The three numbers that matter
Voice is sensitive to timing. Unlike a file download, a late packet is as useless as a lost one. These are the commonly used targets for good quality, based on ITU-T guidance (G.114 recommends one-way delay below 150 ms, and treats delay above 400 ms as unacceptable for general use):
- Latency: one-way delay of 150 ms or less. Above about 400 ms, conversation becomes difficult.
- Jitter: 30 ms or less. Above about 50 ms you will notice choppy or garbled audio.
- Packet loss: 1% or less. Higher loss produces gaps, clipped words, and robotic voices.
Common symptoms and what they mean
- Choppy or robotic audio: usually packet loss or high jitter, often from Wi-Fi or a congested link.
- Delay and talking over each other: high latency, often from long network routes or overloaded connections.
- One-way audio or no audio: typically a firewall, NAT, or SIP ALG problem blocking the media stream.
- Echo: usually speakerphone feedback or a poor headset, not the network.
- Calls dropping after a fixed time: often a NAT timeout or session timer mismatch.
How to improve call quality
- Use wired Ethernet for desk phones and desktop computers wherever possible.
- Enable QoS or voice prioritization on your router or switch so calls are served before bulk traffic.
- Make sure bandwidth covers your peak number of concurrent calls. A G.711 call uses roughly 80 to 90 kbps each way with overhead; G.729 uses roughly 30 to 35 kbps.
- Choose an appropriate codec: G.711 for quality on good networks, a compressed codec when bandwidth is tight.
- Use a quality headset and avoid open speakers in noisy rooms.
- Disable SIP ALG on routers unless your provider recommends it, and check firewall rules for the provider's required ports.
- Keep apps, phone firmware, and operating systems up to date.
- Add a secondary internet connection or failover if phones are business-critical.
How to test and troubleshoot
Start by finding the scope. If one user has the problem, look at their device, headset, and Wi-Fi. If one location has it, check the local network and internet link. If everyone has it, check the provider and carrier routes. Test during business hours rather than at night, compare wired and Wi-Fi results, and run a jitter and packet loss test, not just a speed test. Call logs and reports help you spot patterns by time of day or destination.
Common questions
- How much bandwidth does a VoIP call need?
- It depends on the codec. With overhead, G.711 uses roughly 80 to 90 kbps per call in each direction and G.729 roughly 30 to 35 kbps. Stability matters more than raw speed.
- What is a good jitter value for VoIP?
- Aim for under 30 ms. Sustained jitter above about 50 ms is likely to cause noticeable audio problems.
- Why do I hear echo on VoIP calls?
- Echo usually comes from speakerphone feedback, low-quality headsets, or a device picking up the other party's audio. A headset typically resolves it.
- Why is there one-way audio?
- One-way audio is commonly caused by firewall or NAT settings blocking the media stream. Review SIP ALG, port, and firewall configuration.
- Is Wi-Fi good enough for VoIP?
- It can be with a strong signal and low interference, but wired Ethernet is more consistent for desk phones and desktops.