IMDHMIControls LibraryShop main store

Selection guide

Selecting remote I/O architecture for a safety upgrade

Controllers and I/O Published 8 min read

Short answer

Sort every field signal into ordinary, safety or motion first, then buy the fewest network interfaces your crew can support. The safety controller fixes the safety network — CIP Safety runs over EtherNet/IP or DeviceNet — and ordinary hubs and gateways never carry safety data. Count channels, not functions: 16 stations at 20 inputs and 10 outputs is 480 planning signals, and every dual-channel safety device takes two safety channels.

On this page
  1. Sort the schedule into three classes first
  2. Count channels, not functions
  3. Sizing the schedule for 16 stations at 20 inputs and 10 outputs each
  4. The safety controller fixes the safety network
  5. Compare the architectures
  6. Guarded access is a matrix, not a bigger emergency-stop circuit
  7. Decide case by case
  8. Questions asked

Classify every field signal as ordinary, safety-related or motion-related before anyone quotes an I/O block, then buy the fewest network interfaces your maintenance crew can diagnose and replace. The safety controller you install decides which safety network the remote safety I/O runs on: under CIP Safety that means blocks listed for EtherNet/IP or DeviceNet, while ordinary Ethernet switches, IO-Link hubs and protocol gateways never carry safety data. Count channels, not functions — a dual-channel enabling device is two wired channels, and each monitored contactor adds a feedback channel. This article gives the scheduling method, the channel arithmetic for the 16 stations in scope, an architecture comparison and a decision for each case.

Sort the schedule into three classes first #

The station estimate alone cannot pick hardware. Walk every deck with the electrical drawings in hand and record four things for each device: what it is, how it physically connects, where its wiring lands today, and which machine function it serves. Then tag every signal in one of three classes:

  1. Ordinary control — position sensors, valve manifolds, process signals. Record analog separately from discrete, because they land in different modules.
  2. Safety-related — guard switches, emergency stops, enabling devices, drive safety inputs such as safe torque off. For each one record the channel arrangement, the contact or electronic output type, the reset behavior, and which final switching devices need feedback monitoring.
  3. Motion-related — command and feedback wiring that belongs to the servo system, which is usually home-run to the motion controller rather than parked on deck I/O.

Also mark which existing blocks feed the controller directly and which pass through the legacy FX3U layer. Whatever that layer does — interlocks, scaling, mapping — has to be rebuilt in the ordinary controller's program, with the replacement signal mapping documented, before the layer is deleted.

The schedule sorts the purchase: a deck of mostly ordinary discrete signals is a candidate for field I/O masters, while a deck that concentrates safety functions gets its safety channels sized on their own.

Count channels, not functions #

A dual-channel safety device presents two channels so the safety logic can cross-monitor them: one function, two wired inputs. A safety output function adds a feedback input for each contactor it monitors. Remote safe I/O blocks monitor the safety and non-safety devices wired to them over the safety network, so a properly specified block can carry a deck's ordinary sensors alongside its safety channels — the block's own manual states which channels are safety-rated.

The planning total and the channel total are different numbers, and the next section works both for this upgrade.

Sizing the schedule for 16 stations at 20 inputs and 10 outputs each #

Inputs:

  • 16 stations — the stop count in the upgrade scope.
  • 20 inputs and 10 outputs per station — the rough deck estimate, discrete.
  • Dual-channel safety devices — the guarded-access requirement puts dual-channel enabling devices and guard switches into the schedule.

The rule in words: planning signals = stations × (inputs per station + outputs per station). Safety channel demand = 2 × the number of dual-channel devices + one feedback channel per monitored contactor.

Arithmetic:

  • 16 × 20 = 320 inputs
  • 16 × 10 = 160 outputs
  • 320 + 160 = 480 planning signals

Result: 480 planning signals across the machine.

Decision: order blocks against the channel schedule, never against the 480. Every dual-channel device consumes two safety channels and every monitored contactor one feedback channel before ordinary inputs fill the rest. If a deck's safety channel demand exceeds the channels on the block planned for it, add a block — safety channels are never shared between devices. The 480 sizes cable runs, the network address plan and commissioning time; it is not a purchase quantity.

The safety controller fixes the safety network #

CIP Safety is the safety transport between safety controllers and remote safety I/O on EtherNet/IP and DeviceNet. The standard extends the common protocol with Safety I/O assemblies plus Safety Validator and Safety Supervisor objects, and safety I/O blocks from several vendors attach field devices to it — field-mount blocks in IP67 housings included. So the protocol question is settled from one side: whatever safety network the safety controller supports is the network, and every safety block must be listed for that network and role.

Everything ordinary stays ordinary. Ethernet switches, IO-Link masters and hubs, protocol converters and HMI exchanges move bits but are not safety endpoints; safety devices wire to safety channels on safe blocks, and no amount of configuration turns an ordinary connection into a safety one.

The two controllers still cooperate. The Q-series keeps production control, exchanging mode, status and set-point data with the safety controller over a non-safety link — plain Ethernet, Modbus TCP or an HMI exchange is enough, because no safety function may depend on that traffic. Where such a link runs serial Modbus RTU, one port polls its devices in turn and the cycle stretches; why Modbus RTU drive polling on one serial port is slow walks through the timing.

The robot-mediated EtherCAT path with field boxes is the arrangement this tidy-up exists to question. It commissioned quickly because it is a communications path, and quick commissioning is not low installed cost: it duplicates cabinet I/O, cable and a conversion layer. Check what the deck does when the robot is powered down, stopped or disconnected, and keep the path only where that behavior meets machine requirements.

Compare the architectures #

Five architectures cover the options on this machine:

Architecture Build it when Check before you commit
Remote safety I/O blocks on the safety controller's network (CIP Safety over EtherNet/IP or DeviceNet) Decks concentrate safety functions or are being rewired anyway Block listed for the safety controller's network and role; channel count from the schedule, dual channels doubled
IO-Link masters and sensor blocks for ordinary discrete Sensor, valve, manifold and analog wiring is being re-terminated Ordinary transport only — safety devices wire elsewhere; expect fewer home-run cables and cabinet terminals
Gateway onto existing networked blocks Existing CC-Link equipment is serviceable and meets the revised design The gateway must not preserve the legacy layer the project removes; comms-loss behavior documented
Robot-mediated EtherCAT with field boxes Only where the deck's behavior on robot power-down meets machine requirements Duplicated cabinet I/O and cable; a conversion layer this rebuild otherwise removes
Hardwired safety relay bank Fixed functions and the final emergency-stop path Relays are not fault-free for avoiding a network; mode-dependent logic still needs the safety controller

One machine can mix rows: safety blocks where safety concentrates, IO-Link where ordinary wiring is redone, retained CC-Link blocks where they already work.

Guarded access is a matrix, not a bigger emergency-stop circuit #

Running selected functions near an open door on an enabling device, while indexing stays locked out outside normal running, is mode-dependent safety logic. Write a cause-and-effect matrix per operating mode: which hazards must stop when a guard opens, which limited actions the enabling device authorizes, what happens the moment that device leaves its valid state, and how restart and reset differ from ordinary production commands. The safety controller implements the matrix; ordinary I/O cannot.

Safe torque off sits inside the matrix, not instead of it. STO as standardized for drives in IEC 61800-5-2 removes the drive's ability to generate torque; it does not brake, hold a vertical axis or de-energize anything else. Gravity, stored hydraulic or pneumatic energy, coasting and externally driven motion need their own measures, selected from the machine risk assessment — not from the presence of an STO terminal.

Keep the rating honest in both directions. A safety-rated controller is one element of one function: IEC 61508 carries the electrical, electronic and programmable-electronic safety-related system through its lifecycle, with hardware requirements in IEC 61508-2 and software requirements in IEC 61508-3, and the function runs from input devices such as sensors through the logic to output devices and final elements. Sensors, wiring, communications, contactors and validation each carry part of the claim, and the required performance level comes from the risk assessment under ISO 13849-1. Where the risk assessment calls for it, keep the final hardwired emergency stop that removes control power — it works whatever any network is doing.

Decide case by case #

  • Signals first. Classify and schedule every device before any purchase; the schedule, not the station count, picks block sizes.
  • Safety channels. Dual-channel devices and monitored contactors go to remote safety I/O blocks listed for the safety controller's network, sized at two channels per dual-channel device plus feedback.
  • Ordinary decks. Re-terminated sensor, valve and analog wiring goes to IO-Link masters and blocks; safety devices stay off them.
  • Existing CC-Link blocks. If they are serviceable and meet the revised design, keep them direct to the ordinary controller — no gateway, no replacement.
  • Robot path. Fold the deck into the two networks unless the deck must run with the robot down.
  • FX3U. Rebuild its logic in the ordinary controller's program, verify the signal mapping, then delete it.
  • Inter-controller link. Non-safety data only; no safety function may depend on it.
  • Commissioning. Take each re-terminated station through the remote I/O commissioning checklist for a modular station before it carries production.
  • House standard. Set it once — the same trade-offs drive SCADA and PLC platform selection for plant revamps, so the next machine inherits one architecture instead of a new debate.

Questions asked

Can safety signals cross an ordinary EtherNet/IP network if it is configured carefully?

No. A safety connection exists only between safety endpoints: the validator objects defined for CIP Safety sit in the safety controller and in the safe I/O block and supervise every message. Ordinary switches, gateways, IO-Link hubs and HMI links are not safety endpoints, so safety devices belong on safety channels wired to safe blocks.

Why does a monitored safety output need a feedback input?

So the safety logic can catch a welded contactor. The feedback input watches an auxiliary contact on each monitored switching device, letting the safety program compare the commanded state with the real state and latch a fault on mismatch. Two monitored contactors on one function can therefore cost two block channels beyond the function's own outputs.

What does an ordinary remote I/O station do when the network drops?

It enters its configured communications-loss state: outputs typically hold their last state or move to a defined value, inputs stop updating, and diagnostics flag the loss. Exact behavior is set per device. Machine safety must never depend on any of it — the risk assessment decides whether a loss of ordinary communications must itself trigger a safety stop.

Does adding a safety controller mean replacing the ordinary PLC?

No. The architecture keeps the ordinary controller for production logic and adds a separate safety-rated controller beside it. The legacy conversion layer is the part that goes: its interlocks, scaling and mapping move into the ordinary controller's program first. The only traffic between the two controllers is non-safety data such as mode, status and set points.

How fast must the safety network be?

Fast enough for the stopping time the risk assessment requires. Total reaction time is the sum of sensor, network, logic and output-device times, and it must stay below the time to hazard. A CIP Safety connection updates deterministically and is supervised by the safety connection's watchdog, so the network contributes a known, configured time rather than a random delay.

Keep reading

  • How-to

    Remote I/O commissioning checklist for a modular station

    Commission a modular remote I/O station in this order: match the coupler's protocol and addressing to the controller interface, budget and verify station power, seat modules and connectors against the drawing, then prove cyclic data exchange and every field signal before enabling outputs. A loading web page and lit status LEDs prove the coupler is alive; only a value that changes at the controller proves valid I/O data.

    Controllers and I/O

  • Selection guide

    How to select a CODESYS fieldbus master for industrial Ethernet

    Select the master around the device, not the connector: the 750-340 is a PROFINET IO coupler, and a PROFIBUS station needs a different coupler. Confirm three matches — device protocol, runtime master package, supported network adapter — then size the bus cycle from the process budget; the worked example clears a 10 ms sensor-to-actuator budget at a 2 ms cycle and fails it at 4 ms. Bench-test a laptop NIC only if the runtime documents it.

    Industrial networks

  • Selection guide

    Choosing PLC to SCADA communication over long distances

    For PLC-to-SCADA communication over long distance, choose the transport by reach and connection behavior: fiber Ethernet for a continuous long link (one span rated 2 km), RS-485 where serial fits, extenders or radio on existing pairs, dial-up or leased-line modems on phone service. Direct copper Ethernet fails at 100 m per segment; the SCADA driver or OPC service, not the PLC's ASCII modem commands, carries the protocol.

    Industrial networks