09/04/2024 1:45 pm
Topic starter
Notifications
Clear all
0
I posted a question earlier asking about how auto-config broke sounds into frequency bands across the 16 input channels. To try and get at the answer myself, I played pure tones to see how they binned. I also tried just counting to ten to collect human voice. But in all cases, the raster array is all zeros.
Based on the tutorials, this is how I am attempting to record audio from the device.
afe_hdks = xdu.find_xylo_a2_boards() afe = x.AFESamna(afe_hdks[0], dt=10e-3, auto_calibrate=True) audio_events = afe(np.zeros([0, 1000, 0]))
The resulting array, audio_events, has shape 1000 x 16.
If I powered off the device and tried again. This time, I could get some 1s to appear at column index 2 when playing a pure tone of 880Hz. But this wasn't the case for all samples/rows. Sometimes it will skip one row, sometimes two rows.
If there is a pure tone playing continuously at the same volume, why would so many rows be encoded as all zeros?
And do I need to power cycle the device every debug cycle? Because I get all zeros if i don't.
Thanks
