Calibration for acceleration biases

[mon1020]

Hi Kriss,

We are working on MPU9250. We have referred you code at below link: https://github.com/kriswiner/MPU-9250/blob/master/MPU9250BasicAHRS.ino We use this code calibrate the MPU and then read acceleration values to calculate angle. The values are not coming properly. -1.5205078125,-2,-1.453125,-142.75594569821885 Please advise on the same.

Regards, Monica

[kriswiner]

You will need more than accelerations to get angles.

Anyway, what does this mean?

"The values are not coming properly."

On Tue, Jan 24, 2017 at 1:44 AM, mon1020 notifications@github.com wrote:

Hi Kriss,

We are working on MPU9250. We have referred you code at below link: https://github.com/kriswiner/MPU-9250/blob/master/MPU9250BasicAHRS.ino https://github.com/kriswiner/MPU-9250/blob/master/MPU9250BasicAHRS.ino We use this code calibrate the MPU and then read acceleration values to calculate angle. The values are not coming properly. -1.5205078125 <(520)%20507-8125>,-2,-1.453125,-142.75594569821885 Please advise on the same.

Regards, Monica

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/MPU-9250/issues/111, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qjzfzOfI8IjADPgjLtmY2PpkUn-1ks5rVcgYgaJpZM4LsDQY .

[mon1020]

Hi Kriss,

Thanks for your quick reply.

When we take accel values with default offset registers(not calibrated) the values are as below: -0.04931640625,0.96923828125,0.24755859375 , Angle here is calculated as: angle=atan2(x/y)*(180/pi); For above example: angle= -2.9127896829355713 Through out all the rotation of IMU I get angle between -180 to +180. Which is correct. I want to calibrate IMU for more stable values.

Please correct me if angle calculation is wrong here.

And after calibration, It is giving following values throughout all positions of IMU which give the same angle: -1.520507813,-2,-1.453125 -1.567871094,-2,-1.434082031 -1.566894531,-2,-1.473632813 -1.59375,-2,-1.615722656 -1.568359375,-2,-1.477539063 -1.556640625,-2,-1.330078125 -1.539550781,-2,-1.464355469 -1.6015625,-2,-1.57421875 -1.515136719,-2,-1.39453125 -1.560546875,-2,-1.531738281 -1.594238281,-2,-1.504882813 -1.58984375,-2,-1.546386719 -1.577148438,-2,-1.453125 -1.641113281,-2,-1.483886719 -1.687011719,-2,-1.337890625 -1.538574219,-2,-1.475097656 -1.5859375,-2,-1.513183594 -1.543457031,-2,-1.415039063 -1.627929688,-2,-1.434570313 -1.522949219,-2,-1.408203125 -1.629882813,-2,-1.493164063 -1.612792969,-2,-1.511230469

It is same all around the rotation of IMU. Not sure if calibration is correct. Also, in your code you have mentioned this comment: "Apparently this is not working for the acceleration biases in the MPU-9250"

I am very new to IMU calibration. Your help will be much appreciated.

Thanks, Monica

[kriswiner]

Yes, I wouldmn't use the accelerometer offset bias register to perfrom the calibration, do it in the main loop instead.

On Tue, Jan 24, 2017 at 9:25 AM, mon1020 notifications@github.com wrote:

Hi Kriss,

Thanks for your quick reply.

When we take accel values with default offset registers(not calibrated) the values are as below: -0.04931640625,0.96923828125,0.24755859375 , Angle here is calculated as: angle=atan2(x/y)*(180/pi); For above example: angle= -2.9127896829355713 Through out all the rotation of IMU I get angle between -180 to +180. Which is correct. I want to calibrate IMU for more stable values.

Please correct me if angle calculation is wrong here.

And after calibration, It is giving following values throughout all positions of IMU which give the same angle: -1.520507813,-2,-1.453125 -1.567871094,-2,-1.434082031 -1.566894531,-2,-1.473632813 -1.59375,-2,-1.615722656 -1.568359375,-2,-1.477539063 -1.556640625,-2,-1.330078125 -1.539550781,-2,-1.464355469 -1.6015625,-2,-1.57421875 -1.515136719,-2,-1.39453125 -1.560546875,-2,-1.531738281 -1.594238281,-2,-1.504882813 -1.58984375,-2,-1.546386719 -1.577148438,-2,-1.453125 -1.641113281,-2,-1.483886719 -1.687011719,-2,-1.337890625 -1.538574219,-2,-1.475097656 -1.5859375,-2,-1.513183594 -1.543457031,-2,-1.415039063 -1.627929688,-2,-1.434570313 -1.522949219,-2,-1.408203125 -1.629882813,-2,-1.493164063 -1.612792969,-2,-1.511230469

It is same all around the rotation of IMU. Not sure if calibration is correct. Also, in your code you have mentioned this comment: "Apparently this is not working for the acceleration biases in the MPU-9250"

I am very new to IMU calibration. Your help will be much appreciated.

Thanks, Monica

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/MPU-9250/issues/111#issuecomment-274874271, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qpRO3mFT9fr8nixyoLOelncmgy20ks5rVjPwgaJpZM4LsDQY .

[mon1020]

Hi Kriss,

I am not sure I understand what you are saying here. How to use gyro values for angle calculation? How can I fix the issue of incorrect acceleration values?

Regards, Monica

[TediumRemedy]

In theory the accelerometer gives you the acceleration (d2x/dt2, d2y/dt2, d2z/dt2, not the actual speed dx/dt, dy/dt, dz/dt, and not the actual position x,y,z). So you cannot use just the accelerometer to calculate the angle. With gyro the story is similar: it returns you the speed of change of the angle (d(alpha)/dt, d(beta)/dt, d(theta)/dt), not the actual angle (alpha, beta, theta). [where alpha beta and theta - are the angles of rotation around axes x y and z].

Also the raw returned values are "noisy", you need to perform some kind of filtering upon them before integrating.

So To get the actual angle, you must perform an integration of the gyroscopic measurements by time (i.e., summing of all the values of measured gyroscopic angle speeds (say, d(alpha)/dt) times delta t (this is a small period of time that equals to the time between two nearby measurements, upon which you can consider the angle to be constant)). See http://stackoverflow.com/questions/31278784/integrate-readings-from-a-3-axis-gyroscope

To my understanding, you can additionally use the accelerometer and the magnetometer to improve precision of the initial gyro integration.

By the way, MPU9250 has an internal cpu (called DMP) that can do this integration for you, but in that case you do not use raw measurements from gyro/accelerometer and configure MPU to return this postprocessed data (please correct me if I am wrong).

[kriswiner]

If you want yaw, pitch, and roll why not just use the open-source sensor fusion algorithms.

For accel biases, do not use the hardware bias registers,just subtract the biases in the main loop.

There are several examples of how to do both in the github repository.

What else is missing?

On Thu, Jan 26, 2017 at 5:24 AM, TediumRemedy notifications@github.com wrote:

In theory the accelerometer gives you the acceleration (d2x/dt2, d2y/dt2, d2z/dt2, not the actual speed dx/dt, dy/dt, dz/dt, and not the actual position x,y,z). So you cannot use just the accelerometer to calculate the angle. With gyro the story is similar: it returns you the speed of change of the angle (d(alpha)/dt, d(beta)/dt, d(theta)/dt), not the actual angle (alpha, beta, theta). [where alpha beta and theta - are the angles of rotation around axes x y and z]. So To get the actual angle, you must perform an integration of the gyroscopic measurements by time (i.e., summing the values of measured gyroscopic angle speed (say, d(alpha)/dt) times delta t (this is a small period of time upon which you can consider the angle to be constant)). See http://stackoverflow.com/ questions/31278784/integrate-readings-from-a-3-axis-gyroscope

To my understanding, you additionally use the accelerometer and the magnetometer to improve precision of the initial gyro integration

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/MPU-9250/issues/111#issuecomment-275387152, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qvLjxWbrMIrNIqcQwBtNOzORepFjks5rWJ6DgaJpZM4LsDQY .

[TediumRemedy]

For accel biases, do not use the hardware bias registers,just subtract the biases in the main loop.

@kriswiner could you tell why not? (Sorry if you already answered this question)