You do realise that pin 26 is NOT a master clock output.
I2S is actually a three line data transmission protocol, not a 4 line as is used most of the time with today's devices.
As per 32 bit data, running at 44.1kHz.
Each stereo sample of data is sent down a 64 bit long word if you like. The bit clock, in this case, runs at 64x the sampling frequency (2 32 bit channels), so 2.8224MHz. Each rising edge of the bit clock is used to clock the 64 bit long data word into the audio sink device. The LR clock runs at fs and is used to tell the sink device what part of the 64 bit long data word is meant for the left channel and which is meant for the right. Half of the period of fs is equal to 32 sample periods of the bit clock, or represents the 32 bit data for one channel. The other half period of fs signals the 32 bits for the other channel.
With stereo 24 bit data running at 44.1kHz you're getting 2.1168MHz output on the SCK, which is representing the bit clock. Different nomenclature is often used for I2S inputs and outputs which can get a little confusing. Some use SCK to represent the System Clock, which is synonymous with the master clock. Some use SCK to represent the serial clock, which is used to represent the clock that is used to clock in the data present on the 'serial' data lines and is synonymous with the bit clock, so called the bit clock because it's used for clocking each individual bit into the sink device.
Then you've got even more confusion where there's the LRCLK, meaning, left/right clock, or the Word clock, which is used to represent the clock that switches to tell the sink device which part of the data stream is meant for the left or right channel.
The RN52 provides the standard 3 wire I2S communication required to transmit stereo audio. If your sink device requires a master clock to function then you will need to provide this separately!
Now the cut down data sheet brief that ST provide indicates that the device includes an on board PLL for the generation of a master clock. This will be required to run the noise shaping circuitry and over sampling of the amplifier.
According to the data sheet and as per the Flexiwatt pin out you need to connect the RN52s pin24 to pin12 and/or 13 of the power amplifier chip. Then you need to connect the RN52s pin25 to the power amps pin5 and the RN52s pin26 to the power amplifiers pin11.
It seems like the chip can work in either hardware or software control mode and I do not know which you are using, but this is important.
If you are using a hardware configuration then I'm guessing here that you need to set the configuration using pins 26 and 27 during device power on. You then need to enable the device and PLL via pin 2.
If you are using the power SO package then you will also need to use pins 20 and 21 to tell the device which serial input pin (pin 12 or 13) is going to be used as the data input. In addition to this you will also need to pull the device out of a mute condition by using pin 34.
If you are using the I2C bus to program the device then I do not know what you will need to do to make the thing work as the default configuration could be anything. It says that the PLL will function off of a 50 - 64x SCK, which would indicate that you do need to use 25 to 32 bit data. It is possible that this is only necessary in hardware mode, the PLL might be configurable via I2C to work from lower bit rates.
Two additional things, one related one not.
I have just done some tests of the I2S output myself, simply measuring the clock rates, to get an idea of what the settings do.
Setting the device to S|,0102 yields 44.1k sample rate with 24 bit data.
Setting the device to S|,0112 yields 44.1k sample rate with 24 bit data.
Setting the device to S|,0113 yields 44.1k sample rate with 24 bit data.
The last setting should work at 48, but does not. I wondered if this had anything do with the data being sent, so I loaded up a 48kHz sample rate file on my android device and nothing changed.
Setting the device to S|,0111 should give 32k sample rate and 32 bit data, but it's still stuck at 24 bit and 44.1k.
This is an improvement over the firmware before 1.16 that didn't work at all, but clearly the I2S settings don't work, the device is stuck in 24 bit 44.1kHz. I wonder if this has anything to do with the device sending the data to the RN52?
The second thing is that my trouble with the various S% settings seems to have come from the fact that I was using a Raspberry Pi's UART interface to program the RN52. Programming the same settings via the USB>UART bridge works as expected.
I have been in contact with a microchip representative, over the device with firmware version 1.10 failing to update to version 1.16 and we've yet to come to any kind of solution.