Measurement, troubleshooting and quality of experience

“The Wi-Fi is slow” is a symptom, not a diagnosis. This topic gives you the measurements that tell the likely causes apart, and a way to relate them to what the user feels.

-101 dBmthermal noise floor in 20 MHz

SNR = S – Nsignal minus noise, in dB

150 msone-way delay ITU-T G.114 regards as acceptable for most calls

retriesthe clearest sign of wasted airtime

1. DefinitionWhat to measure

Signal strength (RSSI or dBm): how strong the wanted signal is at the receiver. Different devices scale it differently.
Noise floor: the background level of unwanted energy, including thermal noise and interference.
Signal-to-noise ratio (SNR): signal minus noise floor, in dB. It decides which rate a link can hold.
Data rate and MCS: the physical rate chosen for each frame.
Retry rate and loss: how many frames had to be sent again.
Channel utilisation: how much of the time the channel is busy.
Throughput, latency, jitter, loss: what the application experiences. Together they make up the quality of experience (QoE).

2. PurposeWhy measure

Wi-Fi problems have many causes that feel the same to a user. A low rate can mean weak signal, high noise, a slow client or a busy channel; each needs a different fix. Measuring turns a guess into a short list.

3. MechanismFrom signal to rate

The model shows which rates a client can hold for a given signal and noise floor, using the receiver sensitivities the standard requires. Raise the noise floor and watch the usable rate fall even though the signal is unchanged.

Try this. (1) Set the signal to -60 dBm and the noise floor to -95 dBm. (2) Raise the noise floor to -80 dBm without touching the signal. (3) Notice the margin on the best rate vanishing: a strong signal on a noisy channel is not a good link.

4. VisualDoes the link meet the need?

Users do not feel signal strength; they feel video that stutters or a call that clips. Enter measured throughput, latency, jitter and loss and see which applications are served well.

5. MathematicsNoise, SNR and the budget

Thermal noise = -174 dBm/Hz + 10 log₁₀(bandwidth in Hz). For 20 MHz that is -174 + 73 = -101 dBm; for 40 MHz -98 dBm; for 80 MHz -95 dBm; for 160 MHz -92 dBm. Doubling the width raises the floor by 3 dB.
The receiver adds its own noise (its noise figure). The standard’s minimum sensitivities assume a 10 dB noise figure and 5 dB implementation margin, which is about -91 dBm in 20 MHz.
SNR = signal – noise floor. A signal of -60 dBm on a -95 dBm floor has 35 dB of SNR; the same signal on a -75 dBm floor has only 15 dB.
Retry rate = frames with the retry bit set / data frames. The airtime lost is roughly proportional.
Throughput can be tested with iperf3: iperf3 -c server -t 30 -P 4 for TCP with four parallel streams, -R for the reverse direction and -u -b 20M for a UDP stream at 20 Mbit/s.

6. Real-world behaviourSymptoms and likely causes

Symptom	Likely causes	Check
Slow everywhere	Busy channel, interference, wide channels	Channel utilisation, retry rate, noise floor
Slow in one place	Weak signal or poor SNR, a wall	Signal and SNR on the spot; walk around
Drops when walking	Sticky client, thin coverage between access points	Reassociation frames, signal at the moment of roam
Fine, then very slow at times	Neighbour on the same channel, a non-Wi-Fi source, DFS channel change	Time-based capture, spectrum view, DFS events
Good download, bad calls	Jitter or loss, not throughput	Ping with a long run, retry rate, QoS settings
Fast Wi-Fi, slow internet	Not Wi-Fi at all	Test wired to the same router; check DNS and the uplink

Work from the bottom up. Check the signal and noise first, then the rates and retries, then the protocol steps, then the application.
Change one thing at a time and measure again.
Measure from the client’s side. The access point’s view of the link is not the client’s.

Practical validationProve it in a capture and on a spectrum analyser

Reading about a band is one thing; seeing it in your own evidence is the skill. These checks work with any monitor-mode capture:

Capture on the channel and filter for retries: wlan.fc.retry == 1. Then look at the signal in radiotap.dbm_antsignal and the data rate in wlan_radio.data_rate for the frames that are failing.
Find TCP problems: tcp.analysis.retransmission and tcp.analysis.lost_segment.
Run ping -c 100 gateway and look at the maximum and the spread of the round-trip times, not only the average.
Question to answer: a user has -55 dBm but only 24 Mbit/s. List three measurements you would take, in order.

Check your understandingQuick quiz

ReferencesSources and further reading

Receiver sensitivities are from IEEE Std 802.11-2020 (Tables 17-18 and 21-25) with the noise figure and margin assumptions stated in 17.3.10.2. The QoE targets are planning rules of thumb, not standards; ITU-T G.114 is the only recommendation cited. Check Wireshark field names against your version, since they are updated over time.