Wireless problems: signal, interference and roaming

Why this matters

Wireless faults are the ones users report most and technicians measure least. The reason is that the evidence is invisible: a cable either has a link or it does not, while a radio link degrades continuously and the only thing most people can see is a bar chart that is not a measurement.

Replacing the bars with numbers is most of this lesson. Signal strength in dBm, signal-to-noise ratio, channel occupancy and retry rates are all readable with free tools, and once they are readable a wireless complaint becomes an ordinary fault with an ordinary cause.

The other half is knowing which faults are radio faults at all. A substantial share of wireless tickets are authentication, addressing or profile problems that have nothing to do with signal, and lesson 40's sequence is what separates them.

The lesson

Signal strength against signal quality, and why bars are not a measurement

Two different numbers, and they are not interchangeable.

• Signal strength, in dBm, is how loud the access point sounds at the client. It is negative, and closer to zero is stronger: around -50 dBm is excellent, -65 dBm is good, -70 dBm is workable, -80 dBm is marginal and below that is unusable.
• Noise floor, also in dBm, is everything else in the band. The difference between the two is the signal-to-noise ratio, and that is what actually determines throughput. A ratio above about 25 dB supports high data rates; below 10 dB the link barely functions.

Which is why a strong signal can still perform badly. A client at -55 dBm in a room where the noise floor is -65 dBm has only 10 dB of headroom and will run slowly, while a client at -70 dBm with a -95 dBm noise floor has 25 dB and will run well.

Bars are a vendor-chosen mapping of signal strength alone, with no agreement between devices, and they ignore noise entirely. Two phones side by side routinely show different bars. Using bars as evidence is the wireless equivalent of "it felt slow".

Other numbers worth reading where the tools expose them: the negotiated data rate, which falls as conditions worsen, and the retry rate, which rises. A link holding a high data rate with few retries is healthy regardless of what the icon shows.

Interference from other networks and from things that are not networks at all

Two categories, and they need different fixes.

Other Wi-Fi networks. Co-channel interference means sharing airtime politely, so throughput falls as the neighbours get busier. Adjacent-channel interference — partial overlap — is worse, because the transmissions are not coordinated. Lesson 10 explains why only channels 1, 6 and 11 avoid this at 2.4 GHz.

Everything else in the band, which is the category people forget:

• Microwave ovens, which blanket a large part of 2.4 GHz while running. The signature is unmistakable: wireless fails for two minutes at lunchtime, every day.
• Cordless phones and baby monitors on older 2.4 GHz designs.
• Bluetooth devices, which hop across the band.
• Wireless cameras and video senders, often continuous transmitters.
• Poorly shielded equipment — some lighting, some motors, some display cables.

A Wi-Fi analyser shows the first category. The second category is often invisible to one, because the interference is not a Wi-Fi signal — and the diagnostic is correlation instead: what else happens when the fault happens, and does moving to 5 GHz resolve it. Moving band is the fastest test available, because almost all non-Wi-Fi interference in scope is at 2.4 GHz.

Physical obstruction belongs here too. Concrete, brick, metal, foil-backed insulation, mirrors, water tanks and lift shafts all attenuate heavily. A dead spot that is geographically consistent is an obstruction, not a fault.

Channel overlap and congestion, and the fix that costs nothing

The most common actionable finding in small-office wireless is that everyone is on the same channel, because everyone accepted the default.

The procedure:

1. Survey. Run an analyser where the problem is, and list the networks, the channels and their strengths.
2. Count the occupancy on each candidate channel. At 2.4 GHz only 1, 6 and 11 are candidates.
3. Move to the least occupied, weighting by how strong the neighbours are — a distant network on channel 1 matters less than a strong one.
4. Narrow the channel width to 20 MHz at 2.4 GHz, which stops the network overlapping its neighbours.
5. Retest from the client, with numbers, and compare to the numbers before.

Other no-cost adjustments worth knowing:

• Move devices that can move to 5 or 6 GHz, which is less congested and carries more.
• Reposition the access point. Central, high, away from metal and away from the floor. Moving an access point off a desk and onto a wall regularly produces a larger improvement than any setting.
• Reduce transmit power where several access points cover one area, which is counter-intuitive and correct: it makes clients roam to the nearest access point instead of clinging to a distant loud one.
• Turn off the very lowest data rates, which forces distant clients to move closer or roam rather than consuming airtime at 1 Mbps.

Roaming between access points, and the sessions that break when it happens

Roaming is the client's decision, not the network's. A client holds its association until its own thresholds say to look for something better, and different devices have very different thresholds.

Sticky clients are the characteristic problem: a laptop carried from one end of a building to the other stays associated with the first access point at -85 dBm while a much stronger one is overhead. The symptom is a user whose wireless is terrible in a location where everyone else's is fine, resolved instantly by toggling Wi-Fi off and on — which is the diagnostic as well as the fix.

What can be done about it:

• Reduce access point transmit power, as above, so staying attached becomes less attractive.
• Ensure both access points share the same network name and security settings, or the client cannot roam between them at all — it must fully reconnect.
• Use the same passphrase or authentication, for the same reason.
• Fast roaming standards shorten the reassociation, which matters for voice and video.

The sessions that break during a roam are the ones with no tolerance for a gap: voice calls, video meetings, remote desktop sessions, and some VPN connections. A user who reports that calls drop while walking between rooms is describing a roaming problem, and that is a complete diagnosis rather than a complaint.

Authentication failures that present as a coverage problem

A meaningful share of wireless tickets have nothing to do with radio. They present as "it will not connect", which users interpret as signal.

The causes, and how each is recognised:

• Wrong or changed passphrase. The device reports an authentication error or simply loops. The fix is to forget the network and rejoin, because the saved profile from lesson 5 keeps presenting the old credential.
• A stale profile after an access point replacement, which behaves identically.
• Certificate problems on enterprise wireless — an expired certificate, a clock that is wrong, or a certificate the client does not trust. A machine whose date is wrong, from lesson 27's coin cell, fails wireless authentication for exactly this reason.
• Account problems — expired password, locked account, or the user not in the permitted group. The signature is one user failing on every device while others succeed on the same device.
• MAC filtering, from lesson 14, where a new device is simply not on the list.
• Guest network captive portal not appearing, which is usually a DNS or browser issue rather than a wireless one.

The separating question is the same three-way split as lesson 7: does the fault follow the device, the location, or the account? A fault that follows the user to another device is an account problem; one that follows the device to another location is a device problem; one that stays in a location is a radio problem. Asking it first prevents an afternoon spent surveying a building for a locked account.