Analog Microphone audio source shows `ADC analog - quiet` after enable (audioreactive usermod)

[pjkerly]

What happened?

Analog Microphone audio source after initially staring goes to ADC analog - quiet after starting. The lights are no longer responsive to audio sound / music. This happens for both pre-compiled and well as self-build binaries (multiple versions) for audio reactive. I've already debugged and am working on a Pull Request to submit proposed changes, review, and hopefully incorporated.

To Reproduce Bug

Install esp32dev_audioreactive binary. Configure the hardware. I'm using GPIO 36 R/O (supposed on ADC1). Select Generic Analog and Enable. Refresh the Info screen and see the audio source to go quiet.

Expected Behavior

Expected ADC Audio Source to register Peak (varying in %) in reaction for audio.

Install Method

Binary from WLED.me

What version of WLED?

v0.14.4

Which microcontroller/board are you seeing the problem on?

ESP32

Relevant log/trace output

No response

Anything else?

There are several issues with the current source code. I will Submit a PR soon. Summary 1). I2S_MODE_MASTER is not a suitable mode for I2S ADC configuration. Thie mode is for controlling an external I2S device. 2). The order of initializing the I2S driver and the associated ADC is not created in correct order. Should configure the ADC completely before installing the driver and then enable the I2S and ADC on the driver. The current order seems to be intermittent failure/success. Correct order seems to eliminate this intermittent failure. 3). The I2SSource getSamples tries to down sample to 16 bits when it shouldn't. This consistently causes the symptom of going to quiet.

Code of Conduct

• I agree to follow this project's Code of Conduct

[netmindz]

Hi @pjkerly

Thanks for looking into this common bug, can you try the MoonModules fork of WLED to see if the same issues exist as this is the source of that userMod but the version in this repo is very old version

Do we have the same ordering issue in MM @softhack007 ?

[softhack007]

Hi @pjkerly,

The first thing to try - and this helped in many cases - is to install WLED from this site https://wled-install.github.io/

instead of installing from install.wled.me. My suspicion is that the default WLED bootloader (created by @Aircoookie many years ago) has some problems with initializing the I2S-ADC peripheral unit properly.

There is an old PR that also goes into the same direction atuline#257

Update:

I'm using GPIO 36, [then] select Generic Analog and Enable. Refresh the Info screen

This reminds me of another important solution

Caution

After changing audioreactive analog settings, a power off-on cycle is required - normal reboot is not enough. I2S-ADC will only initialize properly when switching the power off/on. If you don't do this, then "ADC analog - quiet" is expected behaviour.

[softhack007]

About your other findings - please double-check each of them individually before making a PR.

1). I2S_MODE_MASTER is not a suitable mode for I2S ADC configuration.

Where does this insight come from? Can you point to some espressif documentation that supports this statement?

Actually the I2S-ADC init sequence is based on examples from espressif, and it's the same sequence you find in other open source libs that use I2S-ADC for input.

https://github.com/espressif/arduino-esp32/blob/43016390510e23422b0abcf432ceef2479a05f41/libraries/ESP32/examples/I2S/HiFreq_ADC/HiFreq_ADC.ino#L42-L75

and the "modernized" one

• https://github.com/espressif/arduino-esp32/blob/43016390510e23422b0abcf432ceef2479a05f41/libraries/I2S/examples/ADCPlotter/ADCPlotter.ino#L76
• https://github.com/espressif/arduino-esp32/blob/43016390510e23422b0abcf432ceef2479a05f41/libraries/I2S/src/I2S.cpp#L226
• https://github.com/espressif/arduino-esp32/blob/43016390510e23422b0abcf432ceef2479a05f41/libraries/I2S/src/I2S.cpp#L115

(my conclusion from the "modernized" espressif code was - if you want to specify a sample rate, you need to use MASTER mode together with I2S_MODE_ADC_BUILT_IN)

2). The order of initializing the I2S driver and the associated ADC is not created in correct order.

This might be possible, however there are lots of user setups where analogue microphones work without any problems - including my own test setup (but i'm not using analogue mics since almost 2 years - I2S digital is just better). If the sequence was generally wrong, I would expect that no analog microphone had ever worked 🤔

3). The I2SSource getSamples tries to down sample to 16 bits when it shouldn't

Maybe you misunderstood the code? There is no down-sampling involved - the code simply scales the microphone sample (float) which was previously up-scaled by the I2S driver - needed because we are requesting 32bit samples. As this works in many setups, I doubt that the processing code is totally wrong.

Please try if it works better in pure 16bit mode (add -D I2S_USE_16BIT_SAMPLES to build_flags). If this helps, then we should change the ADC audio source driver to always use 16bit mode.

You might want to try these build_flags :

• -D I2S_USE_16BIT_SAMPLES to force 16bit processing instead of 32bit
• -D I2S_GRAB_ADC1_COMPLETELY to enable continuous sampling
• -D MIC_LOGGER for sound input monitoring & debugging (use arduino serial plotter to see a graph)

Plus, try if using a "V4" buildenv - with newer arduino-esp32 version 2.0.9 instead if 1.0.6 - makes a difference. You can use this buildenv
https://github.com/Aircoookie/WLED/blob/076497e14da2fc268f8424809dfc3479f2dda235/platformio.ini#L483

Or this one, if you add the AR build_flags and lib_deps
https://github.com/Aircoookie/WLED/blob/076497e14da2fc268f8424809dfc3479f2dda235/platformio_override.sample.ini#L250

[softhack007]

Also this page is a great help if you cannot get analog microphones to work:

https://github.com/atuline/WLED/wiki/First-Time-Setup#analog-or-i2s-digital

Plus sometimes its sufficient to use a different pin for the analog input.

And it's always good to double-check that the microphone Vcc (3.3V) and GND connections are stable. Even when you can see reaction with a simple a analogread() sketch, it might still be that the microphone signal is too distorted when Vcc or GND are not stable.

[softhack007]

Do we have the same ordering issue in MM @softhack007 ?

Actually I never had problems that could be related to ordering of initializations.

@pjkerly The MoonModules fork is the "master" for audioreactive - please try if your microphone works better in the WLED-MM fork.

The MoonModules fork of WLED (WLED-MM) can be found here https://github.com/MoonModules/WLED

[pjkerly]

Is there a known hardware issue that you are aware of? The comment below makes me even MORE suspicious that the I2S ADC Mode isn't being setup/disabled correctly. As a hardware guy, there is a problem if you have to physically power off the device to get it back in to a know state in order to get the software to work. In my opinion,, either a hardware power reset nor a hardware reboot should not be required. So far with my changes, it has been consistently stable without power cycles or resets required.

This reminds me of another important solution

Caution
After changing audioreactive analog settings, a power off-on cycle is required - normal reboot is not enough. I2S-ADC will only initialize properly when switching the power off/on. If you don't do this, then "ADC analog - quiet" is expected behaviour.

[pjkerly]

I've seen the examples and a lot of code examples that have used th I2S_MODE_MASTER. With my changes, I've tried with it and without. And I saw issues when it was enabled. From the espressif documents, I2S_MASTER_MODE generates the required clocks:
Bit Clock (BCLK): Synchronizes data transmission at the bit level.
Word Select (WS) or Frame Clock: Indicates the beginning of a word/frame in audio transmission.

But in I2S ADC Mode, I'm not even using those pins. It apparently is needed for external digital I2S devices which is not the case in ADC Mode from what I can tell. I did wonder if somehow the ADC Mode might use those signals internally. But removing it seemed to resolve some issues. So I decided to not enable it.

After making all my other changes, I did try enabling and disabling the I2S_MODE_MASTER and it definitely caused issues when enabled. I will again triple check.

[softhack007]

Is there a known hardware issue that you are aware of?

Well, there are known problems with espressif driver software that requires power cycling - see atuline#231 (comment). Unfortunately, the workaround suggested by another user (same thread) did not work. So until espressif fixes their stuff (will not happen) we are stuck with "power cycle needed".

The espressif/arduino-esp32#4782 is relevant when you have analog audio (mic or line-in) and another analog signal, for example potentiometer. In this setup, I2S and analogread() drivers seem to get into conflict, and they both stop working.

Due to espressif/esp-idf#7442, changing analog mic config always requires a hardware reset (power off/on cycle)

espressif/esp-idf#7442 concludes that "only a hard CPU reset can disconnect the I2S signal from built-in ADC".