[fredoh9]

This is the fix for S32_LE/S24_LE format playback.

fixes: #3340

Signed-off-by: Fred Oh fred.oh@linux.intel.com

[fredoh9]

This change increases BUF1.0 and BUF1.1 buffer size from 384 to 768. Tested with S16_LE/S24_LE/S32_LE.

[fredoh9]

pr-device-test result is also positive. check-playback-all-formats.sh is PASS now.

[fredoh9]

alsabat test failed on BSW_CYAN. hum..., hardly imagine buff size affects alsabat testing. @keqiaozhang I tested with one more device with my change, seeing similar glitches. What do you think?

11:40:54  BAT analysis: signal has 65536 frames at 44100 Hz, 1 channels, 2 bytes per sample.
11:40:54  
11:40:54  Channel 1 - Checking for target frequency 997.00 Hz
11:40:54  Amplitude: 5.7; Percentage: [0]
11:40:54  Detected peak at 7.40 Hz of 9.28 dB
11:40:54   Total 9.3 dB from 7.40 to 7.40 Hz
11:40:54  Detected peak at 11.44 Hz of 7.51 dB
11:40:54   Total 12.8 dB from 11.44 to 12.11 Hz
11:40:54  Detected peak at 16.82 Hz of 8.59 dB
11:40:54   Total 14.2 dB from 16.82 to 16.82 Hz
11:40:54  Detected peak at 32.30 Hz of 6.90 dB
11:40:54   Total 14.9 dB from 32.30 to 32.30 Hz
11:40:54  Detected peak at 91.52 Hz of 6.76 dB
11:40:54   Total 15.6 dB from 91.52 to 91.52 Hz
11:40:54  Detected peak at 685.70 Hz of 7.43 dB
11:40:54   Total 16.2 dB from 685.70 to 685.70 Hz
11:40:54  Detected peak at 983.80 Hz of 6.59 dB
11:40:54   Total 16.6 dB from 983.80 to 983.80 Hz
11:40:54  Detected peak at 993.22 Hz of 7.77 dB
11:40:54   Total 17.5 dB from 992.55 to 993.22 Hz
11:40:54   PASS: Peak detected at target frequency
11:40:54  Detected peak at 995.24 Hz of 6.53 dB
11:40:54   Total 17.9 dB from 995.24 to 995.24 Hz
11:40:54   PASS: Peak detected at target frequency
11:40:54  Detected peak at 999.95 Hz of 17.97 dB
11:40:54   Total 23.4 dB from 996.58 to 1002.64 Hz
11:40:54   PASS: Peak detected at target frequency
11:40:54  Detected at least 10 signal(s) in total
11:40:54  
11:40:54  Return value is -1003

[plbossart]

this doesn't seem right @fredoh9, the SSP can only handle 16 bits for Cyan, see the SSP config:
0f37299#diff-ecdb294c36b938b54c1c83d66735f103R104

Also this used to work, see e.g. #3257 where we change the parameters for device 0,1. to support S24/S32_LE. Device 0,0 had no problem with S32/S24_LE.

[fredoh9]

Also this used to work, see e.g. #3257 where we change the parameters for device 0,1. to support S24/S32_LE. Device 0,0 had no problem with S32/S24_LE.

@plbossart check-playback-all-formats.sh was run only in daily build. Recently Chao added to PR test. As far as I know, this test on BSW CYAN never worked.

I'm very narrow sighted. if "the SSP can only handle 16 bits for Cyan" so, how can we support 24bit/32bit for CYAN? I need your help. Or should be remove 24bit/32bit PCM capability from the topology?

[plbossart]

@fredoh9 can you try #3343 instead? I think there's a problem with the deep-buffer size that's not been mirrored for Cyan.

[keyonjie]

alsabat test failed on BSW_CYAN. hum..., hardly imagine buff size affects alsabat testing. @keqiaozhang I tested with one more device with my change, seeing similar glitches. What do you think?

11:40:54  BAT analysis: signal has 65536 frames at 44100 Hz, 1 channels, 2 bytes per sample.
11:40:54  
11:40:54  Channel 1 - Checking for target frequency 997.00 Hz
11:40:54  Amplitude: 5.7; Percentage: [0]
11:40:54  Detected peak at 7.40 Hz of 9.28 dB
11:40:54   Total 9.3 dB from 7.40 to 7.40 Hz
11:40:54  Detected peak at 11.44 Hz of 7.51 dB
11:40:54   Total 12.8 dB from 11.44 to 12.11 Hz
11:40:54  Detected peak at 16.82 Hz of 8.59 dB
11:40:54   Total 14.2 dB from 16.82 to 16.82 Hz
11:40:54  Detected peak at 32.30 Hz of 6.90 dB
11:40:54   Total 14.9 dB from 32.30 to 32.30 Hz
11:40:54  Detected peak at 91.52 Hz of 6.76 dB
11:40:54   Total 15.6 dB from 91.52 to 91.52 Hz
11:40:54  Detected peak at 685.70 Hz of 7.43 dB
11:40:54   Total 16.2 dB from 685.70 to 685.70 Hz
11:40:54  Detected peak at 983.80 Hz of 6.59 dB
11:40:54   Total 16.6 dB from 983.80 to 983.80 Hz
11:40:54  Detected peak at 993.22 Hz of 7.77 dB
11:40:54   Total 17.5 dB from 992.55 to 993.22 Hz
11:40:54   PASS: Peak detected at target frequency
11:40:54  Detected peak at 995.24 Hz of 6.53 dB
11:40:54   Total 17.9 dB from 995.24 to 995.24 Hz
11:40:54   PASS: Peak detected at target frequency
11:40:54  Detected peak at 999.95 Hz of 17.97 dB
11:40:54   Total 23.4 dB from 996.58 to 1002.64 Hz
11:40:54   PASS: Peak detected at target frequency
11:40:54  Detected at least 10 signal(s) in total
11:40:54  
11:40:54  Return value is -1003

@fredoh9 did this happen without your PR applied?

[fredoh9]

@fredoh9 did this happen without your PR applied?

I don't think so, after checking PR's device test result, I checked with different CI BSW, alsabat test failed also.
I can do waveform analysis also later.

[plbossart]

@fredoh9 I just tested this PR on a Cyan device, and while the issue with S32_LE goes away, the output is silent. There is NO sound coming out of the speakers, which seems aligned with my explanation above that the DAI assumes an S16_LE format.

[fredoh9]

@plbossart Thank you double checking with this. All BSW devices are in remote, I couldn't check the sound output.
BTW, our CI is getting reliable to catch regression. :)

[plbossart]

@plbossart Thank you double checking with this. All BSW devices are in remote, I couldn't check the sound output.
BTW, our CI is getting reliable to catch regression. :)

Indeed, now we are dealing with real issues. Progress!

[paulstelian97]

Is this still needed? If so, please resubmit with "main" as PR base branch.