RF Channel Planning When the Site Is Full of Wi-Fi

The 06:00 rehearsal was clean. Every drone took its waypoint on time, the formation closed and opened without a single hesitation, and we packed up assuming the evening would be the same show with a darker sky. It was not. By 20:00 the square was full of people, and two aircraft in the outer ring were arriving a quarter second late, which is enough to smear a transition that looked sharp in the morning.

Nothing had changed in the sequence file. What changed was the radio environment. A few thousand phones, a temporary event network, a broadcast crew's wireless kit and the venue's own access points had turned the 2.4 GHz band into something closer to a busy street than an open field.

The scan we now run before every show

We walk the launch area and the crowd perimeter with a spectrum analyser about ninety minutes before the display, then again twenty minutes out. The first pass tells us which channels are already occupied. The second tells us what has appeared since the gates opened, which is usually the more useful number.

What we log: occupied channel width, the strongest competing signal and where it is coming from. A hotel access point on the far side of the square matters far less than a hotspot sitting ten metres from the ground station.

Choosing a channel, and why the crowd position wins

We pick the quietest channel with enough separation from the loudest neighbour, not simply the emptiest one on the graph. A channel that looks clear but sits one step away from a strong transmitter will degrade as soon as that transmitter ramps up.

The harder lesson was about placement. We had the ground station close to the fleet because that felt logical. It should have been positioned relative to the crowd instead, elevated and back from the densest cluster of phones, with a clear line to the display volume. Moving it four metres and raising the antenna did more for link quality than any channel change that night.

If the link drops mid-show

You do not get a second rehearsal, so the response has to be decided in advance. Our rule now: if link quality falls below the threshold we set during the scan, the operator holds the current formation rather than pushing the next transition. A held shape reads as intentional to the audience. A late transition reads as a fault.

We also keep the backup ground station powered and on a different channel, not just a different laptop. Swapping hardware mid-show is slow; switching to a prepared second link is not.

The setting we no longer trust

Automatic channel selection. It sounds helpful and it behaved reasonably in testing, but on a site where the RF picture shifts every few minutes it made decisions we could not explain afterwards. We set the channel manually now and leave it alone unless the scan says otherwise.

None of this is exotic. It is a scan, a placement decision and a written rule for what happens when the numbers move. The city-centre show taught us that the ground is where the interference lives, and the ground is the part you can actually control.