OPC UA vs Modbus TCP vs MQTT — choosing a protocol for a retrofit
The three protocols solve different problems and are usually used together. A practical decision guide for brownfield plants.
Insights
Each of these answers one question properly — the kind an automation engineer actually searches for at eleven at night. Written by the people who do the work, and published whether or not it leads to an enquiry.
In the pipeline
We would rather show the plan than pad this page with posts that were not written by an engineer. Two per month. If one of these is a question you need answered now, ask us directly and we will answer it before it is published.
The non-intrusive path from an installed PLC to a live dashboard: which protocols to read, how to size poll rates, and where the genuine limits are.
Availability, performance and quality — where the assumptions hide, and how to produce an OEE figure that survives a review meeting.
Current transformers, proximity counters and remote I/O: getting real state and real counts from machines that expose nothing.
Why plain MQTT is not enough, what birth and death certificates give you, and how state awareness changes alarm design.
A decision framework based on latency, bandwidth, availability and cost — with worked examples from vibration and OEE.
A full worked architecture: gateways, protocols, tag model, storage, retention and what it costs to run.
The boundary problem, the person in the middle with a spreadsheet, and how to design the interface properly.
Clamp-on CTs, meter selection, where to install, and how to attribute consumption to machine, shift and product.
Why most reason-code systems are abandoned within a month, and the design constraints that prevent it.
Proving functional equivalence, pre-testing off-line, staged cut-over, and keeping a real way back.
What the levels mean on a real shop floor, where the boundary device sits, and what your IT team will ask for.
An engineer will answer it, whether or not there is a project in it.