WiFi Calling Quality Problems? The Real Culprit Is Usually Not Signal Strength
A video call that stutters and drops despite showing four bars of WiFi. A VoIP call where the other end sounds like they are speaking through a wall, even though a browser-based speed test shows 200 Mbps. These are the symptoms of a WiFi problem that raw signal strength measurement does not capture, because the issue is usually not signal — it is latency, jitter, and packet loss at the levels that voice and video codecs cannot tolerate.
Why Voice Is Different From Everything Else
File downloads, video streaming, web browsing, and most internet applications are delay-tolerant. TCP handles packet loss through retransmission; buffering absorbs latency variation. A web page that takes 150ms instead of 50ms to respond is imperceptible to a user. A file download that briefly stalls and resumes is invisible. These applications adapt to available conditions without degrading user experience.
Real-time voice and video are not delay-tolerant. VoIP codecs sample audio at 20ms intervals. Each 20ms sample is packetized, transmitted, and played back at the far end. If a packet is lost — not retransmitted, just dropped — a 20ms gap appears in the audio stream. If a packet arrives late — after the jitter buffer’s deadline — it is discarded, same effect. If packets arrive with wildly variable delays (high jitter), the receiving codec either extends the jitter buffer (adding perceived latency to the call) or discards late packets (producing gaps).
A VoIP call is unacceptable with 1% packet loss. It is noticeably degraded with 0.5% loss. A file download is barely affected at 5% loss — TCP recovers gracefully. The tolerance bands differ by two to three orders of magnitude.
The Actual Sources of VoIP Degradation Over WiFi
Contention and channel busy time. When the WiFi channel is heavily utilized — by other devices on your network, by neighboring networks on the same channel, by interference sources — your device’s VoIP frames must queue behind other transmissions. The CSMA/CA backoff mechanism serializes access to the medium. During peak congestion, a VoIP frame intended for 20ms delivery may be delayed 50 to 100ms waiting for the channel to clear. The jitter buffer absorbs some of this, but sustained congestion produces sustained jitter that overwhelms the buffer.
Power save behavior. Modern devices aggressively power down WiFi radios between transmissions to extend battery life. Standard 802.11 power save mode wakes the radio to receive buffered frames at each DTIM beacon interval. With a router configured at the default DTIM interval of 2 or 3, the radio sleeps for 200 to 300ms between wake-ups. A VoIP packet arriving during a sleep period is buffered at the AP and delivered at the next beacon — adding 100 to 300ms of delivery latency. The DTIM interval directly sets the worst-case power-save induced latency for incoming VoIP audio.
WMM configuration. WiFi Multimedia (WMM) defines four traffic access categories with different contention parameters — Voice, Video, Best Effort, and Background. Voice traffic marked as AC_VO uses smaller contention window parameters, allowing it to access the channel more quickly than background traffic. This prioritization only functions when both the AP and the client device mark VoIP traffic with the appropriate access category. If VoIP traffic is marked as Best Effort (the default for unmarked UDP) because the application does not set DSCP values and the router does not classify traffic, VoIP frames compete equally with everything else on the network.
Upstream congestion. Most home internet connections have asymmetric bandwidth with upload capacity significantly lower than download. A video call uploading at 2 to 3 Mbps on a 10 Mbps upload connection may seem fine, but if another device simultaneously initiates a cloud backup or a firmware update at the same time, the upload is saturated. VoIP packets queue behind bulk data packets in the router’s upload queue, adding the queuing latency to every transmitted frame. The result is the characteristic one-way latency issue where your voice sounds fine but you sound degraded to the other end — your upload is constrained, their download is fine.
The Interventions That Actually Work
Set DTIM interval to 1. In the router’s advanced wireless settings, DTIM interval (also called DTIM period) controls beacon frequency. Setting it to 1 means buffered frames are delivered at every beacon (every 100ms) rather than every 2 or 3 beacons. Battery-powered devices will drain marginally faster, but VoIP latency from power save behavior is eliminated. This is the single highest-impact change for voice quality.
Enable WMM and mark VoIP traffic. WMM should be enabled by default on all WiFi 5 and WiFi 6 certified hardware. Verify it is not disabled. For the marking to work end-to-end, the VoIP application must mark its traffic with DSCP EF (Expedited Forwarding) or CS6, the router must either pass this marking through or classify traffic itself, and the WiFi radio must map the DSCP value to AC_VO. Most enterprise VoIP clients (Teams, Zoom with correct QoS configuration, Cisco Jabber) mark traffic correctly. Consumer VoIP applications vary.
Implement upstream QoS. Standard consumer router QoS only manages downstream traffic. Upstream shaping — limiting bulk uploads to below the available upload capacity and prioritizing VoIP traffic in the remaining budget — requires either a router with bidirectional QoS (Netgear’s DumaOS, ASUS AI QoS) or replacement of the router with one running OpenWrt with SQM and fq_codel. This is the fix for the “I sound bad on calls when someone else is uploading” problem.
Reduce co-channel interference. As described in channel planning guidance: survey neighboring networks, select the least-congested channel, and verify the choice manually rather than relying on auto-select. Co-channel interference increases channel busy time, which increases CSMA/CA backoff delay, which increases VoIP jitter.
Place the VoIP device close to the AP. At strong signal, the radio negotiates a high MCS rate, reducing the airtime each VoIP frame consumes (and thus the delay while the frame is transmitted). At marginal signal, frame retransmissions increase latency on individual frames by 10 to 50ms — significant for a 20ms sample interval.
The diagnostic sequence for voice quality problems: check DTIM, check WMM, check channel utilization during calls, check upload saturation. Address in that order. Signal strength is usually the last thing to investigate.