Join GitHub today
GitHub is home to over 36 million developers working together to host and review code, manage projects, and build software together.Sign up
Version 2.2.0 and 2.2.1 POSHOLD Skyrocket #4925
Enabling POSHOLD while descending causes QUAD to skyrocket with motors at full throttle until NAV mode is disabled.
Steps to Reproduce
QUAD should stop and stabilize at its current position.
Problem is caused by the new acceleration weight that was added to the navigation_pos_estimator in version 2.2.0. As a temporary solution, I disabled the accWeight calculation in the estimationPredict function in a custom build.
Blackbox log i attached.
Link to video:
I just commented out the accWeight as listed below and my 2.2.1 quad is back-up and working without the skyrocket behavior.
static void estimationPredict(estimationContext_t * ctx)
Listed below is the diff from several days ago. Since this dump, I have changed from DSHOT to MULTISHOT and enabled AIRMODE with the same results.
INAV/OMNIBUSF4V3 2.2.1 Jul 3 2019 / 22:13:58 (a6d8474)
GCC-8.2.1 20181213 (release) [gcc-8-branch revision 267074]
mmix 0 1.000 -1.000 1.000 -1.000
serial 0 64 115200 115200 0 115200
aux 0 0 1 1025 1275
#wp 0 invalid
set looptime = 250
set mc_p_pitch = 30
set vbat_max_cell_voltage = 430
@teckel12 this is 10-inch 3d quad with 8-inch props and the correct value is around 1650. I just did a quick non-3d setup to duplicate the problem using the standard 2.2.1 code. I just set it to 1400 to get it close for just one flight to verify the problem is with the standard build. This quad is now back to the 3d setup with a custom build with the accWeight removed and is now flying great using the base 2.2.1 code.
Accelerometer dynamic weigh safeguards significantly reduced accelerometer weight in height calculation, but not enough to entirely prevent the unexpected behavior. For your flights without the issue - you got lucky
PR #4930 will give insight into accelerometer issues via OSD message.
Another action item for us - look into increasing barometer weight to allow barometer-only flight without accelerometer assistance.
Tested out a new 12" build some days ago. Kakute F7 board with 2.2.1. Fitted a TF Mini rangefinder. Had modes set to position (i) Alt Hold + Surface. (ii) Alt Hold + Pos Hold (Surface off).
Can I be confident please that 2.2.1 on Kakute F7 is unrelated to the problem and all Nav modes will work correctly?
Thanks. I did have a black box log that showed throttle went to max for few seconds after pos hold engaged. Unfortunately don’t have that log anymore..... so doesn’t really help matters. I did have to replace 4 in 1 esc after crash. Seemed a bit fragile to me. It is a newish airbot furling32 airport 50A which is very thin. I will give 2.2.1 a go .... over some long grass.…
Sent from my iPhone
On 13 Jul 2019, at 10:38, Konstantin Sharlaimov ***@***.***> wrote: @neilm06 we can't debug the issue without blackbox log. Flipping in the air (aka "roll of death") is generally a hardware issue, but it's impossible to tell without blackbox. Poshold, althold and surface mode are verified to be working in 2.2.1 — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.