Mandatory support is not a negotiated result
WebRTC audio requirements include Opus and G.711 PCMA/PCMU support. That interoperability baseline does not tell you which codec an individual call selected. Read the accepted SDP and actual stream rather than assuming every browser call uses Opus. A browser-to-PBX leg can use a different codec from the PBX-to-carrier leg. If the PBX must transcode, verify codec availability and resource use under expected concurrency. Passing a codec through and encoding or decoding it are different capabilities.
Model the complete audio path
Include microphone processing, codec rate, packetization, network loss and jitter buffering. A lower payload bitrate can reduce bandwidth while a poor path still produces unusable speech. Opus settings and actual rate can vary; a single advertised maximum is not a measured per-call consumption figure. DTMF may use negotiated RTP telephone events rather than audible tones alone. Test menus through the final carrier path. Echo cancellation and gain control also affect the user experience without necessarily changing SIP signaling.
Compare with controlled evidence
Use the same devices, network and test material when comparing codec configurations. Record the negotiated codec, packet interval, loss counters and subjective observations separately. Avoid publishing a universal quality ranking from a single headset test. For planning, use the bandwidth tool and state which payload and header assumptions apply. Include relay and encryption overhead when relevant. If the destination forces narrowband audio, choosing a wideband browser codec alone cannot make the entire conversation wideband.
Sources & applicability
Primary references for the technical details above. Operational examples and planning checklists are VoIP.info editorial guidance.
Examples require adaptation to your topology. No live PBX or hardware testing is claimed.