Modifying FFVA for Real-time Audio Loopback

[bennychainik]

I want to use the ffva example from the sdk on the xkvoice l71 development kit to create a low-latency audio loopback. My goal is to route the audio from the both the mics on the kit directly back to the DAC and j3 speaker port to monitor the input in real-time.Can you provide guidance on changes needed in the code, particularly concerning the data flow and buffer management? Are there any specific configurations or settings I should be aware of to ensure a true real-time, low-latency loopback?the FFVA example is for low-latency audio capture and playback, which is crucial for a real-time loopback so I want to change it for my need

In the future, I plan to integrate a custom Dynamic Range Compression (DRC) algorithm into this pipeline. I'm looking for advice on where to place this algorithm within the existing FFVA architecture.
Any advice on both the loopback setup and the future DSP integration would be greatly appreciated. Thanks!

[andrewxcav]

Will the monitor be headphones or a speaker? If it is a speaker you may need to incur a latency for AEC/anti howl processing.

Either way, I think you would want to start with an empty pipeline and add your DSP from there:

https://github.com/xmos/sln_voice/tree/ ... ence/empty

Another route would be to follow this app note:

https://www.xmos.com/file/an02031-live- ... d-example/

and fork: https://github.com/xmos/lib_board_support/tree/develop to add support for the L71 kit.

Cheers,
-Andrew