Airtime is finite | OT Wireless

A Wi-Fi channel is one shared, half-duplex medium, and everything you put on it spends from the same fixed budget of time.

That budget is airtime. Controller I/O, fleet manager commands, a robot’s map download, a camera stream and the beacon for an SSID nobody uses all draw from it. Prioritization changes who tends to go first. It does not create more time. As fleets grow, the way to protect control traffic is to know everything that is on the air and to keep bulk traffic off the airtime that control depends on.

One talker at a time

RFC 8325, the IETF’s guidance on carrying DiffServ over 802.11, describes the medium as half-duplex and shared, with every station, the access point included, contending for it on equal terms. One vendor’s high-density design guide puts it in floor terms: only one device can talk at a time, and the others wait their turn.

Every frame pays a toll before any data moves: an interframe space, a random backoff, then the frame and its acknowledgment. RFC 8325 notes that the backoff alone can add as much as 9 ms of jitter per attempt¹ at the largest contention window, and a frame can make several attempts. For a wired controls engineer used to switched Ethernet, that is the first surprise. On Wi-Fi, the client next to yours is on the same wire.

Three things multiply the cost of each client:

Slow clients. Airtime is shared by turns, not by bytes. Heusse and colleagues showed that one 802.11b host at 1 Mb/s pulls hosts at 11 Mb/s below 1 Mb/s², because CSMA/CA gives every host an equal long-term chance at the channel and the slow host holds it longest. A vehicle at the edge of a cell, on a low rate, spends airtime every other client was counting on.
Management frames. Each SSID is a virtual AP with its own beacons and probe responses. One vendor’s guidance says each sends a beacon every 100 ms at the lowest supported data rate³. Two APs with four SSIDs each on one channel put eight beacons on the air every interval.
Low basic rates. Beacons and broadcasts go out at the lowest enabled rate. The same vendor notes that raising the minimum rate sends those frames faster, which shortens every one of them.

Prioritization improves the odds

The obvious objection is to turn on QoS and let control traffic go first. It does help. WMM, the Wi-Fi Alliance certification of 802.11e channel access, sorts traffic into four access categories: voice, video, best effort and background. Each category waits a different fixed interval and draws its random backoff from a different window, so the higher categories get statistically superior service.

Statistically is the word that matters. RFC 8325 states that 802.11 cannot assure that one access category is served with strict priority over another, and cannot assure any category a minimum amount of bandwidth, because contention has a random element. One vendor’s wireless QoS design guidance describes the effect of the EDCA timers as statistical in nature. Prioritization raises the odds that a control frame wins the channel. It reserves nothing. When the channel is busy, a high-priority frame still waits.

Prioritization improves the odds; it does not reserve the channel

Each access category waits a fixed time, then a random one. The ranges overlap, so a lower category sometimes goes first.

Drawn to scale

Default parameters for the first attempt, drawn to scale. A best-effort frame that draws a short backoff (from 3 slots) can beat a video frame that draws a long one (up to 9), which is why 802.11 cannot promise strict priority or a minimum bandwidth to any category. Each failed attempt doubles the window, up to its maximum.

Source: RFC 8325, Figures 3 and 4, which reproduce the IEEE 802.11-2016 defaults.

Wi-Fi Alliance also certifies WMM-Admission Control, which uses bandwidth management to prevent oversubscription. Ask your vehicle vendors whether their radios support it before you plan around it.

Uplink priority belongs to the client

Downlink is the direction you control. The access point classifies and queues what it sends. Uplink is different, because the client decides how to mark its own frames before it contends for the air.

RFC 8325 places the upstream mapping from DSCP to 802.11 user priority inside the client’s operating system. The same vendor’s guidance says WMM gives upstream QoS only to clients that support WMM, and that with downstream-only QoS, client upstream traffic is treated as best effort.

Vehicles talk as well as listen: status to the fleet manager, input data to a controller. If a vehicle’s software leaves those packets unmarked, they contend as best effort, whatever you configured on the AP. RFC 8325 recommends that the AP set DSCP at the wireless-to-wired edge, but by then the frame has already crossed the air. The reverse risk is real too. RFC 8325 warns that any client marking its traffic to the top priorities could monopolize the air.

Check marking end to end with a packet capture on both sides of the AP, not with a configuration screen. The cause page on missing QoS marking covers how it breaks.

Separation is the stronger lever

Prioritization decides who goes first in a shared line. Separation decides who is in the line at all. It is the stronger lever because it does not depend on the odds.

Channels and bands. Access points on different non-overlapping channels do not contend with each other. Adding APs on the same channel with overlapping coverage does not add capacity. Width is the trade: 40 MHz channels halve, and 80 MHz channels quarter⁴, the number of non-overlapping 5 GHz channels. See co-channel contention.
Dedicated radios. A radio or a band that carries only vehicles and controllers keeps your own laptops, phones and cameras from contending with them on that channel.
Two radios on the vehicle. Carried onto the vehicle, separation means two radios with two jobs: one for control and safety traffic and nothing else, one for everything else. The control radio can be Wi-Fi on its own band and channels, or private 5G. A map download on the second radio cannot queue in front of an I/O packet on the first. Keep the control radio lightly loaded; its value is that nothing else rides on it.
SSIDs, used sparingly. An SSID separates policy, not airtime. It shares the radio and the channel, and every one adds beacons. One vendor recommends no more than 3 SSIDs per AP and states that more than 5 consume 20% or more of available bandwidth in management overhead. Treat that as a vendor figure, not a law of physics.
Schedule windows. Map downloads, firmware updates and log uploads can usually wait for a break or a shift change. NIST’s guide to industrial wireless deployments advises against automatic installation of updates and warns that updating any component of factory operation could cause downtime. A window outside production keeps both the airtime load and the update risk off the shift. See bulk transfers competing with control.

Two radios, two jobs

A map download on the second radio cannot queue in front of an I/O packet on the first.

Schematic, not to scale

On one shared channel, 802.11 cannot promise strict priority between access categories or a minimum bandwidth to any of them, so prioritization improves the odds and reserves nothing. Two radios separate the traffic instead. Conceptual: no products, timings, or throughput implied.

Source: RFC 8325.

Fleets are growing into the same channels

The International Federation of Robotics counted 117,500 transportation and logistics robots sold in 2025, up 21%⁵ on the year before. Each one that lands on your floor is another station contending for the same channels.

Field note. Bot and AGV density is rising across industries and is saturating channels. 6 GHz is being deployed to relieve utilization. Where two systems share the same space, channel utilization has averaged 60 to 70 percent at times, with spikes to 80 or 90. Moving to 6 GHz has eased that, for now.

The new band is real relief. The FCC opened 1,200 MHz between 5.925 and 7.125 GHz⁶ for unlicensed use, with low-power indoor access points allowed across the whole band, explicitly to ease existing and anticipated congestion. It only helps the vehicles whose radios support it, but moving every client that can use 6 GHz off 5 GHz also frees 5 GHz airtime for the clients that cannot. The 6 GHz guide covers what the band fixes and what it does not.

Spectrum buys time. It does not stop a growing fleet from filling the new channels too. The cause page on fleet density covers how saturation shows up and how to confirm it.

Inventory everything on the air

Field note. You have to know everything on the air so control and safety traffic are prioritized over large transfers and updates. Firmware downloads, log uploads, and image uploads cause disconnects and flickers, and the evidence shows up in packet captures and logs.

NIST says the same from the planning side. Its deployment guide calls for a spectral occupancy survey of every wireless network in the factory and every source of interference, including microwave ovens and employee-owned devices. It recommends a spectrum management plan that keeps an inventory of all known wireless networks and emitters. For each data flow, it suggests recording how fast the data changes, how much there is, how much delay and loss it tolerates, and how critical it is to the operation.

The inventory has to include transmitters that are not on your network.

Field note. In automotive plants, vehicles run wireless Android Auto and CarPlay. Some use 80 MHz channels on 5 GHz and interfere with production wireless.

Those radios spend airtime on your channels without ever associating to your APs. So do phones and personal hotspots. See vehicle infotainment hotspots and personal hotspots and phones.

Write the inventory as a table. Each row is a traffic class, from safety and control I/O through fleet commands, telemetry, video, updates, log uploads, and corporate and guest traffic. Each column records what the separation decision needs: criticality, direction, rate, latency tolerance, timeout, QoS marking, band, radio and schedule window. The rows that matter most are the ones where a bulk class shares a channel with a control class.

QoS decides who waits in line. Separation decides who is in the line at all. Count everything on the air, then build the shortest line you can for control traffic.