I2s Stereo input slave mode example

[LouisCARRIOU]

Hello,
I'm trying to use your nice i2s lib in slave mode.
I'm using a WN8782 module to analyze right and left sound but iI don't find a working solution .
I also tried with a CS5343 but I have the same problem.
Do you have a working example for i2s stereo input I'm slave mode ?
BR
Louis

[appliedsine]

I have this working in slave mode but the audio is really distorted, it feels like a clock issue to me but it's hard to tell with the scope I have.

Re: your issue, Louis, all you have to do to run this in slave mode is call

i2s_program_start_slaved(pio0, &i2s_config_default, dma_i2s_in_handler, &i2s);

instead of

i2s_program_start_synched(pio0, &i2s_config_default, dma_i2s_in_handler, &i2s);

the default pin connections are as follows:

• i2s data out pin -> GPIO6 (out from pico)
• i2s data in pin -> GPIO7 (in to pico)
• i2s bit clock 'BCLK' -> GPIO8 (if using the default config this needs to be 3.072 MHz)
• i2s frame clock LRCLK -> GPIO9 (if using the default config this needs to be 48 KHz)
• master clock (SCK) out -> GPIO10 (freq is ~12.288 mHz)

I have verified that the above does in fact work, the codec I am using is my test codec, pcm5102A. I get audio to come through but it is heavily distorted. If you get no audio, you may need to look at the codec to see if it is being initialized correctly.

To the author of the code, I am working on a hobby project that I really want to incorporate this into, will pay $100 (+ another $100 if we get it running) for 1 hour of QA/help on this if you're interested.

[malacalypse]

See my response below for a first set of troubleshooting. After looking at the PCM5102A datasheet, I observed a few things:

1. It's not a codec (codec = coder/decoder, e.g. ADC + DAC), it's just a DAC. It also does not appear (after a quick glance through the datasheet) to have a master mode at all. So I'm not sure how you're getting this library to work in I2S slave mode with this DAC.
2. I assume you're using the output master block only, and in full 32-bit (64fs for bck) mode. The code in this library won't work at any fewer bits per frame than 64 (32 bits L, 32 bits R).
3. Make sure the FMT pin on your DAC is pulled low to ground (not left disconnected, definitely not pulled high).
4. Your audio data should be left-aligned in the buffer, signed integer values - see below (not floating point), ranging from full-scale negative -(2^31 - 2) to full-scale positive (2^31-2). Check your math on this - one good way to do it is to have your audio algorithm output a ramp (aka sawtooth) at some frequency from minus full scale to positive full scale and starting over again. You can use the logic analyzer to read the actual data values being sent on the bus and compare them with the computed values in the memory to be sure they're matching up.

If you don't have a good 'scope with logic analysis, a $20 8 channel logic analyzer like the Sparkfun 24MHz unit (under $20) or any of the eBay ones that are similar is an absolute must for troubleshooting I2S issues.

[malacalypse]

Hi all, this isn't a bug so please don't open issues for not-bugs. Start a discussion instead.

@LouisCARRIOU I don't have example code but as @appliedsine said, it works perfectly in slave mode if you follow the directions carefully. I spent a lot of time writing them and unless there's something wrong or unclear there (which, please state specifically so I can fix), they contain the exact, detailed, and necessary information needed to make this work, given that you understand how I2S works (which is also, I hope, thoroughly explained, though it's a very simple protocol and you should be able to find other information which might do a better job elsewhere).

@appliedsine : Have you checked the following:

1. That you are using SIGNED 32-BIT INTEGERS (e.g. int32_t, not int, not uint_32_t, and definitely not int16_t or uint16_t) for your buffers, for the dma setup, etc?
2. Does your codec output signed, left-justified I2S PCM values or is it in a different format (some codecs might output unsigned values with 0 as maximum negative, for instance). You also want to be sure the codec is in I2S mode and not simply left-justified mode, because I2S has an extra offset clock at the beginning of the frame and that can cause distortion too. Use a logic analyzer to look at the actual I2S data - ideally time-synched with the input waveform so you can see if, numerically, the values read by the codec look correct.

Check those aspects first, make sure the data on the bus is truly I2S (re-read my description carefully to note the offset clock at the frame start), and you can try simply piping input to output with no math (e.g. just copy the input buffer to the output buffer) to make sure that the basic data flow is working.

I'm closing this issue because it's not a bug, but if you want to discuss it further, open a discussion about it and I'll chime in as I have time.

[malacalypse]

My apologies, I realized discussions weren't enabled on this repo - I thought they were. They're enabled now!

[appliedsine]

Hey man, thank you for opening this up!

I'm sorry, I wasn't clear, I am using a bluetooth module which outputs I2S, my test signal flow is as follows:
bt module -> pico -> pcm5102, where the bt module is master providing 48KHz frame clock and 3.072MHz bit clock to both the pico and the PCM5102. I believe the bluetooth module is providing the correct alignment (I2S LJ) but I will double check; documentation is a bit scant. I just have this broken out on a breadboard for validation, in my project, I am using the PCM 5242 which is also just a DAC. My hope was to use the rp2040 for some simple processing or possible serve as a microphone interface.

regarding the buffers, I did not change the data types from what was in the example code:

static void process_audio(const uint32_t* input, uint32_t* output, size_t num_frames)

so maybe that's where I messed up. My scope is at my dad's so I will get my hands on it in a day or so.

I was able to get a simple tone to play, albeit in the left channel (likely a data conversion error on my part), I am definitely not a dsp expert but I know how to stay up late and do caffeine.

Thank you very much for your time, sorry I am late to the party. I'll tinker with it today and see what I figure out

[murdo-g]

@flpvsk @appliedsine I had a similar issue with distorted input data when using the i2s_bidi_slave example which appeared to be caused by bit-shifted signed data. I rewrote the PIO program #8 to closely resemble the i2s_in_slave example which solved the distortion for me using an SGTL5000 codec.

If you have a moment, it would be great if you could test the rewritten program with your setups to verify things a bit more.

Thanks!

[flpvsk]

@murdo-g just checked it and can confirm that your patch fixes the i2s_program_start_slaved :) Even with the last instruction in the PIO code removed as per #8 (comment)

[malacalypse]

Thanks flpvsk for the feedback, I'll take a look at that MR again shortly.

[longo92-vrmedia]

Hi guys,
i am using sgtl5000 set to 16bit per sample, 48Khz on microphone. I am using i2s_program_start_synched. What i've observed is that the input buffer (which each value is an int32_t) has following values:
2147450880
8449
2147450880
1217
2147450880
2881
2147450880
4191361
2147450880
5441
2147450880
4929
2147450880
4191169
2147450880
7745
2147450880
6017
2147450880
7425
2147450880
5697
2147450880
4182657
2147450880
4192897
2147450880
4193217
2147450880
4189377
2147450880
4184513
2147450880
14273
2147450880
4184961
2147450880
4185729
2147450880
641
2147450880
4481
2147450880
12737
2147450880
257
2147450880
14017
2147450880
4180289
2147450880
4178497
2147450880
4185601
2147450880
4174849
2147450880
6465
2147450880
4191681
2147450880
4188161
2147450880
4179201
2147450880
4182401
2147450880
....

• why i have 2147450880 alternating?
• if i write the bit representation of 8449, for example, is: 00000000 00111111 11110100 10000001. It seems on 24 bit instead of 16...am i correct?

[malacalypse]

Edit: Your reply is off-topic, by the way. You're not using slave mode, nor are you discussing the slave mode examples. I'll reply once here, but if you want further discussion please open a new discussion with an appropriate title, such as "Using with sgtl5000".

* why i have 2147450880 alternating?

That's the other channel's data. There are always two channels in I2S, if you read the README you will see that they are interleaved.

* if i write the bit representation of 8449, for example, is: 00000000 00111111 11110100 10000001. It seems on 24 bit instead of 16...am i correct?

Errr, the "bit representation" of 8449 is 0010 0001 0000 0001, you've written the binary for 4,191,361.

And you do realize that you're looking at an int32_t, which is 32 bits, not 24, right? Do you understand how I2S works? How the bits are shifted MSB into the appropriate word in the frame? How the various bit depths work in the frame? I detail all of this in the README...

The fact that you're not getting any negative numbers in the buffer implies that you may not have the chip set to I2S mode, but to one of the other modes (pcm? left- or right- justified?). Only true I2S mode will work.