On Fri, Feb 22, 2019 at 4:27 PM Russell King <rmk+ker...@armlinux.org.uk> wrote: > > It appears that TDA998x derives the CTS value using the supplied I2S > bit clock (BCLK) rather than 128·fs. TDA998x uses two constants named > m and k in the CTS generator such that we have this relationship > between the source BCLK and the sink fs: > > 128·fs_sink = BCLK·m / k > > Where BCLK = bclk_ratio·fs_source. > > We have been lucky up to now that all users have scaled their bclk_ratio > to match the sample width - for example, on Beagle Bone Black, with a > 16-bit sample width, BCLK = 32·fs, which increases to 64·fs for 32-bit > samples. 24-bit samples are sent as 32-bit. > > We are now starting to see users whose I2S blocks send at 64·fs for > 16-bit samples, which means TDA998x now breaks. > > ASoC has a snd_soc_dai_set_bclk_ratio() call available which sets the > ratio of BCLK to the sample rate. Implement support for this. > > Signed-off-by: Russell King <rmk+ker...@armlinux.org.uk>
Works with an imx6q ssi, but only if the card driver calls set_bclk_ratio() with the correct value, which is 32/channel for the fsl_ssi in master mode. I will post an RFC patch shortly which adds bclk-ratio support to simple-card. Tested-by: Sven Van Asbroeck <thesve...@gmail.com> _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
