Cannot get I2S sample working

[Elppu]

Hi!

I cannot get I2S echo sample to work on nrf54L15. Tried following overlay

&pinctrl {
	i2s20_default: i2s20_default {
		group1 {
			psels = <NRF_PSEL(I2S_SDIN, 1, 14)>,
					<NRF_PSEL(I2S_LRCK_M, 2, 7)>,
					<NRF_PSEL(I2S_SCK_M, 2, 9)>;
		};
	};
};

i2s_rxtx: &i2s20 {
	status = "okay";
	pinctrl-0 = <&i2s20_default>;
	pinctrl-names = "default";
};

build configuration  nrf54l15pdk_nrf54l15_cpuapp

I verified with oscilloscope that no clock is modulated or wordselect switched. What might be the problem?

Is the i2s driver mature on nrf54 series / something new that has to be done?

[Einar Thorsrud]

Hi,

Wich pins are you using now? The I2S peripheral can only be used on GPIO port P1 (see Configuration). Moreover, MCK an dSCK must be on clock pins (see Clock pins).

[Elppu]

Hi!

Thank you! Yes should read more documentation. I still have more issues.

&pinctrl {
	i2s20_default: i2s20_default {
		group1 {
			psels = <NRF_PSEL(I2S_SDIN, 1, 14)>,
					<NRF_PSEL(I2S_LRCK_M, 1, 11)>,
					<NRF_PSEL(I2S_SCK_M, 1, 12)>;
		};
	};

This is my current pins and now I get data. Next issue is that my left channel data does not seem to be correct. Right channel seem to be same as my test sound when I plot it, but left channel not. Left channel is the first word? The buffer is interleaved?

L0 R0 L1 R1 L2 R2 etc??

Clock pin seem to be now correct. How about wordselect?

Example data:

Left: 1034, Right: -1762
Left: -684, Right: -1628
Left: -680, Right: -1484
Left: 1276, Right: -1086
Left: 664, Right: -898
Left: 1120, Right: -824

Plotted from log prints:

For some reason only right channel seem to be about right maybe? Did I understand correctly how the left and right data is in the buffer? Documentation

how will this be filled with this mic? Is this correct way to get the data?

        int32_t *samples = (int32_t *)mem_block;
        int total_words = size / BYTES_PER_SAMPLE;
        int num_frames = total_words / 2;  // each stereo frame = 2 words
        
        static int frame_count = 0;

        for (int i = 0; i < num_frames; i++) {
            // Left channel is first word in each frame
            int32_t raw_left = samples[2 * i];
            int32_t left = (raw_left >> 8) & 0xFFFFFF;
            if (left & 0x800000) {
                left |= 0xFF000000;  // Sign extend
            }
        
            // Right channel is second word in each frame
            int32_t raw_right = samples[2 * i + 1];
            int32_t right = (raw_right >> 8) & 0xFFFFFF;
            if (right & 0x800000) {
                right |= 0xFF000000;  // Sign extend
            }

[Elppu]

Might have been personal skill issue with bit magic. Now I get stream of values that make more sense.

Next I need to stream 16khz samples to PC to verify, but that might require SEGGER RTT .