Serial PCM Audio to and from AnalogAudioStream ESP32

[wsciaroni]

I am have a project that I am trying to run on the ESP32. I have a couple of fundamental pain points as I learn to use this library.

Project

You can take a look at my embedded code for this project here: https://github.com/wsciaroni/wsciaroni_kv4p-ht/blob/d6fd7c3e870ea084421675657cef2f4eceb9f21a/wsciaroni_kv4p-ht.ino

Currently, I have a GeneratedStream just to put audio through, but that is just for testing.

Project Requirements

• Hardware, I am using the esp32 devkit v1. I am using its built-in ADC and DAC.
• Audio Sample Rate: The audio should have a sample rate of 44100Hz.
• Audio Channels: I only need one audio channel. The audio is mono.
• Serial Settings: I want Serial to be configured with a baud rate of 921600.
• Serial Protocol: Simple command/data protocol with length indication. (Described below)

I want two modes for this. Either rx or tx.

In RX mode, I want to use the AudioTools library to read from the built in Analog to Digital converter. I want to use the AudioTools library to automatically send this audio stream to the serial port in 8 Bit PCM format. No separators.

In TX mode, I want to manually add bytes to an AudioTools BufferedAudioStream. I want to use the AudioTools library to automatically send this audio stream to the built in Digital to Analog converter. I will put the bytes on the stream as I read them from 8 BitPCM coming in from Serial.

When in Tx mode, the rx audio should not be sent.

The Serial communication will come in the following format:

• Other device will send one byte indicating whether the following is a CMD or DATA.
• If it is a CMD, one more byte will be provided indicating which mode to change to.
• If it is a DATA, two bytes will be provided indicating the length of the data to come. Then, the data will be the tx audio stream in 8 bit PCM format.

I think I want to use an AnalogAudioStream to read from and write to the ADC and DAC.
I think I want to use StreamCopy to take data from the incoming stream and write to the outgoing stream.

I don't think I can begin the AnalogAudioStream In in RX_MODE and AnalogAudioStream Out int TX_MODE at the same time due to the underlying implementation.

Questions

1. How should I most efficiently read a chunk of PCM8bit data from Serial and write it to a buffer then have the data written via the internal DAC?

2. How should I most efficiently read from the internal ADC and write the data to Serial in a PCM8bit stream of bytes.

3. What is the setup and teardown required to start and stop the different streams when I switch modes?

[pschatzmann]

You would use an EncodedAudioStream with an 8 bit codec!
You can directly use Serial as source and destination...

I think your sample rate is a bit high for Serial: you need to use > 396900 baud

I think there are some examples in the communction folder!

[wsciaroni]

Also this one: https://github.com/pschatzmann/arduino-audio-tools/blob/main/examples/examples-communication/serial/send-8bit-receive/send-8bit-receive.ino

I think this example solves my question <2> from my original post. However, I'm still stuck on the others:

How should I most efficiently read a chunk of PCM8bit data from Serial and write it to a buffer then have the data written via the internal DAC?

What is the setup and teardown required to start and stop the different streams when I switch modes? I don't think I can set the ADC to be both RX and TX simultaneously can I?

[pschatzmann]

For your first open question: you don't: let the framework handle this for you! Read the chapter about Encoding/Deconding. You want to use L8!

Just follow the Arduino conventions and call end() ! If you have any doubts check in the Class documentation for the available methods.

[wsciaroni]

Maybe I left some details out of my write-up.

The Serial in to the ESP32 contains a simple command/data protocol with length indication. This means that I cannot use the EncodedAudioStream to read from serial (or I would consume my CMD messages as audio data!).

How would you propose using your framework for me to indicate I have a certain number of bytes of audio data available on Serial without the stream consuming my CMD messages?

[pschatzmann]

Use a container: I have the BinaryContainer which does what you want: your commans you can send as metadata.
But I guess it would be more reliable if you just use a separate serial port for the commands.

[wsciaroni]

Okay, I read over https://www.pschatzmann.ch/home/2023/04/28/arduino-audio-tools-codecs-and-containers/

It looks like it requires a fixed length MetaData? Is that correct? It shouldn't be that big of a deal to make that happen. So, this keys off of the ContainerType when it reads its header? Then it determines if the incoming is Metadata or Data?

I don't see an example of pure audio pcm data along with some additional custom (meta) information. Is that correct?

I would love to have been able to use a separate Serial connection for ctrl vs audio. Unfortunately, I can't use another Serial connection (The other one is already in use to control a different chip).

I'm open to suggestions though. This is for an App to send and receive data on androids over USB.

[pschatzmann]

You can use any pin with SoftwareSerial
If I remember right, the meta data is variable length as well: therefore you you need to specifiy the length when writing data
and you can add it anytimes into the stream.

[pschatzmann]

Your sketch from which you pretend it is working, can't work: it will give compile errors.
The second one as well...

The different between GeneratedSoundStream and AnalogAudioStream is the speed with which the audio is generated: the first does it as fast as possible, the second by the defined rate!

The second is tricky when you have multiple systems involved because the clocks are not synchronized: usually filling a big buffer on the receiver side before starting to play will help

[wsciaroni]

That's a verry sharp eye. If you had looked at the link to the full code, you would have seen the full sketch. I simplified the code for you to see the change. If you would look at the full sketch, you would see my AUDIO_USE_SIN_FOR_TESTING that I use to switch between the sin and analog. With only that change, I get broken audio.

[pschatzmann]

see the second part of my answer above

[wsciaroni]

Thanks for your help! I appreciate you taking the time to look at this.

I've already implemented a fairly large buffer on the receiving end, longer than I've used before I switched to AudioTools. I'm starting to suspect it might be related to the ADC resolution. I noticed the audio improves when the signal to the ADC is louder.

I was using ADC-ATTEN_DB_0 before, but I'm not sure how this is being set (I couldn't find it in the docs). I saw how to change it with the ESP32 v3.0 or newer, but I'm using ESP32 v2.0.17. Would you recommend changing to the newer board or sticking with v2? I saw there were large changes in the ADC/DAC driver (via i2s ADC removal from the IDF??). Do you have any suggestions for configuring the ADC attentuation in v2.0.17?

[pschatzmann]

My recommendation would be to use the latest ESP32 Arduino core version.
Analyse the audio from the ADC with the Serial Plotter!