A neighbouring AP on the same channel and one on a nearby channel hurt your network in opposite ways. Move the neighbour's channel, change how strong and how busy it is, and see why sharing a channel is often less damaging than overlapping one.
The neighbour is on your channel. Your radios can decode its frames, so they take turns: you lose airtime, but every frame you do send is clean. Like one lane shared politely.
Adjacent-channel interference (ACI): corrupting
The neighbour is on a nearby channel whose energy leaks into yours. Your radios can't decode it, so unless it is strong they don't wait: both send at once, and the leakage garbles frames on both sides.
1 · How much of the neighbour lands in your channel
Your channel (blue) and a neighbouring AP's channel (orange), drawn as the 802.11 transmit spectrum: flat across the channel, then falling off steeply at the edges. The shaded part is the neighbour's energy that falls inside your channel.
Band
Neighbour's channel
Neighbour's transmitter
Reference: 802.11 transmit spectrum masks
OFDM, 20 MHz (802.11a/g)
The mask (red) slopes: 0 dBr to ±9 MHz, −20 dBr at ±11, −28 dBr at ±20 and −40 dBr from ±30 MHz. An OFDM signal is almost flat across its subcarriers and drops steeply at the edge; its skirt must stay under the mask. This is the shape the tool uses. Wi-Fi 6 and 7 (802.11ax/be) use a slightly different 20 MHz mask (0 dBr to ±9.75 MHz, −20 dBr at ±10.5 MHz, then the same −28 and −40 dBr), which changes the leakage into a neighbouring channel by less than 0.5 dB.
DSSS, 22 MHz (802.11b)
The older 802.11b mask steps down: 0 dBr to ±11 MHz, −30 dBr to ±22 MHz, −50 dBr beyond. The raw DSSS signal has a sin(x)/x shape whose side lobes (only 13 dB down) break the mask, so the radio filters them off. Its 22 MHz width is why 2.4 GHz channels 1, 6 and 11 are the non-overlapping set.
What the mask means
The spectrum mask is the maximum power a transmitter may put at each distance from its channel centre, measured relative to its own in-channel level (dBr). The transmitted spectrum must stay under that outline at every frequency: it is an envelope, not an average.
It covers every kind of out-of-channel energy, not only side lobes. For DSSS that is the sin(x)/x lobes. For OFDM it is a mix: each subcarrier's small side lobes (from the abrupt start and end of every symbol), distortion from the power amplifier ("spectral regrowth", which grows at high transmit power), phase noise and other imperfections. The mask limits the total, whatever causes it.
It is relative. −28 dBr at ±20 MHz means 28 dB below this radio's own in-channel level, so a 30 dBm transmitter may leak more absolute power than a 10 dBm one, as long as its shape stays under the outline.
There is also an absolute floor: in 802.11ax/be the spectrum may always reach −53 dBm/MHz in 2.4 GHz and −39 dBm/MHz in 5 and 6 GHz, even where that is above the relative mask. It only matters for low-power transmitters.
It is measured in fine slices: the power density in a narrow 100 kHz measurement bandwidth, compared with the highest level inside the channel, so a narrow spur can't hide by being averaged away.
Regulators add their own absolute limits at the band edges (FCC, ETSI, NBTC and others), in dBm rather than dBr. On channels next to a band edge, such as 2.4 GHz channels 1 and 11 or 5 GHz channels 36 and 165, that can be stricter than the 802.11 mask, which is why APs often use less power there. See the Wi-Fi Channel Map for the power limits per country.
It is the worst case a compliant transmitter is allowed. Real radios usually stay 5–15 dB below it; the tool's "Typical" option models that.
The receiver matters too. The tool treats your receiver as a perfect filter. The standard's minimum adjacent channel rejection (802.11be Table 36-68: 16 dB at MCS 0 down to −14 dB at MCS 11) works out, for every MCS, to about 23 dB of total isolation from the next 20 MHz channel and 39 dB from the one after, for the worst compliant pair of radios. That is about 3 dB less than the "At the mask limit" setting assumes; good radios do much better.
2 · What it does to your link
Your client is connected to your AP; a neighbouring AP and its clients are busy on the channel chosen above. "Neighbour's signal" is how strong the neighbour is where your devices are. 802.11ax, 20 MHz, 2 spatial streams.
Preamble detect (CCA)
On the air
Your framesNeighbour's framesOverlapping, harmlessOverlapping: slower rateOverlapping: frame lostWaiting (deferring)
3 · 2.4 GHz channel plans
Three APs in a row. Your AP is the middle one; the two either side are about equally strong at your clients. Which channels they use decides whether they share with you (CCI), leak into you (ACI) or leave you alone.
4 · Designing around both
CCI and ACI need different fixes.
Less CCI: fewer radios hearing each other
Reuse channels so that APs on the same channel are far apart; lower the transmit power so cells don't overlap more than needed; use narrower channels (20 MHz) and the extra channels of 5 and 6 GHz; remove low basic rates so frames spend less time on the air. On the same channel, fewer overlapping cells means more airtime each.
Less ACI: keep neighbours off the next channel
In 2.4 GHz use only 1, 6 and 11 (or 1, 5, 9, 13 for OFDM-only networks where 13 channels are allowed, if the neighbours do the same): anything in between overlaps. In 5 GHz, adjacent 20 MHz channels don't overlap, but a strong neighbour still leaks about 26 dB below itself (less with a good radio), so give APs close to each other channels further apart (36 and 44, not 36 and 40).
Close together is where ACI bites
ACI grows with how strong the neighbour is. Two APs mounted back to back, two radios in one box, or a client standing right under someone else's AP: the neighbour arrives at −35 dBm or more, and even 30 dB of filtering leaves interference well above the noise.
Energy detect: when ACI turns into waiting
A radio waits for Wi-Fi on its own channel from −82 dBm (preamble detect), but for energy it can't decode, such as a neighbour on another channel, only from −62 dBm (energy detect: 20 dB higher). Anything below that is ignored but still adds noise: the closer to −62 dBm, the more it corrupts. That band is where ACI does its damage.
Other Wi-Fi devices on exactly your channel. Because they can be decoded, everyone takes turns: the channel's airtime is shared. It costs capacity, not frame quality.
Energy from a transmitter on a nearby channel that spills into yours. It can't be decoded, so it acts like extra noise, and radios often don't wait for it.
2.4 GHz channels are only 5 MHz apart but about 20 MHz wide, so neighbours overlap; 1, 6 and 11 are the usual set that doesn't (it dates from 22 MHz wide 802.11b). For 18 MHz wide OFDM, 1, 5, 9 and 13 also clear each other, where all 13 channels are allowed. In 5 and 6 GHz, 20 MHz channels sit side by side without overlapping.
The maximum power a transmitter may put at each distance from its channel centre, relative to its own in-channel level. It limits all out-of-channel energy (side lobes, amplifier distortion and noise). For 20 MHz OFDM: −20 dBr at 11 MHz from the centre, −28 dBr at 20 MHz and −40 dBr from 30 MHz. Real radios usually stay 5–15 dB below it.
How much stronger a signal on the next channel may be than the wanted signal before frames start failing, in dB. The standard's minimum ranges from 16 dB at MCS 0 to −14 dB at MCS 11 (the fast rates tolerate less); better filters mean more tolerance of strong neighbours.
A radio treats the channel as busy when it detects the start of a Wi-Fi frame on its own channel at −82 dBm or stronger (20 MHz), the minimum sensitivity of the slowest data rate (IEEE 802.11-2024 27.3.22.6.3; IEEE 802.11be-2024 36.3.21.6.3 for Wi-Fi 7). Many radios detect weaker preambles; the −86 dBm option shows one.
An 802.11ax feature: a radio may ignore frames from another network (identified by its BSS colour) up to a higher threshold (the standard allows up to −62 dBm; −72 dBm is used here), and transmit over them at reduced power, so nearby cells on the same channel share less.
For energy it can't decode, such as leakage from another channel or non-Wi-Fi signals, a radio treats the channel as busy only from −62 dBm: 20 dB above the −82 dBm minimum sensitivity (IEEE 802.11-2024 17.3.10.6.1). It stays at −62 dBm whichever preamble-detect option you pick.
A strong nearby transmitter drowning out a weak wanted signal. For ACI it means a neighbour close to your client can hurt even from another channel.
↗ opens a term's full entry in the Wi-Fi Glossary, which explains over 400 Wi-Fi terms with their IEEE 802.11 references.
Notes
Leakage: the neighbour's spectrum follows the 802.11a/g 20 MHz OFDM transmit mask (0 dBr to ±9 MHz, −20 dBr at ±11, −28 at ±20, −40 dBr from ±30 MHz); the 802.11ax/be mask (0 dBr to ±9.75, −20 dBr at ±10.5 MHz) gives leakage within 0.5 dB of it. The energy inside your channel is that shape integrated over your channel's ±9 MHz. Real radios are usually 5–15 dB cleaner than the mask outside their channel ("Typical" subtracts 10 dB). The receiver is treated as a perfect filter; real receivers add some adjacent-channel leakage of their own. The standard's minimum adjacent channel rejection allows about 23 dB of total isolation from the next 20 MHz channel for the worst compliant radios, 3 dB less than "At the mask limit" assumes.
Reference levels: noise floor −90 dBm; preamble detect at the standard's −82 dBm, the level at which a radio must decode a Wi-Fi preamble on its own channel and wait; energy detect at −62 dBm, 20 dB above it, the level at which it waits for energy it can't decode, such as a neighbour on another channel (IEEE 802.11-2024 17.3.10.6.1, IEEE 802.11-2024 27.3.22.6.3). The "−86 dBm (sensitive radio)" option is a radio that decodes preambles 4 dB above this noise floor; "802.11ax spatial reuse" ignores other networks' preambles up to −72 dBm. Energy detect stays at −62 dBm in both (spatial reuse also lowers the transmit power, which this model leaves out). Partly overlapping 2.4 GHz channels are treated as not decodable.
Throughput: 802.11ax 20 MHz, 2 streams, with the MCS chosen from the SINR using the standard's minimum sensitivities (4 dB SNR for MCS 0 up to 34 dB for MCS 11: the 802.11be minimum sensitivities for 20 MHz, −82 to −52 dBm, measured against a −86 dBm reference floor); noise floor −90 dBm; 65% of the PHY rate reaches the user. When your radio waits, you get the airtime the neighbour leaves (half if you are both busy). When it doesn't, a frame is hit if a neighbour frame overlaps it (chance 1 − e−2 × busy, as for random access with equal frame lengths), and rate control picks the MCS with the best average: either a fast rate that fails when hit, or a slower one that survives the interference.
Slower or lost: when frames overlap, rate control can either drop to a rate that survives the interference (purple in the timeline) or stay fast and lose the overlapped frames (red). The model picks whichever gives the higher average throughput, as rate-control algorithms such as Minstrel broadly do. Losing frames is chosen when the surviving rate is very slow: with your signal at −60 dBm and the neighbour 50% busy, that happens once the interference reaches about −76 dBm, until energy detect (−62 dBm) makes your radio wait instead. The average hides the cost of the losses: every lost frame means a retry, extra delay and jitter, which matter more for voice and video than for downloads.
Both sides suffer: when the two cells don't wait for each other, the neighbour's frames get hit too. With CCI, nobody's frames are corrupted; everyone just gets less airtime.
Simplified: one client per cell, no retries counted separately, no hidden nodes, no capture effect. Real numbers vary with the radios, the traffic and the building; use it to compare situations, not to predict exact throughput. For CCI in design, see also the article Design for Capacity vs CCI.