XCore Exchange

Posted: **Fri Nov 25, 2016 11:58 am**

Hi,
here the action on board and resulting print on console with the debug code added just before my patch force dsdMode to OFF:

Code: Select all

    printstr("dsdMode : ");
    printintln(dsdMode);

    #if(DSD_CHANS_DAC == 0)
        dsdMode = DSD_MODE_OFF;
    #endif

1) Run board with USB detached
dsdMode : 0

2) Attach USB to host
dsdMode : 0
dsdMode : 0
dsdMode : 0
dsdMode : 2

3) Run arecord -D hw:0 -r 48000 ...
dsdMode : 2
(so this without patch will record empty audio)

4) Run spekaker-tests -D hw:0 -r 48000 ...
dsdMode : 0
(AudioHwConfig will be executed also if the sample rate is not changed and correctly init ADC and DAC also without patch)

5) Run arecord -D hw:0 -r 48000 ...
(AudioHwConfig was not executed because sample rate is not changed but ADC is just correctly initialized and audio will becorrectly recorded)

6) Run arecord -D hw:0 -r 96000 ...
dsdMode : 0
(Changing samplerate force execution of AudioHwConfig)

7) Detaching and attaching again USB cable result in retarting exactly from point (2)

Hope this help to understand!

Lorenzo

Posted: **Fri Nov 25, 2016 2:11 pm**

That's a helpful report but it displays some some bizarre behaviour. Following the control path from EP0 through to audio, the stream format is set in the descriptors. Each USB interface gets a stream format from the descriptors and will be native DSD if it is initialised to UAC_FORMAT_TYPEI_RAW_DATA. Later on, as this format value is passed through, it reaches decouple.xc where it is translated into a dsd setting and passed to audio.xc, which is where you have been debugging.

Code: Select all

#ifdef NATIVE_DSD
                if(dataFormat == UAC_FORMAT_TYPEI_RAW_DATA)
                {
                    dsdMode = DSD_MODE_NATIVE;
                }
#endif

and

Code: Select all

#define DSD_MODE_OFF          0
#define DSD_MODE_DOP          1
#define DSD_MODE_NATIVE       2

So if you are seeing the value 2, then the only way this can be is if NATIVE_DSD is set at compile time.

This gets set in devicedefines.h:

Code: Select all

/**
 * @brief Number of DSD output channels. Default: 0 (disabled)
 */
#if defined(DSD_CHANS_DAC)
    #if defined(NATIVE_DSD) && (NATIVE_DSD == 0)
        #undef NATIVE_DSD
    #else
        #define NATIVE_DSD       1  /* Always enable Native DSD when DSD mode is enabled */
    #endif
#else
    #define DSD_CHANS_DAC        0
#endif

The only way I could explain the behaviour you are seeing is if DSD_CHANS_DAC is defined, and the linux driver tries to select the DSD interface present in the device and it's descriptors. That would indicate 2 issues - DSD is enabled and Linux incorrectly selects that interface. What may be an issue here is that setting DSD_CHANS_DAC to 0 and NATIVE_DSD is undefined, then NATIVE_DSD will be defined as 1, which will enable it in the ref design...

Does this match your setup?

Posted: **Fri Nov 25, 2016 2:41 pm**

yes, you are right this is exactly wath happen to me:

this is my makefile config, you can see that i define DSD_CHANS_DAC=0, exactly as in other configuration from the reference design:

Code: Select all

XCC_FLAGS_MyConfig  = $(BUILD_FLAGS) -DI2S_CHANS_DAC=4 -DI2S_CHANS_ADC=12 -DNUM_USB_CHAN_OUT=4 -DNUM_USB_CHAN_IN=12 \
								-DMIDI=0 -DSPDIF_TX=0 -DSPDIF_RX=0 -DADAT_TX=0 -DADAT_RX=0 -DDSD_CHANS_DAC=0 \
								-DMAX_FREQ=96000 -DOUTPUT_VOLUME_CONTROL=0 -DINPUT_VOLUME_CONTROL=0 -DMIXER=0 -DHID_CONTROLS=0
INCLUDE_ONLY_IN_MyConfig =

If I simply avoid to define it in configuration all work perfectly without any patch!!

so probably a better fix of the problem can be editing devicedefines.h in this way:

Code: Select all

#if defined(DSD_CHANS_DAC) && (DSD_CHANS_DAC != 0)
    #if defined(NATIVE_DSD) && (NATIVE_DSD == 0)
        #undef NATIVE_DSD
    #else
        #define NATIVE_DSD       1  /* Always enable Native DSD when DSD mode is enabled */
    #endif
#else
    #define DSD_CHANS_DAC        0
#endif

In this way DSD_CHANS_DAC can be undefined or defined as 0 giving the same result!

Lorenzo

Posted: **Fri Nov 25, 2016 6:14 pm**

In this way DSD_CHANS_DAC can be undefined or defined as 0 giving the same result!

Yes, this is a more intuitive behaviour. Thanks very much for raising it - I have submitted an internal bugzilla #17588.

Still unsolved is why the linux USB audio driver tries to select the native DSD interface for playback...

XCore Exchange

xCore200 UAC2 Demo record only silence under Linux

Re: xCore200 UAC2 Demo record only silence under Linux

Re: xCore200 UAC2 Demo record only silence under Linux

Re: xCore200 UAC2 Demo record only silence under Linux

Re: xCore200 UAC2 Demo record only silence under Linux