Does anybody know why I2S master clock is not enabled in the ESP32 port?

[DavideRossi]

It seems like it would be just matter of setting MICROPY_PY_MACHINE_I2S_MCK and adding the pin in line 378 of machine_i2s.c:

 .mclk = I2S_GPIO_UNUSED,  -->  .mclk = self->mck,

But maybe there's a reason why it has not already been done.
Does anybody know about that?

[miketeachman]

There will be a few more additions to the code to add this feature, but they are likely not difficult. Search for self->sck which will give clues to other code changes that are needed. There are specific pins on the ESP32 that support MCK. I'm not sure if those pins vary across ESP32 variants. It would be nice if the MicroPython I2S implementation could detect that the user is using a wrong MCK pin and raise an exception, rather than having a cryptic ESP-IDF message be raised to the user.

[DavideRossi]

It seems trivial nevertheless. As it seems equally trivial to enable slave mode (with the quirks that you have to mess with a standard interface as this is not currently supported by any port).
Oh well, when I have time I will give both a try (and get back to report).

[DavideRossi]

It works just fine, up to 32-bit/96 kHz. At that point, the MCK is already in the 25 MHz range. If I try to push it to 192 kHz, the DAC complains about a bad MCK signal.
I also have little means to verify this, as we’re above the frequencies that both my scope and my logic analyzer can manage.
The scope does show a signal, with expected significant attenuation. The waveform varies a lot, even in amplitude, and the frequency seems roughly around 30 MHz.
If I'm reading that correctly the ESP-S3 defaults to the 160MHz PLL clock as a source (see https://github.com/espressif/esp-idf/blob/6897a7bf40e27cabe7f5ef60e71a7426f4047c00/components/soc/esp32s3/include/soc/clk_tree_defs.h#L300) and Micropython uses the default configuration (

micropython/ports/esp32/machine_i2s.c

Line 375 in ef8282c

.clk_cfg = I2S_STD_CLK_DEFAULT_CONFIG(self->rate),

) so it should not be a hardware limitation.
It must also be said that I'm using a breadboard for these experiments and I guess this does not help.
At this point I just do not know if that is a problem that I can overcome...
I also played a little using different pins for the MCK signal and they all seem to work (ESP32-S3). I've not found a specific reference citing which can be used and, which is worse, I've also not found a way to programmatically check for that. We should ask in the ESP-IDF forum.
Now I'm curious about slave mode...