TSN vs 5G URLLC: Deterministic Networking on the Factory Floor (2026)

The first time you watch a 250 microsecond motion-control cycle wobble by 80 microseconds because a Windows update started chatting on the same VLAN, you become a believer in deterministic networking. TSN vs 5G URLLC deterministic networking has gone from a slide-deck debate to a procurement question this year, because both standards finally have shipping silicon, real integration profiles, and reference deployments you can actually visit. The trouble is that most “TSN vs 5G” comparisons online are written by people selling exactly one of the two. This post is not. It is an engineering walk-through from someone who has helped commission both technologies on live brownfield lines, including the parts where the numbers in the datasheet did not survive contact with the plant. What this post covers: the comparison matrix, how each stack actually works, and when to pick which. We also cover a hybrid TSN-over-5G pattern more common than vendors admit, plus the gotchas that bite in year two.

Context: Where Deterministic Networking Matters Now

Answer-first summary: Deterministic networking matters where a late frame causes a physical event to go wrong — motion control, machine vision triggers, safety interlocks, and synchronized multi-axis drives. Best-effort Ethernet handles the office. The factory cell, the surgical robot, and the substation need bounded jitter and guaranteed delivery, and that is what TSN and 5G URLLC are designed to provide.

Industrial Ethernet has been with us for two decades. PROFINET IRT and EtherCAT solved the determinism problem inside a vendor silo, but they did it with proprietary scheduling or by hijacking the Ethernet frame. The pain showed up the moment a plant tried to mix vendors, push OT data into a unified namespace, or extend the cell across a building.

Two parallel efforts have changed the picture. IEEE 802.1 standardized Time-Sensitive Networking as a set of amendments to standard Ethernet — same MAC, same cables, but with gate scheduling, time sync, and frame replication baked into the bridges. 3GPP, in parallel, rebuilt the radio stack to chase the same goal wirelessly: URLLC service in Release 15, native TSN integration in Release 16, and tighter precision in Release 17 and 18.

Two forces accelerated this convergence. The pandemic-era push for remote and reconfigurable manufacturing exposed how brittle vendor-specific deterministic stacks were when a line had to be reorganized on a quarter’s notice. At the same time, regulators in Germany, Japan, the United Kingdom, and the United States carved out shared or licensed industrial spectrum bands, removing one of the historic barriers to private 5G adoption. The combined effect is that buyers now have a real choice rather than a vendor-imposed one.

By 2026 both standards have crossed the “demonstrable at scale” threshold. The German Plattform Industrie 4.0 reference architecture mentions both as peer transports. The IEC/IEEE 60802 industrial profile finally pins down which TSN features must be present in a “TSN-Industrial” switch, so a buyer can stop guessing. Meanwhile private 5G has dropped from a multi-million-euro custom build to a turnkey cell that fits in a small rack. Silicon for industrial 5G UEs is now a normal line item from several module vendors. The decision is no longer “if” — it is “which, where, and why.”

TSN vs 5G URLLC: The Side-by-Side Decision Matrix

Answer-first summary: TSN delivers single-digit microsecond determinism over wires with a low integration cost on brownfield lines. 5G URLLC trades a little jitter and a higher integration cost for mobility and cable-free deployment. The matrix below summarizes the eight criteria that drive most procurement decisions in 2026.

The numbers here are conservative — quoted from the published standards and from independent measurement studies, not from vendor brochures. Treat the bounds as planning envelopes, not guarantees.

Criterion	TSN (IEEE 802.1 over Ethernet)	5G URLLC (3GPP Rel-17/18, Standalone)
Worst-case jitter (one hop, scheduled)	typically sub-microsecond with 802.1Qbv	typically tens of microseconds, slice-dependent
End-to-end latency (cell-scale)	hundreds of nanoseconds to low microseconds per hop	targets in the 1–10 ms band; sub-ms in tuned URLLC slices
Max nodes per cell (practical)	hundreds, switch-bound	tens, scheduler- and spectrum-bound
Packet loss / reliability target	1e-9 with 802.1CB FRER seamless redundancy	3GPP target up to 1e-6 for URLLC, higher with PDCP duplication
Integration cost on brownfield	medium — drop-in switches, plus tooling	high — RAN, core, spectrum, OT/IT bridge
Brownfield friendliness	high — same Ethernet cabling, profiles map to PROFINET/CIP	medium — needs UEs on every device or a TSN-to-5G bridge
Time-sync source	gPTP per 802.1AS-Rev, sub-microsecond	5G system clock, gated to TSN via DS-TT / NW-TT
Vendor maturity (2026)	mature — Cisco, Siemens, Hirschmann, Moxa shipping IEC/IEEE 60802 profile	maturing — Ericsson, Nokia, Huawei, plus private-5G specialists

Two takeaways. First, the columns are not interchangeable; TSN and URLLC optimize for different physical realities. Second, the cost gap is real today. A TSN refresh for a single cell is typically a low six-figure capex line; a comparable URLLC private cell is multiples of that once you include spectrum, integration, and the OT/IT bridge. That gap will narrow, not disappear.

How TSN Works

Answer-first summary: TSN is a family of IEEE 802.1 amendments that turn standard Ethernet into a deterministic transport. The four load-bearing pieces are 802.1AS-Rev for time, 802.1Qbv for scheduling, 802.1CB for redundancy, and 802.1Qci for ingress policing. Together they bound jitter, guarantee bandwidth, and survive single link failures.

Time Synchronization with 802.1AS-Rev

Everything in TSN starts with a shared clock. 802.1AS-Rev is a profile of IEEE 1588 PTP, tuned for bridged LANs, that delivers sub-microsecond synchronization across a switch tree. A grandmaster — usually a hardened time server or a TSN switch with a GNSS input — distributes time via the Best Master Clock Algorithm. Every bridge timestamps incoming sync messages in hardware, calculates link delay, and corrects local clock skew. Without 802.1AS-Rev, the rest of the stack does not work, because gate schedules and stream offsets are all expressed in absolute time.

Scheduled Traffic with 802.1Qbv

This is the centerpiece. 802.1Qbv defines a Time-Aware Shaper, in which each egress port has a gate per traffic class. A pre-computed schedule — the Gate Control List — opens and closes those gates at offsets within a repeating cycle, often 31.25, 62.5, or 125 microseconds. During a scheduled window only the chosen traffic class can transmit, so cyclic motion frames see no contention. Best-effort traffic uses the other windows. Calculating the schedule is the hard part; it is an offline optimization problem usually solved by a central network configurator and pushed via 802.1Qcc Stream Reservation Protocol.

Reliability with 802.1CB and Policing with 802.1Qci

802.1CB Frame Replication and Elimination for Reliability (FRER) duplicates critical streams across disjoint paths, tags each frame with a sequence number, and eliminates duplicates at a downstream merge point. A failed link does not drop a single frame — that is what makes TSN viable for safety-rated applications. 802.1Qci adds per-stream filtering and policing at ingress, so a misbehaving device cannot starve the schedule. Together they turn TSN from a “best-jitter” play into a fault-tolerant one.

Brownfield Migration: TSN as a Superset

A common misunderstanding is that adopting TSN means ripping out PROFINET or CIP. It does not. TSN sits below those protocols, in the bridging layer, and the relevant industrial profiles — PROFINET over TSN and CIP over TSN — are defined so existing controllers, drives, and engineering tools continue to work. The path most large brownfield estates take is gradual. Replace switches first. Run mixed standard-Ethernet and TSN traffic side by side. Then enable scheduling once the configurator tooling is trusted. The conformance label that matters during this migration is IEC/IEEE 60802. A TSN switch without 60802 conformance might still pass a Qbv test but fail interoperability with another vendor’s gPTP grandmaster or stream reservation tool. That kind of silent mismatch is the single most expensive surprise we see in year-one TSN deployments. Plan a multi-vendor interoperability test before commercial commitment, not after.

How 5G URLLC Works

Answer-first summary: 5G URLLC is a service category in 3GPP standards that targets 1 ms latency at 99.999% reliability for short payloads. It depends on mini-slot scheduling on the radio, PDCP duplication, a Standalone core, edge-deployed UPFs, and network slicing to isolate the URLLC traffic from best-effort users sharing the same gNodeB.

Mini-Slots and Short Transmission Time Intervals

5G New Radio inherits OFDM from LTE, but URLLC adds two critical changes. The Transmission Time Interval shrinks from one millisecond to as little as a few symbols — the “mini-slot” — so a frame can hit the air much faster after generation. The scheduler is also work-preserving with preemption, meaning an URLLC packet can interrupt an in-flight eMBB transmission. The cost is wasted radio resources, which is why URLLC is a deliberate trade rather than a free upgrade.

PDCP Duplication and HARQ Tuning

The Packet Data Convergence Protocol layer can duplicate URLLC packets across two carriers or two cell legs, so a single fading event does not lose the frame. HARQ retransmission targets are tightened, and the modulation and coding scheme is biased toward robustness — typically a lower MCS than an eMBB user would tolerate. The net effect is the 1e-6 block-error rate target that headline URLLC specifications quote.

Standalone Core, Edge UPF, and Network Slicing

URLLC presumes a 5G Standalone core; the non-standalone EPC fallback cannot deliver the latency budget. The User Plane Function is pushed to the edge — typically into a multi-access edge computing node inside the plant — so packets never traverse a public backhaul. A dedicated network slice, identified by an S-NSSAI, carries QoS policy from the core down to the radio scheduler. Without slicing, URLLC SLAs cannot survive a busy stadium of best-effort UEs on the same radio.

TSN Integration: DS-TT and NW-TT

3GPP Release 16 introduced TSN translators at each end of the 5G system — the Device-Side TSN Translator and the Network-Side TSN Translator. A TSN Application Function in the 5G core programs these translators with stream identifiers, schedules, and time offsets, so the entire radio network looks like a single logical TSN bridge to the wired estate. Release 18 tightens this further with improved hold-and-forward semantics and survival-time-aware HARQ.

What the Lab Numbers Hide

Published URLLC numbers — the “1 ms at 99.999%” line that appears in every 5G slide — come from controlled test conditions. Real plants change the environment in three ways the lab does not. First, metal everywhere: drives, racks, ducting, all of which create multipath that the channel estimator has to track in real time. Second, electromagnetic noise from variable-frequency drives and welding equipment can degrade signal quality unpredictably in industrial bands. Third, occupancy is dynamic: a single shift change can multiply the active UE count, and a poorly tuned scheduler will quietly burn URLLC budget on retransmissions. The fix is rigorous site survey, a tightly bounded slice configuration, and end-to-end jitter measurement on the actual production payload. Headline numbers are necessary but not sufficient evidence during procurement.

When to Choose Which

Answer-first summary: Choose TSN when devices are fixed, cycle times are below one millisecond, and the brownfield is PROFINET, EtherCAT, or CIP. Choose 5G URLLC when devices rotate, move on AGVs, or live in dangerous zones where cables fail. Choose both when you need cross-cell reach without trenching new cable.

The factory-floor decision is rarely “TSN or URLLC” in isolation. It is “which deterministic transport for this workload, given everything else I already run.” A motion-control loop on a fixed press is not the same problem as a teleoperated forklift across a forty-thousand-square-meter warehouse.

TSN-First Workloads

Sub-millisecond motion control on fixed equipment. Multi-axis drives, presses, robotic arms anchored to a frame, and synchronized servo loops. The cable is already there; jitter must be tight; the schedule is mostly static.
Large device counts per cell. Modern paint lines, large assembly carriers, or print bars can host hundreds of IO modules. TSN switches scale to this; a single gNodeB usually cannot serve hundreds of cyclic-traffic UEs without contention.
Existing industrial-Ethernet investments. When you already own PROFINET, you keep your engineering tooling, your safety story, and your spare-parts kit by migrating to TSN. The IEC/IEEE 60802 profile is explicitly designed to be a superset that those protocols can ride on.

5G URLLC-First Workloads

Mobile or rotating equipment. AGVs, AMRs, swing arms, rotating tables, overhead cranes, and reconfigurable cells. Trailing cables are a maintenance nightmare; slip rings are worse.
Greenfield with no wired plan. Logistics yards, outdoor mining sites, harbors. The capex to trench fiber dwarfs a private-5G deployment.
Sites where IT/OT convergence is a stated strategy. A 5G slice carries plant traffic and corporate WiFi backhaul on the same physical network, governed by policies, with proper QoS isolation. Some organizations want exactly that.
Hazardous-zone applications. Petrochemical, mining, automotive paint booths. Cabling in Atex/IECEx zones is expensive and dangerous to maintain.

When to Re-Scope Instead

If your cycle time is one to ten milliseconds and devices are mostly fixed, be honest. With no in-house networking team, a well-engineered industrial Ethernet with QoS and PRP-RSTP may serve you for another five years. Buying TSN or URLLC for a problem you do not have is the most common deterministic-networking mistake of 2024–2026.

Hybrid TSN-over-5G Patterns

Answer-first summary: The most interesting deployments are not pure plays. Hybrid TSN-over-5G uses the wireless system as a TSN bridge — wired TSN inside each cell, 5G linking cells across the plant. The pattern is endorsed by 3GPP Release 16 and refined in Release 18. It is how mobile robotics and large-footprint plants will likely be networked by 2028.

In this model each TSN island runs at sub-microsecond determinism internally — that is where the press lives, where the vision system lives, where you need 802.1Qbv windows. Between islands, the 5G system carries TSN streams using the DS-TT and NW-TT translators. The TSN Application Function in the 5G core gets the same Stream Reservation Protocol input as a wired TSN backbone would, programs the radio scheduler accordingly, and reports back its own residence-time bounds.

There are three practical reasons to prefer the hybrid pattern:

Cost. You only spend 5G capex on the parts of the plant that actually move; the static cells stay on cheap copper.
Risk. Wired TSN is a more mature technology in 2026; you concentrate it where the consequences of jitter are highest.
Phased rollout. A line manager can deploy TSN this year, add 5G island links next year, and only retire legacy switches when budget allows. Both technologies share gPTP-compatible time, so the schedule stays consistent.

A fourth, less-discussed pattern is using 5G URLLC as the redundancy partner to a wired TSN backbone rather than its replacement. 802.1CB FRER will happily duplicate a critical stream across two disjoint paths; one of those paths can be a TSN-bridged 5G slice. When a forklift severs a cable run during shift change — which happens — the second leg survives. The same idea works in reverse: a 5G-primary cell can keep a thin wired TSN drop as a hot standby for the most safety-relevant streams. Either pattern is more economical than provisioning a fully duplicated wired plant and is materially more reliable than betting everything on a single transport.

The catch is configuration complexity. Stream reservations have to be translated at the boundary. Time sync must be carefully designed so neither side picks the other as a worse-quality grandmaster. Observability tooling has to handle two very different telemetry stacks. Vendors are converging on a single network configurator, but the joint operations story is still being written.

Trade-offs and Gotchas

Answer-first summary: Both standards have failure modes that are invisible in the spec sheet — schedule fragility, electromagnetic interference, spectrum politics, and silent profile mismatches. A serious comparison has to call these out, because they dominate real-world cost in year two.

TSN Schedule Fragility

802.1Qbv schedules are computed offline. Add one new vibration sensor to a fully scheduled cell and the central network configurator has to recompute and push gate lists to every bridge in the path. Some configurators do this gracefully; others trip on edge cases like asymmetric link speeds or differing cycle-time domains. Without disciplined CI for network configurations, the schedule decays as field changes accumulate. Treat it like infrastructure-as-code.

URLLC Spectrum Reality

URLLC numbers assume contiguous licensed spectrum, low cell load, and no neighboring interferers. Inside a factory you frequently get none of those for free. National regulators have carved out shared bands in many countries, but the bands and licensing rules differ — n78 in much of Europe, n79 in some Asian deployments, CBRS in the United States. Site survey results in 2024 measurements occasionally show URLLC jitter doubling under bursty eMBB load, even with slicing, if the gNodeB scheduler is not properly tuned.

Profile Mismatches

A switch advertised as “TSN-capable” may support 802.1Qbv but not 802.1CB or 802.1AS-Rev. A 5G chipset advertised as “URLLC-ready” may support short TTIs but not PDCP duplication. The IEC/IEEE 60802 profile is the test that matters on the wired side; on the wireless side, look for explicit 3GPP Release version compliance and a working DS-TT integration test report. Datasheet bingo is not procurement.

Cybersecurity and the Bigger Picture

Determinism is one of three legs of an industrial network — the other two are security and interoperability. A flawless schedule is no defense against a compromised PLC. Both stacks need a credible security overlay, segmentation strategy, and audit story. The OPC UA FX field-level communications work is one of the most thoughtful answers to the upper-layer half of this problem and is worth reading alongside this post.

Vendor Lock and the “Single-Pane-of-Glass” Mirage

Both technology stacks are sold with promises of unified management. In practice 2026 deployments still run two parallel tool chains — one for TSN switches, one for the 5G network — even when buying from a single vendor. Plan for two operations consoles, two upgrade paths, and two on-call rotations. Honest planning beats optimistic procurement.

The Skills Question

Both standards require skills that most plant networking teams have not had to carry before. TSN demands disciplined network configuration management, including offline schedule computation, central network configurator tooling, and a willingness to treat gate control lists as code. URLLC demands a familiarity with cellular concepts — slicing, RAN sharing, QoS flows, PDU sessions — that has historically lived in telecom carriers, not in factories. The talent gap is real and it shows up as project delays. Nearly every deployment that does not budget for training, a systems integrator with bench depth, or a managed-service contract for two years feels this pain. Treat the skills plan as a deliverable, not an afterthought, and revisit it at every architecture review.

Practical Recommendations

Answer-first summary: Start with the cycle-time and mobility requirement, then check brownfield estate, then decide. Pilot one cell before you industrialize. Insist on IEC/IEEE 60802 conformance for TSN gear and explicit Release-17-or-later compliance for 5G. Measure jitter end-to-end, not at vendor demo conditions.

A short checklist used on real engagements:

Inventory your cycle-time tiers. Hard real-time at sub-millisecond, firm real-time at one to ten milliseconds, soft real-time above that. Do not buy determinism for the soft tier.
Map mobility honestly. A device that “could move” but never does is fixed equipment. Cabling is cheap; deploying URLLC for imagined mobility is not.
Demand conformance. TSN gear without IEC/IEEE 60802 conformance is a procurement risk. 5G UEs without an explicit 3GPP Release tag are a worse one.
Pilot one cell. Burn three months on a single live cell with real workloads before extending. Both technologies expose surprises only under genuine traffic.
Buy the observability before the bridges. Tooling for TSN stream visibility and for URLLC slice telemetry is the long pole on incident response in year two.
Plan the time-sync architecture early. Both stacks reduce to “is the clock right?” when things break. A GNSS-disciplined grandmaster, redundancy plan, and holdover spec belong in the first design review.
Document the schedule. Treat 802.1Qbv gate lists and URLLC slice configs as production artifacts in version control, peer-reviewed, with a rollback path.

Buying TSN or URLLC Equipment: A Procurement Lens

A purchase order should ask three questions that vendor brochures rarely volunteer. First, what is the worst-case jitter under a stated load profile, measured on real hardware, not in a simulator? Second, which exact IEC/IEEE 60802 features and which 3GPP Release are conformant, and is there an independent test report? Third, what is the upgrade story when the standards revise — both stacks are still evolving, and a switch or radio bought today should not be a stranded asset in three years. Ask for a written commitment on firmware support windows. The cheapest gear is rarely the cheapest gear after factoring in the cost of a forced refresh during a future expansion phase.

For a broader view of how these transports plug into the higher-level data plane, the DDS, MQTT, and OPC UA messaging-protocol comparison covers the layer above. The Sparkplug B vs OPC UA pub/sub comparison gets into the application-layer trade-offs that often shape transport choice.

FAQ

Answer-first summary: These are the six questions we hear most often in deterministic networking workshops in 2026. They reflect the friction points that procurement, architecture, and operations teams hit at different stages of a TSN or 5G URLLC project, and the answers below distill the engineering judgement, not the marketing position.

Is TSN faster than 5G URLLC?

For point-to-point latency, yes — a single TSN hop typically delivers sub-microsecond jitter and microsecond-scale latency, while URLLC targets the one-millisecond band even in well-tuned slices. But “faster” is the wrong question. TSN wins on raw determinism for fixed equipment; URLLC wins where mobility matters. Comparing them on latency alone is like comparing a freight train to a delivery drone on top speed — the contexts of use are different, and so the right answer depends on workload.

Do I still need TSN if I have a private 5G network?

Often yes, inside each cell. Most production deployments use TSN for the wired backbone within a cell — drives, vision, IO — and use 5G URLLC for between-cell links or for mobile devices like AGVs. The hybrid pattern is described in 3GPP Release 16 and refined in Release 18, and it sidesteps the need to put a 5G modem on every PLC. Treat the two technologies as complementary unless a strict greenfield brief says otherwise.

What is the IEC/IEEE 60802 profile and why does it matter?

IEC/IEEE 60802 is the industrial TSN profile that pins down which subset of TSN features a switch or end station must implement to be safely used in industrial automation. It picks specific values for cycle times, sync accuracy, and security baseline behavior. Without it, two vendors could both honestly claim “TSN support” and still fail to interoperate. In 2026 it is the conformance label to look for during procurement and is a standard reference in IEC 61784 industrial communications profiles.

Can 5G URLLC handle motion control at 250 microseconds?

In tightly engineered cells with high-end 5G silicon, recent measurements have shown sub-millisecond latencies and low single-digit-microsecond jitter under controlled conditions. But guaranteeing 250 microseconds end-to-end across radio plus core plus translator, day after day, is still at the edge of what shipping Release-17 systems do reliably. Release 18 narrows this gap. For 2026 procurement, plan motion control on TSN unless your specific vendor has an independently verified URLLC reference for the cycle time you need.

Does TSN replace PROFINET IRT or EtherCAT?

Not in 2026, and probably not by 2030 for installed bases. The migration story is that PROFINET over TSN is the official forward path for PROFINET sites, and CIP over TSN is the equivalent for Rockwell estates. EtherCAT will likely remain its own ecosystem for some time because its scheduling model is tighter than vanilla 802.1Qbv. The future is “TSN as the shared backbone, with vendor profiles riding on top,” not a forced replacement.

Is private 5G worth the integration cost for a small plant?

If your only requirement is deterministic Ethernet within a single cell and your devices do not move, almost certainly no — a TSN refresh is faster and cheaper. Private 5G earns its cost on plants with mobile assets, multiple cells separated by long distances, or compliance constraints that demand wireless. A useful gut check: if you cannot list at least three workloads that genuinely require mobility, you do not need private 5G yet.

References

Answer-first summary: Authoritative primary sources for the claims, standard names, and architecture descriptions used in this post. Each link goes to the standards-body, vendor, or research-body home page for the relevant technology, so you can verify version numbers and read the underlying specifications directly.

IEEE 802.1 TSN Task Group — official charter and amendment list: https://1.ieee802.org/tsn/
3GPP Release 18 description and URLLC enhancements: https://www.3gpp.org/specifications-technologies/releases/release-18
IEC/IEEE 60802 industrial TSN profile overview (IEC project page): https://www.iec.ch/dyn/www/f?p=103:38:::::FSP_ORG_ID,FSP_APEX_PAGE,FSP_PROJECT_ID:1252,23,103832
ETSI multi-access edge computing (MEC) specifications: https://www.etsi.org/technologies/multi-access-edge-computing
Cisco TSN technology overview and IE-series switch documentation: https://www.cisco.com/c/en/us/solutions/industries/manufacturing/industrial-networking.html
Ericsson industrial 5G and TSN integration whitepapers: https://www.ericsson.com/en/industrial-iot