Small four-axis 9-axis data Fusion success

Two days before the second generation of hardware finally to QI, roommate Fengfenghuohuo welding out two new hardware, did not think that the magnetic field generated on the circuit board than the magnetic field is also strong, resulting in the magnetic meter data changes are not obvious, we think it is welding problems (after all, datasheet said not recommended manual welding and temperature can not exceed 260 degrees), Is the pit of a roommate a hehe ~, in fact, a correction on the good. The problem of the magnetometer resolved, decisively replaced the early written nine-axis fusion code, and duly completed. November 22, a memorable day, is the completion of the previous plan to settle the attitude of the December before the plan. A picture of it ~

Copyright Notice
Original article, reprint please indicate source: sheng-blog.cn
Original source

Spit Trough the second generation buy new blades, hard and ugly and inefficient, a generation of blade motor to 40% duty-free to fly up, new blades to be added to 70% ~ and vibration than the old oar higher than n times, resulting in the attitude of the result has a 1 degree of jitter. The most important thing is that the card can't be pulled down on the motor ....
Illustrated with the nine-axis fusion algorithm:

Gradient Descent method versions are also available:

Add: Later found happy too early, when the motor is turned on, due to the magnetic field generated by the motor, the magnetometer can not distinguish the direction, because the motor magnetic field is changed, so can not be corrected to eliminate interference. In the subsequent third edition of the hardware has tried to use a separate small board to place the magnetometer, and the plane from the motor surface (because the motor plane is the most affected, usually using a rack-off treatment, such as the DJI series will hide the magnetometer in the tripod), but in the acceptable height range, still can not get valid data. So on the four-axis PCB should be unable to use the magnetometer, and currently do not see the market has used a magnetometer on the small four-axis PCB.
Another: The small four-axis program is mainly two parts, one is the attitude of the calculation, one is PID control. Because the PID control has nothing to summarize, the invariable formula set is, mainly is the parameter adjustment compares torture. So directly here to add the PID diagram is as follows:
Single-loop PID control, the parameter is good, but the effect is not stable:

Cascade PID, the inner ring is difficult to adjust, the effect is great: