Repository navigation
[ADL/TGL/MTL NOCODEC] Frequency response is different between left and right channel #7184
Description
Found alsabat failure with ADLP_RVP_NOCODEC_IPC4ZPH.
https://sof-ci.01.org/linuxpr/PR4212/build3371/devicetest/index.html?model=ADLP_RVP_NOCODEC_IPC4ZPH&testcase=check-alsabat-headset-playback
- Kernel PR: 4212
- SOF Branch: main
- SOF Commit: a74ad93
- Zephyr Commit: zephyr-v3.3.0-200-g9af2789cad17
- tplg: sof-adl-nocodec.tplg
alsabat test only complains about right channel.
BAT analysis: signal has 65536 frames at 48000 Hz, 2 channels, 2 bytes per sample.
Channel 1 - Checking for target frequency 821.00 Hz
Amplitude: 31128.9; Percentage: [95]
Detected peak at 821.04 Hz of 41.56 dB
Total 42.4 dB from 818.85 to 823.24 Hz
PASS: Peak detected at target frequency
Detected at least 1 signal(s) in total
Channel 2 - Checking for target frequency 821.00 Hz
Amplitude: 30990.5; Percentage: [94]
Detected peak at 815.92 Hz of 20.72 dB
Total 20.7 dB from 815.92 to 815.92 Hz
Detected peak at 821.04 Hz of 38.11 dB
Total 39.5 dB from 817.38 to 822.51 Hz
PASS: Peak detected at target frequency
Detected peak at 823.97 Hz of 20.52 dB
Total 39.5 dB from 823.97 to 823.97 Hz
PASS: Peak detected at target frequency
Detected at least 3 signal(s) in total
Return value is -1003
Attached: ADLP_RVP_NOCODEC_alsabat_wav_captured.zip
- bat.wav.9LAB5b ; alsabat test output
- left.wav ; only left channel
- right.wav ; only right channel-
Left and Right channels are separated. Did frequency analysis.
One caveat is I couldn't reproduce this manually with same device.
Activity
fredoh9 commentedon Feb 24, 2023
Found same failure with different pr test on different device.
Intel internal test: planresultdetail/21323?model=ADLP_RVP_NOCODEC_IPC4ZPH&testcase=check-alsabat-headset-playback
[-]Frequency response is different between left and right channel[/-][+][ADLP_RVP_NOCODEC_IPC4ZPH] Frequency response is different between left and right channel[/+]fredoh9 commentedon Feb 24, 2023
I scanned through TGLU_RVP_NOCODEC_IPC4ZPH test result and found same failure.
Intel internal test: planresultdetail/21269?model=TGLU_RVP_NOCODEC_IPC4ZPH&testcase=check-alsabat-headset-playback
Now this is more making sense.
[-][ADLP_RVP_NOCODEC_IPC4ZPH] Frequency response is different between left and right channel[/-][+][ADL/TGL NOCODEC] Frequency response is different between left and right channel[/+]juimonen commentedon Feb 24, 2023
actually got the same with link clock fix PR on mtl: https://sof-ci.ostc.intel.com/#/result/planresultdetail/21290
Like the channels are "mis-aligned" and right channel has gaps (also left, but much less).
[-][ADL/TGL NOCODEC] Frequency response is different between left and right channel[/-][+][ADL/TGL/MTL NOCODEC] Frequency response is different between left and right channel[/+]juimonen commentedon Feb 28, 2023
tried to reproduce with tgl, but just could not....
singalsu commentedon Feb 28, 2023
There's a glitch in the second channel at about 0.466s time:
singalsu commentedon Feb 28, 2023
One sample is from 1ms ago:
juimonen commentedon Feb 28, 2023
yeah, so there is the phase difference from smart amp between channels, but also some "gaps/peaks" in the channels. I think I actually also spotted them also in the left channel, but much more rarely than right channel. So I suspect something is fishy with buffer copy, somewhere.
ranj063 commentedon Feb 28, 2023
@aiChaoSONG can you please look into this?
lgirdwood commentedon Mar 2, 2023
Unlikely DMA as glitch would happen on both channels at the same time. Is SSP loopback mode used here ? SSP reads one word in FIFO at a time.
juimonen commentedon Mar 8, 2023
I tried with mtl device and ssp1 and ssp2 are clean with the alsabat test. This has to be something to do with the complex pipeline with smart amp feedback.
smart_amp_test: process same frames from all inputs