On Wed, Jun 29, 2016 at 09:50:24PM +0300, Alexander Shiyan wrote:

...

Среда, 29 июня 2016, 21:26 +03:00 от Nicolin Chen nicoleotsuka@gmail.com:

On Wed, Jun 29, 2016 at 10:52:01AM +0300, Alexander Shiyan wrote:

...

...

your failed case. Could you please tell me what were the FSclk, BITclk and size of your case?

...

Here is my oscillograms with/without patch: Bclk=12 MHz, sound is just a "cat /dev/urandom > /dev/dsp", i.e. FS=8 KHz, MONO. Upper (yellow) line is TXDATA, Lower (blue) is FS.

Measurement without patch: https://cloud.mail.ru/public/Rcsv/xr88V5Sa1 Measurement with patch: https://cloud.mail.ru/public/NCwn/DQnzioQEL

Thanks for the reply. So in the case without patch, it looks like SSI just sent the data ignoring the slot size right?

I could be wrong, but think that WL (word length) still point to number of valid data bits within 32bit chunk (as defined for I2S-Master mode), but since DC=0, this cause to send next word to the channel again and again.

I see. Thanks for the fix.