In standard Ethernet, you cannot guarantee when a critical frame will leave a switch; you can only measure the average. TSN (Time-Sensitive Networking) is a family of standards developed under IEEE 802.1 to close this gap: deterministic latency, bounded jitter, a common time base and redundancy through frame replication. It is not a single protocol. It is a toolkit of components, each solving a separate problem, chosen according to need. Below, for automation and network engineers planning a factory network investment, we separate the components by function and set out what to look for in today’s switch and controller decisions.
Why standard Ethernet is not deterministic
A classic switch receives a frame and places it in the egress port’s queue. If another frame is being transmitted on the egress port at that moment, the new frame waits whatever its priority; a transmission that has started is not interrupted. Putting critical traffic in a high-priority queue with the 802.1Q priority tag (PCP) reduces the waiting time. It does not eliminate it.
Example calculation: on a 100 Mbit/s port, the largest standard frame (1522 bytes with a VLAN tag), together with the preamble and inter-frame gap, occupies the line for the time of about 1542 bytes; this is roughly 123 µs. A critical frame can wait this long at every switch hop. On a line with ten switches, this effect alone produces a worst-case variability of more than a millisecond. At 1 Gbit/s the same time drops to about 12.3 µs; the problem does not go away, it only shrinks in scale.
The second gap is time. Every device has its own clock, and applications requiring timestamps, motion synchronisation or scheduled transmission do not work without a common time base. Industrial protocols have been solving these problems with their own methods for years; TSN moves the solution to IEEE level, into vendor-independent standards.
The TSN standards family and the function of each
TSN includes many standards and amendments. A significant number of these were incorporated into later revisions of the main 802.1Q standard after publication; that is why product documentation shows both amendment names such as “802.1Qbv” and a plain “802.1Q” reference. The most relevant for the factory are:
| Standard | Function | Problem it solves |
|---|---|---|
| 802.1AS | Time synchronisation (gPTP) | Common time base |
| 802.1Qbv | Time-based scheduling (time-aware shaper) | A transmission window reserved for critical traffic |
| 802.1Qbu + 802.3br | Frame preemption | A large frame holding up critical traffic |
| 802.1CB | Frame replication and elimination (FRER) | A copy reaching its destination via the other path when a link fails |
| 802.1Qci | Per-stream filtering and policing | A faulty device flooding the network |
| 802.1Qcc | Stream reservation and centralised configuration models | Distributing the schedule to the network |
“TSN-capable” in a datasheet does not mean all these standards are supported. The phrase “TSN-ready” does not provide an end-to-end timing guarantee either; the guarantee depends on all devices on the path supporting the relevant functions, being time-synchronised and the schedule loaded onto the network. In quotes and specifications, ask one by one which standard is supported, on which port and at which speed; get the answer in writing.
Time synchronisation with 802.1AS
802.1AS is the profile of the IEEE 1588 Precision Time Protocol defined for bridged networks and is known as gPTP (generalized PTP). A “grandmaster” clock on the network becomes the reference; the Best Master Clock Algorithm (BMCA) selects this clock. Peer delay is measured on every link, and each bridge adds the time the frame spends inside it to the correction field. As a result, time is distributed with high accuracy regardless of the number of switches in between.
The practical impact is two-fold. Applications (motion control, event timestamps, simultaneous measurement) gain a common clock. More importantly, time-based scheduling mechanisms such as 802.1Qbv depend on this clock. Without synchronisation, there is no gate schedule either. That is why 802.1AS is the first layer to verify during commissioning: grandmaster selection, delay measurement on each port and clock offset logs are checked before the scheduling settings. For the effect of time synchronisation on production data quality, the article timestamps and data quality in production data is a useful complement.
Time-based scheduling and preemption
802.1Qbv: the gate control list
802.1Qbv places a “gate” in front of each traffic class queue on every egress port. The gate control list defines, on a cyclic timetable, which queue may transmit in which interval. During the window reserved for critical traffic, the gates of the other queues are closed; the critical frame meets an empty line.
To prevent a large frame from starting transmission just before the critical window and overrunning into it, a “guard band” is needed. This band lasts as long as the transmission time of the largest frame, and the line waits idle for that time. According to the example calculation above, at 100 Mbit/s this is a loss of about 123 µs; as the cycle time shortens, the proportional loss of bandwidth increases.
802.1Qbu and 802.3br: preemption
Preemption reduces the guard band loss. 802.3br splits the MAC layer into “express” and “preemptable”. If an express frame arrives while a preemptable frame is being transmitted, the large frame is fragmented, the express frame is sent, and then the large frame resumes where it left off. 802.1Qbu, on the bridge side, manages which traffic classes are express and which are preemptable. For preemption, both ends of the link must support this feature; having it enabled at one end only gains nothing.
Redundancy and protection: 802.1CB and 802.1Qci
802.1CB replicates the frames of a stream at a point close to the source and sends them over disjoint paths, then eliminates the duplicates by sequence number at a point close to the destination. When one path breaks, the other copy is already on its way; there is no recovery time to wait for. Logically it is similar to IEC 62439-3 HSR/PRP, but it can be applied per stream and at any point in a bridged network. For a comparison of seamless redundancy options, see the article zero-interruption redundancy with HSR and PRP.
802.1Qci performs per-stream filtering and rate limiting on the ingress port. When a misconfigured or faulty device sends traffic exceeding its scheduled window, this traffic is stopped before it spreads into the network. Anyone who has seen a device start a broadcast storm in the field because of a firmware bug knows the value of this feature. Protection is needed as much as scheduling.
802.1Qcc and the IEC/IEEE 60802 industrial profile
The time schedule is not something to be written into each switch by hand. 802.1Qcc defines three configuration models: fully distributed, centralised network/distributed user, and fully centralised. In the fully centralised model, the CUC (Centralized User Configuration) collects application requirements, and the CNC (Centralized Network Configuration) calculates paths and gate schedules and distributes them to the bridges. Centralised models come to the fore in industrial applications. In other words, the engineering tool and network management software are now a link in the TSN chain; the version of that software also goes into the project documentation.
When there are many options, there is a risk that different manufacturers choose mutually incompatible subsets. IEC/IEEE 60802 is the joint IEC and IEEE standard defining TSN profiles for industrial automation; according to the IEEE 802.1 project page, it was published as IEC/IEEE 60802-2026 on 29 June 2026. The profile specifies which TSN features, options and defaults are to be used for bridges and end stations. Verify the profile’s current publication status and version for project decisions on the IEEE 802.1 TSN task group page; if a product document says “60802 compliant”, ask which draft or version it refers to.
The relationship between TSN, PROFINET IRT and EtherNet/IP
Existing industrial protocols have already solved some of the problems TSN addresses with their own methods:
| Approach | Time synchronization | Deterministic transmission | Hardware requirement |
|---|---|---|---|
| PROFINET RT | Not required | With VLAN priority | Standard switches |
| PROFINET IRT | PTCP | Reserved time phase | IRT-capable ASIC |
| EtherNet/IP + CIP Sync | IEEE 1588 | Priority and QoS | 1588-capable switches recommended |
| TSN-based protocols | 802.1AS | 802.1Qbv/Qbu | TSN-capable switches |
PROFINET IRT achieves a similar result by dividing the cycle into reserved and open time phases, but it does so with dedicated IRT-capable hardware. PROFIBUS & PROFINET International (PI) is carrying out specification work for running PROFINET over TSN; ODVA is also working on TSN integration for EtherNet/IP. The OPC Foundation’s OPC UA FX work envisages TSN as the transport layer for controller-to-controller communication. Check on the relevant organisation’s website which specification version has been published and which devices are certified; a roadmap in a brochure is not a certificate.
The practical conclusion is this: a PROFINET RT or EtherNet/IP line running today does not have to be redesigned to move to TSN. TSN’s value shows when different protocols and IT traffic share the same physical network with guaranteed behaviour. In existing PROFINET networks, the main issue is still the right topology and redundancy; for this, see the article ring topology with PROFINET MRP.
Criteria for TSN readiness in today’s investments
For a plant investing in switches, controllers or cabling today, both extremes are wrong: ignoring TSN completely, and basing every decision on it. A checklist for a balanced approach:
- Backbone speed: Choose gigabit or faster for new cell and backbone ports; guard band and scheduling flexibility are directly related to speed.
- Hardware readiness: Ask in writing whether the switch has hardware support for 802.1AS, 802.1Qbv and 802.3br, and whether the features will be enabled later in software.
- Time synchronisation: If you have applications that need IEEE 1588 today, plan the clock architecture (grandmaster location, redundancy) now.
- Management model: Ask whether standard management interfaces such as NETCONF/YANG are supported for centralised configuration.
- Protocol roadmap: Find out your controller manufacturer’s product roadmap for TSN-based PROFINET, EtherNet/IP or OPC UA FX.
- Mixed operation: Verify in a pilot cell that TSN switches carry existing RT and IT traffic without problems.
Make TSN a priority investment item only if there is a concrete requirement, such as shared infrastructure or deterministic communication between controllers. If there is no requirement, readiness is enough. For an assessment of your existing network in this respect and for cell design, see our industrial Ethernet and factory networks and PROFINET and PROFIBUS communication service pages.
Frequently asked questions
Will TSN replace PROFINET or EtherNet/IP?
No. TSN is a family of standards providing timing, synchronisation and redundancy mechanisms at Layer 2; it is not an application protocol. Protocols such as PROFINET, EtherNet/IP and OPC UA FX can use TSN as a transport layer, but the device model, configuration and diagnostic functions still come from these protocols’ own definitions.
Do existing switches become TSN-capable with a software update?
Usually not. 802.1Qbv gate scheduling and 802.3br preemption require timestamping and queue gate support in the port hardware. In some products the hardware is ready and the features are enabled later in software; confirm this in writing with the manufacturer for each product.
What is the difference between 802.1AS and IEEE 1588 PTP?
802.1AS is the profile of IEEE 1588 defined for bridged networks (gPTP). It runs at Layer 2 and requires all bridges to be time-aware and to measure peer-to-peer delay. Its behaviour is therefore more constrained and predictable; general 1588 profiles leave more options open.
Is upgrading the network to 1 Gbit/s mandatory for TSN?
The standards do not mandate a specific speed. However, because the guard band and frame transmission time shorten as speed increases, time scheduling gains flexibility: at 1 Gbit/s, the time the largest frame occupies the line is ten times shorter than at 100 Mbit/s. Choosing gigabit ports in new backbone investments therefore makes sense.