Recommended RRM Power Profile

15 m density, dual‑band OpenWiFi deployment. Position‑based power planning with no scanning.

Profile settings

Heuristic Tx Power. Distances are in metres. Power values are attenuation from full, not absolute dBm.

On dual‑band hardware the 6 GHz fields have nothing to act on. Leaving them blank is correct and is also the explicit way to say “no plan for this band” — see the note on blank fields below.

Channel Mapper. Leave the entire section empty on this profile.

What this produces at 15 m spacing

The profile is written in attenuation, but the value sent to the AP is an absolute dBm between 1 and 30, computed as base + attenuation. Both are shown, because the two are coupled and the attenuation alone does not tell you what the radio does — see the conversion base below.

The thresholds deliberately straddle 15 m so the nominal case lands in Medium rather than at either extreme. That is what produces a spread instead of a uniform result, and it means the plan tightens by itself wherever the build ended up denser than the design intent.

2.4 GHz is cut harder than 5 GHz for two reasons: it has three non‑overlapping channels against 5 GHz’s dozen or more, so co‑channel reuse pressure is far higher; and it propagates further, so APs that are comfortably separated at 5 GHz are still interfering at 2.4. If the operator also wants to steer clients off 2.4, -12dB Low / -9dB Medium is a reasonable second step — but commit the conservative set first and look at the result.

Why “By Channel” is unchecked, even with a channel plan committed

The option changes “nearest neighbour” to mean nearest neighbour on the same channel. A good channel plan deliberately separates same‑channel APs, so at 15 m nominal spacing the nearest same‑channel neighbour is much further away:

When the planner finds no same‑channel neighbour it assigns Full. So enabling this option while using thresholds sized for 15 m yields a plan that is almost entirely Full — the feature appearing to do nothing, for a reason that is not visible in the output.

Thresholds could instead be sized to same‑channel reuse distance, but the profile would then depend on the channel plan’s reuse pattern and would behave differently at a property with a different usable‑channel count (DFS enabled or not, 40 vs 80 MHz). For a profile meant to be reused across installs, physical geometry is the stable input and the channel plan is not.

If it is enabled anyway, note that whether planned channels are used at all is decided by inspecting only the first AP in the list. If that one AP happens to have no committed channel plan, the whole run silently falls back to current channels.

The conversion base — a pending change

The rXg stores power as attenuation from full and converts to absolute dBm for the AP as base + attenuation, clamped to 1–30. The base is a parameter of that conversion with a default of 30. A base of 30 asserts that Full means 30 dBm — 1 W conducted, which these radios do not do. That is the reason for the change: the top of the scale is currently fiction, not merely offset.

Why this matters for the profile above. The AP clamps a config value it cannot achieve down to its own maximum, in the region of 23 dBm. At base 30, every attenuation from Full to roughly -7dB converts to 23 dBm or more and therefore clamps to the same radiated power. The top third of the dropdown is inert, and the profile’s Medium and Full levels are indistinguishable on air.

So the base correction is not a tidy‑up that happens to accompany this profile — it is what makes the profile take effect. Until it lands, expect the plan to differentiate only the deeper levels.

Two things to settle before the change ships. First, 30→24 is a 6 dB reduction; if the intent is exactly 7, the base is 23. Second, the clamping behaviour above is inferred from the hardware’s capability, not measured. It is directly testable without any scanning: set one radio to Full and another to -6dB, then compare the radio_power_level each reports in telemetry. Identical readings confirm the clamp, and give the merge request an evidence base for its impact assessment — including the useful finding that APs at Full are expected to see no change at all.

Related, now fixed. The lower clamp in the conversion was dead code: the guard holding results at or above 1 dBm was written inside a tap block, and tap returns its receiver, so the block’s value was discarded and only the upper bound ever applied. This was invisible at base 30 — the deepest attenuation available still lands at 12 dBm, and Minimum returns early, so no valid input reached the guard. It would have become reachable with the base change: at a base of 17, -17dB and -18dB produce 0 and −1 dBm, and an AP rejects a configuration carrying an out‑of‑range power rather than clamping it, so the symptom would have been radios failing to come up rather than a quietly wrong power. Replaced with a single two‑sided clamp, verified to produce identical output for every reachable value at base 30 — so it carries no field impact on its own and removes the trap ahead of the base change.

Prerequisites

1. Floorplan scale correct on every floor. This is the one input that silently invalidates everything downstream: it converts image pixels to metres, and if it is wrong every distance is wrong by a constant factor while the plan still completes and looks plausible. Verify against one known dimension — a corridor length — before trusting any plan.
2. APs placed, meaning both an area assignment and x/y coordinates. Any AP missing a coordinate is skipped silently and simply receives no plan.
3. Zone TX power optimization disabled for this site. It is a separate, measurement‑driven system that writes the same TX power columns and has no awareness of the committed plan, so leaving it enabled invites the plan to drift. On a non‑scanning site it has no data to act on regardless, but the toggle should be explicit.
4. Profile named for the standard, not the property. The Default flag is a singleton, and any area with no profile attached inherits whichever profile holds it — so a property‑named profile carrying Default silently becomes the house standard for every future site.

Acceptance check — before you commit

Create the power plan, then inspect the planned level across the placed APs. This needs no scanning and no client traffic.

The all‑Full case is worth separating carefully. On a corridor build at 15 m it should not happen with these thresholds, so suspect the scale before adjusting the numbers — a scale error produces a credible‑looking plan computed from fictional distances, which is considerably worse than a plan that obviously does nothing.

Limits of this approach