BMI160 Values wrong, offset error?

[ALackOfNumbers]

I am using a BMI160 with a TI MSP430FR6989. I am able to communicate with the BMI160 without problems, reading chip ID, registers, etc. However, when I read data from the X, Y, and Z accelerometer registers I get strange values, even after I run fast offset compensation. My code is available here

https://github.com/ALackOfNumbers/MSP430-BMI160.git 

When I run my code, the output from the main function is 6 bytes. These are the three 2 byte outputs of the three data registers for the accelerometer. When I have the sensor sitting so that the Z axis is parallel to gravity and the fast offset compensation is compensating for the Z axis, I get this output.

11111001 11111111 01010011 11111111 00011111 00000111

This is the X, Y, and Z registers, in LSB, MSB format because I am burst reading through registers 0x12-0x17. Therefore, the decimal values for X, Y, and Z respectively are 65529, 65363, and 1823. With a range of 8g, I divide by 4096 to get values of 16g, 16g, and 0.45g respectively. I do not understand how this could even occur. Offset notwithstanding, the values should never be that wrong. I am expecting to get 0g, 0g, and 1g if I am calibrating around the Z axis. This makes me think that I have initialized the accelerometer wrong. Can anyone tell what might be wrong or things I can try to fix this?

I've tried calibrating around different axes, such as setting the BMI160 so that X is parallel to gravity, then running fast offset compensation around the X axis. This provides the same results, where the axis parallel to gravity always reads 0.45g~0.5g, and the other two axis perpendicular to gravity read 15.5g~16g. 

[BSTRobin]

Hello ALackOfNumbers,

There was BMI160 example code in github, you could refer and run it first. First run the function normally, then optimize it to your platform.

https://github.com/BoschSensortec/BMI160_driver/blob/master/README.md

[zgg · Accepted solution]

Hello

the value you read out from the chip is 16bits in 2's complement.

So 65529, 65363, and 1823 would actually mean:

0xFF29, 0xFF53, 0x71F (16 bits)

and thus respective true values are: -215, -173, 1823

and likely (my guess) you have set the g range to +/- 16G (instead of 8G), to confirm this, register dump of 0x40 would clarify things up.

and in that case (assuming range of 16G), the g values are:

-215/2048 = -0.1g (close to 0)

-173/2048 = -0.08g (close to 0)

1823/2048= 0.89g (which is close to 1g)

like the other user commented, it would be helpful to use the APIs on the github (https://github.com/BoschSensortec/BMI160_driver) for easy translation/convertion of the register values in to g-values, also it would help with the proper set up of the sensors to avoid surprises, it should have some examples on how to use as well.