BMI270 gyroscope x axis zero-rate offset out of spec

[sebmadgwick]

The BMI270 datasheets states that the gyroscope zero-rate offset is ±0.5 dps.  I have tested three parts (all marked "5P 057") and found the x axis to be around -3 dps on all three parts.

I am using the API to read sensor data. The attached plot shows the gyroscope zero-rate output for 10 seconds with an ODR of 400 Hz and range of 2000 dps. Tests confirm that the zero-rate offset in dps is the same for all ranges. All parts pass the self-test as implemented by the API.

Has anyone else observed this?  What is the explanation...

• User error?
• Bad batch / damaged parts?
• API bug?
• Data sheet error?

[sebmadgwick]

I suspect the parts are in spec and that issue is in the datasheet.

• The BMI160 datasheet states "Zero-rate offset" is ±3 dps for "TA=25°C, fast offset compensation".
• The BMI270 datasheet states "Zero-rate offset" is ±0.5 dps for "TA=25°C, nominal VDD soldered, over life time"

I suspect the same parameter means very different things in each datasheet; That for the BMI160 it means the absolute zero-rate offset, and for the BMI270 it means the relative change in zero-rate offset over life time.

Can anyone one from Bosch comment on this?

[o_o]

Hello,

The datasheet typical values are specified for "1 sigma" deviation from from the mean value. Which means it is expected that 68% of parts fall inside of that range. When provided, min/max values are "3 sigma" which corresponds to 99.7% of parts.

There is no significant difference in the way BMI160 and BMI263 are spec'd.

Since all parts are factory calibrated to 0dps, I find it odd that all the parts that you tested have a deviation in the negative x axis of similar amplitude. Which PCB are you using?
Did you follow the HSMI guide for IMUs?


I would suspect that the offset is caused by an external factor, such as PCB bending.

[sebmadgwick]

The parts were soldered using a hot air rework station and low melting point solder. Care was taken to minimise thermal stresses. However, this assembly process cannot guarantee the recommended thermal profile.

I will update this thread once we have tested boards soldered in a reflow oven with controlled thermal profile.

[o_o · Accepted solution]

Hi sebmadgwick,

Thanks for reporting this issue. After investigation, we found that the part was in spec with the datasheet, but found an issue in our driver with regards to offset compensation. Please update to the latest Sensor API version.

Since in most applications the gyroscope offset needs to be compensated before use, we think that the real-world performance to be unchanged for most users.