When I make VOIP calls, the other person always hears an echo. Some people are quite annoyed to the edge of their tolerance. But everything sounds fine to me. They often say “turn off speakerphone”. I never use that. Only headsets. We can usually function well enough but conversations are kept short.

What about robots? Robots have zero tolerance for feedback. Even though I sit there in silence the bot sounds like a broken record: “press 1 to … press 1 to … press 1 … press … I don’t understand … press 1 to…” until it gives up and redirects to an operator.

The bots are completely dysfunctional because they hear themselves echo and they don’t know the echo is of their own transmission. I really needed to talk to a bot because I had to renew a license and the human operators had a policy to refuse to accept payment details. Only the bot could process payments.

An operator said “turn off speakerphone”. I said it cannot be feedback because I’m on a headset. She said: “it’s that… you cannot use a headset”. I thought buuullshit!, she doesn’t know how feedback works. My piezo-based headset even had rubber gaskets so there was no chance of speaker output reaching the mic. But I tried a different headset anyway. It worked! She was right - it was the headset.

WTF? Apparently the stock headset that came with my cheap phone is so shitty that the input signal bleeds into the output signal electromagnetically via induction or something.

VOIP makes the feedback much worse than POTS because of the extra latency due to compression+decompression, etc… the longer the delay the more annoying an echo is.

What we need are apps to test headsets for feedback. In principle an app should be able to play sample audio and listen to the mic to see if any of it comes back through. It should probably be a built-in feature of VOIP software. I checked f-droid.org, found nothing.