Power consumption in normal usage (Wi-Fi off, running audio DSP code with mics active) = 100 mA, with Audio Codec in standby mode = 52.4 mA, low power usage (Audio Codec in standby mode and ESP32 ...

IMHO the chosen audio frequencies are less than ideal. It would be interesting how this 96-level FSK compares to V.21, DTMF and Eurosignal, especially on distorted or noisy channels. Report comment ...

For the Arduino demonstration, ... If the sound hits both microphones at once, ... and replacing the headphone speakers with omni electret capsules.