    Evidence: Repeated runs produce the same bounded result.

    Avoid: Changing several parameters before a baseline exists.

  4. 04

    Challenge assumptions

    Test short pulses, immediate i/o, multiple tasks, asynchronous data, first scan and watchdog limits without changing the acceptance contract.

    Evidence: Limits, timing and restart behavior reach defined states.

    Avoid: Testing only one ideal sequence.

  5. 05

    Isolate one failure

    Introduce or analyse a missed edge, overwritten output, race, stale input or task-order assumption and locate the first disagreement.

    Evidence: The proving action distinguishes the leading hypotheses.

    Avoid: Resetting, forcing or replacing before evidence is retained.

  6. 06

    Close the evidence loop

    Complete the reduced timing case reproduced with target traces and official task documentation and repeat the affected regression cases.

    Evidence: Reference use is complete when inputs, assumptions, units or initial conditions are recorded and the result is independently checked at a useful boundary.

    Avoid: Treating an acknowledged message or one successful rerun as handover.

Diagnostic matrix / 04

Symptoms, proving points and next actions

The table is a reasoning aid, not a parts-replacement chart. Preserve the initial symptom, inspect the named boundary and use the interpretation to choose the next controlled test. Site safety procedures and equipment manuals remain authoritative.

Diagnostic symptoms, inspection points, interpretations and next actions for PLC scan cycle explained: implementation, evidence and troubleshooting
Observed symptomInspectInterpretationNext proving action
The expected result is unclearRequirement, initial state, actor, stimulus, units and pass conditionThe technician, programmer and reviewer may be solving different versions of the task.Rewrite one observable acceptance case before continuing.
Internal state changes but the outcome does notRequest, final owner, output or service boundary and independent feedbackA software or interface indication proves intent at one layer, not the complete outcome.Trace the first boundary after the changing state.
Normal case passes but an edge case failsLimits, timing, simultaneous events, reset and restart assumptionsThe implementation contains a hidden assumption exposed by the changed condition.Add the failed boundary as a permanent regression case.
The failure disappears after resetOriginal symptom, histories, diagnostics, timestamps and active causeReset changed evidence or state without proving the initiating cause.Reproduce under a controlled condition and preserve pre/post-event data.
Simulator and target disagreeModel boundary, software version, task timing, I/O behavior, data types and configurationA learning model and the intended target do not share one of the recorded assumptions.Reduce the case and verify against current target documentation.
The result cannot be explainedPrediction, observation, proving action, alternative hypotheses and limitationsActivity occurred but the evidence is not yet transferable or reviewable.Have the learner defend the signal path and repeat a changed case.

Product evidence / 05

What the browser practice can actually demonstrate

The page connects definitions and worked examples to runnable tools, explicit assumptions and repeatable checks so a formula or pattern can be challenged.

Where simulation stops

Real controllers can use periodic, event, interrupt and communication tasks; exact timing and I/O update behavior must be verified for the selected platform.

Commissioning notebook / 06

Six cases that turn the concepts into evidence

Use these as written briefs rather than click-through instructions. For every case, state the expected condition before acting, retain the first useful observation and explain why the final result proves the requirement. A different program or component choice can still be correct when it produces the same bounded behavior and evidence.

Case 01

predict → observe → prove

Prove define the operating contract

Engineering context. input sampling, program or task execution, output update, communications and housekeeping in the selected controller context. For PLC scan-cycle behavior, record the initial condition, actor, requested change, observable result and stopping condition before selecting a tool or implementation. Begin with a written normal condition and identify which request, state, physical result or communication value will provide independent confirmation. Do not begin by changing the configuration; the initial state is part of the evidence and should remain reproducible.

Controlled setup. Use the “Write the acceptance case” stage of the workflow: convert input sampling, program or task execution, output update, communications and housekeeping in the selected controller context into initial conditions, one stimulus and observable pass criteria. The acceptance record should show this result: another person can repeat the case without guessing the intended result. Record initial conditions, the exact stimulus and the observation point so another learner can repeat the case without relying on your memory.

Fault challenge. Introduce or analyse “The expected result is unclear” as one bounded deviation. Inspect requirement, initial state, actor, stimulus, units and pass condition The working interpretation is that the technician, programmer and reviewer may be solving different versions of the task. The next proving action is to rewrite one observable acceptance case before continuing. Change only one condition before observing the result, and preserve timestamps or measurements where timing matters.

Review and recovery. The most common trap here is using page completion or an animation as the acceptance criterion. After restoring the cause, repeat the normal case and at least one stop, timeout, disconnect or restart boundary relevant to this topic. Remove temporary forces and bypasses, return the model to a known state and retain the evidence that both operation and recovery are deliberate.

Explain it aloud: What should I learn first about PLC scan-cycle behavior? A defensible short answer is: Start with the operating contract and evidence path: input sampling, program or task execution, output update, communications and housekeeping in the selected controller context, followed by field state to input image, rung or code order, output image, physical output and feedback. Add advanced features only after the baseline is predictable.

Case 02

predict → observe → prove

Prove map the evidence path

Engineering context. field state to input image, rung or code order, output image, physical output and feedback. Separate request, internal state, output or service, physical or user-visible result and independent feedback so each boundary can be inspected. Begin with a written normal condition and identify which request, state, physical result or communication value will provide independent confirmation. Do not begin by changing the configuration; the initial state is part of the evidence and should remain reproducible.

Controlled setup. Use the “Build the map” stage of the workflow: document field state to input image, rung or code order, output image, physical output and feedback and name who owns each state or decision. The acceptance record should show this result: every request and result has a source, destination and useful inspection point. Record initial conditions, the exact stimulus and the observation point so another learner can repeat the case without relying on your memory.

Fault challenge. Introduce or analyse “Internal state changes but the outcome does not” as one bounded deviation. Inspect request, final owner, output or service boundary and independent feedback The working interpretation is that a software or interface indication proves intent at one layer, not the complete outcome. The next proving action is to trace the first boundary after the changing state. Change only one condition before observing the result, and preserve timestamps or measurements where timing matters.

Review and recovery. The most common trap here is using the same value as command, status and independent feedback. After restoring the cause, repeat the normal case and at least one stop, timeout, disconnect or restart boundary relevant to this topic. Remove temporary forces and bypasses, return the model to a known state and retain the evidence that both operation and recovery are deliberate.

Explain it aloud: How do I practise PLC scan-cycle behavior effectively? A defensible short answer is: Use short cases with known initial conditions, a written prediction, one action and an observable result. Then alter a boundary or fault and explain why the evidence changed.

Case 03

predict → observe → prove

Prove prove normal operation

Engineering context. a one-scan state transition observed repeatedly with timestamps or trace state. Run more than one cycle from a known state and retain the values, timings or artifacts that demonstrate repeatability. Begin with a written normal condition and identify which request, state, physical result or communication value will provide independent confirmation. Do not begin by changing the configuration; the initial state is part of the evidence and should remain reproducible.

Controlled setup. Use the “Run the baseline” stage of the workflow: apply a one-scan state transition observed repeatedly with timestamps or trace state from a clean start and record the expected evidence. The acceptance record should show this result: repeated runs produce the same bounded result. Record initial conditions, the exact stimulus and the observation point so another learner can repeat the case without relying on your memory.

Fault challenge. Introduce or analyse “Normal case passes but an edge case fails” as one bounded deviation. Inspect limits, timing, simultaneous events, reset and restart assumptions The working interpretation is that the implementation contains a hidden assumption exposed by the changed condition. The next proving action is to add the failed boundary as a permanent regression case. Change only one condition before observing the result, and preserve timestamps or measurements where timing matters.

Review and recovery. The most common trap here is changing several parameters before a baseline exists. After restoring the cause, repeat the normal case and at least one stop, timeout, disconnect or restart boundary relevant to this topic. Remove temporary forces and bypasses, return the model to a known state and retain the evidence that both operation and recovery are deliberate.

Explain it aloud: What counts as proof of competence? A defensible short answer is: A repeatable artifact or system result plus an explanation of the signal path is stronger than time spent, screenshots or a copied answer. Physical competence requires separate supervised evidence.

Case 04

predict → observe → prove

Prove exercise a boundary case

Engineering context. short pulses, immediate I/O, multiple tasks, asynchronous data, first scan and watchdog limits. Choose minimum, maximum, simultaneous, delayed or restart conditions that reveal assumptions hidden by the happy path. Begin with a written normal condition and identify which request, state, physical result or communication value will provide independent confirmation. Do not begin by changing the configuration; the initial state is part of the evidence and should remain reproducible.

Controlled setup. Use the “Challenge assumptions” stage of the workflow: test short pulses, immediate i/o, multiple tasks, asynchronous data, first scan and watchdog limits without changing the acceptance contract. The acceptance record should show this result: limits, timing and restart behavior reach defined states. Record initial conditions, the exact stimulus and the observation point so another learner can repeat the case without relying on your memory.

Fault challenge. Introduce or analyse “The failure disappears after reset” as one bounded deviation. Inspect original symptom, histories, diagnostics, timestamps and active cause The working interpretation is that reset changed evidence or state without proving the initiating cause. The next proving action is to reproduce under a controlled condition and preserve pre/post-event data. Change only one condition before observing the result, and preserve timestamps or measurements where timing matters.

Review and recovery. The most common trap here is testing only one ideal sequence. After restoring the cause, repeat the normal case and at least one stop, timeout, disconnect or restart boundary relevant to this topic. Remove temporary forces and bypasses, return the model to a known state and retain the evidence that both operation and recovery are deliberate.

Explain it aloud: Why test faults and restart behavior? A defensible short answer is: Because a missed edge, overwritten output, race, stale input or task-order assumption or short pulses, immediate i/o, multiple tasks, asynchronous data, first scan and watchdog limits can expose assumptions that never appear during ideal startup and steady operation.

Case 05

predict → observe → prove

Prove diagnose a controlled fault

Engineering context. a missed edge, overwritten output, race, stale input or task-order assumption. Preserve the first symptom, divide the system at a measurable boundary and change one condition only after predicting the result. Begin with a written normal condition and identify which request, state, physical result or communication value will provide independent confirmation. Do not begin by changing the configuration; the initial state is part of the evidence and should remain reproducible.

Controlled setup. Use the “Isolate one failure” stage of the workflow: introduce or analyse a missed edge, overwritten output, race, stale input or task-order assumption and locate the first disagreement. The acceptance record should show this result: the proving action distinguishes the leading hypotheses. Record initial conditions, the exact stimulus and the observation point so another learner can repeat the case without relying on your memory.

Fault challenge. Introduce or analyse “Simulator and target disagree” as one bounded deviation. Inspect model boundary, software version, task timing, I/O behavior, data types and configuration The working interpretation is that a learning model and the intended target do not share one of the recorded assumptions. The next proving action is to reduce the case and verify against current target documentation. Change only one condition before observing the result, and preserve timestamps or measurements where timing matters.

Review and recovery. The most common trap here is resetting, forcing or replacing before evidence is retained. After restoring the cause, repeat the normal case and at least one stop, timeout, disconnect or restart boundary relevant to this topic. Remove temporary forces and bypasses, return the model to a known state and retain the evidence that both operation and recovery are deliberate.

Explain it aloud: Can browser practice replace official software or hardware? A defensible short answer is: No. It can build concepts and diagnostic reasoning. Exact firmware, I/O electrical behavior, networking, safety and commissioning require current official tools, documentation and target equipment.

Case 06

predict → observe → prove

Prove transfer and hand over

Engineering context. the reduced timing case reproduced with target traces and official task documentation. Restore normal state, remove temporary changes, repeat affected checks and document which claims remain limited to the learning environment. Begin with a written normal condition and identify which request, state, physical result or communication value will provide independent confirmation. Do not begin by changing the configuration; the initial state is part of the evidence and should remain reproducible.

Controlled setup. Use the “Close the evidence loop” stage of the workflow: complete the reduced timing case reproduced with target traces and official task documentation and repeat the affected regression cases. The acceptance record should show this result: reference use is complete when inputs, assumptions, units or initial conditions are recorded and the result is independently checked at a useful boundary. Record initial conditions, the exact stimulus and the observation point so another learner can repeat the case without relying on your memory.

Fault challenge. Introduce or analyse “The result cannot be explained” as one bounded deviation. Inspect prediction, observation, proving action, alternative hypotheses and limitations The working interpretation is that activity occurred but the evidence is not yet transferable or reviewable. The next proving action is to have the learner defend the signal path and repeat a changed case. Change only one condition before observing the result, and preserve timestamps or measurements where timing matters.

Review and recovery. The most common trap here is treating an acknowledged message or one successful rerun as handover. After restoring the cause, repeat the normal case and at least one stop, timeout, disconnect or restart boundary relevant to this topic. Remove temporary forces and bypasses, return the model to a known state and retain the evidence that both operation and recovery are deliberate.

Explain it aloud: How should progress be documented? A defensible short answer is: Keep the requirement, initial state, program or configuration, observed values, fault hypothesis, proving action, recovery result and a concise limitations statement.

Answer surface / 07

Questions people ask about PLC scan cycle explained

These concise answers define the operating, training and product boundaries most often missed in broad summaries. The full workflow and diagnostic table above provide the evidence behind them.

What should I learn first about PLC scan-cycle behavior?

Start with the operating contract and evidence path: input sampling, program or task execution, output update, communications and housekeeping in the selected controller context, followed by field state to input image, rung or code order, output image, physical output and feedback. Add advanced features only after the baseline is predictable.

How do I practise PLC scan-cycle behavior effectively?

Use short cases with known initial conditions, a written prediction, one action and an observable result. Then alter a boundary or fault and explain why the evidence changed.

What counts as proof of competence?

A repeatable artifact or system result plus an explanation of the signal path is stronger than time spent, screenshots or a copied answer. Physical competence requires separate supervised evidence.

Why test faults and restart behavior?

Because a missed edge, overwritten output, race, stale input or task-order assumption or short pulses, immediate i/o, multiple tasks, asynchronous data, first scan and watchdog limits can expose assumptions that never appear during ideal startup and steady operation.

Can browser practice replace official software or hardware?

No. It can build concepts and diagnostic reasoning. Exact firmware, I/O electrical behavior, networking, safety and commissioning require current official tools, documentation and target equipment.

How should progress be documented?

Keep the requirement, initial state, program or configuration, observed values, fault hypothesis, proving action, recovery result and a concise limitations statement.

What should I do when the answer differs from a guide?

Check assumptions, version, units and initial state first. Reduce the case, compare one boundary at a time and prefer current primary documentation for target-specific behavior.

When is a PLC scan-cycle behavior exercise finished?

Reference use is complete when inputs, assumptions, units or initial conditions are recorded and the result is independently checked at a useful boundary.