What is OPC UA, and how do bioreactors use it?

OPC UA is the open, vendor-neutral standard that industrial equipment uses to expose its live data to other software. The letters stand for Open Platform Communications Unified Architecture. In a bioprocess suite, it is how a bioreactor controller publishes pH, dissolved oxygen, temperature, agitation, gas flows, pump rates, and events while the run is in progress. Any authorized client can read them, with no vendor-specific driver and no file export.

This explainer covers what OPC UA is, how it differs from OPC DA, MQTT, and Modbus, and which bioprocess equipment speaks it. It also covers how Invert uses OPC UA, alongside the other interfaces your systems expose, to bring every signal into one place as the run happens.

What OPC UA is

OPC UA is maintained by the OPC Foundation and standardized as IEC 62541. Four properties matter for bioprocess data.

• Server and client. The equipment side runs an OPC UA server. Sometimes it sits in the vessel controller; more often it sits in the vendor's control software (Sartorius BioPAT MFCS, Eppendorf DASware) or a DCS such as DeltaV. Whatever reads the data runs a client, which browses the server, reads values, or subscribes to changes.
• An information model, not a register list. Each value is a node carrying engineering units, a source timestamp, and a status code, and nodes are grouped into objects. A vessel shows up as a vessel, with its sensors, setpoints, and alarms attached.
• Security in the specification. X.509 certificates identify server and client, messages can be signed and encrypted, and users authenticate. The server sets the policy and listens on port 4840 by default, which is why your automation or IT group is part of every OPC UA connection.
• Platform independence. It runs on Windows, Linux, and embedded controllers, and supports both client-server and publish-subscribe communication. Publish-subscribe runs over UDP or MQTT.

Laboratory equipment has a companion specification, OPC UA for Laboratory and Analytical Devices, or LADS. Published in 2023 as OPC 30500, it standardizes how instruments describe themselves. Its worked example of a device split into functional units is a bioreactor vessel with two containers, each with a temperature sensor, a motor, and a program. Most equipment in service predates LADS and exposes a generic OPC UA server, which is enough to stream process values.

OPC UA vs OPC DA

OPC DA (Data Access), often shortened to DA, is the original OPC specification from the 1990s, now called OPC Classic. It is built on Microsoft's COM and DCOM, so it runs only on Windows, and configuring DCOM across firewalls and domains is a well-known source of pain. It exposes a flat tag list of current values, and security is whatever Windows provides.

OPC UA replaces all of that in one platform-independent specification and is the right choice for anything new. The complication in bioprocess is the installed base. As of 2026, Cytiva's UNICORN software, which runs ÄKTA systems and Xcellerex XDR bioreactors, still speaks only DA, and it is not alone. UA and DA cannot talk to each other directly, so a client either speaks both or sits behind a gateway.

OPC UA vs MQTT vs Modbus

These three come up together, but they answer different questions.

In practice the three are layered rather than chosen between. Modbus gives you a number at an address. MQTT moves messages without saying what they mean. OPC UA defines the meaning and the security, and it can publish over MQTT.

Which bioprocess equipment speaks OPC UA

Most current bioreactor platforms do, sometimes at the controller and sometimes only through the vendor's SCADA. Invert is a bioprocess data platform that ingests data from equipment, historians, ELNs, and LIMS and runs monitoring, analysis, and reporting on top of it. The table lists the systems it connects to and the protocol each exposes. The full catalog, including ELN, LIMS, MES, and analytical instruments, is on the integrations page.

When the server sits in the SCADA rather than in the vessel, you read what the SCADA has logged, at its sample rate. That is still live data.

OPC UA and the data historian

Bioprocess scientists rarely interact with OPC UA directly. Where a historian such as AVEVA PI, Canary, or DeltaV Continuous Historian sits between the equipment and the scientist, process data is typically collected continuously into the historian. It is then accessed downstream, often as an export after the run. Offline results arrive through a LIMS once the sample has been analyzed. In many labs there is no historian, so the path is a CSV pulled from the vendor software, a USB stick, or transcription.

A historian does its job well. It is the plant's long-retention record, it collects continuously, often over OPC itself, and its own trend tools show live values. The friction is not the historian. It is getting a complete, batch-labeled record of a run out of wherever the data landed and into the tool where the analysis happens. That is where the time goes, for the scientist who needs the data and for whoever gets asked to pull it.

• Batch context. Historian data is organized by tag, and a run is not a native object in it. Batch context can be added, for example with PI Event Frames or a batch historian. At many sites it is not configured, or it is only visible inside the vendor's own viewer. Someone then cuts the timeline into runs by hand, and ordinary mistakes creep in: one team's run zero-point was off by a day because pH logging began before inoculation. That cutting is the step that eats afternoons.
• Resolution. Storage settings decide what survives. Many historians compress or store by exception out of the box, and data handed over by a CDMO is often downsampled. An MSAT lead receiving bioreactor data from a CDMO put it this way: "I want to say it's like a five minute interval. So there's definitely some gaps, but I think that's just what their scan rate is." Whatever pH or DO did between those points is gone.
• Handling. Every export, transcription, or manual import is a place where units get dropped and timestamps shift. One fermentation group we spoke with assembles run comparisons by hand dozens of times a week. One upstream team spent two days on a single manual import.

Whichever path the data takes, the goal is the same: all of it, from every system, available while the run is still going, labeled by batch.

How Invert connects to bioprocess equipment

Invert connects to whatever interface a system already exposes. For bioreactor controllers and SCADA that is usually OPC UA or OPC DA. For a historian it is the PI Web API, SQL, or OPC UA. For analyzers and chromatography software it is the files they write, picked up by a file watcher. For ELN, LIMS, and MES it is their REST APIs. The connectors are pre-built for the common systems, so there is no protocol to choose and no middleware to build.

For live equipment data, your IT installs an edge connector on a machine that can reach the equipment network. That connector is the OPC UA client. It subscribes to the tags you select and streams them to Invert as the run proceeds. Where equipment speaks only OPC DA, or the data already sits in a PI historian, the same connector reads those interfaces instead. The stream does not know which batch it belongs to. Data assignment rules link it to the run by time window, so values land on the batch rather than in a tag pile. Alerts run on the live stream. Latency runs from about thirty seconds on Sartorius and Eppendorf controllers to about fifteen minutes on file-based paths such as UNICORN.

The connectors for Sartorius, Eppendorf, Distek, Cytiva, and the common historians are pre-built, so the connector work takes hours. The calendar time, typically about two weeks, is mostly IT: a host on the equipment network, outbound network access, and the OPC UA server trusting the client's certificate. On a locked-down manufacturing network that part can take longer. Two limits are worth knowing up front. The edge connector needs IT to install it and open access, and an instrument outside the connector catalog needs engineering work before it can stream.

One clinical-stage cell therapy company ran five bioreactor systems across two sites, from 2 L to 200 L, behind a historian that left scientists matching online data to batches by hand. It connected all five to Invert over OPC UA and OPC DA and kept batch definitions in its ELN; run monitoring now happens in Invert. Batch analysis prep went from days to near real time, and the company avoided up to $400K in integration and custom development costs. The case study has the details.

What this means for you

Every signal your equipment produces lands in one place, live, at the rate it was logged. Units and timestamps arrive intact, on the batch they belong to. Comparisons across runs, scales, and vendors start from tags mapped once to common metric names, which saves another round of reconciling metric names. Live view, alerts, and later trending in Invert run on the same data, alongside whatever your historian already does. OPC UA is one of the ways that data arrives. It is not something you have to operate.

If you want to see live data from your own bioreactors in Invert, book a demo, or browse the integrations catalog.