Hi,
After 2 months having a C6746 DSP running ECU from VoLIB in real time in our custom board (working perfectly), we decided to do several more tests in other boards. We're facing some problems in our E1 boards with ECU running. We've noticed that the echo was NOT cancelling on those boards, We recorded some audio files and tried to run them in simulation mode and see if there was, in fact, a problem with our algorithm. After running the simulation, the echo was NOT cancelled either. But, when I changed the echo tail from 64ms to 128ms, the echo canceled.
The delay in the files I used was around 53ms, so it wasn't a problem of high delay. These are the files I used:
(Please visit the site to view this audio)
(Please visit the site to view this audio)
I have other samples too, but this one will suffice, for now, I believe.
This is the output of the echo canceller running these files in simulation mode:
When I changed the configurations in the simulation to 128ms, the echo was cancelled. I also changed the DSP firmware to 128ms and I got positive results (in real time), but this will affect in our channel density, making it impossible to use it commercially (I got a 10 channel reduction, from 30 to 20).
Why this change solved the problem? Why is the 64ms filter not capable of canceling this echo?
This project is very important to us, as we plan to ship it in 5 months. This setback may compromise our whole project.
Thank you very much.
Best Regards,
Leonardo Trierveiler