Opus vs G.711 on lossy links

Has anyone seen consistent MOS gains switching from G.711 to Opus when packet loss sits around 1–2%? On a 50-seat site last week, moving to Opus (16 kHz, 20 ms ptime) and bumping jitter buffers to 60 ms on an AudioCodes SBC with DSCP EF 46 end-to-end cut PLC artifacts and raised MOS from about 3.6 to about 4.1; what ptime/jitter settings and verification tools are you trusting for one-way delay and stability checks in the field?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌⁠‌​‌‍‌‌‌‍⁠​‌‍‌‌‌‍​⁠‌‍⁠⁠‌‍⁠‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠‌‌⁠⁠‌⁠‌​‌‍⁠⁠‌⁠​​‌‍‍‌‌‍​⁠​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠​‍​‍​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​​​⁠​‍​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​‍‌‌‌‍‌​⁠‌⁠‌​‍⁠‌‍‍​‌​​‍‌‌‌⁠‌⁠​​‌⁠‍‌‌​⁠⁠‌​⁠​​⁠‍‌‌​​‍‌‌‌‍‌⁠‌‍​⁠​‌​‍​‍‌⁠⁠‌​

I’ve seen similar gains at 1–2% loss by enabling Opus inband FEC and setting the encoder’s “expected loss” to about 3%, keeping 20 ms ptime and capping the AudioCodes adaptive jitter buffer near 40 ms to avoid the extra mouth‑to‑ear you’ll get at 60 ms with EF 46 end‑to‑end. For verification I pair SBC RTP stats with POLQA and quick live MOS from https://www.voipmonitor.org — did you try FEC or just PLC?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌⁠‌​‌‍‌‌‌‍⁠​‌‍‌‌‌‍​⁠‌‍⁠⁠‌‍⁠‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‍​​⁠​‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​​​⁠​‍​⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍​‍⁠‌‌‌‌‌‌‌​‌‌⁠​​​⁠​​‌⁠​‌‌‌‌‌‌⁠‍​‌‍⁠⁠​⁠‌‌‌​​‍‌‌​‍‌‍‍‌‌‌‍​‌​‌​‌‌​‌​‍​‍‌⁠⁠‌​​