kriswiner
commented
6 years ago The idea is to calibrate the sensor well once then put the offset biases in the code so you don;t have to calibrate every time.
So the "hard-coded" numbers are for one particular MPU9250 that I calibrated. These numbers should not be used for any other MPU9250.
Calibrate the magnetometer once empirically and use these numbers in the code. But remember, the offset biases will depend on location and temperature, so it is always best to calibrate again before use.
On Fri, Feb 16, 2018 at 4:41 PM, Yang Liu notifications@github.com wrote:
Hi Kris, thanks for your contribution, have learnt a lot from it. I've read many issues, and hope this is a new one.
You have tried two ways to set the biases for the magnetometer. One is hard coded values in milliGauss in "MPU9250BasicAHRS_t3.ino". https://github.com/kriswiner/MPU9250/blob/1594ddd5735e65f35f9d5a5414a067 a8d2757a9c/MPU9250BasicAHRS_t3.ino#L414-L422 The other is from experiment that the user needs to rotate the component to all directions to sample the maximum and minimum on all axes, as in all "*_t3.ino" with the following function. https://github.com/kriswiner/MPU9250/blob/1594ddd5735e65f35f9d5a5414a067 a8d2757a9c/MPU9250_MS5637_AHRS_t3.ino#L1064
I am wondering which is better? You said "MPU9250_MS5637_AHRS_t3.ino" is the most modern and correct version, which uses the experiment one. But after a few tried, I found the biases are not consistent, and often varies a lot. I guess it is due to not enough sampling. Personally I found this method kind of tedious, because I need to calibrate it every time, or when move to a new location. And rotating the component definitely needs some skills to cover the extreme values on all axes.
On the other side, the hard coded values are much easier to implement, and the result looks reasonable too! My only concern for this one is, where are these numbers com from? I didn't seem to find them in the datasheet, or are they simply empiric numbers?
Thank you for your time. Your opinion is highly appreciated.
— 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/235, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qk7W9Hq52r8KY0PU-xsiltK7aT_Tks5tViBQgaJpZM4SJG9b .
yangliu28
Hi Kris, thanks for your contribution, have learnt a lot from it. I've read many issues, and hope this is a new one.
You have tried two ways to set the biases for the magnetometer. One is hard coded values in milliGauss in "MPU9250BasicAHRS_t3.ino". https://github.com/kriswiner/MPU9250/blob/1594ddd5735e65f35f9d5a5414a067a8d2757a9c/MPU9250BasicAHRS_t3.ino#L414-L422 The other is from experiment that the user needs to rotate the component to all directions to sample the maximum and minimum on all axes, as in all "*_t3.ino" with the following function. https://github.com/kriswiner/MPU9250/blob/1594ddd5735e65f35f9d5a5414a067a8d2757a9c/MPU9250_MS5637_AHRS_t3.ino#L1064
I am wondering which is better? You said "MPU9250_MS5637_AHRS_t3.ino" is the most modern and correct version, which uses the experiment one. But after a few tried, I found the biases are not consistent, and often varies a lot. I guess it is due to not enough sampling. Personally I found this method kind of tedious, because I need to calibrate it every time, or when move to a new location. And rotating the component definitely needs some skills to cover the extreme values on all axes.
On the other side, the hard coded values are much easier to implement, and the result looks reasonable too! My only concern for this one is, where are these numbers com from? I didn't seem to find them in the datasheet, or are they simply empiric numbers?
Thank you for your time. Your opinion is highly appreciated.