MPU-9150 DMP quaternions

Option 1:
You need to use the Embedded Motion Processing Library (currently only really available in the MotionFit SDK without source code for the MSP430F5528 microcontroller) to load the DMP, do the magnetometer calibration, and perform fusion of 6-axis quaternion with magnetometer data to get 9-axis (i.e. heading corrected w.r.t North) quaternion.

Option 2:
Use a hack like: http://www.i2cdevlib.com/devices/mpu6050#source. This loads the DMP but does not do the heading correction. This is based on MotionFit SDK 4.1. You can recreate it for SDK 5.1 by listening in on the I2C writes (using a digital analyzer, e.g. Salae Logic...see examples at link above), and recreating them all in your code.

Option 3:
Wait for Embedded Motion Driver. Supposedly InvenSense is releasing a platform independent release "soon" that will allow you to load the DMP from any platform and will give guidelines for integrating the magnetometer data

Option 4:
Wait until InvenSense actually releases a DMP version in which magnetometer data is fused with 6-axis quaternion data within the DMP itself. As far as I know, InvenSense has not given a timeline for when this would be available for either the MPU-9150 or the brand new MPU-9250

Any update regarding this matter?
Have they released the 9-axis DMP version?

Regards,

Víctor.

A 9-axis DMP has not been released and there is no update at this time.

For anyone interested, we have posted some easy to use demonstration code that performs the 9-axis data fusion (DMP for accelerometer and gyro plus software for the final magnetometer integration) and provides simple calibration tools. There are two versions, https://github.com/Pansenti/MPU9150Lib for Arduinos and https://github.com/Pansenti/linux-mpu9150 for Linux-based systems (tested mainly on the Raspberry Pi).

Hi richardbarnett,
I used calibration tools of arduino version. get accl the following data ,
I don't understand this, the data is raw? maybe is wrong? Thanks!
x: 23275.08 y: -4357.08 z: 18092.53
x: 25182.03 y: -1591.71 z: 21937.88
x: 28454.59 y: -528.16 z: 17491.31
x: 28085.01 y: -2337.66 z: 21387.71
x: 26442.65 y: -1487.28 z: 18240.13
x: -32634.82 y: -1593.54 z: 20986.55
x: -27269.43 y: -15951.43 z: -1091.63
x: -30100.82 y: 424.63 z: 22832.57
x: 18138.38 y: 19039.37 z: 22332.95
x: 29329.74 y: 2126.41 z: 22282.59
x: -29018.79 y: -1170.79 z: 19884.19

Yes, that's definitely wrong! Must be some sort of wiring issue...

Hi richardbarnett,
Hardware circuit problem?But I can read the original data using other programs ?

Please be aware that Embedded MotionDriver has been released on the Developer Corner download page, and includes a supported version of DMP sensor fusion. It is released in C source, so it can be ported between platforms.

Does anyone know the difference between the hex provided with Motion Driver 6.12, and that provided with I2Cdevlib (MotionFit SDK 4.1). In my application using a MPU-9250 and a small µC, I have the RAM to use the older code, but not nearly enough to use the current stuff. Does that old hex even run right on a MPU-9250?