LSM6DS3 + LIS3MDL calibration

[pierre63]

Hello, 

Our teams is making a proof of concept about dead reckoning using a STM WESU board. 

The goal is to get pedestrian position onyl using ACC/GYR/MAG for a short period of time and small moves.

First results in certains conditions are pretty good. Without any calibration/correction, we manage to get approx 3m absolute error on a 20m walk (30seconds). We expect same error on a 100m walk with good AHRS algorithm parameters and sensor calibration.

We are ready to put more effort on the IMU. 

Do you have calibration source code/protocol for WESU sensor ? 

* LSM6DS3 = acc value + gyro drift

* LIS3MDL = soft and hard iron calibration

We also have other more advanced IMU, but we want to keep it simple and low cost for now and WESU allow us to be pretty fast on POC steps.

Thanks a lot for your future help.

Best regards.

[Federica Bossi]

Hi @pierre63 ,

The WeSU board is not a recent board and therefore integrates the sensors that were available when it was designed, being an integrated solution obviously there is no way to update the sensors with more current ones.

The fw package integrates a calibration routine for the magnetometer only, built into the MotionFX package. In the firmware, the reference is the MotionFX_manager_MagCal_run function.

[pierre63]

Hello Federica, 

Thanks a lot for your reply.

Yes, this demo board is a bit old, but pretty well integrated and pretty good to make demo.

I checked this function, but there is no the source code available.

Also, for gyroscope or accelerometer, do I only need to apply offset or there is another correction to apply ?

Thanks,

[Federica Bossi]

Hi @pierre63 ,

For gyroscope and accelerometer we don't have any example code but yes, you can use the offset when the sensor is in a steady state position.

[pierre63]

Hello Federica, 

Before closing the topic, I have few questions :

* Since our last discussion, we want now to focus on activity detection (walking, carrying, steady, driving) instead of raw data measurement. Could you please recommand which eval board can I use for the project ?

* We need to finish the POC and I need to add memory on the board, but I can identify test point on the PCB (Gerber file don't mention it). Could you send me the Test point location if possible ?

Thanks a lot,

Regards,

[Andrea VITALI]

MotionFX, the sensor fusion library, does include a robust version MotionMC, which is the magnetometer calibration library, but which only performs sphere fitting hence computing hard-iron to be subtracted. MotionFX also includes gyroscope bias estimation. For reference see UM2220 for MotionFX.

MotionFX is fed with raw accel and gyro data, and compensated mag data (compensated by subtracting the hard-iron coming from the embedded magnetometer calibration library).

Alternatively, one can include and run the full blown MotionMC (X-Cube-MEMS1 package) which performs ellipsoid fitting and therefore also computes the soft-iron matrix for the magnetometer compensation. This WESU software package can indeed be modified and customized by including MotionMC.

If this is done then running the mag calib embedded in MotionFX is no more necessary. MotionFX will need to be fed with compensated mag data: first subtract hard-iron, then apply the soft-iron matrix coefficients.

(for more details on mag calibration one can also check design tip DT0059)

[Andrea VITALI · Accepted solution]

Forgot to mention: the latest and best evaluation platform today is SensorTile.Box PRO for the consumer market, or the STWIN.Box for industrial applications.

[pierre63]

Hello Andrea,

Thanks a lot for this informations. I checked MotionMC and that exactly what I'm looking for about magnetometer correction. 

Also, I will buy a SensorTile.Box PRO for datastream and logging instead of customize the WESU. 

I'm closing the toping and if needed I will open another one.

Regards,