Very interesting' I placed the file on the same google drive since quite large, really appreciate your assistance here
drive.google.com
, thanks
This is another case where I cannot see any explanation for the yaw error. It's clearly visible in the DAT data at 720 seconds:
While we don't know the exact details of the DJI flight control algorithms, we do know that the yaw is computed by transforming the rate gyro output into the earth frame of reference and biasing with the compass data. The only data that the FC has is from the rate gyros, accelerometers and compass, so if the two IMUs disagree wildly, as they do here, one would expect to find a significant difference in some or all of those input data. I won't post all the graphs, but in this case the rate gyros and accelerometers in the two IMUs agree extremely well with each other, and both take their input from the same compass.
Note the slight disagreement at 720 seconds between IMU0 and IMU1 magnetic yaw, which generally means an attitude disagreement that results in a different interpretation of the measured magnetic field. Sure enough, looking at the pitch and roll data:
Both pitch and roll show some disagreement, after having been in very close agreement for the flight up to that point. But while this accounts for the second (CCW) blip in IMU1 yaw at. 728 seconds, the large IMU1 yaw excursion occurs 5 seconds previously.
724.986 : 40125 [L-FDI]NS(0) FUSION(1): fault on , magn_heading_err_large
If we accept the logged data as accurate, I can only conclude that these errors, that I've seen several examples of recently, are computational errors. Looking at the DAT event stream, while there are no entries that seem directly relevant, I have noticed that each of the events I've looked at has been accompanied by a large number of the following preceding entries:
717.580 : 39753 [L-VISION][VIO]receive vio data befor current ts||
717.662 : 39758 [L-FMU/LED]action changed. Normal Flash:Ctrl forearm LED off by cam(2)
717.662 : 39758 [L-FMU/LED]type:1, normal flash action:
717.662 : 39758 [L-FMU/LED]c0:12,2;c1:0,0;c2:0,0;c3:0,0;c4:0,0;c5:0,0;c6:0,0;c7:0,0
Pure guesswork at this stage, but I wonder if the FC is being overloaded.
@BudWalker?