OPC UA vs Modbus: differences, trade-offs and when to use each
Part of the OPC UA guide
OPC UA and Modbus both move values between industrial devices, at very different levels of description. Modbus transfers numbered bits and 16-bit registers and leaves their meaning to a register map in a manual. OPC UA transfers named, typed nodes with a status and a timestamp, over a session that can be signed and encrypted. In the OPC UA vs Modbus choice, the device usually decides: use what it offers, and where a device offers both, the answer depends on how much the reader needs to know about the data without a document beside it.
This page is part of the OPC UA guide. The Modbus side is covered in depth in the Modbus guide.
OPC UA vs Modbus at a glance
| Modbus | OPC UA | |
|---|---|---|
| Specification | Modbus Organization, free | OPC Foundation, IEC 62541 |
| First published | 1979 | 2006 |
| Transports | RS-485 serial (RTU), TCP port 502 | TCP port 4840 (opc.tcp), HTTPS, WebSockets |
| Data model | Four tables of coils and registers | Address space of typed nodes and references |
| Address of a value | Table plus a 16-bit address, such as holding register 40108 | NodeId, such as ns=2;s=Temperature |
| Data types | Bit and 16-bit word; anything larger is a convention | Boolean, integers of 8 to 64 bits, Float, Double, String, DateTime, arrays, structures |
| Byte order | Big-endian per register; word order varies by vendor | Defined by the encoding |
| Quality and time | None | StatusCode and timestamps with every value |
| Discovery | None; a register map is required | Browse the address space; GetEndpoints |
| Getting changes | Client polls | Client polls, or subscribes with sampling interval and deadband |
| Security | None | Certificates, signing, encryption, user authentication |
| Connection setup | None on serial; a TCP connect on Ethernet | Hello, secure channel, session |
| Typical device cost | Runs on the smallest microcontrollers | Needs memory and processing for the stack and cryptography |
The data model is the real difference
To read a temperature over Modbus, the client has to know in advance that it is in holding register 108 of unit 17, that it is a signed 16-bit integer, and that it is in tenths of a degree. None of that is on the wire. The request and response for it on Modbus TCP are 12 and 11 bytes:
Request: 00 01 00 00 00 06 11 03 00 6B 00 01
Response: 00 01 00 00 00 05 11 03 02 00 EBThe value 0x00EB is 235. Whether that means 23.5 degrees, 235 rpm or a status word is a matter for the manual, and the register addressing rules (the off-by-one between register numbers and addresses, the word order of 32-bit values) are where most commissioning time goes.
Over OPC UA, the client can connect knowing only the endpoint URL. It browses from the Objects folder, finds a variable named Temperature, and reads its attributes: the DataType attribute says Float, the EngineeringUnits property (when the server provides it) says degrees Celsius, and the Value attribute returns this:
Value 23.5 (Float)
StatusCode Good (0x00000000)
SourceTimestamp 2026-11-21T14:03:07.250Z
ServerTimestamp 2026-11-21T14:03:07.262ZThe cost of that description is setup. Before the first Read, an OPC UA client has exchanged Hello and Acknowledge, opened a secure channel and created and activated a session. For a device read once a minute by one client, that overhead buys little. For a line with thousands of tags read by several systems, it replaces a register spreadsheet that would otherwise be maintained by hand in each of them.
Polling and subscriptions
Modbus is strictly request and response. A value that changes between two polls is missed, and the only way to see fast changes is to poll fast, which on a serial bus quickly runs into the scan time of the bus.
OPC UA allows the same polling with the Read service, and adds subscriptions. The server samples each monitored item at its own interval and reports only values that changed by more than a deadband. Traffic then follows the process: nothing while values are steady, a notification when something moves. The client and server page lists the subscription parameters.
Error reporting
A Modbus server that cannot answer returns one of a few exception codes, such as 02 (illegal data address), for the whole request. There is no way for it to say that a register was read successfully but the sensor behind it has failed; devices work around this with status registers or sentinel values such as 0x7FFF.
OPC UA reports a status per value. A read of fifty nodes can return forty-nine Good values and one BadNodeIdUnknown, and a variable whose sensor is disconnected can carry a Bad or Uncertain status while the node still exists. A client has to check those statuses for the benefit to be real.
Security
Modbus has no authentication and no encryption. Any host that can reach port 502 or the RS-485 pair can read and write. The protection is the network.
OPC UA authenticates applications with certificates, can sign and encrypt every message, and identifies users. That makes it the usual choice when data crosses from the control network to an IT network. It also means more to configure: certificates have to be trusted on both sides, they expire, and a server left with security mode None and anonymous access is no better protected than a Modbus device. The OPC UA security page covers the settings.
An OPC UA server in front of Modbus devices
The two protocols are more often layered than chosen between. Power meters, drives and small controllers speak Modbus because it is cheap to implement. A PLC or a software server polls them and exposes the results as OPC UA nodes with names, types and units applied once, at that server. Everything above it reads OPC UA.
Three points to check in that arrangement:
- The OPC UA status is only as good as the mapping. If the Modbus poll behind a node times out, the server should mark the node Bad. Some configurations keep serving the last value as Good.
- Timestamps come from the polling server, since Modbus carries none. The SourceTimestamp is the time of the poll that fetched the value.
- Update rate is bounded by the Modbus scan. A 100 ms sampling interval on an OPC UA monitored item returns nothing new if the server polls the RS-485 bus every two seconds.
OPC UA vs Modbus: when to use each
Read Modbus directly when the device offers nothing else, when there are few values and a register map exists, when the device is on an RS-485 run, or when no intermediate server is already in place. Adding a server only to translate a dozen registers adds a component that can fail.
Read OPC UA when the controller or an existing server already exposes it, when tags are named in the PLC program and change with it, when more than one system consumes the data, when the status and source time of each value matter, or when the connection leaves the control network and has to be authenticated.
On a mixed site the OPC UA vs Modbus question is answered device by device, and both end up in use: OPC UA to the PLCs and the plant’s existing servers, Modbus to the meters and packaged equipment nobody has mapped into a PLC.
Reading both with ControlCom
The ControlCom Edge Server, the on-site component of ControlCom Connect, offers both in the same Add New Client dialog: Modbus TCP, Modbus RTU and OPC UA. A Modbus tag is configured with an address, a data type, a scale and an offset. An OPC UA item is configured with a Node ID such as ns=2;s=Temperature and an Attribute ID that defaults to 13, and its data type follows the platform variable it is mapped to. Either way the result is a variable on the gateway, and a saved configuration takes effect on the Edge Server’s next restart.
The OPC UA client documentation and the Modbus client documentation list the fields for each, and the page for controls engineers describes the commissioning workflow.
PLC tags read over OPC UA and meter registers read over Modbus end up as variables in the same ControlCom Connect account, collected by one Edge Server.