I2s audio debugging with buspirate

Funcan · May 22, 2025, 5:22pm

Has anybody done anything with i2s audio and the bus pirate? I’ve just started experimenting with trying to add audio expansion onto the EMF2024 badge, and I’m failing miserably. I will dig out a scope and at least check that clocks look like clocks and bitstreams look like bitstreams, but I’m rather stuck for debugging beyond that currently.

ian · May 22, 2025, 5:36pm

I bought i2s mics and amps for exactly that. Haven’t done much with it yet though.

Would be happy to start a new mode and see what pio programs we might be able to implement. Maybe a simple mic to wav and wav to amp? Or just mic to amp test?

ian · May 22, 2025, 5:42pm

github.com/raspberrypi/pico-extras

src/rp2_common/pico_audio_i2s/audio_i2s.pio

master

;
; Copyright (c) 2020 Raspberry Pi (Trading) Ltd.
;
; SPDX-License-Identifier: BSD-3-Clause
;

; Transmit a mono or stereo I2S audio stream as stereo
; This is 16 bits per sample; can be altered by modifying the "set" params,
; or made programmable by replacing "set x" with "mov x, y" and using Y as a config register.
;
; Autopull must be enabled, with threshold set to 32.
; Since I2S is MSB-first, shift direction should be to left.
; Hence the format of the FIFO word is:
;
; | 31   :   16 | 15   :    0 |
; | sample ws=0 | sample ws=1 |
;
; Data is output at 1 bit per clock. Use clock divider to adjust frequency.
; Fractional divider will probably be needed to get correct bit clock period,
; but for common syslck freqs this should still give a constant word select period.

This file has been truncated. show original

There is an example in the SDK now. I’ll check it out tomorrow.

ian · May 23, 2025, 11:20am

These are the I2C parts I ordered a while ago.

Decoders:

Microphones:

INMP441
MSM261S4030H0

A new firmware branch i2s_mode has the framework for a new I2S mode. Nothing is implemented, but there is a playground to test things. I’m going to see if I can make the decoder screech a bit.

ETA: pushed an update with the I2S data out PIO installed. It makes crackling on my I2C amplifier, so I assume something is happening.

ian · May 24, 2025, 11:51am

This is simply alternating periods of low and high (0x00000000, 0xffffffff). The signals look good. the data is high then low for left and right channels. Word Select (Left/right) is changing at ~44.1kHz.

Stole a sine wave function from a random project and sent the output to the I2S bus.

We do indeed get a poorly formed sine wave in the auto output.

ian · May 24, 2025, 12:31pm

    // 10 cycles of 1kHz at 44.1kHz = 10 * 44.1 = 441 samples
    #define TABLE_SIZE 441
    #define AMPLITUDE 32767.0  // Max amplitude for 16-bit signed
    #define CYCLES 10
    #define SAMPLE_RATE 44100
    #define FREQ 1000

    int16_t sine_table[TABLE_SIZE];

    for (int i = 0; i < TABLE_SIZE; i++) {
        // Each index represents: i / SAMPLE_RATE seconds
        // For 10 cycles: phase = 2*pi*FREQ*(i/SAMPLE_RATE)
        // But for exactly 10 cycles in 441 samples: phase = 2*pi*CYCLES*i/TABLE_SIZE
        double phase = 2.0 * M_PI * CYCLES * i / TABLE_SIZE;
        sine_table[i] = (int16_t)(AMPLITUDE * sin(phase));
    }

    for(int i=0; i<100; i++) {
        // Send the sine wave samples to the PIO
        // The PIO will handle the timing and output
        for(int j=0; j<TABLE_SIZE; j++) {
            // Send each sample, shift left to fit in 32 bits
            pio_sm_put_blocking(pio_config.pio, pio_config.sm, sine_table[j] << 16 | (sine_table[j] & 0xFFFF));
        }
    }

Calculate 1khz sine wave, 10 cycles so it’s smooth (44.1*10=441). It goes into a lookup table for speed, this could be DMAed for the best performance.

It is generally reliable. Occasionally the signal is a bit garbled.

Nifty to note: I2S uses signed values where 0 is the midpoint.

Mode config option to set I2S frequency (and mono-stereo?)
sine command to output sine waves. Options to configure frequency. Some nifty math to calculate the right lookup table size and iterations for a smooth wave.
Handle incoming I2S, perhaps with a loopback function run by DMA (eg microphone to DAC).

ian · May 24, 2025, 2:26pm

I2S mode configuration is working. Sample frequency 4K-96K, data bits is limited to 16 by the PIO.

Still a bit of a hack, but [ is attached to a sine wave table generator. The table is customized for the current sample frequency. The sine wave freq is configurable, but that feature isn’t exposed yet.

The very common audio CD sample rate of 44.1kHz does not divide equally (44100/1000 = 44.1 samples per cycle), so the sine wave is incomplete.

The sine wave generator tests for a remainder (44100/1000 = 44.1).
Then tests if a 10 cycle table also has a remainder ((44100/1000) * 10 = 441).
If not, it generates a 10 cycle sine wave in the table, assuming it will fit.
Max sine wave table is currently 500 samples.

This won’t be perfect for every combination of sample rate and sine wave frequency, but it will provide a useful range.

Working pretty solid now. The problems I heard before were due to my noise canceling headphones freaking out a bit at the volume.

This was an interesting and unexpected weekend project Maybe tomorrow I’ll have a look at the microphone.

ian · May 25, 2025, 3:13pm

sine is now a command that accepts a frequency and duration options. Default is 1000Hz for 2 seconds.

Previously the wave fit test tried 1 and 10 cycles. Now it tries 1, and then 10 to 2 cycles. This will open up some previously unavailable frequencies.

Added an I2S input PIO program and attempted to sample from the INMP441 microphone and output on the UA1334A decoder. So far I only get static, but the waveform looks ok.

Pushed to the wrongly named i2c_mode branch of the repo