Can FLAC Extension play 96kHz/24bit track without downsampling? #6643

ivan2311 · 2019-11-12T15:53:43Z

I'm using FLAC Extension to play .flac files and I have really basic ExoPlayer setup:

ExoPlayer exoPlayer = new SimpleExoPlayer.Builder(this).build();

public void playUrl(String url) {
            MediaSource mediaSource = new ProgressiveMediaSource.Factory(new DefaultDataSourceFactory(this, agent))
                    .createMediaSource(Uri.parse(url));
            exoPlayer.prepare(mediaSource);
            exoPlayer.setPlayWhenReady(true);
    }

.flac files are playing ok, but when I look at the audio output it shows that audio signal can't go above the 24kHz, it always gets downsampled. For testing purposes I have a 96kHz/24bit .flac file that is just simple audio signal that goes from 1kHz to 32kHz and I'm checking audio output with Audacity. And what I'm seeing is that signal never goes above 24kHz.
Is this a bug in the ExoPlayer library since there is a commit that states that 192kHz should be supported: 0b08688?
Does that mean that ExoPlayer downsamples everything that goes above 48kHz?
Am I missing something here to support 96kHz?

Also, I've performed same test using Android MediaPlayer and it's ok, signal goes to 32kHz without any issues.

EDIT:
ExoPlayer version: current dev-v2 version, with latest commit: 87003b3 (I couldn't build FLAC extension with the latest release version release-v2 ). Previously I've used ExoPlayer v2.10.1 and it had the same issue
Testing device: Sony Xperia XZ
Other Audio equipment: external USB sound card (Steinberg UR28) to test output of analog (via on-device 3.5mm stereo jack) and digital audio signal (via USB audio through Steinberg UR28)

The text was updated successfully, but these errors were encountered:

kim-vde · 2019-11-13T17:51:11Z

I tried to reproduce on a Pixel 2 on Android 10 with a flac test file containing a sweep going from 1Hz to 96kHz sampled at 192kHz.

I first examined the PCM data outputted by the FLAC extractor using ExoPlayer TeaAudioProcessor.java. The frequency still goes to 96kHz, which means that the FLAC extension is not responsible for decreasing the frequency.

I then examined the frequency received by the audio engine (audioflinger) and can confirm it received a 192kHz stream. As I do not have a USB sound card supporting 192kHz, I was not able to confirm whether the whole output chain is at 192kHz when such sound card is connected.

Note that the AOSP audio system does not open USB sound cards higher than 96kHz.

It is possible that Sony modified their MediaPlayer to force output at 192kHz when the source is at such sampling rate but did not change the code of AudioTrack. If that is the case, Exoplayer will not be able to workaround the issue as it uses AudioTrack and not MediaPlayer.

In order to better understand the issue, could you please generate 2 bug reports while playing your 96kHz audio FLAC on your USB sound card:

one while playing the audio through Exoplayer
the other while playing it through the MediaPlayer

Make sure not to stop the playback until the bug report is fully generated (takes ~2 minutes).

How to take a bug report: https://developer.android.com/studio/debug/bug-report

krocard · 2019-11-14T10:37:54Z

Also, I've performed same test using Android MediaPlayer and it's ok, signal goes to 32kHz without any issues.

To clarify, have you observed a 32kHz signal on both the analog (on-device 3.5mm stereo jack) and the digital (USB) or just the digital interface?

ivan2311 · 2019-11-15T15:20:00Z

@kim-vde since you've tried this on your end and had no issues, a dig a little bit further into this problem and created even more simpler app that just plays sample .flac. And there was no issues, so it turns out it wasn't because of the ExoPlayer, but apparently something inside my app, which I'm still investigating.

@krocard I've observed on both interfaces.

@kim-vde @krocard Thanks for investigating this and helping me to eliminate ExoPlayer as the source of the problem.
I'll close this issue for now...

ivan2311 · 2019-12-04T13:17:32Z

I'm reopening this issue with new info:

As it turns out, issue on my side was because of the Equalizer. If Equalizer is attached to the player, output always gets downsampled. So, removing Equalizer completely, solved problem but only on some devices...

Audio signal is OK (not downsampled):

Sony XZ (over 3.5mm stereo jack)
Samsung (S5, S6 Edge, S8) devices (over 3.5mm stereo jack)

Audio signal is downsampled:

on every device without 3.5mm stereo jack, so I've used USB output.
on devices with 3.5mm stereo jack:
- Huawei (P9, Nexus 6P, P Smart, Mate 20 Lite)
- LG (G6, V30, Nexus 5X)
- Xiaomi Redmi Note 6

As it turns out, only Sony and Samsung devices with 3.5mm stereo jack doesn't downsample.

krocard · 2019-12-04T13:53:36Z

Audio signal is OK (not downsampled):

Are those test done with Exoplayer or Mediaplayer?

To be able to tell if the issue has something to do with Exoplayer, or those phone just do not support 96kHz on their USB output, we need a bug report.

Could you please play your 96kHz test sound on USB and generate a bug report, please try to not change the audio state (aka do not pause, reach the EOF...) during the bug report generation (~3min).
Thanks

ivan2311 · 2019-12-04T14:04:14Z

Are those test done with Exoplayer or Mediaplayer?

With ExoPlayer.

I'll generate bug report...

krocard · 2019-12-04T14:07:53Z

Note that outputting audio at 96kHz on an analog jack is probably not very compatible with cheap earphones.
Speakers (especially cheap ones) are not design to handle such high frequencies (>24kHz) and they might produce unwanted non linearities in the audible frequency range (<20kHz).
As a result playing > 24kHz on an analog output may diminish the sound quality instead of improving it.
That is one of the reason most phone only support 24kHz (48kHz SR) on their analog output. So it is not surprising you only found support in a few flagship.

ivan2311 · 2019-12-06T13:42:25Z

Bug reports

With ExoPlayer: exoplayer_report_06122019.zip

With MediaPlayer: mediaplayer_repory_06122019.zip

This bug reports were generated while playing .flac test file via simple sample app. Sample app can play that test file via ExoPlayer or MediaPlayer, whichever is currently chosen.
Test file is 96kHz/24bit .flac containing a sweep going from 1kHz to 32kHz. Sweep is repeating few times so that we have playback that lasts more that 3 minute, which is required to successfully report the bug.
For testing device, I've used Pixel 3 XL, with OS 10. Sound was played trough USB-C output. To connect device to the audio equipment, I've used USB-C to 3.5mm adapter, that comes with the device.

Test results: when playing on USB output with this device, downsampling at around 24kHz happens on both ExoPlayer and MediaPlayer

krocard · 2019-12-06T14:22:48Z

Thanks for the bug reports, I will take a look later today.

To connect device to the audio equipment, I've used USB-C to 3.5mm adapter, that comes with the device.

Why use the USB-C to 3.5mm adapter and not your Steinberg UR28 USB input?

ivan2311 · 2019-12-06T14:48:11Z

@krocard sorry I forgot to mention that this new tests were all done using Scarlett Focuscrite 2i2 which doesn't have USB input, so I need to use USB-C to 3.5mm adapter.

But, also as I remember, Steinberg UR28 also didn't have USB input, so whenever I was testing output from the devices that only have USB-C, I had to use USB-C to 3.5mm adapter. Sorry if there was some misunderstanding.

krocard · 2019-12-09T09:38:03Z

From the bug report section (adb shell) dumpsys media.audio_policy:

tag name: USB-Audio - USB-C to 3.5mm-Headphone Adapte
type: AUDIO_DEVICE_OUT_USB_DEVICE
address: card=0;device=0;
name: USB-Audio - USB-C to 3.5mm-Headphone Adapte
Profiles:
Profile 0:[dynamic format]
- format: AUDIO_FORMAT_PCM_24_BIT_PACKED
- sampling rates: 48000
- channel masks:0x0003, 0x80000003

The only supported sample rate of your USB audio device is 48kHz, thus the framework downsamples the track that was originaly at 96kHz (dumpsys media.audio_flinger):

Output thread 0x7a72a4d2c0, name AudioOut_1D, tid 1360, type 0 (MIXER):
  Sample rate: 48000 Hz // The output SR
   12-06 12:32:50.319 AT::add       (0x7a6332a700)          58     no   4338      25      35 A  0x000 00000001 00000003  \
96000  3   1  0  -inf     0     0     0  00000000  24000   24000 f         0        0       new
^~~~ The track sample rate

Please test with an USB sound card supporting 96kHz.

ivan2311 · 2019-12-10T13:31:16Z

@krocard thank you for checking. I can see that this USB-C to 3.5mm adapter supports only 48kHz.

Further more, I've created bug report with one device that is working (Sony Xperia XZ) and one that isn't working with ExoPlayer, but works with MediaPlayer:

Sony Xperia XZ

OS version: 8
device 3.5mm output directly connected to the Scarlett Focuscrite 2i2
working correctly on ExoPlayer and MediaPlayer
bug report:

Output thread 0xe5003480, name AudioOut_1D, tid 1575, type 0 (MIXER):
  I/O handle: 29
  Thread Type: 0 (MIXER)
  Output device: 0x8
  Input device:  0
  Standby: yes
  Sample rate: 192000 Hz
  HAL frame count: 7680
  HAL format: 0x4 (AUDIO_FORMAT_PCM_8_24_BIT)
  HAL buffer size: 61440 bytes
  Channel count: 2
  Channel mask: 0x00000003 (front-left, front-right)
  Processing format: 0x4 (AUDIO_FORMAT_PCM_8_24_BIT)
  Processing frame size: 8 bytes
  Pending config events: none
  Output device: 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE)
  Input device: 0 (AUDIO_DEVICE_NONE)
  Audio source: 0 (default)
  Normal frame count: 7680
  Last write occurred (msecs): 61365
  Total writes: 5074
  Delayed writes: 0
  Blocked in write: no
  Suspend count: 0
  Sink buffer : 0xe55d4000
  Mixer buffer: 0xe55e5000
  Effect buffer: 0xe5354000
  Fast track availMask=0xfe
  Standby delay ns=3000000000
  AudioStreamOut: 0xe589a060 flags 0x8 (AUDIO_OUTPUT_FLAG_DEEP_BUFFER)
  Frames written: 38968320
  Suspended frames: 0
  Hal stream dump:
  Thread throttle time (msecs): 53
  AudioMixer tracks: 0x00000002
  Master mono: off
  No FastMixer
  Stream volumes in dB: 0:-7.4, 1:-33, 2:-42, 3:-26, 4:-26, 5:-42, 6:0, 7:-42, 8:-27, 9:-96, 10:-26, 11:0, 12:0
  Normal mixer raw underrun counters: partial=0 empty=0
  1 Tracks of which 0 are active
    Name Active Client Type      Fmt Chn mask Session fCount S F SRate  L dB  R dB  VS dB    Server Main buf  Aux buf Flags UndFrmCnt  Flushed
       1     no   7934    3 00000001 00000003      73  30752 I 0 96000     0     0     0   00000000 E55D4000 00000000 0x000         0        0
  0 Effect Chains
  Local log:
   12-10 10:31:17.869 CFG_EVENT_CREATE_AUDIO_PATCH: old device 0x2 (AUDIO_DEVICE_OUT_SPEAKER) new device 0x2 (AUDIO_DEVICE_OUT_SPEAKER)
   12-10 10:31:26.401 CFG_EVENT_RELEASE_AUDIO_PATCH: old device 0x2 (AUDIO_DEVICE_OUT_SPEAKER) new device 0 (AUDIO_DEVICE_NONE)
   12-10 10:32:09.199 CFG_EVENT_CREATE_AUDIO_PATCH: old device 0 (AUDIO_DEVICE_NONE) new device 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE)
   12-10 10:32:09.221 addTrack_l    (0xe4a25180)    0     no   6241    3 00000001 00000003       9  30752 A 1 96000     0     0     0   00000000 E55D4000 00000000 0x000         0        0
   12-10 10:32:24.700 removeTrack_l (0xe4a25600)    1     no   6828    3 00000001 00000003      41  30752 T 0 96000     0     0     0   00000000 E55D4000 00000000 0x000         0        0
   12-10 10:35:29.362 removeTracks_l(0xe4a25180)    0     no   6241    3 00000001 00000003       9  30752 S 1 96000     0     0     0   0124F800 E55D4000 00000000 0x600         0        0
   12-10 10:36:05.890 removeTrack_l (0xe4a25180)    0     no   6241    3 00000001 00000003       9  30752 T 1 96000     0     0     0   0124F800 E55D4000 00000000 0x600         0        0

LG V30

OS version: 9
device 3.5mm output directly connected to the Scarlett Focuscrite 2i2
working correctly on MediaPlayer, downsampled on ExoPlayer
bug report:

Output thread 0xe7403740, name AudioOut_15, tid 1575, type 0 (MIXER):
  I/O handle: 21
  Standby: yes
  Sample rate: 48000 Hz
  HAL frame count: 1920
  HAL format: 0x1 (AUDIO_FORMAT_PCM_16_BIT)
  HAL buffer size: 7680 bytes
  Channel count: 2
  Channel mask: 0x00000003 (front-left, front-right)
  Processing format: 0x1 (AUDIO_FORMAT_PCM_16_BIT)
  Processing frame size: 4 bytes
  Pending config events: none
  Output device: 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE)
  Input device: 0 (AUDIO_DEVICE_NONE)
  Audio source: 0 (default)
  Normal frame count: 1920
  Last write occurred (msecs): 72738
  Total writes: 5289
  Delayed writes: 0
  Blocked in write: no
  Suspend count: 0
  Sink buffer : 0xe7a39000
  Mixer buffer: 0xe7a72000
  Effect buffer: 0xe7a77000
  Fast track availMask=0xfe
  Standby delay ns=3000000000
  AudioStreamOut: 0xe8eac550 flags 0x8 (AUDIO_OUTPUT_FLAG_DEEP_BUFFER)
  Frames written: 10154880
  Suspended frames: 0
  Hal stream dump:
  Thread throttle time (msecs): 90
  AudioMixer tracks: 
  Master mono: off
  No FastMixer
  Stream volumes in dB: 0:-10, 1:-53, 2:-53, 3:-30, 4:-23, 5:-49, 6:-inf, 7:-18, 8:-29, 9:-96, 10:-30, 11:0, 12:0
  Normal mixer raw underrun counters: partial=0 empty=0
  0 Tracks
  0 Effect Chains
  Local log:
   12-10 10:53:03.504 CFG_EVENT_CREATE_AUDIO_PATCH: old device 0x2 (AUDIO_DEVICE_OUT_SPEAKER) new device 0x2 (AUDIO_DEVICE_OUT_SPEAKER)
   12-10 10:53:03.547 CFG_EVENT_RELEASE_AUDIO_PATCH: old device 0x2 (AUDIO_DEVICE_OUT_SPEAKER) new device 0 (AUDIO_DEVICE_NONE)
   12-10 10:53:04.441 CFG_EVENT_CREATE_AUDIO_PATCH: old device 0 (AUDIO_DEVICE_NONE) new device 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE)
   12-10 10:53:04.443 AT::add       (0xe6fb6000)      0     no    945       9 A  0x000 00000001 00000003  48000  1     0     0     0  00000000  24000   24000 f         0        0 E7A39000 00000000
   12-10 10:53:07.681 CFG_EVENT_CREATE_AUDIO_PATCH: old device 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE) new device 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE)
   12-10 10:53:09.482 CFG_EVENT_CREATE_AUDIO_PATCH: old device 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE) new device 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE)
   12-10 10:53:10.186 AT::remove    (0xe6fb6000)      0     no    945       9 S  0x600 00000001 00000003  48000  1     0     0     0  00041D41  24000       0 f         0        0 E7A39000 00000000
   12-10 10:53:11.449 removeTrack_l (0xe6fb6000)      0     no    945       9 T  0x600 00000001 00000003  48000  1     0     0     0  00041D41  24000       0 f         0        0 E7A39000 00000000
   12-10 10:53:41.626 AT::add       (0xe8eed780)      0     no   7353      57 A  0x000 00000001 00000003  96000  3     0     0     0  00000000  30752   30752 f         0        0 E7A39000 00000000
   12-10 10:57:01.780 AT::remove    (0xe8eed780)      0     no   7353      57 S  0x600 00000001 00000003  96000  3     0     0     0  0124F800  30752       0 f         0        0 E7A39000 00000000
   12-10 10:57:15.224 removeTrack_l (0xe8eed780)      0     no   7353      57 T  0x600 00000001 00000003  96000  3     0     0     0  0124F800  30752       0 f         0        0 E7A39000 00000000



Output thread 0xe7783cc0, name AudioOut_D, tid 1569, type 0 (MIXER):
  I/O handle: 13
  Standby: yes
  Sample rate: 48000 Hz
  HAL frame count: 192
  HAL format: 0x1 (AUDIO_FORMAT_PCM_16_BIT)
  HAL buffer size: 768 bytes
  Channel count: 2
  Channel mask: 0x00000003 (front-left, front-right)
  Processing format: 0x5 (AUDIO_FORMAT_PCM_FLOAT)
  Processing frame size: 8 bytes
  Pending config events: none
  Output device: 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE)
  Input device: 0 (AUDIO_DEVICE_NONE)
  Audio source: 0 (default)
  Normal frame count: 960
  Last write occurred (msecs): 50076
  Total writes: 1361
  Delayed writes: 0
  Blocked in write: no
  Suspend count: 0
  Sink buffer : 0xe8ec0000
  Mixer buffer: 0xe8efe000
  Effect buffer: 0xe7a3e000
  Fast track availMask=0xf8
  Standby delay ns=3000000000
  AudioStreamOut: 0xe8eac730 flags 0x6 (AUDIO_OUTPUT_FLAG_PRIMARY|AUDIO_OUTPUT_FLAG_FAST)
  Frames written: 1306560
  Suspended frames: 0
  PipeSink frames written: 1306560
  Hal stream dump:
  Thread throttle time (msecs): 0
  AudioMixer tracks: 
  Master mono: off
  FastMixer thread 0xe7703500 tid=1563  FastMixer command=COLD_IDLE writeSequence=13650 framesWritten=1310400
            numTracks=1 writeErrors=0 underruns=4 overruns=1
            sampleRate=48000 frameCount=192 measuredWarmup=17.1 ms, warmupCycles=5
            mixPeriod=4.00 ms
  Simple moving statistics over last 7.8 seconds:
    wall clock time in ms per mix cycle:
      mean=4.00 min=0.65 max=9.75 stddev=0.33
    raw CPU load in us per mix cycle:
      mean=283 min=0 max=1204 stddev=193
  Distribution of mix cycle times in ms for the tails (> ~3 stddev outliers):
    left tail: mean=0.65 min=0.65 max=0.65 stddev=nan
    right tail: mean=9.75 min=9.75 max=9.75 stddev=nan
  Fast tracks: sMaxFastTracks=8 activeMask=0x1
  Index Active Full Partial Empty  Recent Ready    Written
      0    yes  656       0     0    full  1152    1305408
      1     no  281       1    11   empty     0      53980
      2     no  117       9   109   empty     0      18508
      3     no   16       1    12   empty     0       3120
      4     no    0       0     0    full     0          0
      5     no    0       0     0    full     0          0
      6     no    0       0     0    full     0          0
      7     no    0       0     0    full     0          0
  Stream volumes in dB: 0:-10, 1:-53, 2:-53, 3:-30, 4:-23, 5:-49, 6:-inf, 7:-18, 8:-29, 9:-96, 10:-30, 11:0, 12:0
  Normal mixer raw underrun counters: partial=0 empty=0
  2 Tracks of which 0 are active
    T Name Active Client Session S  Flags   Format Chn mask  SRate ST  L dB  R dB  VS dB   Server FrmCnt  FrmRdy F Underruns  Flushed Main Buf  Aux Buf
    FS   1     no   2700      17 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  0000D2DC  53980       0 f         0        0 E8EC0000 00000000
    FS   2     no   1679      49 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  0000484C   2644       0 f         0        0 E8EC0000 00000000
  0 Effect Chains
  Local log:
   12-10 10:53:03.480 CFG_EVENT_CREATE_AUDIO_PATCH: old device 0x2 (AUDIO_DEVICE_OUT_SPEAKER) new device 0x2 (AUDIO_DEVICE_OUT_SPEAKER)
   12-10 10:53:03.546 CFG_EVENT_RELEASE_AUDIO_PATCH: old device 0x2 (AUDIO_DEVICE_OUT_SPEAKER) new device 0 (AUDIO_DEVICE_NONE)
   12-10 10:53:27.329 CFG_EVENT_CREATE_AUDIO_PATCH: old device 0 (AUDIO_DEVICE_NONE) new device 0x8 (AUDIO_DEVICE_OUT_WIRED_HEADPHONE)
   12-10 10:53:27.333 AT::add       (0xe6fb6a00) FS   1     no   2700      17 A  0x000 00000001 00000003  48000  1    -6    -6     0  00000000  53980   53980 F         0        0 E8EC0000 00000000
   12-10 10:53:28.590 AT::remove    (0xe6fb6a00) FS   1     no   2700      17 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  0000D2DC  53980       0 f         0        0 E8EC0000 00000000
   12-10 10:53:30.086 AT::add       (0xe8eed500) FS   2     no   1679      33 A  0x000 00000001 00000003  48000  1     0     0     0  00000000   2644    2644 F         0        0 E8EC0000 00000000
   12-10 10:53:30.193 AT::remove    (0xe8eed500) FS   2     no   1679      33 S  0x601 00000001 00000003  48000  1  -inf  -inf     0  00000A54   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:53:31.716 AT::add       (0xe8eed500) FS   2     no   1679      33 A  0x201 00000001 00000003  48000  1     0     0     0  00000A54   2644       0 F         0*       0 E8EC0000 00000000
   12-10 10:53:31.833 AT::remove    (0xe8eed500) FS   2     no   1679      33 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  000014A8   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:53:34.811 threadLoop_standby: framesWritten:356160  suspendedFrames:0  monoPipeWritten:356160  monoPipeLeft:1152
   12-10 10:53:34.811 threadLoop_standby: BOOTTIME offset 0
ExtendedTimestamp[0]  position: 0  time: -1
ExtendedTimestamp[1]  position: 356160  time: 38636144203
ExtendedTimestamp[2]  position: 353808  time: 38649354359
ExtendedTimestamp[3]  position: 356160  time: 38636144203
ExtendedTimestamp[4]  position: 353616  time: 38645351078
   12-10 10:53:34.922 AT::add       (0xe6fb6c80) FS   3     no   1679      41 A  0x000 00000001 00000003  48000  1     0     0     0  00000000   3120    3120 F         0        0 E8EC0000 00000000
   12-10 10:53:34.927 removeTrack_l (0xe8eed500) FS   2     no   1679      33 T  0x600 00000001 00000003  48000  1  -inf  -inf     0  000014A8   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:53:35.109 AT::remove    (0xe6fb6c80) FS   3     no   1679      41 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  00000C30   3120       0 f         0        0 E8EC0000 00000000
   12-10 10:53:38.084 threadLoop_standby: framesWritten:504960  suspendedFrames:0  monoPipeWritten:504960  monoPipeLeft:1536
   12-10 10:53:38.084 threadLoop_standby: BOOTTIME offset 0
ExtendedTimestamp[0]  position: 0  time: -1
ExtendedTimestamp[1]  position: 504960  time: 41914187223
ExtendedTimestamp[2]  position: 502224  time: 41919843837
ExtendedTimestamp[3]  position: 504000  time: 41894233420
ExtendedTimestamp[4]  position: 501840  time: 41911850764
   12-10 10:53:38.381 AT::add       (0xe8eed500) FS   2     no   1679      49 A  0x000 00000001 00000003  48000  1     0     0     0  00000000   2644    2644 F         0*       0 E8EC0000 00000000
   12-10 10:53:38.384 removeTrack_l (0xe6fb6c80) FS   3     no   1679      41 T  0x600 00000001 00000003  48000  1  -inf  -inf     0  00000C30   3120       0 f         0        0 E8EC0000 00000000
   12-10 10:53:38.576 AT::remove    (0xe8eed500) FS   2     no   1679      49 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  00000A54   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:53:41.261 AT::add       (0xe8eed500) FS   2     no   1679      49 A  0x200 00000001 00000003  48000  1     0     0     0  00000A54   2644       0 F         0*       0 E8EC0000 00000000
   12-10 10:53:41.373 AT::remove    (0xe8eed500) FS   2     no   1679      49 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  000014A8   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:53:44.357 threadLoop_standby: framesWritten:785280  suspendedFrames:0  monoPipeWritten:785280  monoPipeLeft:1728
   12-10 10:53:44.357 threadLoop_standby: BOOTTIME offset 0
ExtendedTimestamp[0]  position: 0  time: -1
ExtendedTimestamp[1]  position: 785280  time: 48186688991
ExtendedTimestamp[2]  position: 782352  time: 48194687012
ExtendedTimestamp[3]  position: 784320  time: 48166398418
ExtendedTimestamp[4]  position: 781776  time: 48182703158
   12-10 10:57:15.161 AT::add       (0xe8eed500) FS   2     no   1679      49 A  0x200 00000001 00000003  48000  1     0     0     0  000014A8   2644       0 F         0*       0 E8EC0000 00000000
   12-10 10:57:15.365 AT::remove    (0xe8eed500) FS   2     no   1679      49 S  0x601 00000001 00000003  48000  1  -inf  -inf     0  00001EFC   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:57:18.342 threadLoop_standby: framesWritten:933120  suspendedFrames:0  monoPipeWritten:933120  monoPipeLeft:1344
   12-10 10:57:18.342 threadLoop_standby: BOOTTIME offset 0
ExtendedTimestamp[0]  position: 0  time: -1
ExtendedTimestamp[1]  position: 933120  time: 262171830680
ExtendedTimestamp[2]  position: 930576  time: 262180596566
ExtendedTimestamp[3]  position: 932160  time: 262152053284
ExtendedTimestamp[4]  position: 930000  time: 262168777711
   12-10 10:57:19.410 AT::add       (0xe8eed500) FS   2     no   1679      49 A  0x201 00000001 00000003  48000  1     0     0     0  00001EFC   2644       0 F         0*       0 E8EC0000 00000000
   12-10 10:57:19.585 AT::remove    (0xe8eed500) FS   2     no   1679      49 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  00002950   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:57:21.793 AT::add       (0xe8eed500) FS   2     no   1679      49 A  0x200 00000001 00000003  48000  1     0     0     0  00002950   2644       0 F         0*       0 E8EC0000 00000000
   12-10 10:57:21.905 AT::remove    (0xe8eed500) FS   2     no   1679      49 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  000033A4   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:57:22.992 AT::add       (0xe8eed500) FS   2     no   1679      49 A  0x200 00000001 00000003  48000  1     0     0     0  000033A4   2644       0 F         0*       0 E8EC0000 00000000
   12-10 10:57:23.105 AT::remove    (0xe8eed500) FS   2     no   1679      49 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  00003DF8   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:57:24.174 AT::add       (0xe8eed500) FS   2     no   1679      49 A  0x200 00000001 00000003  48000  1     0     0     0  00003DF8   2644       0 F         0*       0 E8EC0000 00000000
   12-10 10:57:24.285 AT::remove    (0xe8eed500) FS   2     no   1679      49 S  0x600 00000001 00000003  48000  1  -inf  -inf     0  0000484C   2644       0 f         0        0 E8EC0000 00000000
   12-10 10:57:27.272 threadLoop_standby: framesWritten:1306560  suspendedFrames:0  monoPipeWritten:1306560  monoPipeLeft:960
   12-10 10:57:27.272 threadLoop_standby: BOOTTIME offset 0
ExtendedTimestamp[0]  position: 0  time: -1
ExtendedTimestamp[1]  position: 1306560  time: 271091675312
ExtendedTimestamp[2]  position: 1304400  time: 271107270208
ExtendedTimestamp[3]  position: 1306560  time: 271091675312
ExtendedTimestamp[4]  position: 1304016  time: 271099270989

I can see in the LG V30 logs, that supported sample rate is only 48kHz so it should downsample, but why it works on the MediaPlayer?

krocard · 2019-12-10T13:51:18Z

I'm sorry but those dumpsys extract are not enough for me to understand the LG V30 audio configuration. I'm critically missing the Audio Policy dumpsys.

Could you please post 2 full bug reports (zip file like you did in #6643 (comment)):

generated while playing with media player at >48kHz
generated while playing with Exoplayer at the same frequency (ideally the same file)
I should then be able to compare the audio platform configuration to understand why the AudioPolicy chooses the 48kHz sample rate for Exoplayer output but not for MediaPlayer.

Also, it would be nice to disable the touch sounds, as they produce lots of audio event unrelated to our problem (settings/sound/touch sounds)

ivan2311 · 2019-12-10T15:23:58Z

@krocard here are 2 full bug reports:

krocard · 2019-12-10T17:09:18Z

Thanks for the dumpsys @ivan2311, it let me figure out what is going on.

Without entering into too much details, MediaPlayer uses a special way of playing audio called "offload". It consists of not decoding the audio in the app and instead have the platform DSP decode it. It is a completely different audio path then the regular PCM playback.
It has many restriction and is not suitable for most playback, but in your case, for this specific phone, audio format and sample rate, it allows the MediaPlayer to bypass the limitation of the phone PCM interface.

Basically LG's PCM interface (contrary to Sony's) has an artificial limitation to 4800kHz even though the HW supports other sample rates. This is not something Exoplayer or even the android framework can change, its forced by LG's configuration.
By using offload, encoded FLAC is sent to a low level HW component (DSP) that decodes it without the 48kHz limitation as it does not go through LG's PCM interface but LG offload interface.

Unfortunately, this low level behavior is phone specific and can not be relied on across the ecosystem.

Note that offload APIs were made public in Android 10, so starting with Android 10, non platform player like Exoplayer can support it. Though Exoplayer currently does not.

krocard · 2019-12-10T17:15:33Z

To summarize:

LG's PCM path is limited to 48kHz (contrary to Sony's) when the Offload path is not.
Only the media player can use the offload path until Android 10.
Exoplayer do not support offload using the new Android 10 APIs but we are working on it!

ivan2311 · 2019-12-10T22:09:02Z

@krocard thank you for resolving this issue and the explanation!

Just one more quick question: if I'm playing HiRes 24bit track, should I enable conversion from HiRes int pcm to float inside DefaultAudioSink, since when it's set to false it converts 24bit pcm to 16bit, but I would like to have an unchanged 24bit at the output?

krocard · 2019-12-11T11:17:20Z

This question is unrelated to the main issue. Next time please open a new issue.
Quick answer: playing high res samples is not currently supported by Exoplayer with MediaCodec. I juste open a feature request for it: #6749. Refer to its Alternatives considered section for workarounds.

ivan2311 · 2019-12-11T11:50:38Z

Thanks again @krocard !

ivan2311 added bug needs triage labels Nov 12, 2019

marcbaechinger assigned kim-vde Nov 12, 2019

kim-vde added question and removed bug needs triage labels Nov 13, 2019

kim-vde added the need more info label Nov 13, 2019

ivan2311 closed this as completed Nov 15, 2019

google-oss-bot added needs triage and removed need more info labels Nov 15, 2019

ivan2311 reopened this Dec 4, 2019

krocard added the need more info label Dec 4, 2019

google-oss-bot removed the need more info label Dec 4, 2019

kim-vde assigned krocard and unassigned kim-vde Dec 6, 2019

kim-vde removed the needs triage label Dec 6, 2019

krocard added the need more info label Dec 9, 2019

google-oss-bot added the needs triage label Dec 10, 2019

google-oss-bot removed the need more info label Dec 10, 2019

krocard removed the needs triage label Dec 10, 2019

krocard added the need more info label Dec 10, 2019

google-oss-bot added needs triage and removed need more info labels Dec 10, 2019

krocard removed the needs triage label Dec 10, 2019

krocard closed this as completed Dec 10, 2019

krocard added the bug: device specific label Dec 10, 2019

krocard mentioned this issue Dec 11, 2019

Support MediaCodec audio Float output #6749

Closed

krocard removed the bug: device specific label Dec 11, 2019

google locked and limited conversation to collaborators Feb 9, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Can FLAC Extension play 96kHz/24bit track without downsampling? #6643

Can FLAC Extension play 96kHz/24bit track without downsampling? #6643

ivan2311 commented Nov 12, 2019 •

edited

Loading

kim-vde commented Nov 13, 2019

krocard commented Nov 14, 2019 •

edited

Loading

ivan2311 commented Nov 15, 2019

ivan2311 commented Dec 4, 2019

krocard commented Dec 4, 2019 •

edited

Loading

ivan2311 commented Dec 4, 2019

krocard commented Dec 4, 2019

ivan2311 commented Dec 6, 2019

krocard commented Dec 6, 2019 •

edited

Loading

ivan2311 commented Dec 6, 2019

krocard commented Dec 9, 2019 •

edited

Loading

ivan2311 commented Dec 10, 2019 •

edited

Loading

krocard commented Dec 10, 2019 •

edited

Loading

ivan2311 commented Dec 10, 2019

krocard commented Dec 10, 2019

krocard commented Dec 10, 2019 •

edited

Loading

ivan2311 commented Dec 10, 2019

krocard commented Dec 11, 2019 •

edited

Loading

ivan2311 commented Dec 11, 2019

Can FLAC Extension play 96kHz/24bit track without downsampling? #6643

Can FLAC Extension play 96kHz/24bit track without downsampling? #6643

Comments

ivan2311 commented Nov 12, 2019 • edited Loading

kim-vde commented Nov 13, 2019

krocard commented Nov 14, 2019 • edited Loading

ivan2311 commented Nov 15, 2019

ivan2311 commented Dec 4, 2019

krocard commented Dec 4, 2019 • edited Loading

ivan2311 commented Dec 4, 2019

krocard commented Dec 4, 2019

ivan2311 commented Dec 6, 2019

Bug reports

krocard commented Dec 6, 2019 • edited Loading

ivan2311 commented Dec 6, 2019

krocard commented Dec 9, 2019 • edited Loading

ivan2311 commented Dec 10, 2019 • edited Loading

krocard commented Dec 10, 2019 • edited Loading

ivan2311 commented Dec 10, 2019

krocard commented Dec 10, 2019

krocard commented Dec 10, 2019 • edited Loading

ivan2311 commented Dec 10, 2019

krocard commented Dec 11, 2019 • edited Loading

ivan2311 commented Dec 11, 2019

ivan2311 commented Nov 12, 2019 •

edited

Loading

krocard commented Nov 14, 2019 •

edited

Loading

krocard commented Dec 4, 2019 •

edited

Loading

krocard commented Dec 6, 2019 •

edited

Loading

krocard commented Dec 9, 2019 •

edited

Loading

ivan2311 commented Dec 10, 2019 •

edited

Loading

krocard commented Dec 10, 2019 •

edited

Loading

krocard commented Dec 10, 2019 •

edited

Loading

krocard commented Dec 11, 2019 •

edited

Loading