getting I2S realtime volume data from INMP411 microphone

[Rumidom]

Hi I'm trying to get realtime data from a INMP411 I2S microphone, to make it trigger the start of a recording after a certain volume level I was able to run the examples from this repository: https://github.com/miketeachman/micropython-i2s-examples , however I was unable to convert the data I got from I2S.readinto(buf) into a usable format, more specifically I'm trying to convert the bytebuffer into a list of ints that I can use to measure volume level. Just to try to make same sense of the data I tried printing the buffer on the console and converting it to a numpy array on my computer and plot it with matplotlib:

dt = np.dtype("uint16")
dt = dt.newbyteorder('>')
plt.plot(np.frombuffer(bytearray_sample, dtype=dt))

I got this from a sample with music playing near by:

and this from recording silence:

not sure why it ramps like that, am I using the wrong data type?

[Rumidom]

I changed the byte order from python dt.newbyteorder('>') to python dt.newbyteorder('<') and python np.dtype("uint16") to python np.dtype("int16") and got something that makes more sense, but its still unusable:

music:

silence:

[ricksorensen]

I have similar results from a INMP441- with the little dip at beginning and the low signal noise. Using both RP2040 and ESP32C3.

I used a fixed (purish) tone signal out a small speaker 2" from microphone- and if I cut the first chunk out and amplify the remainder I can clearly see the wave form, and the spectral analysis matches the expected frequency. But the level is very very low as in your example.

[Rumidom]

level is low because I'm using my laptop speakers, it seems like this dip is some sort of initialization of the mic, if I call audio_in.readinto() a few times before using the buffer it goes away, thanks!

[ricksorensen]

try signed int16 ('h'?)