TOPIC:

Hi all, team,

We're currently playing with the AEC settings available in firmware 1.5.

The echo cancellation itself is pretty efficient. To those interested in that feature, it's really easy: just feed the UMA-8 "sink" device with the reference signal you want.

Some AEC settings are pretty self explanatory once you have a basic grasp of the algorithms involved, some are definitely not, especially the ones that pertain to doubletalk management.

In our tests, using a simple microphone at the far-end and the UMA-8 at the near-end, we could confirm that using the default settings, while the far-end is talking the AEC DT simply cuts the near-end signal input: the far-end doesn't hear the near-end anymore, or (in some cases) the signal is completely garbled.

You can imagine what happens when both sides use an UMA-8 with AEC on, and try to talk at the same time: no one can hear a sensible word...!

We tried and play with the AEC DT settings, but couldn't find a way to really tell what does what. What we are trying to achieve is a setup that:
- efficiently cancels outs echos
- doesn't mute or garbles the recorded voice if the far-end is talking

Any inputs would be highly appreciated...!