search

kriswiner / MPU9250

Arduino sketches for MPU9250 9DoF with AHRS sensor fusion
1.03k stars 472 forks source link

MPU9250 extreme strange behaviours #210

Closed fschramka closed 6 years ago

fschramka commented 6 years ago

Hi Kris

I build my own uP board with bluetooth (rn42), mpu (9250), a atmega32u4 @ 8MHz and a battery charging chip. My goal is to have a sensor for rotations for a research project. Here a small movie of what we have:

https://www.dropbox.com/s/dnt5pyxb6nzgmva/Bike%20Simulator.mp4?dl=0

The sensors we have so far are built from a Adafruit 32u4 Proto with a MPU6050 and a EZ-link for bluetooth. For the steering we still use a HTC Vive controller and therefore the whole HTC Lighthouse system is needed. The new sensor should replace it to overcome this issue.

Now my problem. When I use your lib for the MPU9250 I dont get any nice rotations. I receive data and if I rotate the board I can see that something happens (I wrote a small Unity visualisation). But it's far away from a nice rotation and it has a huge drift. This happens if I dont move the sensor:

https://www.dropbox.com/s/1p0nfurofmsrun5/Unity%202017-11-24%2018-50-51-84.avi?dl=0

This when I rotate the sensor around its Y axis:

https://www.dropbox.com/s/9g1sxipmhgil5zu/Unity%202017-11-24%2018-53-20-16.avi?dl=0

I dont need any EulerAngles / roll-pitch-yaw angles, so I don't go for the last conversion in your code (from if (!AHRS) on). I just update the Time, and go for the MadgwickQuaternionUpdate. Then I send it over bluetooth to the computer (x,z,y,w). xzy because of the directX coordinate system.

I also use your function magcalMPU9250(MPU9250magBias) to get the magBias which I can later use to correct the error. I rly dont know what i'm doing wrong...

Hope you can help me :)

P.s. If you have time to check the code 👍 MPU9250Board.txt

kriswiner commented 6 years ago

Maybe many things, but you are not feeding the Madgwick filter the sensor data in the correct order. It expects NED order.

On Fri, Nov 24, 2017 at 10:03 AM, FSchramka notifications@github.com wrote:

Hi Kris

I build my own uP board with bluetooth (rn42), mpu (9250), a atmega32u4 @ 8MHz and a battery charging chip. My goal is to have a sensor for rotations for a research project. Here a small movie of what we have:

https://www.dropbox.com/s/dnt5pyxb6nzgmva/Bike%20Simulator.mp4?dl=0

The sensors we have so far are built from a Adafruit 32u4 Proto with a MPU6050 and a EZ-link for bluetooth. For the steering we still use a HTC Vive controller and therefore the whole HTC Lighthouse system is needed. The new sensor should replace it to overcome this issue.

Now my problem. When I use your lib for the MPU9250 I dont get any nice rotations. I receive data and if I rotate the board I can see that something happens (I wrote a small Unity visualisation). But it's far away from a nice rotation and it has a huge drift. This happens if I dont move the sensor:

https://www.dropbox.com/s/1p0nfurofmsrun5/Unity%202017- 11-24%2018-50-51-84.avi?dl=0

This when I rotate the sensor around its Y axis:

https://www.dropbox.com/s/9g1sxipmhgil5zu/Unity%202017- 11-24%2018-53-20-16.avi?dl=0

I dont need any EulerAngles / roll-pitch-yaw angles, so I don't go for the last conversion in your code (from if (!AHRS) on). I just update the Time, and go for the MadgwickQuaternionUpdate. Then I send it over bluetooth to the computer (x,z,y,w). xzy because of the directX coordinate system.

I also use your function magcalMPU9250(MPU9250magBias) to get the magBias which I can later use to correct the error. I rly dont know what i'm doing wrong...

Hope you can help me :)

P.s. If you have time to check the code 👍 MPU9250Board.txt https://github.com/kriswiner/MPU9250/files/1502320/MPU9250Board.txt

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/MPU9250/issues/210, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qtoIxpUHNdFTdwO-IThW-00d37HGks5s5wTbgaJpZM4QqFlc .

fschramka commented 6 years ago

Well okay, according to page 38 of the datasheet the Acc & Gyro are ENU, the Magnetometer is, like you said, NED.

So I call the filter (a.x, a.y, a.z, g.x, g.y, g.z, M.Y, M.X, -M.Z).

Or do you mean the filter itself needs NED for all inputs? So it would be:

filter(a.y, a.x, -a.z, g.y, g.x, -g.z, m.x, m.y, m.z) ?

I guess I still fail somewhere, both ways still show strange results...

https://www.invensense.com/wp-content/uploads/2015/02/PS-MPU-9250A-01-v1.1.pdf

kriswiner commented 6 years ago

Either of these should work, so maybe the sensors aren;t fully calibrated? Maybe there is an error in the hardware design?

On Mon, Nov 27, 2017 at 1:34 AM, FSchramka notifications@github.com wrote:

Well okay, according to page 38 of the datasheet the Acc & Gyro are ENU, the Magnetometer is, like you said, NED.

So I call the filter (a.x, a.y, a.z, g.x, g.y, g.z, M.Y, M.X, -M.Z).

Or do you mean the filter itself needs NED for all inputs? So it would be:

filter(a.y, a.x, -a.z, g.y, g.x, -g.z, m.x, m.y, m.z) ?

I guess I still fail somewhere, both ways still show strange results...

https://www.invensense.com/wp-content/uploads/2015/02/PS- MPU-9250A-01-v1.1.pdf

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/MPU9250/issues/210#issuecomment-347126681, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qj2BcyiOK1sdUVR9HoeBQYVgqRSBks5s6oJAgaJpZM4QqFlc .

fschramka commented 6 years ago

After long searching I found a solution for my problem, but I had to use another library:

https://github.com/jrowberg/i2cdevlib

I used the code of the MPU9150, the HW registers are the same, hence it worked. The chip is in FIFO mode and I get directly a Quaternion from the MPU9250. It is extremely stable, I guess there is some filter algorithm implemented in hardware.

Thanks for your answers 👍