FSMA: Gateway-Controlled MAC for LoRa NTN

• FSMA is a gateway-controlled MAC protocol for LoRa-based NTN that uses the FreeChirp signal to indicate channel idleness and assess link quality.
• It reduces collisions by narrowing the access window from full packet airtime to roughly 6 symbols, resulting in up to 2.5× throughput improvement in mobile scenarios.
• FSMA offers energy efficiency and scalability, achieving up to 5× node energy savings and supporting thousands of devices with firmware updates on commercial LoRa hardware.

Free Signal Multiple Access (FSMA) is a synchronization-free, gateway-controlled MAC protocol for LoRa-based IoT networks in non-terrestrial networks with mobile gateways, especially Low Earth Orbit satellites and drones. Its defining mechanism is the FreeChirp, a single LoRa up-chirp transmitted by the gateway when the channel is free, so that nodes transmit only when the channel is idle and when links are reliable. In the formulation reported for LoRa NTN, FSMA is designed to reduce collisions, exploit link variability caused by gateway motion, avoid synchronization or complex scheduling, and remain deployable on commercial LoRa hardware through firmware updates rather than new hardware (Vennam et al., 3 Oct 2025).

1. Operating context and design motivation

FSMA is motivated by two simultaneous properties of non-terrestrial LoRa deployments. The first is the very large coverage footprint of a satellite or drone gateway, which causes many devices to contend for the same uplink opportunity. The second is that the gateway moves, so link quality changes rapidly with time. The reported measurements and traces describe both effects directly: a Norby-2 packet was received over more than 3000 km of coverage, and link SNR can vary by up to 10 dB during visibility windows, with complete outages between passes (Vennam et al., 3 Oct 2025).

These conditions make several standard random-access strategies inadequate. In large footprints, ALOHA is highly collision-prone because nodes transmit as soon as they have data. CSMA is weakened by hidden terminals, since LoRa CAD sensing range is on the order of 5–15 km while satellite footprints span thousands of kilometers; the sensing region can be less than 0.1% of the footprint. BSMA reduces hidden terminals by letting the gateway indicate channel occupancy, but it incurs large gateway-side energy overhead, may require a separate channel or full duplex capability, and does not address whether a node’s link to a moving gateway is currently reliable. FSMA is defined against this background as a mechanism that addresses both contention and fast-changing link reliability through a gateway-issued free signal rather than through node-side carrier sensing or continuous busy signaling (Vennam et al., 3 Oct 2025).

A central implication is that FSMA treats access control as a gateway-observed property of the wide-area channel, not as a neighborhood-level sensing problem. This distinguishes it from terrestrial LoRa MAC behavior, where local carrier sensing can still be informative.

2. FreeChirp and protocol operation

The core signaling primitive in FSMA is the FreeChirp, which is a single LoRa up-chirp transmitted by the gateway when the channel is free. The gateway continuously monitors whether a packet is being detected or received; for SX127x hardware, the reported implementation uses bit 0 of RegModemStat as the signal/preamble detection indicator, accessed through an external trigger pin. If the channel is idle, the gateway sends one FreeChirp of duration $t_{chirp}$, then waits for

$t_{wait}=6\times t_{nSym},$

where $t_{nSym}$ is the node packet symbol duration. The interval between successive FreeChirps is

$t_{interval}=t_{chirp}+t_{wait}.$

If packet detection occurs during the waiting period, the gateway stops chirping and enters a longer defer period, reported as a long backoff of $4\times t_{wait}$ (Vennam et al., 3 Oct 2025).

Node behavior is also explicitly constrained by FreeChirp detection. When a node has data buffered, it does not transmit immediately. Instead, it senses for FreeChirp using LoRa CAD over a duration

$t_{nSense}=t_{interval}.$

Detection is two-step. A node first keeps doing CAD until a positive detection occurs or the sensing window expires, then immediately performs another CAD. A positive CAD followed by a negative CAD is interpreted as a single FreeChirp; a positive CAD followed by a positive CAD is interpreted as a nearby node’s longer preamble or transmission. Only the first pattern authorizes transmission. If no valid FreeChirp is confirmed, the node backs off using exponential backoff with initial window equal to packet length, doubling on each miss, and reset after $>100\times$ the initial window (Vennam et al., 3 Oct 2025).

In operational terms, a node is allowed to transmit only if two implicit conditions hold simultaneously: it receives a valid FreeChirp, meaning the gateway judged the channel free, and it can detect that FreeChirp, which the protocol uses as a link-quality filter.

3. Timing structure, collision reduction, and capture-aware access

FSMA’s collision-control logic is expressed through timing. The protocol defines the FSMA collision window as

$T_{collision\ FSMA}=t_{nSense},$

whereas baseline schemes use

$T_{collision\ Baselines}=t_{packet\ length}.$

For SF10 packets, the reported baseline packet durations are 20.25 symbols for 0-byte payload and 404.25 symbols for 192-byte payload. Accordingly, FSMA reduces the collision window from packet airtime to roughly 6 symbols, described as about $4\times$ to $t_{wait}=6\times t_{nSym},$0 reduction, and in the introduction as roughly $t_{wait}=6\times t_{nSym},$1–$t_{wait}=6\times t_{nSym},$2 depending on payload (Vennam et al., 3 Oct 2025).

The protocol is also designed to improve the usefulness of the LoRa capture effect. When multiple nodes respond to the same FreeChirp, their packet arrival times at the gateway differ mainly because of propagation-delay differences, modeled as

$t_{wait}=6\times t_{nSym},$3

where $t_{wait}=6\times t_{nSym},$4 and $t_{wait}=6\times t_{nSym},$5 are the propagation delays of the first and last packet. In baseline schemes, the corresponding spread is

$t_{wait}=6\times t_{nSym},$6

The paper summarizes prior capture results as follows: the gateway can decode a packet during collision if there is at least 1 dB signal strength difference, the receiver locks within about 4 symbols, and if packets arrive within that 4-symbol locking period, the stronger packet can be decoded in 98% of cases. FSMA helps this mechanism because packets triggered by the same FreeChirp arrive close together in time (Vennam et al., 3 Oct 2025).

FreeChirp also serves as a link-awareness mechanism through spreading-factor asymmetry. The gateway can transmit FreeChirp at a lower spreading factor than the nodes’ uplink packets; the paper’s main configuration is FreeChirp at SF9 and uplink packets at SF10. The intended interpretation is that if a node can detect the lower-SF chirp, its higher-SF uplink is likely reliable. This makes FSMA not only collision-aware but also link-aware under gateway motion (Vennam et al., 3 Oct 2025).

4. System assumptions, implementation, and evaluation framework

The reported FSMA implementation assumes a LoRa IoT network with a single moving gateway in the main design description, although multi-gateway operation is identified as future work. End devices generate packets according to a Poisson arrival process, may be duty-cycle constrained, and use LoRa CAD capability. The protocol is explicitly synchronization-free and does not require strict slot synchronization, explicit localization, or preloaded satellite schedules for operation. It also does not require prior explicit link knowledge; instantaneous link viability is inferred from whether the node can hear FreeChirp (Vennam et al., 3 Oct 2025).

A major design goal is compatibility with off-the-shelf commercial LoRa hardware. The hardware prototype uses 25 commercial LoRa devices with a drone-mounted moving gateway. Nodes include Adafruit Feather M0 with SX1276 and Mbed STM32 boards with SX1272. The node LoRa configuration is SF10, 125 kHz bandwidth, 20-byte payload, CR 4/8, explicit header, and CRC enabled. The gateway uses two Adafruit Feather RP2040 devices: one operates as the transceiver and stores received data locally, while the other periodically checks the trigger from the receiver and decides when to send FreeChirp. Because commercial LoRa devices do not natively send a single chirp, the implementation uses OS callback functions and interrupt timers to stop transmission after exactly one chirp duration (Vennam et al., 3 Oct 2025).

Two experimental environments are described. The mobile setup uses 25 nodes distributed across a campus-scale area greater than 1 km, with a drone-mounted gateway flying a loop in 4 minutes at 10 m/s. A static setup uses 16 nodes and attenuators to emulate low-SNR links, with offered load increased up to 30% per node, corresponding to as much as 500% offered network load. In addition to hardware experiments, the study uses a custom Python-based simulator called NTNLoRa, which models PHY and MAC behavior end to end, packet arrivals, LoRa waveform generation, satellite channel dynamics, detection and reception, capture effect during collisions, and TLE-based LEO trajectories. Reported metrics include throughput, normalized throughput, packet reception ratio, channel usage, node energy per successful packet, gateway energy overhead, gateway failure ratio, and node wait times (Vennam et al., 3 Oct 2025).

5. Reported performance and scalability

The reported results are consistently favorable to FSMA. In hardware, baseline ALOHA/CSMA-like schemes achieve only about 30% throughput in mobile scenarios and up to 40% in static scenarios, whereas FSMA achieves $t_{wait}=6\times t_{nSym},$7 higher throughput in static setups and up to $t_{wait}=6\times t_{nSym},$8 higher throughput in mobile scenarios. The abstract states the headline more conservatively as up to $t_{wait}=6\times t_{nSym},$9 higher throughput. Reliability gains are also substantial: the abstract reports $t_{nSym}$0 to $t_{nSym}$1 better packet reception ratio, while the hardware section reports over $t_{nSym}$2 higher PRR at 100% offered load and nearly $t_{nSym}$3 improvement at 500% offered load. Channel usage reaches around 80% in experiments (Vennam et al., 3 Oct 2025).

The simulator extends these trends to large-scale NTN settings. In static large-scale simulations, FSMA yields about $t_{nSym}$4 more received packets than baseline and about $t_{nSym}$5 better PRR. In moving-gateway simulations, it shows about $t_{nSym}$6 throughput improvement and $t_{nSym}$7–$t_{nSym}$8 PRR improvement, and is reported to sustain high throughput beyond 2000 nodes while avoiding throughput collapse observed in other approaches. The abstract states scalability to 5000+ devices per satellite pass. Elsewhere, the evaluation overview also reports scalability to over 10,000 devices at 0.1% duty cycle within a 10-minute visibility window. The more conservative repeated headline is 5000+ devices per satellite pass (Vennam et al., 3 Oct 2025).

Energy behavior is another major reported outcome. FSMA achieves up to $t_{nSym}$9 improvement in node energy efficiency compared with ALOHA, including sensing overhead. Relative to BSMA, gateway energy overhead is reduced by up to $t_{interval}=t_{chirp}+t_{wait}.$0, and one figure caption reports up to $t_{interval}=t_{chirp}+t_{wait}.$1 lower overhead. Gateway failure ratio is reduced by up to $t_{interval}=t_{chirp}+t_{wait}.$2 under high contention. In moving-gateway scenarios, average node wait times remain around 1–2.5 minutes, approximately 150 seconds, within a typical 10-minute visibility window. The simulator also indicates that SF9 FreeChirp provides reliable detection out to 2000 km, while SF10 remains favorable for uplink packet decoding with low enough airtime (Vennam et al., 3 Oct 2025).

6. Scope, nomenclature, and adjacent research directions

FSMA, in the sense defined for LoRa NTN, is a gateway-controlled permission-based MAC centered on FreeChirp rather than a generalized umbrella term for all non-orthogonal access. The paper itself leaves several extensions open: multi-gateway deployments, concurrent decoding and interference cancellation, adaptive backoff, sensitivity to FreeChirp coverage choice, and fuller treatment of Doppler and latency optimization. It also notes that the current implementation depends on CAD-capable nodes, gateway access to packet-detection state, and firmware-level customization of chirp generation and CAD behavior (Vennam et al., 3 Oct 2025).

The acronymic landscape around similar-looking terms is nontrivial. In adjacent literature, “SFMA” denotes semantic feature multiple access, not Free Signal Multiple Access (Wang et al., 10 Apr 2026). Likewise, “FAMA” denotes fluid antenna multiple access, including cell-free and fast fluid-antenna variants, rather than the LoRa-NTN FreeChirp protocol (Han et al., 29 Apr 2025). A broader grant-free access literature studies structurally related problems such as blind active-user detection with non-orthogonal signatures (Sivalingam et al., 2021), deterministic slot-pattern design for latency and reliability (Kotaba et al., 2022), and joint device activity detection, channel estimation, and signal detection via BiGAMP (Zhang et al., 2023). This suggests that FSMA belongs to a larger family of grant-free and receiver-centric access mechanisms, but its distinctive identity is its concrete realization as a synchronization-free, gateway-controlled LoRa NTN MAC built around a single-chirp free signal rather than around spreading, message passing, or fluid-antenna port selection.

A common misconception is to treat FSMA as merely a lighter busy-tone protocol. The defining distinction is that FreeChirp is sent only when the channel is idle, not while it is occupied, and that its lower spreading factor also acts as a reciprocity-based link-quality probe. Another misconception is to read FSMA as a slotted or fully scheduled system; the defining description is instead synchronization-free access mediated by gateway-issued permission signals. Under that interpretation, FSMA is best understood as a LoRa-native NTN access protocol that compresses contention into short gateway-observed permission windows and couples channel idleness with instantaneous link viability.