Yaw miscalculated [Solved by calibrating]

[Zarbizaure]

Hey, I recently bought a cheap GY-91 sensor, which does wiork well as an acceleromer, gyro and barometer despite its price. However, the yaw value returned by the MPU9250BasicAHRS.ino program is really weird.

First of all, pitch and roll value are correct, and they are not influenced by any change in the other direction. The accuracy is not exeptional (~5degree) but that's enough for what I want to do.

The yaw is apparently (kind of) calibrated with the earth magnetic shield; there is a 180 degree difference between the higer and the lowest value. Between these two angles, the mesured value goes up and down again. It looks like a sinusoïd: see attached file. Yaw.xlsx

Do you have any idea? I though that the conversion from quaternion to euler angle for yaw was wrong, but I have no idea how to get it correct.... Thanks!

[kriswiner]

How did you calibrate the sensors?

-----Original Message----- From: Zarbizaure [mailto:notifications@github.com] Sent: August 18, 2016 1:43 AM To: kriswiner/MPU-9250 Subject: [kriswiner/MPU-9250] Yaw miscalculated (#70)

Hey, I recently bought a cheap GY-91 sensor http://chipskey.en.alibaba.com/product/60439909912-218419143/GY_91_MPU9250_ BMP280_10DOF_9_axis_Accelerometer_Gyroscope_Compass_Sensor_Module.html#! , which does wiork well as an acceleromer, gyro and baromter despite its price. However, the yaw value returned by the MPU9250BasicAHRS.ino https://github.com/kriswiner/MPU-9250/blob/master/MPU9250BasicAHRS.ino program is really weird.

First of all, pitch and roll value are correct, and they are not influenced by any change in the other direction. The accuracy is not exeptional (~5degree) but that's enough for what I want to do.

The yaw is apparently (kind of) calibrated with the earth magnetic shield; there is a 180 degree difference between the higer and the lowest value. Between these two angles, the mesured value goes up and down again. It looks like a sinuso�d: see attached file. Yaw.xlsx https://github.com/kriswiner/MPU-9250/files/424446/Yaw.xlsx

Do you have any idea? I though that the conversion from quaternion to euler angle for yaw was wrong, but I have no idea how to get it correct.... Thanks!

� 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/70 , or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qtjf1ME13IpkcrdB2FfSn zgmTee8ks5qhBr4gaJpZM4JnQyU . https://github.com/notifications/beacon/AGY1qkYv5397GuUqjtJVYQVFGYEDFvB3ks5 qhBr4gaJpZM4JnQyU.gif

[Zarbizaure]

I didn't. I have no idea how to calibrate them, but the gyro/accelerometer results are fine and precise. I doubt calibrating would change anything to this sort of sinusoïdal result, but I could still try. What do you do to calibrate it? I guess I'll have to change these lines, but with which values?

    magbias[0] = +470.;  // User environmental x-axis correction in milliGauss, should be automatically calculated
    magbias[1] = +120.;  // User environmental x-axis correction in milliGauss
    magbias[2] = +125.;  // User environmental x-axis correction in milliGauss

[kriswiner]

Read this:

https://github.com/kriswiner/MPU-6050/wiki/Simple-and-Effective-Magnetometer -Calibration

and then copy the calibration functions from here:

https://github.com/kriswiner/MPU-9250/blob/master/MPU9250_MS5637_AHRS_t3.ino

If you don't calibrate your mag you will never get proper heading.

It makes no sense to use the hard coded calibration value from my particular magnetometer, no wonder you are getting a sinusoid.

-----Original Message----- From: Zarbizaure [mailto:notifications@github.com] Sent: August 18, 2016 3:07 PM To: kriswiner/MPU-9250 Cc: Kris Winer; Comment Subject: Re: [kriswiner/MPU-9250] Yaw miscalculated (#70)

I didn't. I have no idea how to calibrate them, but the gyro/accelerometer results are fine and precise. I doubt calibrating would change anything to this sort of sinuso�dal result, but I could still try. What do you do to calibrate it? I guess I'll have to change these lines, but with which values? magbias[0] = +470.; // User environmental x-axis correction in milliGauss, should be automatically calculated magbias[1] = +120.; // User environmental x-axis correction in milliGauss magbias[2] = +125.; // User environmental x-axis correction in milliGauss

� You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/MPU-9250/issues/70#issuecomment-240872114 , or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qgvd1gewYS8xTuK-JAF3s sPMaIeLks5qhNeYgaJpZM4JnQyU . https://github.com/notifications/beacon/AGY1qrXACdeacSP6WePPuOukaf-lxAhSks5 qhNeYgaJpZM4JnQyU.gif

[Zarbizaure]

Thanks! It seems better now but I still have some issues. First of all, only mag scales value are returned; magBias returns 0.0 for the three value (which I guess is wrong). Plus, the compass is fine for a moment then start to "drift": the value increase really fast and starts to be totally different. I guess my magBias matrix is wrong somewhere, I will check...

[Zarbizaure]

Finaly working! Hurray! With precision close to the degree :-) I used your latest file (with the MS5637 display) and changed some stuff...calibration with a "8" was really wrong, but I make it spin on the z-axis during calibration and it worked perfectly! Thanks for your amazing work! I think I'll make a library from your edited file (in order to simplify my code), if it does work well may I submit a PR? (it will be suited for serial only though).

[kriswiner]

If calibration with figure eight motion was wrong it just means you didn't sample the response surface enough. You can increase the time or increase the range of your motion. Figure eight is the better motion, otherwise you will have yaw error when you introduce pitch or roll.

If you want to modify the code I will consider a pull request but it is unlikely I will accept if there is a large rewrite; I like the code spread out in one file so I can see everything.

-----Original Message----- From: Zarbizaure [mailto:notifications@github.com] Sent: August 18, 2016 5:17 PM To: kriswiner/MPU-9250 Cc: Kris Winer; Comment Subject: Re: [kriswiner/MPU-9250] Yaw miscalculated (#70)

Finaly working! Hurray! With precision close to the degree :-) I used your latest file (with the MS5637 display) and changed some stuff...calibration with a "8" was really wrong, but I make it spin on the z-axis during calibration and it worked perfectly!

Thanks for your amazing work! I think I'll make a library from your edited file (in order to simplify my code), if it does work well may I submit a PR? (it will be suited for serial only though).

You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/MPU-9250/issues/70#issuecomment-240894983 , or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qlRYjWr-HTbPcT5hYgcGQ Y5FfoTuks5qhPX3gaJpZM4JnQyU . https://github.com/notifications/beacon/AGY1quSq_3AWLxOTnn-8glYpeK6TOuBxks5 qhPX3gaJpZM4JnQyU.gif

[Zarbizaure]

I will try with larger movements but my better results are still induced with a mix from "8" withouth z-rotating and z-rotation (because doing both movement at once, considering that my sensor is housed inside an RC plane and that I need to move the whole plane to calibrate, it is not easy).

No, that was just to create new library files for those who want to use it as a library (it easier as part as a larger project). I won't change others files.

[ajaymills]

Hi, I also have an issue with Yaw and I suspect it is due to miscalibrating. Based on the fact that my plot regions have sort of "two elongated lobes" it's evident I'm not moving it in an optimal way. What should the figure 8 motion look like? It seems like an 8 motion will mainly cover two arbitrary axis or do you have to do multiple 8's in different directions? Do you fully invert the device while going through the motion?

[kriswiner]

You need to sample every point on the response surface many times in your motion.