The physical layer can send bits; the MAC decides who sends them, when, with how much power, and when the radio can sleep. It is the control layer that turns a raw PHY into an efficient, shared and — for a battery device — long-lived link.
A radio channel is shared, so the first job of the MAC is deciding who may transmit and when. Two families. Scheduled access: a controller (a base station) grants each device specific time and frequency resources, so there are no collisions and quality of service can be guaranteed — the cellular approach. Contention access: devices listen and transmit opportunistically (listen-before-talk, CSMA, slotted ALOHA), which is simple and needs no coordinator but collides under load — the Wi-Fi and LPWA approach.
The choice shapes everything else. Scheduled access buys determinism at the cost of signalling and a scheduler; contention buys simplicity at the cost of collisions and unpredictable latency. Many systems blend them — contention to get on the channel, then a grant for the real traffic.
The PHY will sometimes fail to deliver a block, so the MAC owns retransmission. In modern systems this is HARQ — the receiver keeps the failed copy and combines it with the retransmission rather than discarding it — and the MAC runs the acknowledgement loop and several HARQ processes in parallel so the link does not stall waiting for one answer.
It also owns power control: setting each device's transmit power to just enough to be heard reliably, no more. Too much power wastes battery and interferes with neighbouring cells; too little drops the link. In a cell full of devices at different distances, power control is what lets them share the channel without the near ones drowning out the far ones.
For a battery device the dominant cost is keeping the radio on, and the MAC is what turns it off. Discontinuous reception (DRX) lets a device sleep between brief listening windows, waking only on a schedule the network agrees — so a sensor that talks for milliseconds a day can sleep for the other 86,399 seconds. The MAC negotiates the DRX cycle, balancing how long the device sleeps against how quickly it can be reached.
This is where the MAC decides battery life. A longer sleep cycle means years on a coin cell but higher latency to reach the device; a shorter one means responsiveness at the cost of power. Getting that trade-off right for the use case is a core MAC design decision, not an afterthought.
The MAC is where a radio becomes a usable network. It is also where a lot of real-world performance lives: a scheduler that is fair and efficient under load, HARQ timing that keeps the pipe full, power control that converges quickly without oscillating, and a DRX scheme that hits the battery and latency targets together. None of it shows up in a datasheet peak rate, but all of it decides how the system behaves when it is busy.
For anyone building a modem or a stack, the MAC is where the PHY's raw capability is turned into throughput, latency, fairness and battery life that the application actually sees. It is unglamorous, tightly coupled to the PHY's timing, and decisive.
It sits between the upper layers and the physical layer and controls how the shared radio is used: channel access (who transmits when), scheduling, retransmission (HARQ), transmit power control, and discontinuous reception (DRX) for low power. It turns a raw PHY into an efficient, shared link.
Scheduled access has a controller grant each device specific resources, avoiding collisions and guaranteeing quality of service (cellular). Contention access has devices listen and transmit opportunistically (CSMA, ALOHA), which is simple and coordinator-free but collides under load (Wi-Fi, LPWA).
Through HARQ: the receiver keeps a failed block and combines it with the retransmission instead of discarding it, and the MAC runs the acknowledgement loop and several HARQ processes in parallel so the link does not stall waiting for one answer.
Setting each device's transmit power to just enough to be heard reliably. Too much wastes battery and interferes with other cells; too little drops the link. In a cell with devices at different distances, it is what stops near devices drowning out far ones.
Discontinuous reception lets a device sleep between brief listening windows, waking only on an agreed schedule. A sensor can listen for milliseconds and sleep the rest of the time, which is what makes years of battery life possible. The MAC negotiates the cycle, trading sleep depth against latency.
Because the PHY only sends bits; the MAC turns that raw capability into throughput, latency, fairness and battery life. A good scheduler, HARQ timing, power control and DRX scheme are what decide how the system behaves under load, none of which appears in a peak-rate datasheet.
A scheduler that is fair under load, HARQ timing that keeps the pipe full, power control that converges, and a DRX scheme that hits battery and latency together are the real work. Tell us the use case.