When several radios share one frequency, they face the same problem as people in a crowded room: if two speak at once, neither is understood. Listen-before-talk, also called carrier sense, is the polite rule that keeps a shared radio channel usable, requiring each radio to check whether the channel is already busy and hold off until it goes quiet before transmitting. This guide explains how carrier sensing avoids collisions on point-to-multipoint and mesh telemetry networks, how it interacts with the polling schedules that SCADA masters impose, and why contention on a shared channel sets a ceiling on effective throughput and adds variability to latency.
Listen-before-talk in one line: Listen-before-talk, or carrier sense, is a medium-access rule in which a radio checks whether the shared channel is currently in use and waits for it to fall idle before it transmits, rather than transmitting blindly. It exists to prevent collisions, the overlapping transmissions that corrupt each other when two radios talk at the same time on the same frequency. It is the basis of CSMA schemes used in point-to-multipoint and mesh telemetry, and because radios must sometimes wait their turn, contention on a busy channel reduces usable throughput and makes latency less predictable.
A radio channel is a shared medium: only one transmission can occupy a given frequency in a given area at a time, and if two radios transmit simultaneously their signals collide and interfere so that a receiver gets garbled energy instead of either message. Listen-before-talk avoids this by having a radio sample the channel first. It listens for the energy of an ongoing transmission, the carrier, and if it detects one, it treats the channel as busy and refrains from transmitting. Only when it senses the channel is clear does it go ahead and send, dramatically reducing the chance that it steps on someone already talking.
This carrier-sense behaviour is the core of the family of schemes called CSMA, carrier-sense multiple access, which underpins how many unlicensed and shared-band radios coordinate access without a central controller telling each one when to speak. Because there is no master timetable in these schemes, every radio makes its own decision to defer based purely on what it hears, which is elegant precisely because it needs no coordination. On some unlicensed bands, listening before transmitting is also a regulatory requirement designed to keep the shared spectrum fair for everyone using it.
Sensing is not perfect, and its main weakness is the hidden node: two radios that can both reach a central point but cannot hear each other. Neither senses the other's carrier, so both may transmit at once and collide at the receiver despite each having listened first. Real protocols mitigate this with additional handshaking or acknowledgements, but the possibility is inherent to any purely sense-based scheme, and it is one reason telemetry networks often combine carrier sense with some scheduling to keep contention manageable.
SCADA masters traditionally gather data by polling: the master addresses each outstation in turn and asks for its data, so on a well-run polled network only one device transmits at a time by design. Polling is effectively a schedule that arbitrates the channel from the top, and where the master strictly controls the sequence, listen-before-talk becomes a safety net rather than the primary arbiter, catching the occasional unsolicited transmission such as an alarm that a remote sends without being asked.
The picture changes when devices transmit on their own initiative, for example by reporting on exception or raising alarms asynchronously, because those transmissions are not scheduled by the master and can arrive at any moment. Here carrier sense does real work, forcing an outstation with something to say to wait until the channel is idle before it speaks, so its unsolicited report does not collide with the master's polling or with another remote's alarm. The more devices share the channel and the more they talk without being polled, the more often they contend for the same idle moments, and the more time is spent waiting rather than sending.
This interaction shapes how a telemetry network is engineered. A tightly polled network trades responsiveness for orderliness, because a remote must wait for its turn in the poll cycle even when it has urgent data, while a more contention-based, report-by-exception network is more responsive but risks congestion as the population grows. Many designs blend the two, using scheduling to keep routine traffic collision-free and carrier sense to accommodate the occasional unscheduled message, tuning the balance to the number of sites and the mix of routine versus urgent data.
Every moment a radio spends listening and deferring is a moment it is not sending, so contention imposes a real cost on how much data a shared channel can actually carry. As the number of radios and the volume of traffic rise, the channel spends an increasing share of its time occupied or being waited on, and the effective throughput available to any one device falls below the raw radio speed. There is a point where adding more talkative sites to a shared channel yields diminishing returns, because the additional traffic mostly increases the time everyone spends waiting for a clear slot.
Contention also makes latency variable rather than fixed. A message that finds the channel idle goes out immediately, but a message that arrives while the channel is busy must wait, and if a collision does occur it must back off and retry after a random delay, extending the time before the data actually gets through. For most telemetry this jitter is harmless because the data is not time-critical to the millisecond, but it does mean that on a busy channel the delay between an event and its report can stretch, which matters when a design assumes prompt delivery of alarms.
A cloud SCADA platform sits at the receiving end of these shared-channel networks and can make their contention visible in ways that help operators manage it. When radios funnel through a gateway into Merobix, patterns such as rising delivery delays, retries, or missed reports on a particular channel show up as degrading data timeliness across the sites that share it, which is a signal the channel is becoming congested. Seeing that trend lets engineers rebalance the network, move some sites to another channel, lean more on scheduling, or reduce chatty reporting, before contention starts to erode the very visibility the telemetry exists to provide.
It makes a radio check whether the shared channel is already carrying a transmission before it sends its own, and defer until the channel is idle if it hears one in progress. This prevents collisions, where two overlapping transmissions on the same frequency corrupt each other so a receiver gets neither message. It is the foundation of carrier-sense multiple access, or CSMA, and on some unlicensed bands it is also a regulatory requirement to keep shared spectrum fair.
Because a radio must wait whenever the channel is busy, time spent listening and deferring is time not spent transmitting, so contention lowers the usable throughput below the radio's raw speed. As more radios share the channel and traffic grows, the channel spends more of its time occupied or being waited on, and each device's effective share falls. It also makes latency variable, since a message arriving while the channel is busy is delayed until a clear slot appears.
A strictly polled network mostly arbitrates the channel through the poll schedule itself, since the master addresses one outstation at a time and only that device replies. There, carrier sense acts as a safety net for the occasional unsolicited transmission, such as an alarm a remote raises without being asked. It becomes far more important on networks where devices report on exception or send alarms asynchronously, because those unscheduled transmissions can otherwise collide with polling or with each other.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.