I wrote an ADXL345 driver. Its readings are off by a scale of 12%.

[liudr]

I wrote my own ADXL345 driver based on the data sheet and some github code I read. It's pretty basic, with only reading values at a fixed +-16g range and 100Hz without FIFO. It's just for that purpose, reading and printing out accelerometer values. What I noticed was that with my breakout board, the z-axis values span about 1.92g range (ideally 2.00g), 8% off in scale, which was not too bad considering I didn't seriously align the accelerometer and am bound to read a less value range. But x and y values both span about 2.25g (ideally 2.00g) so they are about 12% off in scales, larger than expected. This is a bigger issue since it really shouldn't happen if I don't try very hard to align the accelerometer I should get a smaller range of values, not larger range. This isn't an offset issue since offsets will be cancelled out if I subtract + and - readings on the same axis. I wonder if anyone has ideas what may have gone wrong. It could be the accelerometer itself. I'll acquire a new one in my next batch purchase. Thanks.

from struct import unpack
from time import sleep

_POWER_CTL_REG_ADDR=0x2D
_POWER_CTL_MEAS=0x08
_DATA_FORMAT_REG_ADDR=0x31
_DATA_FORMAT_FULL_RES_16g=0x0B
_DATAX0_REG_ADDR=0x32
_LSB_per_g=256 # LSB per acceleration in g
_LSB_per_mss=26.1 # LSB per acceleration in m/s/s
class ADXL345:
    defaultAddr=0x53 # default 7-bit address (pull SDO/ALT to GND)
    def __init__(self,i2c,addr=defaultAddr):
        self.i2c=i2c
        self.addr=addr
        self.i2c.writeto_mem(self.addr, _POWER_CTL_REG_ADDR, bytearray([_POWER_CTL_MEAS])) # power on and start measurements.
        self.i2c.writeto_mem(self.addr, _DATA_FORMAT_REG_ADDR, bytearray([_DATA_FORMAT_FULL_RES_16g])) # full resolution mode, +-16g range
    def read(self): # Read and return a tuple of (x,y,z) in bits, each LSB=3.906mg or 256LSB/g
        data = self.i2c.readfrom_mem(self.addr,_DATAX0_REG_ADDR, 6)
        (x,y,z)=unpack('<hhh',data)
        return (x, y, z)
    
    def read_mss(self): # Returns acceleration tuple in m/s/s units.
        lsb=self.read()
        return (lsb[0]/_LSB_per_mss,lsb[1]/_LSB_per_mss,lsb[2]/_LSB_per_mss)
    def read_g(self): # Returns acceleration tuple in g.
        lsb=self.read()
        return (lsb[0]/_LSB_per_g,lsb[1]/_LSB_per_g,lsb[2]/_LSB_per_g)

if __name__ == '__main__':
    from machine import I2C, Pin
    i2c=I2C(id=0,sda=Pin(12),scl=Pin(13),freq=100000)
    sleep(0.01)
    adxl=ADXL345(i2c)
    while(True):
        ret=adxl.read_g()
        print('ax=(g)={:.3f}\tay=(g)={:.3f}\taz=(g)={:.3f}'.format(ret[0],ret[1],ret[2]))
        sleep(1)

[liudr]

OK I found the issue. I was using and RPI pico and accidentally connected the VBUS pin to my power bus on the breadboard instead of 3V3OUT. So the accelerometer received a voltage higher than its operating voltage. I tested the accelerometer with a sparkfun pro micro that runs on 3.3V on a different breadboard and its readings were correct. I was using an old library by Jeff Rowberg. So I checked my wiring on pico and discovered the wiring error. Good thing this is just for tinkering.