I believe that there is a mistake in the definition of TUD_AUDIO_HEADSET_STEREO_DESCRIPTOR. The nitfs in the audio descriptor is always 3 regardless of ITF_NUM_TOTAL.