This post is part of a series on designing a wireless microphone system for hybrid online meetings, i.e. with some people present in person and others present online. See also the previous post in this series.
So far I have built and experimented with 4 wifi microphones, including an on/off switch and a rechargeable LiPo battery. I also added a magnetic name tag holder like this to the back of each of the microphones, allowing them to be mounted on a shirt or the the lapel of a jacket. The most relevant parts comprise an INMP441 microphone connected to a Lolin32 lite board. I have a few more wired up with just the Lolin32 board and the microphone to allow testing a larger number.
I have also implemented a Python based server that is running on a Raspberry Pi zero W, which also functions as Wifi access point. The audio server buffers and mixes the incoming signal from the different microphones and plays it on a HifiBerry DAC+ zero audio card. The output is a line level voltage, strong enough to drive a headphone, and with some attenuation also suitable to feed into the microphone input of a low-cost USB headset adapter. The whole system works as expected, although the noise level of the microphones is higher than I had hoped. My guess is that it is in part due to the microphone being so close to the ESP8266 antenna. Also, the wires between the microcontroller and the microphone run over the Lolin32 board without any shielding, probably picking up EM interference.
The Arduino source code, the Python audio server code, and the Fusion360 CAD design files are available from the wifimic repository on Github.
The fact that it works with an USB headset adapter like this, i.e. a miniature external sound card, demonstrates that the device can also be connected to the standard Windows laptop “pink” microphone input.
My MacBook has a TRRS combined audio input/output and the TRS (stereo) cable that comes from the HifiBerry DAC audio card is not recognized as microphone when I plug it in, but over the USB headset adapter it works fine. There are Y-adapters to split the TRRS input into TRS for the headphone and a TS for the microphone that would allow connecting it. However, the Python audio server also works fine on macOS, which has the advantage that I can investigate the microphone audio signals in full quality. Rather than first converting the sound to a analog line-out on the Raspberry Pi, and then back into a digital representation by the USB headset adapter, I can use BlackHole or Soundflower to get the digital audio stream as it is generated by the microphone. A cool feature of BlackHole and Soundflower is that they support many channels. With some modifications to the Python server script, it will also be possible to stream the audio output of each microphone to each own channel, and record them with Audacity.