Expose MPU-6050 FIFO and some helper functions

[CTho9305]

I'm using the MPU-6050 to measure vibration, which requires rapid acceleration sampling. I achieve that by configuring the FIFO to store acceleration samples at 200Hz, accumulate 128 samples (power of 2 for FFT; 256 doesn't fit in the 1024b FIFO since each sample is 6B), then reading them out of the FIFO for processing (calculate magnitude, then FFT or actually ulab.utils.spectrogram). The reading needs to happen fast so that the FIFO doesn't overflow during the readout.

The FIFO itself is highly configurable, so it isn't reasonable for the library to track exactly how the user has configured it.

I also expose helper functions for scaling the samples that are read out of the FIFO, though the user still has to know how to unpack the raw bytes.

My surrounding code, very roughly, is:

mpu = adafruit_mpu6050.MPU6050(i2c)
# We want to sample 200Hz, so divide 8kHz by 40 to get 200Hz
mpu.sample_rate_divisor = 39
mpu.fiforst = True
mpu.accel_fifo_en = 1
mpu.fifo_en = 1 # I'm not sure this is strictly necessary

hz_per_bin = 128/200
while mpu.fifo_count < (6*128):
    time.sleep(0.01)
fifo_bytes = mpu.read_whole_fifo()[:128*6]
data = [None] * 128
for i in range(0, 128):
    # annoying user-side unpack,
    (x, y, z) = unpack_from(">hhh", fifo_bytes, i*6)
    data[i] = mpu.scale_accel([x,y,z])
def root_of_sum_of_squares(vals):
    sum_of_squares = sum(x**2 for x in vals)
    return math.sqrt(sum_of_squares)
mags = list(root_of_sum_of_squares(x) for x in data)
arr = ulab.numpy.array(mags)
spec = ulab.utils.spectrogram(arr)
ispec = ulab.numpy.array(spec, dtype=ulab.numpy.int16)
# the second half is a mirror of the first half
#print(f'{ispec.tolist()[:65]=}')
max_bin = ulab.numpy.argmax(ispec[1:65])+1
max_power = ispec[max_bin]
max_bin_hz = max_bin * hz_per_bin
print(f'%3.1f Hz, %3d RPM, %3d' % (max_bin_hz, max_bin_hz*60, max_power))

[tannewt]

Thanks for the PR! One possible bug.

[tannewt]

Thank you!