Spent some time on the ESPHome Discord server tonight, getting my new build to compile successfully. Mega-THANKS to @robgough1970 for his extremely kind assistance resolving my ESP32-S3 build issues. My build is slightly different, using this ESP32-S3 Development Board. My reasons for using this particular board:
Pin headers are not pre-installed, allowing direct wire connections
Newer S3 revision of ESP32 processor
8191 KB of PSRAM
These are the (current) YAML updates necessary to get this board working:
Current pictures donāt give enough visibility to all the wire connection. Any chance youāll post better picture that could be used by others to re-create this, along with the esp code or links that give some insight into what is required to get this going?
Thanks. Somehow it seems I didnāt click on the first video when I ran through this thread. For others that might be as slow as me, the details are with the video over on the youtube site.
Hey, I tested some parts from your yaml in my voice setup and I have a lot of voice_assist_thinking_phase_id events. Do you also have constantly flashing LEDs? Next to TV, this is not niceā¦
But itās a really great idea to do it your way and define the states and use same Lambda for all, instead of my previous way to declare each individually. I might steal this for my project.
Where do you got the enum from? Is it a list or from esphome repository?
Pretty much the same YAML, no functional changes whatsoever. I added substitution values for all of the GPIO pin assignments, moving them all up to the top. I added comment lines to most sections of the YAML, including the parts links you provided. Two external links that helped me decide the GPIO assignments and better understand the I2S code and wiring are:
I still have to work on a static/buzz problem. After getting the ESPHome YAML to successfully compile, I left the speaker powered on all night to burn in. When I woke, it was making noise. I plan to spend most of the day tomorrow trying to resolve that issue. It might be the power supply, but right now thatās a powered USB 3.0 hub which has been reliably in use for months. I think itās more likely a loose Wago connection or too long wiring picking up interference.
Yes, I compared āyoursā and mine, besides the triggers, itās the same. For example:
Iām using on_wake_word_detected and on_listening and on_stt_vad_end events to trigger the LEDs. But in the code it used the response codes. So this could be nicer way to get less false led.
I couldnāt find a list which explains what is behind id(voice_assistant_phase)ā¦ Iām actually just interested in:
After wake word detected: scan
When progressing: slow pulse blue
When ok: pulse green
When error: pulse red
Your approach looks better, but I have light voice_assist_thinking_phase_id almost all the time while watching TV.
Maybe tomorrow Iām done cleaning up my code and can show you. Itās a big mess at the moment while working on status LEDs for presence detection and air quality.
Strange behavior. Double checked wiring, all connections good, amp board is definitely working, but yielding static. Bottom RGB LED is lighting dimly while static plays. Set logging level to VERBOSE, and VA logic is running as expected, setting LED status and waiting for VAD voice activation from the microphone. Iāll probably disconnect the amp while testing the microphone for proper operation, just to maintain some sanity.
@Rich37804 - A dumb question for you. Do you have GPIO18 directly connected to the center wire of the LED strip? Thatās what I have and Iām not absolutely certain about that.
Yes, directly connected.
LEDās act funny when something is amiss somewhere else.
I do have the positive/negative directly connected to the source power.
Thank you! Everything sources power from the Onn speaker USB power board at two common connection points (5VDC and GND). Iām guessing it might need a filter capacitor somewhere. Just have to try different things today, and hope something cures it.
EDIT: Found two problems that pretty much explained many of the issues. First, the GROUND connection on the INM441 I2S microphone had broken loose at the back of the microphone board. One of the hazards of shrink-wrapping the connections. Also found a wiring issue (canāt remember which lead it was at the moment) on the LED strip. The wire had not made a solid connection in the Wago lever-nut. Finally, I had not updated the ESPHome YAML to tell it there were three (3) LEDs in the string. After correcting the connections and LED code, the Voice Assistant is initializing much better. Ran into problems re-wiring the microphone, but thatās getting worked right now. Should have wake word recognition later tonightā¦
Well, the LEDs are working as expected now, and WOW is the initial white setting BRIGHT. Still appear to have a microphone issue, but decided not to troubleshoot it tonight.
i have the same issue with static. The interesting thing is when i first plug the unit into a wall socket i get static. When i say the wake word the static disappears completely and seemingly does not come back. Anytime the unit reboots though the static reappears until again i say the wake word. IF anyone has a solution for this please let me know. I have to admit this little project is no Alexa but it isnāt bad at all. Sure does beat the Atom Echo. I used the code that JLo worked on in his demo BTW and configured the same GPIO pins.
I was using two GPIO pins immediately adjacent to each other, and the static was worse. I moved it to another pin five spots away, and while itās not as bad, itās still an issue. I finally pulled the DIN pin from the amp, and itās STILL growling out loud. Definitely going to place a capacitor across the VCC and GND pins, just as soon as they arrive.
EDIT: It seems the length of the power wiring to the amplifier (creating resistance) could be a factor in noise from this amp.
It does! Thank you. My power wires are longer than six inches, supplying all of the ESP32, the amplifier, the microphone, and the LEDs. Iāll shorten those this evening, and see how much it may help. It might also help to use larger gauge stranded wire to minimize the resistance.
EDIT: Shortened the leads down to three inches, applied a tank capacitor across Vcc and GND, noise is lessened somewhat. However, what Iām now seeing is, the noise is being picked up by the amplifier board. Whether itās RF/Wifi from the ESP32, or from my iPhone, or even if I pass my hand over/around the amp board, it makes noise. I wonder if the Adafruit boards are any more quiet?
After getting the ESPHome YAML to successfully compile, I left the speaker powered on all night to burn in. When I woke, it was making noise. 
Itās a big mess at the moment while working on status LEDs for presence detection and air quality.
Still appear to have a microphone issue, but decided not to troubleshoot it tonight.
Definitely going to place a capacitor across the VCC and GND pins, just as soon as they arrive.
Thank you. My power wires are longer than six inches, supplying all of the ESP32, the amplifier, the microphone, and the LEDs. Iāll shorten those this evening, and see how much it may help. It might also help to use larger gauge stranded wire to minimize the resistance.