TGL: RT5682-SDW S16_LE SoC Loop Playback is Noisy

[naveen-manohar]

Issue Description:
RT5682-SDW Playback was Noisy and on debug observed that SOC output source has L/R channel phase error to Codec (PFA: )

Further tried experiments wrt SoC Loop Playback using simultaneous arecord(muted) and aplay to observe issue happens only with S16_LE(2ch) playback file. PFA Tabulated Experiments

Recipe Used: Based on TGL_SoF_V0.3
Kernel: https://github.com/naveen-manohar/linux/commits/chrome_bp_sof-dev_debug_noise
Topology: naveen-manohar/sof@5f11032
DSP FW: SOF TGL v0_3 Beta release
Platform: TGL Reference Chromebook with Daughter Card - RT5682+Max98373 - SDW
Coreboot(SNDW): PFA coreboot_nondebug-SNDW.zip

Validation Steps to Reproduce the Issue:
Step#1: Flash the Coreboot & OS Recipe mentioned on TGL Reference Chromebook with Daughter Card - RT5682+Max98373 - SDW
Step#2: Plug Jack Headset
Step#3: Run following script for SoC Loop playback
s16-soc-output.sh.txt

Observation: Noise observed with S16_LE SoC Loop Playback File PFA
original-vs-soc_loop_playback.zip

[plbossart]

@naveen-manohar can you please explain what the issue is?

"SOC output source has phase error to Codec" is rather unclear to me.

if you are doing a loopback inside of the codec, well of course the data will be delayed on the capture side...

[keqiaozhang]

@naveen-manohar
Could you please send the PRs of machine driver and topology to github? I will try to reproduce your issue on my TGL device.

[naveen-manohar]

Hi @plbossart ,
RT5682-SDW was observing headset playback noise and Realtek found that headphone L/R channel has phase error. On further analysis they found RT5682's SDW interface loop-back to check source from SOC to Codec has this phenomenon.

To check above scenario, tried SoC Loop Playback(below code) using simultaneous arecord(muted) and aplay to observe issue happens only with S16_LE(2ch) playback file

+++ b/drivers/soundwire/generic_bandwidth_allocation.c
@@ -140,8 +141,8 @@ static void _sdw_compute_port_params(struct sdw_bus *bus,
                                           sdw_compute_master_ports(m_rt, &params[i],
                                                                                         port_bo, hstop);
-                                            block_offset += m_rt->ch_count *
-                                                                          m_rt->stream->params.bps;
+                                           //block_offset += m_rt->ch_count *
+                                           //                           m_rt->stream->params.bps;
                                           port_bo = block_offset;
                            }

@keqiaozhang : Sure, I'm working on it, shall send it asap, ETA:ww11.4.

[naveen-manohar]

Hi @keqiaozhang :
Please check following PRs based on TGL_SoF_V0.3, let me know if you observe any issues.
#1885
thesofproject/sof@d58fb7c

[bardliao]

@naveen-manohar @plbossart I just checked with Albert from Realtek. He confirmed that the phase error was originally from the testing source. I.e. They use a wav with phase error between L and R channels to do the loopback test. So, that is not from SoC

[naveen-manohar]

Thanks for the clarification @bardliao , Lets re-do the validation and close then close this issue.

[naveen-manohar]

Realtek shared additional settings in codec-driver(rt5682-sdw) with this Headset Audio Playback is Good and no-artifact is observed.
0001-Sync-with-upstream.txt

Appreciate the efforts of all to analyse the issue and share inputs. Closing it for now. Thanks

[bardliao]

Realtek shared additional settings in codec-driver(rt5682-sdw) with this Headset Audio Playback is Good and no-artifact is observed.
0001-Sync-with-upstream.txt

Appreciate the efforts of all to analyse the issue and share inputs. Closing it for now. Thanks

I guess the key point is below :)

• regmap_update_bits(rt5682->regmap, RT5682_HP_CHARGE_PUMP_1,

•   	RT5682_PM_HP_MASK, RT5682_PM_HP_HV);