IO-Link is the first globally standardized point-to-point IO technology for sensors and actuators, codified as IEC 61131-9, with three defined communication speeds of 4.8 kBaud, 38.4 kBaud, and 230.4 kBaud over a standard unshielded three-wire cable up to 20 m [S2]. It is explicitly described as a further development of the traditional three-wire sensor connection, not a fieldbus [S2][S4].
Because IO-Link is fieldbus-independent, the question of "master versus gateway" actually describes two different layers of the same stack: the IO-Link master terminates the point-to-link and exposes the device to a higher-level network, while a fieldbus gateway translates that upstream traffic into PROFINET, EtherNet/IP, Modbus TCP, or OPC UA for the controller and the cloud [S2][S5].
What each device actually does in the stack
An IO-Link master physically terminates IO-Link ports on one side and exposes a standard fieldbus (PROFINET, EtherNet/IP, EtherCAT, Modbus TCP) on the other, acting as a converter between the IO-Link message format and the fieldbus telegram [S6]. The OPC UA specification for IO-Link (OPC 30120) confirms this by defining IOLinkMasterType, IOLinkPort, and IOLinkDeviceType node classes for the master, port, and downstream device [S4].
A protocol gateway, by contrast, sits between two heterogeneous fieldbus or Ethernet networks and is normally selected when the upstream controller, the downstream sensors, or both speak protocols that the IO-Link master cannot directly bridge, such as legacy PROFIBUS, serial Modbus RTU, or vendor-proprietary buses [S3][S5]. In practice the line is blurry: most modern masters ship with PROFINET or EtherNet/IP uplinks and are themselves gateway devices, so the decision is more about port count, IP67 mounting, and protocol coverage than about "master versus gateway" as separate boxes [S2][S6].
Selection criteria: when a master is the right call
An IO-Link master is the right call when the controller speaks a supported industrial Ethernet protocol, the cabinet or field layout is dominated by sub-20 m sensor runs, and the goal is replacing discrete and analog wiring with a single 3-wire drop carrying flow, temperature, and diagnostics on the same cable [S1][S2].
Three numbers anchor the master decision: 4.8/38.4/230.4 kBaud for the IO-Link link, up to 20 m standard cable length, and the IEC 61131-9 standard plus a 50+ million installed-node ecosystem across more than 500 manufacturers and 20,000 products [S2]. For greenfield machines and small cells (under 50 sensors, single controller, one fieldbus protocol), the master is almost always the lowest-risk path: IODD files let the PLC or configuration tool auto-recognise the device, automatic device replacement transfers parameters in seconds, and only one 3-wire drop is needed even when the sensor returns multiple process values [S1][S5].
Selection criteria: when a gateway is the right call
A fieldbus or protocol gateway is the right call when the controller is locked to a protocol the IO-Link master does not natively expose (legacy PROFIBUS DP, RS-485 Modbus RTU, CANopen, BACnet, IEC 61850), when the plant has hundreds of sensors from many vendors, or when the project requires routing diagnostics to a remote I/O module layer, an MES, or a cloud broker via OPC UA [S3][S5]. Bihl+Wiedemann's published architecture uses an ASi-5/ASi-3 PROFINET gateway (BWU3847) feeding an IP67 IO-Link master with 4 ports (BWU4067) to integrate IO-Link sensors into PROFINET cells without losing the ASi topology, illustrating the gateway-to-master cascade pattern [S5].
The threshold that drives the call to a dedicated gateway is around 50+ IO-Link nodes per line, multi-protocol plants, or the need to publish IO-Link diagnostics as an independent data stream to IT/OT systems: in that scenario, an edge gateway separates process data from diagnostic data and forwards the latter to OPC UA, while a protocol gateway handles the legacy fieldbus side [S3][S5]. If the plant has fewer than roughly 20 sensors and runs on one modern protocol, a multi-port IO-Link master is usually cheaper, faster to commission, and simpler to maintain than a separate gateway-and-IO-Link block.
Comparison: IO-Link master vs fieldbus gateway on four decision axes
On port density and field mounting, a master typically offers 4 or 8 IO-Link ports in IP67 M12 form, while a gateway tends to be a 35 mm DIN-rail device in the cabinet with one or two upstream fieldbus channels and serial or sub-bus ports on the field side [S2][S5].
On protocol coverage, a master natively exposes the supported fieldbus (PROFINET, EtherNet/IP, EtherCAT, Modbus TCP) on its uplink, but a gateway adds the legacy and exotic protocols the master does not, including PROFIBUS DP, Modbus RTU, CANopen, BACnet, and IEC 61850 [S3][S5]. On data granularity, the master gives full IODD-based access to parameters, diagnostics, and process data, while a generic gateway only forwards cyclic process data unless an OPC UA or IO-Link-specific companion specification is implemented [S4].
Limitations, failure modes, and standards to check
Three real constraints push projects off the pure-master path. First, cable length is capped at 20 m on standard unshielded 3-wire, so distributed plants with long trunk runs must insert a fiber converter or a sub-bus gateway to reach remote IO [S2]. Second, IO-Link is not a fieldbus: it does not carry process data from many devices on one wire, so any network that depends on multi-drop digital communication (PROFIBUS PA, Foundation Fieldbus, HART multidrop) still needs a gateway or wireless IO-Link extension to talk to those devices [S3][S4].
Third, safety requires the IO-Link Safety extension, which uses the same physical layer but rides on PROFIsafe, CIP Safety, or FSOE above the master, so functional-safety segments must be planned against IEC 61508/IEC 62061 SIL targets rather than against IO-Link alone [S2]. For the IODD and device-replace workflow, the OPC 30120 specification defines 17 named use cases, from UC.001 Configure IO-Link Master through UC.014 Faulty Device Replacement, which is the practical checklist to validate a vendor's "IODD + auto-replace" claims [S4].
Recommended deployment pattern in 2026
For a single machine or skid, specify an IP67 IO-Link master with PROFINET or EtherNet/IP uplink, 8 ports, and 4.8/38.4/230.4 kBaud support, then verify that the vendor publishes IODD files for every device and supports OPC UA for IO-Link per the OPC 30120 companion spec [S2][S4]. For a brownfield multi-line plant, keep cabinet-mounted industrial routers and protocol gateways between the legacy fieldbus and the new IO-Link islands, and make sure the gateway implements IODD upload and at least the UC.001-UC.014 use cases from OPC 30120 so commissioning and replacement remain scripted [S3][S4][S5].
Trackable signals for the next planning cycle: (1) the rollout of IO-Link Wireless for rotary and tool-changing applications, which extends the 20 m limit and may remove some gateway hops, and (2) the ongoing expansion of IO-Link Safety profiles into PROFIsafe, CIP Safety, and FSOE, which determines whether the master-only path is acceptable up to the target SIL [S2].