Help with troubleshooting a Voice Assistant with micro_wake_word

Hello,

I came across some examples using an ESP32 S3 with an onboard tiny wake word detection. So I followed this example: How To Setup On-Device Wake Word Detection For Voice Assistant using Micro Wake Word | Smart Home Circle

But it is not working. There was always a loud noise from the speaker. So I found some other examples where the I2S for the microphone and the speaker are split. I tried this and the tiny wake word was working once. But then never again. I have no idea how to troubleshoot this and hope for some hints or help.

What is working: I can switch the Microphone Mute on or off in HA and this also is shown by my LED ring.

What is not working: No matter how I say the wake word, the LED ring is not responding to it.
If I change the LEDs (color, brightness, etc.) there is also the loud noise on the speaker.
But this seems to be a minor issue which may can be fixed later with a muting either in hardware or maybe in software while the light is changed.

What I have tried:
I have rewired everythin several times, measured for the correct connections and shorts. In my last try yesterday I also soldered every connection to be sure I have no loose contact.

Here is my ESP-Home File:

The wiring of the data pins is also commented in the code below. The LRCLK Pin of the Microphone is connected to GND as shown in the example above, the supply of the microphone is 3,3V from the ESP32, the amp supply is 5,0V from the ESP32 5V Pin.

esphome:
  name: voice-assistant
  friendly_name: Voice Assistant
  platformio_options:
    board_build.flash_mode: dio

esp32:
  board: esp32-s3-devkitc-1
  variant: esp32s3
  framework:
    type: esp-idf

    sdkconfig_options:
      CONFIG_ESP32S3_DEFAULT_CPU_FREQ_240: "y"
      CONFIG_ESP32S3_DATA_CACHE_64KB: "y"
      CONFIG_ESP32S3_DATA_CACHE_LINE_64B: "y"
      CONFIG_AUDIO_BOARD_CUSTOM: "y"

# Enable logging
logger:

# Enable Home Assistant API
api:
  encryption:
    key: "N9N+F3hmN6Sv+8KulwdNhMy9dVVFhdbDnZtQlpHWqws="

ota:
  - platform: esphome
    password: "cadad1e70e9ab95a8b7f90312264e019"

wifi:
  ssid: !secret wifi_ssid
  password: !secret wifi_password

captive_portal:

web_server:

psram:
  mode: octal
  speed: 80MHz

light:
  - platform: esp32_rmt_led_strip
    id: led_bar
    rgb_order: GRB
    chipset: ws2812
    pin: GPIO5
    num_leds: 24
    rmt_channel: 0
    name: "LED bar"
    effects:
      - pulse:
      - addressable_scan:
          name: scan
          move_interval: 100ms
          scan_width: 1 

switch:
  - platform: template
    id: mute
    name: "Mute microphone"
    optimistic: true
    on_turn_on: 
      - micro_wake_word.stop:
      - voice_assistant.stop:
      - light.turn_on:
          id: led_bar
          red: 100%
          green: 0%
          blue: 0%
          brightness: 30%

    on_turn_off:
      - micro_wake_word.start:
      - delay: 2s
      - light.turn_off:
          id: led_bar 

     
i2s_audio:
  - id: i2s_in
    i2s_lrclk_pin: GPIO3  #MIC WS
    i2s_bclk_pin: GPIO2 #MIC SCK
  - id: i2s_out
    i2s_lrclk_pin: GPIO6  #AMP LRCLK
    i2s_bclk_pin: GPIO7 #AMP BLCK

microphone:
  - platform: i2s_audio
    id: va_mic
    adc_type: external
    i2s_din_pin: GPIO4 #MIC SD
    channel: left
    pdm: false
    i2s_audio_id: i2s_in
    bits_per_sample: 32bit
    
speaker:
    platform: i2s_audio
    id: va_speaker
    i2s_audio_id: i2s_out
    dac_type: external
    i2s_dout_pin: GPIO8   #AMP DIN
    mode: mono

micro_wake_word:
  models:
    - model: alexa
  on_wake_word_detected:
    - voice_assistant.start:
        wake_word: !lambda return wake_word;      
    - light.turn_on:
        id: led_bar
        red: 0%
        green: 0%
        blue: 100%
        brightness: 80%

voice_assistant:
  id: va
  microphone: va_mic
  speaker: va_speaker
  noise_suppression_level: 2.0
  volume_multiplier: 4.0
  on_stt_end:
       then: 
         - light.turn_off: led_bar
  on_error:
          - micro_wake_word.start:  
  on_end:
        then:
          - light.turn_off: led_bar
          - wait_until:
              not:
                voice_assistant.is_running:
          - micro_wake_word.start:

How do I troubleshoot this? Can I test the speaker from Home Assistant? How do I check if the Microphone is recognizing anything at all?

best regards
Andreas