Skip to content

[BUG] no sound coming from speakers when playing 16-bit clips on DeepBuffer PCM dev #6595

Closed
Closed
[BUG] no sound coming from speakers when playing 16-bit clips on DeepBuffer PCM dev#6595
Assignees
RanderWang
Labels
MTLApplies to Meteor Lake platformP1Blocker bugs or important featuresbugSomething isn't working as expected
Milestone
v2.4

Description

@yongzhi1
yongzhi1
opened on Nov 15, 2022

Describe the bug
16-bit playback from DeepBufer dev generates no audible sound.

To Reproduce
Run the following cmd and check the sound from speakers:

aplay -Dhw:0,11 -c2 -r48000 -fS16_LE /home/sample.wav -vv

Reproduction Rate
100%

Expected behavior
Expect to hear sound as if we are playing from sofrt5682 [sof-rt5682], device 1: Speakers

Impact
What impact does this issue have on your progress (e.g., annoyance, showstopper)

Environment

  1. Branch name and commit hash of the 2 repositories: sof (firmware/topology) and linux (kernel driver).
  2. Name of the topology file
  3. Name of the platform(s) on which the bug is observed.
    • Platform: MTL-RVP

Screenshots or console output

sof-mtl-max98357a-rt5682

Activity

  1. yongzhi1
    added
    bugSomething isn't working as expected
    MTLApplies to Meteor Lake platform
    on Nov 15, 2022
  2. yongzhi1
    changed the title [-][BUG] no sound coming from speakers with 16-bit playback on DeepBuffer PCM dev[/-] [+][BUG] no sound coming from speakers when playing 16-bit clips on DeepBuffer PCM dev[/+] on Nov 15, 2022
  3. kfrydryx
    assigned
    RanderWang
    on Nov 15, 2022
  4. kfrydryx
    added
    P1Blocker bugs or important features
    on Nov 15, 2022

  5. RanderWang commented on Nov 16, 2022

    @RanderWang
    RanderWang
    on Nov 16, 2022
    Collaborator

    @yongzhi1 it is a sdw platform ? do you have kernel log ? And is it caused by 16bits ?

  6. yongzhi1 commented on Nov 16, 2022

    @yongzhi1
    yongzhi1
    on Nov 16, 2022
    ContributorAuthor

    Hi @RanderWang , it's I2S rt5682+max, 32-bit playback works fine although I only tested once, here is the log for 16-bit playback:
    dmesg-deepplay-16b.txt

  7. RanderWang commented on Nov 16, 2022

    @RanderWang
    RanderWang
    on Nov 16, 2022
    Collaborator

    It works on SDW and HDA platform. I only tested with nocodec tplg since we don't have a platform with I2S codec. It is not caused by deep buffer setting, instead about SSP format selection in driver. I will debug it. Currently SSP uses 24bit valid sample bits /32bit container size, but it should use 32/32

    static int sof_ipc4_pcm_dai_link_fixup(struct snd_soc_pcm_runtime *rtd,
				       struct snd_pcm_hw_params *params)
{
        ....

	switch (ipc4_copier->dai_type) {
	case SOF_DAI_INTEL_SSP:
		ipc4_ssp_dai_config_pcm_params_match(sdev, (char *)rtd->dai_link->name, params);
		break;
	default:
		break;
	}

	return 0;
}

  8. RanderWang commented on Nov 16, 2022

    @RanderWang
    RanderWang
    on Nov 16, 2022
    Collaborator

    @yongzhi1 The mixout-gain-dai-copier-playback.conf is for HDA by default and you override copier at https://github.com/yongzhi1/sof/blob/deepbuffer/tools/topology/topology2/cavs/cavs-rt5682.conf#L165, can you also override the format ? use only 32/32 ?
    [update] this way can't work since driver can't use it.

  9. lgirdwood
    added this to the v2.4 milestone on Nov 16, 2022

  10. yongzhi1 commented on Nov 16, 2022

    @yongzhi1
    yongzhi1
    on Nov 16, 2022
    ContributorAuthor

    Hi, @RanderWang, thanks for the analysis, made the following changes, but did not see improvement.

    --- a/tools/topology/topology2/cavs/cavs-rt5682.conf
+++ b/tools/topology/topology2/cavs/cavs-rt5682.conf
@@ -163,12 +163,29 @@ Object.Pipeline {
                        copier_type "SSP"
                        stream_name "SSP0-Codec"
                        node_type $I2S_LINK_OUTPUT_CLASS
+
+                       num_audio_formats 1
+                       # 16-bit input format 48KHz 2ch. Output sample format is always 32-bit for playback
+                       Object.Base.audio_format.1 {
+                               out_bit_depth           32
+                               out_valid_bit_depth     32
+                               dma_buffer_size "$[$ibs * $DEEPBUFFER_FW_DMA_MS]"
+                       }

    Also the volume level from Spk sound fine w/o above changes:

    card 0: sofrt5682 [sof-rt5682], device 1: Speakers () [] <= vol OK
    card 0: sofrt5682 [sof-rt5682], device 11: Deepbuffer Amps (    ) [] <= silent playing observed

    Let me try to override the deepbuffer pipleline.

  11. RanderWang commented on Nov 17, 2022

    @RanderWang
    RanderWang
    on Nov 17, 2022
    Collaborator

    @yongzhi1 can you share the kernel msg ? we don't have a platform with I2S codec so we can't try. And how about my PR 4016 ? Thanks!

  12. yongzhi1 commented on Nov 17, 2022

    @yongzhi1
    yongzhi1
    on Nov 17, 2022
    ContributorAuthor

    Second try, still did not make any difference.dmesg-deep-override.txt

            }
+
+       deepbuffer-playback.15 {
+               index   1
+
+               Object.Widget.copier."1" {
+                       node_type $I2S_LINK_OUTPUT_CLASS
+                       num_audio_formats 2
+
+                       Object.Base.audio_format.1 {
+                               out_bit_depth           32
+                               out_valid_bit_depth     32
+                               dma_buffer_size "$[$obs * $DEEPBUFFER_FW_DMA_MS]"
+                       }
+                       Object.Base.audio_format.2 {
+                               in_bit_depth            32
+                               in_valid_bit_depth      32
+                               out_bit_depth           32
+                               out_valid_bit_depth     32
+                               # 100ms buffer
+                               dma_buffer_size "$[$ibs * $DEEPBUFFER_FW_DMA_MS]"
+                       }
+               }
+       }
 }
 
 Object.PCM {

    Log for PR4016, dmesg-pr4016.txt, no sound still.

  13. RanderWang commented on Nov 17, 2022

    @RanderWang
    RanderWang
    on Nov 17, 2022
    Collaborator

    @yongzhi1 thanks ! Format 24/32 is used in dmesg-deep-override.txt but 32/32 is used by dmesg-pr4016.txt. Both no sound ? this PR #6583 can work ?

  14. RanderWang commented on Nov 17, 2022

    @RanderWang
    RanderWang
    on Nov 17, 2022
    Collaborator

    Can you try closed source fw ?

  15. RanderWang commented on Nov 17, 2022

    @RanderWang
    RanderWang
    on Nov 17, 2022
    Collaborator

    @yongzhi1 please try this. I tried your change #6595 (comment) and driver reported a error that no matched format

    --- a/tools/topology/topology2/cavs/cavs-nocodec.conf
+++ b/tools/topology/topology2/cavs/cavs-nocodec.conf
@@ -168,6 +168,16 @@ Object.Pipeline {
                        copier_type "SSP"
                        stream_name "NoCodec-0"
                        node_type $I2S_LINK_OUTPUT_CLASS
+
+                       num_audio_formats 2
+                       # 24-bit input format 48KHz 2ch. Output sample format is always 32-bit for playback
+                       Object.Base.audio_format.1 {
+                               in_bit_depth    32
+                               in_valid_bit_depth 24
+                               out_bit_depth           32
+                               out_valid_bit_depth     32
+                               dma_buffer_size "$[$ibs * $DEEPBUFFER_FW_DMA_MS]"
+                       }
+                       Object.Base.audio_format.2 {
+                                in_bit_depth    32
+                                in_valid_bit_depth    32 
+                               out_bit_depth           32
+                               out_valid_bit_depth     32
+                               dma_buffer_size "$[$ibs * $DEEPBUFFER_FW_DMA_MS]"
+                       }
                }

  16. RanderWang commented on Nov 17, 2022

    @RanderWang
    RanderWang
    on Nov 17, 2022
    Collaborator

    we need to set output bits to 32 according to be compatible with SSP. Why 24bit is used in deep buffer ? 24bits is faster than 32bits in FW implementation. The original non-deepbuffer use 32bit is not a optimal way and I will refine them later

    	SSP."0" {
		id 		0
		dai_index	0
		direction	"duplex"
		name		SSP0-Codec
		default_hw_conf_id	0
		sample_bits		32  //  set valid bit to 32
		io_clk		$MCLK

		Object.Base.hw_config."SSP0" {
			id	0
			mclk_freq	$MCLK
			bclk_freq	3072000
			tdm_slot_width	32
		}
	}

  17. yongzhi1 commented on Nov 17, 2022

    @yongzhi1
    yongzhi1
    on Nov 17, 2022
    ContributorAuthor

    Hi, @RanderWang, tried closed src FW, it did not help.
    Here is the log with PR4016 + tplg changes, still no sound from 'Deepbuffer Amps' dev.
    dmesg-pr4016+tplg.txt
    0001-tplg-changes.txt

    With #6583, yes I can hear sound, dmesg-pr6583.txt.

  18. RanderWang commented on Nov 18, 2022

    @RanderWang
    RanderWang
    on Nov 18, 2022
    Collaborator

    @yongzhi1 do you use dmic to check sound ? And can I use your device ?

  19. RanderWang commented on Nov 18, 2022

    @RanderWang
    RanderWang
    on Nov 18, 2022
    Collaborator

    with tplg and 4016 change, why the ouput is still 0x1802 ( 2ch 24 bit), it should b 0x2002 (32bit 2ch)

    [ 233.152302] sof-audio-pci-intel-mtl 0000:00:1f.3: copier copier.SSP.4.1, type 27
    [ 233.152312] sof-audio-pci-intel-mtl 0000:00:1f.3: sample rate: 48000 sample width: 32 channels: 2
    [ 233.152322] sof-audio-pci-intel-mtl 0000:00:1f.3: dai index 1 nhlt type 3 direction 0
    [ 233.152335] sof-audio-pci-intel-mtl 0000:00:1f.3: matching audio format index for 48000Hz, 24bit, 2 channels: 0
    [ 233.152344] sof-audio-pci-intel-mtl 0000:00:1f.3: Init input audio formats for copier.SSP.4.1
    [ 233.152351] sof-audio-pci-intel-mtl 0000:00:1f.3: #0: 48000KHz, 32bit (ch_map 0xffffff10 ch_cfg 1 interleaving_style 0 fmt_cfg 0x1802)
    [ 233.152362] sof-audio-pci-intel-mtl 0000:00:1f.3: Init output audio formats for copier.SSP.4.1
    [ 233.152369] sof-audio-pci-intel-mtl 0000:00:1f.3: #0: 48000KHz, 32bit (ch_map 0xffffff10 ch_cfg 1 interleaving_style 0 fmt_cfg 0x1802)

  20. yongzhi1 commented on Nov 22, 2022

    @yongzhi1
    yongzhi1
    on Nov 22, 2022
    ContributorAuthor

    Hi @RanderWang , compared with your nocodec.conf, changed DEEP_BUFFER_PCM_ID from 11 to 31, now it works as expected:dmesg-deep-tplg-override-02.txt, thanks!!

    With same kernel and tplg, when I switched to mtl-002-drop-stable, now it failed for a different reason:

    [  118.656628] sof-audio-pci-intel-mtl 0000:00:1f.3: matching audio format index for 48000Hz, 24bit, 2 channels: 0
[  118.656638] sof-audio-pci-intel-mtl 0000:00:1f.3: Init input audio formats for copier.SSP.4.1
[  118.656644] sof-audio-pci-intel-mtl 0000:00:1f.3:  #0: 48000KHz, 32bit (ch_map 0xffffff10 ch_cfg 1 interleaving_style 0 fmt_cfg 0x1802)
[  118.656655] sof-audio-pci-intel-mtl 0000:00:1f.3: Init output audio formats for copier.SSP.4.1
[  118.656663] sof-audio-pci-intel-mtl 0000:00:1f.3:  #0: 48000KHz, 32bit (ch_map 0xffffff10 ch_cfg 1 interleaving_style 0 fmt_cfg 0x2002)
[  118.656674] sof-audio-pci-intel-mtl 0000:00:1f.3: copier copier.SSP.4.1, IPC size is 164
[  118.656688] sof-audio-pci-intel-mtl 0000:00:1f.3: pipeline: 15 memory pages: 5
[  118.656697] sof-audio-pci-intel-mtl 0000:00:1f.3: Create widget pipeline.15 instance 0 - pipe 15 - core 0
...
[  118.667396] sof-audio-pci-intel-mtl 0000:00:1f.3: ipc tx done : 0x130f0004|0x0
[  118.667458] sof-audio-pci-intel-mtl 0000:00:1f.3: ipc4 set pipeline 4 state 3
[  118.667475] sof-audio-pci-intel-mtl 0000:00:1f.3: ipc tx      : 0x13040003|0x0
[  118.667886] sof-audio-pci-intel-mtl 0000:00:1f.3: ipc tx reply: 0x33000000|0x0
[  118.667960] sof-audio-pci-intel-mtl 0000:00:1f.3: ipc tx done : 0x13040003|0x0
[  118.667996] sof-audio-pci-intel-mtl 0000:00:1f.3: ipc4 set pipeline 4 state 4
[  118.668013] sof-audio-pci-intel-mtl 0000:00:1f.3: ipc tx      : 0x13040004|0x0
[  118.668842] sof-audio-pci-intel-mtl 0000:00:1f.3: ipc tx reply: 0x33000007|0x0
[  118.668864] sof-audio-pci-intel-mtl 0000:00:1f.3: FW reported error: 7 - Unsupported operation requested
[  118.679527] sof-audio-pci-intel-mtl 0000:00:1f.3: ipc error for msg 0x13040004|0x0
[  118.688009] sof-audio-pci-intel-mtl 0000:00:1f.3: failed to set state 4 for pipeline 4
[  118.696864] sof-audio-pci-intel-mtl 0000:00:1f.3: ASoC: error at soc_component_trigger on 0000:00:1f.3: -22
[  118.707763]  Deepbuffer Amps: ASoC: trigger FE cmd: 1 failed: -22

    mtrace-deep-play-failure.txt

  21. RanderWang commented on Nov 22, 2022

    @RanderWang
    RanderWang
    on Nov 22, 2022
    Collaborator

    @yongzhi1 it worked with closed source fw but failed with mtl-002 ?
    There is a known issue: #6598

  22. kfrydryx commented on Nov 22, 2022

    @kfrydryx
    kfrydryx
    on Nov 22, 2022
    Contributor

    We are encountering issues with unsupported formats in internal testing. AFAIK someone on our team will be working on this now.
    @RanderWang Please help to mitigate the issue for now, proper support for currently unsupported formats should be coming soon.

  23. yongzhi1 commented on Nov 23, 2022

    @yongzhi1
    yongzhi1
    on Nov 23, 2022
    ContributorAuthor

    Double checked the same with our backported kernel series, with DEEPBUFFER_FW_DMA_MS=100, failed to run playback on either Speakers or Deepbuffer Amps, but both works fine with DEEPBUFFER_FW_DMA_MS=50. Since the original silent playing issue was fixed by format override, let's close this one, thanks all for the great support!!

  24. yongzhi1
    closed this as completedon Nov 23, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

MTLApplies to Meteor Lake platformP1Blocker bugs or important featuresbugSomething isn't working as expected

Type

No type

Projects

No projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions