Hello,
we want to build an intercom unit with the C5535 DSP. We are using http://software-dl.ti.com/libs/aer/latest/index_FDS.html algorithms for echo cancellation in handsfree mode. We use agc of aic3254 codec and not agc of aer library. We don't use drc, hlc and equalizer. Now we started tuning of parameters as mentioned in the quick tuning guide. In the first step of chapter 5 "Convergence verification", the y2x_delay parameter should be set to 0. But this is not possible, since the parameter must be positive as mentioned in the header file. The aerControl() function reports error aer_ERR_INVALIDPAR when setting the parameter to 0. So we set the parameter to 1. The output of aerGetPerformance() shows values of 0 or 1 for max_canc_l2 and and curr_canc_l2 and coherence_state has a value of 1. When we set y2x_delay to 40, the values of max_canc_l2 and and curr_canc_l2 are between 5 and 8. The coherence_state is first 0 and then changes to 1 and stays at 1. So there is no real convergence, right?! We already adjusted gains by approximation and removed dc offset from mic input data via high pass filter. Does it make sense to continue with chapter 6 "Main NLP parameter tuning"? Our main goal is echo cancellation. Second goal is to achieve as much as full duplex performance as possible. We are already working with the hardware/housing of the end product. We can access raw data of txout path (only every second sample due to bandwidth limitations) and performance data from aerGetPerformance() function. Additionally we can access filter coefficient info from aerGetfilter() function. What is the best approach for tuning the parameters?
Tanks in advance
Marc