Audio quality: understanding sound problems

Drops, choppy voice, echo, one-way audio… In the vast majority of cases a sound problem does not come from the PBX: it comes from the network or a third-party device. This guide helps you recognise the symptoms, understand their causes and isolate them before opening a ticket.

🎧 Listen and compare

To speak the same language, nothing beats your ears. These excerpts are demonstrations we produced ourselves (synthetic test voice) to illustrate each symptom. Listen to the reference first, then compare. Tip: click an excerpt's title to jump to its explanation.

Demonstration excerpts produced in-house (no third-party files). For the full theory, see the Cisco reference source at the bottom of the page.

Recognise the symptoms

• Choppy / robotic voice — chunks of speech are missing. Typically packet loss or network jitter.
• Full drops — silences during the call: unstable link, weak Wi-Fi, bandwidth saturation.
• Echo — you hear yourself with a slight delay: often a phone or headset, sometimes a level/latency mismatch.
• One-way audio — one party can't hear the other: the classic signature of a NAT or firewall blocking the audio (RTP).
• Latency / gaps — you talk over each other: network delay too high (routing, saturated link, VPN).
• Crackle / static — continuous noise: cabling, power, or hardware interference.

Causes, from most to least common

1. The network (the main culprit)

Voice over IP is very sensitive to the network. Four indicators matter:

• Packet loss — beyond ~1%, voice degrades clearly.
• Jitter — the variation in delay between packets; it makes voice choppy.
• Latency — beyond ~150 ms one-way, conversation becomes uncomfortable.
• Bandwidth / QoS — a saturated link with no priority for voice degrades any call as soon as a large transfer happens.

2. NAT and firewall

SIP carries signalling, but audio flows separately (RTP, over a range of UDP ports). If NAT is poorly translated or the firewall blocks the RTP ports, you get a call that "rings" but has no audio, or one-way audio. The SIP ALG in many consumer routers, meant to help, often does the opposite: it rewrites packets badly. It is usually better to disable it.

3. Third-party equipment

Phones and ATAs, headsets, consumer routers and boxes, Wi-Fi access points, or the carrier trunk: each can introduce echo, drops or crackle. Wi-Fi in particular is rarely suited to voice; prefer wired for testing.

4. Codecs

A codec mismatch or needless transcoding can distort voice, as can poorly transmitted DTMF (key tones). Setting a common, simple codec (e.g. G.711 on the LAN) helps isolate the problem.

Diagnose efficiently

1. Cross-test — another phone, another network (4G, other site), wired vs Wi-Fi, another SIP account. Isolate the faulty element.
2. Check NAT / firewall — RTP ports open, SIP ALG disabled, public address seen correctly.
3. Measure the network — packet loss, jitter and latency to the server/carrier, especially at peak times.
4. Reduce variables — a single codec, a local call then an external one, with and without VPN.

Network best practices for voice

• Give priority (QoS) to voice traffic on the network.
• Prefer wired for fixed phones; a properly sized Wi-Fi otherwise.
• Disable SIP ALG on consumer boxes/routers.
• Ensure enough bandwidth and avoid saturating the link.
• Open the necessary RTP ports and check NAT translation.

Reference source. This guide summarises, in our own words, the main principles of voice-quality troubleshooting. For a deeper technical reference, Cisco's documentation is authoritative: Troubleshooting Voice Quality — Cisco (original article). Credit: Cisco Systems.