White noise when doing 16 bit playback/record via I2S on ICL

[DarrenxWang]

Summary:
White noise when doing 16 bit playback/record by I2S on firware branch icl-001-drop-stable. 24 bit/32bit playback/capture works well w/o this issue.

And this issue seems to be a branch-specific firmware issue for icl-001-drop-stable.
Because with the same kernel & topology, this issue cannot be reproduced on firmware master branch (commit 33e4b99) plus the patch below 0001-DMIC-Enable-power-for-DMIC-for-ICL-platform.patch

Test steps:
aplay -D hw:0,0 Music/48California_Dreaming.wav -d 20 & arecord -Dhw:0,0 -f S16_LE -r 48000 -c 2 test.wav -d 20

Test env:
Sof- icl-001-drop-stable: 40c054a

Soft-Soft- master: 937032f + 1 Patch
0001-topology-sample-code-for-icl-topology.patch.txt

Linux kernel: topic/sof-dev: af44aa9 + 3 Kernel patches
0001-ASoC-Intel-hdac_hdmi-add-Icelake-support.patch.txt
0002-ASoC-intel-add-icelake-dummy-codec-match-table.patch.txt
0003-ASoC-Intel-add-ICL-machine-driver.patch.txt

Firmware patch：
0001-DMIC-Enable-power-for-DMIC-for-ICL-platform.patch.txt

Log&Wav:
There is no obvious error in kmesg.
linux_dmesg.log
logger.txt
wav.zip

[mrajwa]

The only significant difference between the commit that is fine 33e4b99 and the icl-001-drop-stable is this commit 4fa9a7b. @singalsu, could you please check if you see reported noise issue caused by your commit changes?

[singalsu]

@mrajwa I don't have ICL. Do you know which bits -> bits conversion is the suspected volume controller doing?

[singalsu]

I made just PR #555. That's an obvious error, sorry for that. If the playback / capture pipelines in ICL did s24 -> s16 sample format conversion in volume that should impact.

[DarrenxWang]

There is white noise when doing 16 bit playback/record by I2S.
There is no white noise when doing 24/32 bit playback/record by I2S.

[singalsu]

@DarrenxWang I don't have device to test this conversion. Can you please test if PR #555 fixes this.

Edit: no need to get PR, it's merged to git master now.

[bardliao]

@mrajwa @singalsu This issue is fixed by PR #555 . Thanks

[singalsu]

@bardliao Excellent, thanks!

[singalsu]

Closing this issue.