nrf5340_audio I2S AUDIO_BIT_DEPTH_32 breaks streaming

[MStreet]

I've got two nRF5340 AUDIO DKs with one programmed as a gateway and one as a headset using v2.3.0 and the nrf5340_audio sample application. I've added the following to prj.conf since my application will be interfacing with I2S:

    CONFIG_STREAM_BIDIRECTIONAL=y

    CONFIG_HW_CODEC_CIRRUS_LOGIC=n

    CONFIG_AUDIO_SOURCE_I2S=y

After fixing some errors that cropped up related to the hw_codec, everything builds and streaming starts up and locks OK with I2S running at 16-bits/sample:

However, my project requires 32 bits/sample and if I add CONFIG_AUDIO_BIT_DEPTH_32=y to prj.conf things fall apart:

I can see that I2S is running properly with 32 clock cycles per sample and it appears the FIFOs are being set up properly with 4 octets per sample and lengths adjusted accordingly. I don't know where else to look to try to fix this. 

[dejans]

Hi,

Unfortunately, setting CONFIG_AUDIO_BIT_DEPTH_32=y is not working at the moment.
I have reported the issue internally.

Best regards,
Dejan

[MStreet]

This is a showstopper and gating issue for us at this point. I hope you can resolve it soon! Thanks.

[dejans]

Hi,

I am sorry to hear that it is a showstopper for you at this point. I hope this issue will be resolved soon.

Best regards,
Dejan

[MStreet]

Hi Dejan,

The I2S FIFOs were being hard coded for 16-bit samples in audio_datapath.c. I've added the following conditional definitions at line #126:

This fixes the locking problem. I'll be testing audio next week. Thanks.

[dejans]

Hi,

MStreet said:

This fixes the locking problem. I'll be testing audio next week. Thanks.

Thank you for sharing the way you resolved audio bit depth issue. 

Best regards,
Dejan

[Alexander Svensen]

This has now been fixed and merged: https://github.com/nrfconnect/sdk-nrf/pull/10587. Thank you for bringing this to our attention and sorry for the inconvenience 

[MStreet]

Two more conditional defines are needed in audio_datapath.c to get I2S stereo output working properly with CONFIG_AUDIO_BIT_DEPTH_32=y:

Test tones are also broken without more fixes, but that is something I don't need.

[Alexander Svensen]

These changes are in the fix I posted just above this and has been merged to the main branch. And for the fix for the first one is not by using a conditional define, but using BLK_STEREO_NUM_SAMPS instead of BLK_STEREO_SIZE_OCTETS. As for the test tone, I'll create an internal issue for fixing it, thank you for pointing that out.