When a 10% Duty-Cycle Mesh Hits Its Airtime Limit
When a 10% Duty-Cycle Mesh Hits Its Airtime Limit
Ten percent can be a transmitter's compliance ceiling without being a network's usable capacity. Relays, collisions and protocol overhead must be budgeted separately.
A busy LoRa relay can exhaust its permitted transmit time and become a bottleneck. That useful warning does not justify the stronger claim that every European 868 MHz mesh must either collapse at 10% or become illegal above it. The exact SRD sub-band, access method, traffic model and national implementation decide the regulatory boundary; the RF topology and protocol decide capacity.
Regulatory note: technically reviewed 29 August 2026. This is engineering guidance, not legal advice. European harmonisation is not a substitute for checking the current national interface and the conformity evidence for the exact equipment and access technique.
Three Limits, Not One “Airtime Wall”
The exact SRD entry sets frequency, category, radiated power, bandwidth and an access-technique or duty-cycle condition.
Overlapping transmissions, hidden nodes, capture, modulation settings and interference determine successful channel use.
Relaying, flooding, acknowledgements, retries and control traffic multiply the application load.
A device can comply with its own duty-cycle condition while contributing to a congested channel. Conversely, a lightly loaded mesh can provide useful service well below 10%. A percentage in a regulatory table neither reserves channel time nor predicts delivered messages.
EU868 Is Not One Band with One Duty Cycle
The current harmonised EU framework for non-specific short-range devices contains several entries between 868 and 870 MHz. The following examples come from Commission Implementing Decision (EU) 2025/105; they are not interchangeable:
| Frequency range | Maximum e.r.p. | Harmonised access condition |
|---|---|---|
| 868.0–868.6 MHz | 25 mW | Qualifying spectrum-access and interference-mitigation techniques, or duty cycle ≤1% |
| 868.7–869.2 MHz | 25 mW | Qualifying techniques, or duty cycle ≤0.1% |
| 869.4–869.65 MHz | 500 mW | Qualifying techniques, or duty cycle ≤10% |
| 869.7–870.0 MHz | 5 mW without an additional access requirement; 25 mW conditionally | At 25 mW: qualifying techniques, or duty cycle ≤1% |
The familiar 10% figure belongs to the 869.4–869.65 MHz entry, used by common EU_868 mesh profiles. It is an alternative compliance route, not the universal rule for everything marketed as “868 MHz.” Belgium's current BIPT interface specification B01-23 implements the same 500 mW e.r.p. entry: spectrum-access and interference-mitigation requirements apply, with duty cycle not exceeding 10% as the alternative.
“Carrier sense” is not a self-declared exemption. A product relying on the access-technique route needs an implementation and conformity case that meet the applicable standard. The current ETSI EN 300 220-2 V3.3.1 polite-spectrum-access method includes clear-channel assessment, deferral and transmit-time controls. It is not permission for continuous transmission.
For example, the standard's polite-access parameters include at least 160 µs of clear-channel assessment, a 1 s maximum single transmission, a 4 s maximum dialogue or polling sequence and no more than 100 s of cumulative transmission in one hour per 200 kHz of spectrum. The exact applicable route and test evidence still depend on the radio and its operating mode.
What 10% Means for One Transmitter
ETSI defines duty cycle as cumulative transmitter-on time divided by an observation period and observation bandwidth. Unless another value is specified, the observation period is one hour and the observation bandwidth is the permitted frequency band.
d_i = Σ T_TX,i / T_obs
For T_obs = 3600 s and d_i ≤ 0.10, transmitter i has at most 360 seconds of on-air time per hour under the duty-cycle route.
That account must include every emission from the transmitter: user messages, advertisements, telemetry, route maintenance, forwarded packets, acknowledgements and retries. It is not a shared allowance assigned to the mesh, and unused airtime at one node cannot simply be transferred to a busy relay.
Nor is 360 seconds equal to 360 messages. A packet's time on air depends on payload length, preamble, bandwidth, spreading factor, coding rate, header mode and low-data-rate optimisation. One-second packets would permit at most 360 transmissions before any control traffic; 250 ms packets would give a different count. Calculate with the settings actually transmitted.
What “Ten Nodes at 10%” Really Calculates
For nodes whose transmissions occupy the same receiver's collision domain, same channel and relevant modulation resources, define the normalised attempted load:
G = Σ λ_i T_i
where λ_i is attempted transmissions per second and T_i is on-air seconds per transmission. With varying packet types, sum every type, relay copy and retry.
If ten nodes each actually transmit for 10% of the hour under those assumptions, their duty fractions sum to G = 1. That is one channel-time of attempted load per unit time. It does not mean a scheduler has filled ten neat, non-overlapping slots. Random start times overlap, so some airtime is lost to collisions and the union of busy periods is not simply 100%.
A bounded collision illustration
In the classic pure unslotted ALOHA model—equal packet durations, Poisson attempts, no capture and a collision destroying every overlap—the successful normalised throughput is:
S = G e−2G
It peaks at G = 0.5, where S ≈ 0.184. At G = 1, S ≈ 0.135.
This is an illustration, not a MeshCore or Meshtastic prediction. Carrier sensing, random deferral, hidden terminals, capture, unequal received powers, different spreading factors, multiple channels, packet lengths and duplicate suppression all change the result. The value of the model is narrower: attempted airtime and successfully delivered airtime are different quantities, especially near saturation.
Experimental LoRa capacity work by Bor, Roedig, Voigt and Alonso shows why a one-number limit is unsafe: centre frequency, spreading factor, bandwidth, coding, capture, transmitter selection and receiver placement all affect scalability. The corrected LoRa scalability paper found that dynamic transmission settings and multiple sinks can materially change capacity.
Relays Multiply Load Before Collisions Do
A payload delivered across h hops requires at least h successful transmissions when one forwarding node is selected at each hop. Flooding or redundant forwarding can require more. If r_m nodes transmit message m, its minimum channel-time cost in the affected collision domains is:
A_m = r_m T_m
A relay's duty account is d_relay = (own + control + forwarded + retries) / T_obs.
A high site can therefore become a hot spot: it hears more nodes, is selected for more paths and may share an interference domain with a larger population. But “higher always means worse” is also too simple. Protocol rules may suppress duplicates, a well-placed relay may reduce retransmissions, and distant groups can reuse the same spectrum when they do not interfere at the relevant receivers.
The foundational wireless-capacity analysis by Gupta and Kumar formalises the deeper issue: multihop nodes share a channel locally, every forwarded bit consumes transport capacity, and spatially separated transmissions may coexist. A nationwide mesh is therefore not one universal collision domain, but neither does adding relays create free capacity.
A Capacity Budget That Can Be Audited
- Freeze the regulatory profile. Record country, exact frequency entry, e.r.p., bandwidth and whether the equipment relies on duty cycle or a standards-compliant access technique.
- Measure packet time on air. Use the actual firmware, payload sizes and radio parameters. Include preamble and headers, not only application bytes.
- Build a traffic matrix. Record sources, destinations, periodic traffic, bursts and the busiest credible hour.
- Expand every message into transmissions. Apply the routing, forwarding, hop, acknowledgement, retry and duplicate-suppression rules actually used.
- Map collision domains. Include hidden nodes and high-site receivers. Do not assume every node hears every other node—or that hearing implies successful decoding.
-
Budget both duty and offered load. Check each transmitter's regulatory account separately from
Gat every important receiver. - Leave headroom. Legal maximum duty is a ceiling, not a normal design target. Growth, fading and collision-driven retries create positive feedback near saturation.
- Validate with logs. Compare attempted packets, decoded packets, duplicates, retries, latency, queue depth and per-node airtime during a controlled load test.
The useful question is not “How many nodes?” Ask how many offered packet-seconds arrive in each collision domain, how many transmissions each delivered message consumes, and what success probability and latency the application requires.
Does Moving to 70 cm Remove the Problem?
Moving a ham-only mesh to an amateur allocation can remove the specific SRD duty-cycle framework only when the operation is actually authorised under the amateur service. It does not remove shared-medium capacity, collisions, relay load, identification, content, station-control, coordination, emissions or interference obligations.
The linked MeshCore RFC for IARU Region 1 is a living coordination proposal, not a band-plan decision or station authorisation. Its current text asks for re-designation around 434.890 MHz, notes existing 434.900 MHz repeater conflicts in Finland and Norway, identifies proximity to the 435 MHz amateur-satellite segment and requires country-by-country checks for automatic stations. Those are substantive unresolved coordination questions.
Belgium adds a specific legal obstacle. The current BIPT station-licence FAQ says that, apart from APRS, an application cannot concern an unmanned fixed amateur station that continuously retransmits received signals or transmits without the user physically present. A proposed unattended MeshCore backbone therefore cannot be described as already authorisable in Belgium merely because its operators hold amateur certificates. Regulatory clarification or change may be required.
An RFC is not permission. IARU coordination, the national amateur allocation, an operator certificate, equipment compliance and an automatic-station authorisation are separate gates. Capacity engineering remains necessary after every legal gate passes.
The original geometric estimates based on Belgium's land area and a nominal amateur population cannot establish feasibility. Operators are unevenly distributed, only a fraction may participate, terrain and clutter determine links, relay sites need permission, and connectivity does not prove that a shared channel meets the traffic requirement. A propagation survey and measured traffic plan must replace average-density arithmetic.
The Practical Verdict
A 10% duty-cycle mesh does not “always hit the wall.” A low-traffic network may remain useful with ample headroom; a public flooding backbone can overload one relay or collision domain long before every node reaches 10%. The boundary is conditional and measurable.
Use the exact national SRD entry, prove the equipment's access route, calculate time on air, expand application messages into all transmissions, and measure delivery under the busiest credible load. If amateur spectrum is proposed instead, solve its authorisation and coordination questions explicitly. Changing service can change the legal envelope; it cannot repeal airtime physics.
Primary and project references checked
- Commission Implementing Decision (EU) 2025/105 — harmonised SRD conditions
- ETSI EN 300 220-2 V3.3.1 — non-specific SRD spectrum access
- BIPT B01-23 — Belgian 869.4–869.65 MHz interface
- BIPT — station-licence conditions for unmanned fixed amateur stations
- MeshCore official CLI documentation — radio and duty-cycle controls
- Meshtastic official EU_868 radio settings
- Abramson — The ALOHA System: another alternative for computer communications
- Bor et al. — corrected LoRa scalability study
- Gupta and Kumar — The Capacity of Wireless Networks
- RF.Guru MeshCore RFC for IARU Region 1 — current proposal and open issues
Mini-FAQ
- Is every EU868 device limited to 10% duty cycle? No. Conditions vary by frequency range and device category. In 869.4–869.65 MHz, qualifying spectrum-access techniques or duty cycle not exceeding 10% are alternative routes under the harmonised non-specific SRD entry.
- What does 10% duty cycle mean? Under the usual one-hour observation period, it means no more than 360 seconds of cumulative transmitter-on time per device. Every packet, relay copy, acknowledgement and retry transmitted by that device counts.
- Do ten nodes at 10% fill a channel? Their attempted duty fractions sum to a normalised load of one only when they share the relevant channel and collision domain. Random overlaps cause collisions, while spatial reuse and different radio resources prevent this from being a universal 100% occupancy claim.
- Does compliant polite spectrum access provide unlimited airtime? No. The applicable ETSI method includes clear-channel assessment, deferral and transmit-time controls, and the equipment needs evidence that it implements the qualifying method.
- Does moving MeshCore to 70 cm solve capacity and legality? No. It may change the legal framework if amateur operation is authorised, but capacity, coordination and national automatic-station rules remain. The current RFC is a proposal, and Belgium's present unattended-station rule is unresolved for this use.