USB UAC garbled sound randomly (AEGHB-1103)

[digitalLumberjack]

Answers checklist.

• I have read the documentation ESP-IDF Programming Guide and the issue is not addressed there.
• I have updated my IDF branch (master or release) to the latest version and checked that the issue is present there.
• I have searched the issue tracker for a similar issue and not found a similar issue.

IDF version.

v6.0-dev-429-g36a5a71d5c, 5.3, 5.4,

Espressif SoC revision.

ESP32-s3

Operating System used.

Windows

How did you build your project?

Command line with CMake

If you are using Windows, please specify command line type.

None

Development Kit.

Custom Board

Power Supply used.

USB

What is the expected behavior?

Runs audio stream perfectly.

What is the actual behavior?

Hello, I need to use my esp32-s3 to act as a proxy between my host and an i2s DAC

Host -> USB -> EPS32-S3 -> i2s DAC -> speaker

I am working on a custom dev board that I use with esp-adf, which reads opus files and transmits the data in I2S to the DAC (PCM5102a). This allows me to be sure that the problem does not come from the board design.

I used the usb_device_uac component, and I am able to get sound with code based on the example https://github.com/espressif/esp-iot-solution/tree/master/examples/usb/device/usb_uac

Using espressif/esp_codec_dev: 1.3.4 to send i2s data to my PCM5102.

Everything works fine, I have good quality sound on the speaker, but randomly, I get crackling sound that is inaudible. This suggests a buffer underrun (see attachment).

audio_crsipy.zip

On the host side, PulseAudio does not output any logs that could indicate that there is a problem.
The UAC device is detected as intended by usb descriptors:

Sink #0
	State: SUSPENDED
	Name: alsa_output.usb-xxxxxxx_000000001-00.analog-stereo
	Description: xxxxx Analog Stereo
	Driver: module-alsa-card.c
	Sample Specification: s16le 2ch 48000Hz
	Channel Map: front-left,front-right
	Owner Module: 53
	Mute: no
	Volume: front-left: 9619 /  15% / -50.00 dB,   front-right: 58409 /  89% / -3.00 dB
	        balance 0.84
	Base Volume: 65536 / 100% / 0.00 dB
	Monitor Source: alsa_output.usb-xxxxxx_000000001-00.analog-stereo.monitor
	Latency: 0 usec, configured 0 usec
	Flags: HARDWARE HW_MUTE_CTRL HW_VOLUME_CTRL DECIBEL_VOLUME LATENCY 
	Properties:
		alsa.resolution_bits = "16"
		device.api = "alsa"
		device.class = "sound"
		alsa.class = "generic"
		alsa.subclass = "generic-mix"
		alsa.name = "USB Audio"
		alsa.id = "USB Audio"
		alsa.subdevice = "0"
		alsa.subdevice_name = "subdevice #0"
		alsa.device = "0"
		alsa.card = "3"
		alsa.card_name = "xxxxxxxx"
		alsa.long_card_name = "xxxxxxxxx at usb-xhci-hcd.1-1, full speed"
		alsa.driver_name = "snd_usb_audio"
		device.bus_path = "platform-xhci-hcd.1-usb-0:1:1.0"
		sysfs.path = "/devices/platform/axi/1000120000.pcie/1f00300000.usb/xhci-hcd.1/usb3/3-1/3-1:1.0/sound/card3"
		device.bus = "usb"
		device.vendor.id = "1452"
		device.vendor.name = "xxxx"
		device.product.id = "5487"
		device.product.name = "xxxxx"
		device.serial = "xxxxx"
		device.string = "front:3"
		device.buffering.buffer_size = "384000"
		device.buffering.fragment_size = "192000"
		device.access_mode = "mmap+timer"
		device.profile.name = "analog-stereo"
		device.profile.description = "Analog Stereo"
		device.description = "xxxxxxxxxxxx Analog Stereo"
		module-udev-detect.discovered = "1"
		device.icon_name = "audio-card-usb"
	Ports:
		analog-output: Analog Output (type: Analog, priority: 9900, availability unknown)
	Active Port: analog-output
	Formats:
		pcm

In the esp32-s3 code, I have these logs:

I (1775) usb_uac_main: uac_device_set_volume_cb: 100
I (1796) TUSB_DESC: tud_hid_descriptor_report_cb for 0
I (1818) TUSB_DESC: tud_hid_descriptor_report_cb for 1
Clock get is valid 1
Clock get current freq 48000
Clock get is valid 1
Speaker interface 1-1 openedSpeaker interface 1-1 opened
D (1975) usbd_uac: Feedback method: 4, sample freq: 48000
Clock get is valid 1
Clock get current freq 48000
Clock get is valid 1
Speaker interface 1-1 openedSpeaker interface 1-1 opened
D (2015) usbd_uac: Feedback method: 4, sample freq: 48000
Clock get is valid 1
Clock get current freq 48000
Clock get is valid 1
Speaker interface 1-1 openedSpeaker interface 1-1 opened
D (2047) usbd_uac: Feedback method: 4, sample freq: 48000
Set speaker channel 1 volume: -50 dB (0)
I (2051) usb_uac_main: uac_device_set_volume_cb: 0
Set speaker channel 2 volume: -3 dB (94)
I (2058) usb_uac_main: uac_device_set_volume_cb: 94

1. I tried to see if the tud buffer could be undersized compared to the expected bytes by adding

    if (bytes_remained < bytes_require)
        ESP_LOGI(“uac”, “underrun”);

here:

esp-iot-solution/components/usb/usb_device_uac/usb_device_uac.c

Line 404 in 8aecb33

s_uac_device->spk_data_size = tud_audio_read(s_uac_device->spk_buf, bytes_require);

Nothing, I even checked and bytes_remained is always full (its size depends on the configuration)

2. I checked that the audio buffer was always written completely when calling esp_codec_dev_write, which is the case.

3. I checked that software volume management could not impact performance by disabling it.

4. If I set CONFIG_UAC_SPK_INTERVAL_MS to 50 or 100ms, it seems to reduce the times issue occurs (not sure)

5. I'm running the tusb task from the usb_uac component, no other "non system" tasks are running

I don't know where else to look. Do you have any ideas?

Here are my configuration files:

tusb_configuration.zip

Steps to reproduce.

.

Debug Logs.

More Information.

No response

[lijunru-hub]

Did you build your own USB project based on the UAC component and enable the CONFIG_USB_DEVICE_UAC_AS_PART macro?

I noticed there’s also a CDC interface present—are any other USB interfaces active during testing? If the USB bus is heavily loaded, the audio stream might become starved.

Also, CONFIG_UAC_SPK_INTERVAL_MS should be kept at its default value. To reduce pop noise, consider adjusting the audio buffer size via CONFIG_UAC_SPK_NEW_PLAY_INTERVAL.

[digitalLumberjack]

Hello thank you very much for taking time to answer.

Did you build your own USB project based on the UAC component and enable the CONFIG_USB_DEVICE_UAC_AS_PART macro?

Yes but I just retried a very simple project starting from scratch with CONFIG_USB_DEVICE_UAC_AS_PART disabled, so using the descriptors embedded, and the issue is definitely the same.

For my project, I removed the CDC and HID interfaces that I want to add after I fixed this issue. So the USB descriptor is like this :

#define CONFIG_TOTAL_LEN        (TUD_CONFIG_DESC_LEN + \
                                                            CFG_TUD_AUDIO * TUD_AUDIO_DEVICE_DESC_LEN)

enum
{
    ITF_NUM_AUDIO_CONTROL = 0,
    ITF_NUM_AUDIO_STREAMING_SPK,
    ITF_NUM_TOTAL
};

uint8_t const desc_configuration[] = {
    // Config number, interface count, string index, total length, attribute, power in mA
    TUD_CONFIG_DESCRIPTOR(1, ITF_NUM_TOTAL, 0, CONFIG_TOTAL_LEN, 0x00, 500),
    // Interface number, string index, EP Out & EP In address, EP size
    TUD_AUDIO_DESCRIPTOR(ITF_NUM_AUDIO_CONTROL, 4, EPNUM_AUDIO_OUT, EPNUM_AUDIO_IN, EPNUM_AUDIO_FB),
};

I added a log in espressif__usb_device_uac/usb_device_uac.c in tud_audio_rx_done_post_read_cb to be sure that we do not have buffer underrun:

    if (bytes_remained < 500)
        ESP_LOGI("uac", "buffer seems low: bytes_remained < 500");

    if (log)
    {
        ESP_LOGI("uac", "remains mean %lld, require mean %lld", bytesremaintotal/loops , bytesrequiredtotal/loops);
        bytesremaintotal = loops = bytesrequiredtotal = 0;
    }

That show following logs :

I (10512) uac: tud_audio_rx_done_post_read_cb loops by seconds : 1000
I (10512) uac: remains mean 1078, require mean 192
I (11512) uac: tud_audio_rx_done_post_read_cb loops by seconds : 1004
I (11513) uac: remains mean 1078, require mean 192
I (12512) uac: tud_audio_rx_done_post_read_cb loops by seconds : 1000
I (12513) uac: remains mean 1078, require mean 192
I (13512) uac: tud_audio_rx_done_post_read_cb loops by seconds : 998
I (13513) uac: remains mean 1078, require mean 192

What I can say it's that in this callback, called each millisecond, there is never buffer underruns detected, and nothing seems different when the sound is ok and crispy.

I already checked that when we send the i2s data to i2s through the driver, the whole buffer is written.

To be certain of my design, I flashed back an other OPUS player, it decodes and play a 96kbits/2 channels/48kHz/16bit flawlessly.

I'll try with other sample rates, as i'm really out of ideas here :)

[digitalLumberjack]

With the tusb configuration from the usb_uac component project (no CONFIG_USB_DEVICE_UAC_AS_PART):

Same issue with :

• 48kHz 1ch 16bits
• 16kHz 2ch 16bits
• 44.1kHz 2ch 16 bits the glitch appears every 2~3 secondes, for 0.5sec

[digitalLumberjack]

Hello :)

I just checked on ESP32-p4 ev board with the example from the repository, vanilla no modifications except adding I2C old driver support.

Same issue, sound garbled a few seconds, each 2 or 3 minutes.

As I'm on linux, I checked on windows 11 too, same issue.

I have no more idea for cross testing.

Edit: there is one thing that could help us understand what is happening, when the sound is garbled, if I set the volume up/down quiclky it will come back to normal really quickly, like in a second.
If I don't do that the sound gets garbled for 3 to 10 secondes.

Do you have an idea ? Can any of you double check ?

[digitalLumberjack]

Hello, still searching for the cause of this issue.

Anyone else is using the uac component to stream for a pc to an i2s dac ?

[lijunru-hub]

You can test whether the BCLK output by I2S matches the configured frequency. In some configurations, the output frequency of I2S may fail to reach the configured value.

[digitalLumberjack]

Hello thank you for your answer.

I see no differences in LRCK and BCLK frequencies when the sound is OK and when it's not.

(don't mind the peaks it's because I had to use quite long wires to get the signal)

I think the issue might be before I2S transmission.

I tried to bump tinyusb to 0.18.0~2 but still the same issue.