Back to all posts
Control Systems7 min read·October 25, 2026

How to Choose an Industrial Protocol Gateway

TB
ThynkBlox Automation Team
Automation Engineering

Choose an industrial protocol gateway by checking whether it preserves the data meanings and functions your application needs at both ends. Confirm the supported protocol roles, address mapping, byte order, diagnostics and failure behaviour before comparing price or connector compatibility. Make the purchase conditional on a written interface map and a demonstration against your acceptance tests.

What must the gateway actually do?

Start with the plant requirement: which information must cross the boundary, who needs it, and what action depends on it?

Separate monitoring from control. Reading a temperature for a dashboard needs a different acceptance test from writing a setpoint or issuing a machine command. For each signal, specify whether it is read, written or exchanged in both directions.

Record these requirements in the functional design specification. Give each value an owner, a meaning and a defined response when it becomes unavailable.

Ask suppliers to mark every requirement as supported, unsupported or conditional. A conditional answer should name the configuration, licence, firmware or external logic required.

Does it support the right protocol roles and functions?

Ask what role the gateway performs on each interface. Depending on the protocol, the documentation may describe roles such as client and server, controller and device, or publisher and subscriber. Require the supplier to identify the exact role needed for your connection.

Then check the operations within that role. Your selection checklist should cover:

  • Reading the required data areas and data types.
  • Writing only the approved values.
  • Supporting the required message or function types.
  • Handling the planned connections and communication schedule.
  • Exposing the status information the receiving system needs.

Do not accept “supports the protocol” as the complete answer. Ask for the relevant product documentation and a demonstration of your actual operations. If a function is unavailable, decide whether the resulting limitation is acceptable before ordering.

How will addresses and data meanings be mapped?

Prepare a signal map that both equipment owners can review. It should explain each value without requiring someone to reverse-engineer the gateway configuration.

Mapping itemWhat to recordWhat to verify
Source and destinationDevice, data area, address and tagThe intended signal reaches the intended destination
Address conventionHow addresses are written and countedThe first and last mapped points are correct
Data representationType, width and signednessNegative and boundary values retain their meaning
Engineering meaningUnits, scale and valid rangeThe displayed value matches the agreed physical quantity
AccessRead-only or authorised writeUnapproved destinations reject writes
Status and timeValidity, freshness and timestamp meaningThe receiver can judge whether to use the value

Include bit meanings and enumerated states. A destination value labelled “running” should correspond to the agreed source state, rather than whichever non-zero value happens to arrive.

Also define invalid values, reserved states and out-of-range readings. Require an explicit mapping or rejection rule for each. Avoid silently replacing an unknown state with a normal operating state.

How should you check byte order and multi-part values?

Ask for separate confirmation of byte order within each word and word order across a multi-word value. Record the source representation, the destination representation and the transformation applied between them.

Do not approve this mapping using one normal operating reading. Choose test values whose parts are distinguishable, then check the raw source data, gateway interpretation and destination result.

For multi-part values, also ask how the gateway obtains a consistent reading. Require the supplier to explain whether the application needs additional measures to prevent parts from different updates being treated as one value.

What would a worked acceptance example look like?

Illustrative example: a controller stores pressure as an unsigned integer in tenths of a bar. The source test value is 125, and the receiving display must show 12.5 bar.

The acceptance record should show:

  1. The agreed source address and raw value.
  2. The configured data type, byte order and scaling.
  3. The destination tag, value and unit.
  4. The validity indication while communication is healthy.
  5. The displayed value and validity indication after the source is disconnected.

Add tests for the application's valid limits and invalid states. If the real signal is signed, include a negative value. Treat every number in this example as illustrative, not as a recommended process setting.

What timing and write behaviour must be agreed?

Specify the maximum acceptable age of each important value, then ask how the proposed configuration meets it. Include source updates, gateway reads, destination updates and the planned communication load in that discussion.

For writes, define what success means. Distinguish a request accepted by the gateway from a value accepted by the destination and an action confirmed by the machine.

Agree who owns each writable value. If an operator screen and another system can both change it, specify which instruction takes priority and how conflicts are reported.

For commands, document acknowledgement, expiry and retry behaviour. Decide explicitly whether a command may be repeated after a timeout or reconnection. Include these cases in the demonstration rather than relying on a successful write during normal operation.

What should happen when communication fails?

Define the required behaviour for each failure case before choosing a product.

Failure caseDecision to make
Source device unavailableShould the last value remain visible, and how is it marked invalid?
Destination unavailableShould writes be rejected, discarded or retained under a defined rule?
Gateway restartWhat values and status appear before fresh communication is established?
Connection restoredWhich reads resume, and what happens to pending commands?
Invalid mapping or rejected requestWhere is the error shown, and who must respond?

Agree a stale-data limit and a visible validity indication. If the destination cannot receive quality information directly, specify an alternative status signal and how the application uses it.

For any safety-related function, have a qualified assessor define and assess the design, including communication-loss behaviour. Keep that assessment separate from proving ordinary data exchange.

What diagnostics will maintenance need?

Ask the supplier to demonstrate fault finding from the information the maintenance team will actually have.

The acceptance test should require the team to distinguish a disconnected device, an invalid address, a rejected write and a value that has stopped updating. Check which connection states, error details, counters and logs are available, and whether they identify the affected interface or signal.

Agree how communication alarms reach the operator and what action each alarm requires. The guide to alarm management and why operators ignore alarms can help frame that review.

Also require an exportable configuration, a documented restore procedure and clear access permissions. Test restoration using the handover instructions.

What common mistakes should you avoid?

Avoid buying from the connector list alone. Use connector and electrical compatibility as one selection check, then prove the required exchange.

Other mistakes to exclude from acceptance are:

  • Approving values without checking units, signedness and scaling.
  • Testing reads while leaving writes and acknowledgements untested.
  • Treating a retained value as current without checking freshness.
  • Leaving startup, timeout and reconnection behaviour unspecified.
  • Accepting diagnostics that require unavailable specialist access.
  • Finishing commissioning without a configuration backup and restore test.

A dedicated gateway should also have a clear purpose. If the existing systems already provide an acceptable connection, compare that option against adding another configured device to maintain.

What should you require before placing the order?

Request the completed signal map, supported-role evidence, configuration requirements and witnessed acceptance results. Include normal operation, invalid data, communication loss, restart and recovery.

Record limitations plainly. Make the decision against your agreed functions and failure behaviour, with any unresolved requirement visible to the people approving the purchase.

Frequently asked questions

Is matching the protocol name enough?

Require more evidence. Ask the supplier to confirm the roles, functions, data representations and status handling your application needs, then demonstrate the complete exchange.

Should the gateway hold the last value after a fault?

Decide per signal and use. If you retain it for reference, require the receiving application to identify it as stale or invalid and apply the agreed response.

Can one successful bench test justify purchase?

Use it as one part of acceptance. Also test representative data, the planned communication load, writes, faults and recovery against written criteria.

What should maintenance receive at handover?

Require the approved signal map, configuration backup, access arrangements, diagnostic instructions and a tested restore procedure. Include any operating limitations and the agreed response to communication faults.

Next step

Want to apply this to your plant?

Every project starts with a fixed-fee assessment, and you can stop after the report.