Why calls sound bad: a diagnostic guide

Not every bad call has the same cause. Choppy audio, delay, a robotic voice, and total silence in one direction each point at a completely different problem.

Choppy audio, missing words

This is almost always packet loss, or jitter severe enough to overwhelm the jitter buffer. Packets that never arrived, or arrived too late to be useful, leave real gaps in what should have been continuous audio.

A delay before you hear a reply

This is latency, plain and simple, the raw travel time from one side to the other. A large jitter buffer can also add to this, since it is deliberately holding audio back to smooth it out, trading some of this exact delay for fewer dropouts.

Robotic or metallic sounding voice

This points at the codec rather than the network. A codec compressing too aggressively for the available bandwidth, or a poor implementation of it, can leave audio sounding processed and artificial even when the network itself is behaving fine.

One person can hear, the other cannot

This is a completely different category of problem, not a quality issue at all. It almost always traces back to NAT traversal failing in one direction, media being sent to an address that does not actually work.

Putting a number on it

Whatever the symptom, it eventually shows up as a lower MOS score, the one number that tries to summarise all of the above into a single measure of how the call actually sounded.