Open app

Full signal, no usable throughput

very common

“Phone shows full bars standing right under the AP but nothing loads. Speed test crawls or just times out.”

Likely causes

  • Channel is busy: co-channel contention or a neighbour eating the airtime

    30%

    RSSI -45 dBm but channel utilization above 60-70% and retries over 20%. Signal is strong, airtime is gone. Utilization stays high even with your client idle.

  • Low SNR from a raised noise floor (non-Wi-Fi energy)

    20%

    Noise floor well above the usual -90 to -95 dBm on 5 GHz (say -80 dBm) while Wi-Fi channel utilization is low. Spectrum view shows energy that is not 802.11: continuous, swept, or fixed duty-cycle blocks.

  • Uplink-only failure: AP shouting louder than the client can answer

    20%

    Download is fine, upload dies. The AP reports the client much weaker than the client reports the AP, by more than the two transmit powers can explain. Radio sitting at 23 dBm against a 10-13 dBm handset.

  • Wired side, not RF: saturated AP uplink, 100 Mbps negotiation, CRC errors, choked WAN

    15%

    iperf to a server on the AP's own VLAN is fine, anything past the AP switchport is slow. Switchport shows 100M or climbing CRCs. Every client on every AP on that switch is equally bad.

  • Client stuck at a legacy rate or one spatial stream

    10%

    Controller client view shows 6-24 Mbps and 802.11a/g/n at -45 dBm. A second device in the same spot gets full rate, so it is the client, not the cell.

  • Associated but not actually on the network: DHCP, DNS, captive portal or 802.1X

    5%

    Association succeeds and the client has a 169.254 address or no DNS; ping by IP works, by name does not. If the association survives but the EAP exchange never completes, it is authentication: go to Cannot connect: 802.1X, RADIUS, certificates.

What to bring

  • Wi-Fi analyzer app on phone or laptop
  • hand-held spectrum analyzer covering 2.4/5/6 GHz
  • laptop plus an iperf server on site
  • controller or AP GUI credentials
  • test patch cords

Steps

  1. Read the radio's own stats before touching anything

    Pull channel utilization, noise floor and client retry rate off the controller for that radio. Healthy: utilization under 40-50%, noise near -90 to -95 dBm on 5 GHz, retries under 10-15%. Anything over 70% utilization is your answer and you are done diagnosing.

    If that doesn’t do it

    All three healthy: skip to the wired test, step 4.

  2. Test up and down separately

    Run iperf or a speed test and watch each direction on its own. Download good with upload dead is an asymmetry problem, not a coverage problem, and no amount of extra TX power fixes it; take that result into step 3.

    If that doesn’t do it

    Both directions equally bad: the asymmetry check in step 3 has nothing to tell you, so skip it and take the RF out of the picture at step 4.

  3. Compare the two RSSI readings against the two transmit powers

    Read the AP's current TX power and the client's TX power (or its published maximum) and subtract. That difference is the gap you should expect: an AP at 23 dBm talking to a 10-13 dBm handset is legitimately 10-13 dB louder one way with nothing wrong. Only the excess over that predicted gap, roughly 6 dB or more, is evidence of anything. Client-reported RSSI is uncalibrated and vendor-specific, so lean on the direction of the asymmetry and your step 2 upload result, not the absolute figures. The fix is bringing AP TX power down toward what the client can answer with (indoors that usually lands 5 GHz in the low-to-mid teens dBm) rather than chasing a number.

    If that doesn’t do it

    Gap is no wider than the TX-power difference predicts and both sides are strong: move to the wired path.

  4. Take the RF out of the picture

    Patch a laptop into the AP's own drop, or a spare port on the same switch, and rerun the identical test. Full speed wired proves the fault is RF. Slow wired proves it never was.

    If that doesn’t do it

    Wired is slow too: check the switchport for 100M negotiation, CRC/error counters, uplink utilization and WAN load, and stop chasing the antenna.

  5. Look at the spectrum, not the SSID list

    Hand-held analyzer parked on that channel for a full 60 seconds. Non-802.11 energy that comes and goes with the ovens, a camera or a wireless HDMI sender, with the noise floor lifting alongside it, is your interferer. Usual suspects: microwave ovens and video senders at 2.4 GHz, 5.8 GHz sensors and cordless gear at the top of 5 GHz.

    If that doesn’t do it

    Spectrum clean: the airtime is being spent by other Wi-Fi. Go to APs fighting each other for airtime.

  6. Move the channel and re-measure in the same spot

    Put the radio on the quietest 20 MHz you found and repeat the test standing where you started. Going from 20 Mbps to 300+ confirms it. Record before and after so the customer sees the same numbers you do.

    If that doesn’t do it

    No change on a clean channel with 25 dB SNR and a healthy wired path: suspect the client. Prove it with a second, known-good device before you re-terminate, re-aim or re-quote anything.

References

  • IEEE 802.11-2020
  • TIA TSB-162-A

More Wireless & DAS guides

All 54 guides