RF design - coverage, capacity, high density, voice, location
Requirement / business driver
Design the RF so it meets the actual goal - coverage, capacity, voice, or location - which pull in different directions.
Coverage vs capacity (the core tension)
- Coverage design - fewer APs, larger cells, max range; fine for warehouses/low-density.
- Capacity / high-density design - more APs, smaller cells, lower power, tighter channel reuse to put more airtime per client (lecture halls, stadiums, offices). High density is about managing airtime and co-channel interference, not raw signal.
Channel and cell planning
Non-overlapping channels, channel reuse without co-channel interference, right channel width (narrower = more reusable channels in dense areas), and band steering to 5/6 GHz. AP groups / RF profiles (7.2.e) apply different power/channel/data-rate settings per area (e.g., a dense auditorium vs an open lobby).
RRM runs the RF - and the high-density lever set
Static channel/power plans do not survive 40/80 MHz primary-channel interactions, DFS behaviour (a listen-before-use wait; some channels effectively impractical), and uneven client channel support - so the RF is run by RRM, a set of independent algorithms scoped by an RF group and, inside it, RF neighbourhoods (APs mutually audible around -80 dBm), so optimisation runs only on APs that actually share air. TPC lowers power to curb co-channel interference (v1 reacts when the third-loudest neighbour exceeds a threshold, default -70 dBm; v2 computes cell boundaries); set min/max power guard rails where APs hear each other loudly but clients are far below (high ceilings), or TPC will whisper into the void. DCA builds the channel plan and biases away from persistent interferers reported by CleanAir (a dedicated spectrum chip that classifies non-Wi-Fi sources by duty cycle, not just RSSI). CHDM is the counterweight: if enough clients sit below the signal floor long enough (defaults: 3 clients and 25% of the AP's clients for 90 s), power steps up - and since detection is per-WLAN but mitigation is global, disable detection on guest/sticky-client WLANs so they cannot inflate power for everyone. FRA measures the 2.4 GHz coverage-overlap factor and converts redundant 2.4 radios to 5 GHz serving or monitor roles. Off-channel scanning is the data feed and a cost: the scan-list scope trades airtime against rogue blindness on unscanned channels (Wireless security).
High density adds a receive-side lever set: prune low mandatory rates so distant clients stop decoding beacons (shrinks the effective downlink cell and cuts management airtime); raise RX-SoP so the AP treats weak arrivals as noise (shrinks the uplink cell and sheds sticky clients); prefer ATF (airtime budgets per SSID/client) over Mbps bandwidth contracts, because one low-rate client burns far more air than a fast heavy one. Channel width is itself a sensitivity trade - decoding 80 MHz costs roughly 5-8 dB more signal than 20 MHz, and more spatial streams cost more SNR - so wide channels shrink usable cells; design cell edges to the worst-performing client, since RSSI is locally significant and vendor-dependent. Finally, AP power is a design dependency: 802.3af (15 W) leaves modern APs degraded (radio chains down or radios off), 802.3at/UPOE/802.3bt tiers restore full capability, and the closet PSE power budget silently caps how many full-capability APs a refresh can light.
6 GHz changes how clients even find the network
Discovery, not throughput, is the first 6 GHz design problem. The band has 59 x 20 MHz channels and forbids blind wildcard probing, so passively scanning everything would take a client seconds - instead, active probing is confined to every fourth channel (PSC - Preferred Scanning Channels), and the primary discovery path is out-of-band: RNR (Reduced Neighbour Report) elements in the 2.4/5 GHz beacons advertise the co-located 6 GHz BSS. In-band assists (FILS short beacons, unsolicited broadcast probe responses) exist and are mutually exclusive; most clients simply follow RNR. Design consequences: keep a 2.4/5 GHz presence on the same APs (an RNR anchor) even in 6 GHz-first designs, place 6 GHz BSSs on PSCs, and drop the hidden-SSID habit - it never bought security (probe and association requests carry the SSID in clear), and in 6 GHz it is actively broken: hidden networks force probing, non-PSC channels become unfindable, and some clients will not enable their 6 GHz radio at all without first seeing a beacon.
Voice and location
- Voice: needs coverage overlap for roaming, higher minimum signal (e.g., ~-67 dBm), low retransmits, and QoS (QoS design - models and strategy) + fast roaming (Roaming - L2 vs L3 mobility).
- Location: accuracy needs AP density and geometry (APs on the perimeter, not just centre) - a different placement than coverage.
Location - the technology ladder (7.1.a.iv)
"Location" spans an accuracy/cadence ladder, and the chosen rung changes the whole design. Presence / cell-of-origin resolves to the strongest AP: visit counts and dwell times, no map coordinates - and it counts Wi-Fi devices, not people. Probe-RSSI trilateration places clients at roughly 5 m (50%) / 10 m (90%), but only with the right geometry: every point heard by three or more APs at -75 dBm or better (plan for -67), APs on the perimeter and staggered along corridors rather than centre-only or in-line - a different layout than coverage design, usually more APs, with the extras in monitor mode. Its weakness is cadence and identity: updates ride on client probing (a minute or worse, and less once associated) and MAC randomisation fragments tracks. FastLocate switches to data-frame RSSI - real MACs, and the AP can solicit a frame - refreshing every ~6-8 s at the cost of an extra synchronised scanning radio per AP (NTP becomes critical). Active RFID tags invert control: infrastructure-managed emitters give guaranteed cadence plus telemetry (battery, panic button, motion-adaptive beaconing) for asset tracking. AoA/Hyperlocation reads the angle of incidence on a circular antenna array for roughly 1-3 m accuracy - same AP density (it buys accuracy, not fewer APs), strict antenna-orientation discipline, and the AP must be serving (local or Flex central switching), not monitor. BLE beacons flip the model entirely: the infrastructure is the lighthouse and the client's app does the computing - engagement and wayfinding, useless without the app. Match the rung to the requirement: footfall analytics -> presence; "find the wheelchair now" -> RFID tags or FastLocate; "2 m in the operating theatre" -> Hyperlocation with its serving-AP constraint priced in.
Wireless QoS is statistical (EDCA), and the marking path
Wired QoS schedules queues; wireless QoS tilts a lottery. 802.11's medium access is contention-based (CSMA/CA), and 802.11e/WMM EDCA retrofits priority by giving each of four access categories (voice, video, best effort, background) different contention parameters - shorter AIFS waits, smaller contention windows, longer TXOP transmit bursts for higher categories - so priority traffic wins the medium more often, not always: the service is probabilistic, never guaranteed, which is why WLAN voice leans on TSpec-based admission control and on RF design (overlap, signal floor) rather than on queuing alone. The marking path across the CAPWAP system is the other design fact: upstream, the client's 802.11e UP is translated onto the outer CAPWAP DSCP while the inner IP DSCP rides unchanged; downstream, the inner DSCP (not the 802.1p CoS) is copied to the CAPWAP outer and mapped back to UP at the AP - so DSCP is the only end-to-end currency, and honoring a client-set UP upstream is a trust-boundary decision exactly like a wired access port (QoS design - models and strategy).
Recommendation / justification
Design to the dominant goal: high density -> small cells/low power/airtime focus; voice -> overlap + signal + QoS + 11r/k/v; location -> perimeter AP geometry. Indoor vs outdoor (mesh/bridging) changes AP modes/antennas. Justify by what the customer must actually achieve, not a generic "good coverage."
What would change this (mid-scenario twist)
- Goal shifts from coverage to capacity (more users) -> add APs, shrink cells, lower power.
- Add real-time voice -> raise min signal, ensure overlap + QoS + fast roaming.
Validation checks
- Is the design sized to the right goal (coverage vs capacity vs voice vs location)?
- Is co-channel interference controlled (cell size, power, channel plan, AP groups)?
IPv6 / dual-stack note
RF design is L1/L2 and address-agnostic; the dual-stack impact is on the wired VLANs/segmentation behind the SSIDs.
Related
Spaced repetition
Coverage vs capacity RF design: coverage = [...]; capacity/high-density = [...].
Coverage vs capacity RF design: coverage = fewer APs, larger cells, max range; capacity/high-density = more APs, smaller cells, lower power, tight reuse (manage airtime/CCI).
High-density design is fundamentally about [...], not boosting signal strength.
High-density design is fundamentally about managing airtime and co-channel interference, not boosting signal strength.
Location-services accuracy depends on [...], which differs from a coverage layout.
Location-services accuracy depends on AP density and geometry (perimeter placement), which differs from a coverage layout.
Why can WLAN QoS never guarantee what wired LLQ guarantees?
EDCA/WMM only tilts contention odds - per-access-category AIFS, contention-window, and TXOP settings make priority traffic win the medium more often, not always - so wireless voice needs TSpec admission control and RF design, not just queuing.
Across CAPWAP, the only end-to-end QoS marking currency is [...] - upstream UP maps onto the outer CAPWAP DSCP (inner preserved); downstream the inner DSCP (not CoS) drives the outer and the AP's UP.
Across CAPWAP, the only end-to-end QoS marking currency is the inner IP DSCP - upstream UP maps onto the outer CAPWAP DSCP (inner preserved); downstream the inner DSCP (not CoS) drives the outer and the AP's UP.
How do clients discover 6 GHz WLANs, and what does that force in the design?
Mainly out-of-band via RNR in the 2.4/5 GHz beacons (active probing only on PSC channels) - so keep a legacy-band anchor on the same APs, put 6 GHz BSSs on PSCs, and avoid hidden SSIDs (unfindable off-PSC; some clients will not enable 6 GHz without seeing a beacon).
Name the receive-side high-density levers and what each shrinks.
Prune low mandatory rates (downlink cell + management airtime), raise RX-SoP (uplink cell - the AP ignores weak arrivals), and ATF airtime budgets (per-SSID/client air - fairer than Mbps contracts).
CHDM's design trap: detection is [...] - disable detection on guest/sticky WLANs so they don't inflate power for everyone.
CHDM's design trap: detection is per-WLAN but mitigation (power-up) is global - disable detection on guest/sticky WLANs so they don't inflate power for everyone.
Probe-RSSI vs FastLocate vs Hyperlocation?
Probe-RSSI: ~5 m @50% needing >=3 APs at -75 dBm with perimeter geometry, cadence hostage to client probing and MAC randomisation; FastLocate: data-frame RSSI, ~6-8 s, real MACs, extra synchronised radio; Hyperlocation: angle-of-arrival ~1-3 m, same AP density, serving-AP and antenna-orientation constraints.
Sources
- Cisco Press, CCIE Wireless v3 Study Guide (Alcantara, Darchis, Henry, Jimenez, Ziliotto) - RRM algorithms, RX-SoP/ATF, CleanAir, location technologies, PoE tiers.
- Cisco Live BRKEWN-2104 (Krischer, 2024): 6 GHz discovery (PSC/RNR/FILS), hidden-SSID implications.
- Cisco Press, End-to-End QoS Network Design, 2nd Ed. (Szigeti et al.) - WLAN QoS, EDCA/WMM, CAPWAP marking.
- Cisco Press, CWNA-aligned high-density and voice RF design guides.