Incompatibility between WM8960 & Hailo RPi Accelerator

[dlw18]

I have the VoiceBonnet from Adafruit and the Raspberry Pi AI HAT+
. Both are connected physically, and I can get each to work independently on my RPi 5/16GB.

For the Hailo accelerator, I get it to work with the following:

sudo apt install -y hailo-all meson ninja-build
sudo reboot

Verify installation with:

hailortcli fw-control identify

Now verify it works

rpicam-hello -t 0 -v 2 -n --post-process-file /usr/share/rpi-camera-assets/hailo_yolov6_inference.json

Now create python examples

mkdir hailo
cd hailo
git clone https://github.com/hailo-ai/hailo-rpi5-examples
cd hailo-rpi5-examples
source setup_env.sh
./install.sh

Verify that one of the demos runs:

python basic_pipelines/depth.py

For the VoiceBonnet, I get everything set up with:

sudo apt install git -y
sudo apt install -y --upgrade python3-setuptools
sudo apt install -y python3-dev
pip install --upgrade adafruit-python-shell
pip install adafruit-circuitpython-dotstar
git clone https://github.com/HinTak/seeed-voicecard
cd seeed-voicecard
git checkout v6.12
sudo ./install.sh
sudo reboot

Verify it works (speaker test, mic test, dotstars, etc.):

speaker-test -c2

But the Hailo accelerator depth.py now throws segmentation faults, with very little else to go on. It does not matter the sequence of installation - I've started with a clean RPi image, and installed the accelerator first (which works) then the VoiceCard (accelerator stops working), and the opposite ... in which case the accelerator never starts working.

And to be clear, it seems that it's specifically the python bindings on the accelerator that break with the VoiceCard install - rpicam-hello still works, but just not any of the python tests (like depth.py).

If it's relevant:

(.venv) pi@raspberry:~/seeed-voicecard $ uname -r
6.12.25+rpt-rpi-2712

Thank you.