While flying today I noticed a horrible horizon change in the matter of seconds. Anyone know of a fix?
Horrible horizon change in seconds
- Thread starter Birds-iView
- Start date
Just the usual IMU and Gimbal calibrations. Ah, but you know, it can be straighten out from the app, while in flight too. Before bringing it back home.
RedHotPoker
RedHotPoker
I have personally experienced this issue since day one of my P3A ownership and could not remedy it by any means practical. I tried IMU and gimbal calibrations as well as software updates but nothing corrected it. I ended up sending my bird to DJI for service. This problem is being experienced by others and some have overcome it through calibrations and others are unable to correct it. DJI received my unit the other day but I have not heard any diagnosis or estimate for the repair. I assume it is a warranty repair since they sent me a prepaid shipping label and the unit is only 2 months old. For what it's worth, my unit has the updated motors and body.
Oso
Premium Pilot
I also have had a similar issue for the life of my P3 with no fix found. Sometimes it's not as bad as your video, but sometimes it is. It also drifts from level to slanted to level to slanted the other way as I yaw. @hunch speculated that it may be a rare IMU issue.
New Camera / Gimbal - How to match FW | DJI Phantom Forum
New Camera / Gimbal - How to match FW | DJI Phantom Forum
Well, I just received my P3A back from DJI and Guess what? It morphed into a P3 Pro. I guess they had to replace the camera and must have run out of Advanced cameras. Merry Christmas to me. I have not flown yet but let's hope the horizon issue is solved.
This is normal behavior for a small cheap IMU if you're flying in a long curve.
IMU:
It has three gyros that requires constant adjustment against the accelerometers and heading sensor since they drift quite fast, 1° /sec maybe?
When the quad is making a turn they will stop updating the gyros with data from the accelerometers and rely on gyros alone, which means that things will start to drift.
During a long turn they have two choices, either continue to just use the gyros and let them drift or start using the accelerometers again, but when you fly in a curve the centrifugal acceleration will make the accelerometers point in the wrong direction which means that the gyros will be adjusted the wrong way.
The result of this will be a horizon that isn't level!
They use a Kalman filter to fuse data together from gyros, accelerometers compass but you can't solve this problem without good gyros, and they cost a lot!
A side effect of this is also that the compass will show the wrong heading since the IMU has the wrong idea of the horizon.
You cannot calibrate away this drift, but you can make it smaller with an IMU calibration at different temperatures.
They map how the gyros drift at different temperatures at steady state and then use this info to cancel out that error during flight to make the drift smaller.
You can minimize this problem by changing the way you fly the quad and have a good calibration.
Cheers
Hakan
IMU:
It has three gyros that requires constant adjustment against the accelerometers and heading sensor since they drift quite fast, 1° /sec maybe?
When the quad is making a turn they will stop updating the gyros with data from the accelerometers and rely on gyros alone, which means that things will start to drift.
During a long turn they have two choices, either continue to just use the gyros and let them drift or start using the accelerometers again, but when you fly in a curve the centrifugal acceleration will make the accelerometers point in the wrong direction which means that the gyros will be adjusted the wrong way.
The result of this will be a horizon that isn't level!
They use a Kalman filter to fuse data together from gyros, accelerometers compass but you can't solve this problem without good gyros, and they cost a lot!
A side effect of this is also that the compass will show the wrong heading since the IMU has the wrong idea of the horizon.
You cannot calibrate away this drift, but you can make it smaller with an IMU calibration at different temperatures.
They map how the gyros drift at different temperatures at steady state and then use this info to cancel out that error during flight to make the drift smaller.
You can minimize this problem by changing the way you fly the quad and have a good calibration.
Cheers
Hakan
- Joined
- Oct 18, 2015
- Messages
- 1,617
- Reaction score
- 747
- Age
- 70
Thanks for that. Can you expound on 'calibrating the IMU at different temperatures'? Do you mean to do subsequent cold, then perhaps warmer serial calibrations of the IMU? Or are you talking about an IMU that is more advanced than the Phantoms.
Very helpful analysis but also a bit disconcerting. Does that mean I should not expect any real improvement even though the camera module has been replaced? This issue does not appear to afflict everyone. I also own a P2 with the Zenmuse H3-3D gimbal and it remains level quite well. Does this mean they used better gyros on that gimbal? I guess I'll just have to hope I see some improvement when I get to fly. Installing the P3P 1.05.003 firmware now.
ianwood
Taco Wrangler
This is mostly true but the horizon drift can be corrected. Because the gyroscope drifts, it is not used for absolute positioning over a long duration. It measures the rate of angular change over a short duration. Sensor fusion (Kalman, Madgwick AHRS, etc.) is used to incorporate absolute reference data from the compass which provides the gravitational vector. This vector is the critical component.
If the compass calibration is off, the gimbal will exhibit this sort of exaggerated horizon drift. If the calibration is really bad, the drone will also exhibit toilet bowl effect as result of conflict between GPS and compass. My suggestion to the OP is to read up on getting a good compass calibration in the link below and then try again:
Compass Calibration, A Complete Primer | DJI Phantom Forum
Good point with the compass but I'm not sure how they use it in the math.
You only have one magnetic vector so you can't solve the full position of the quad but it could help more to use the vector info together with knowledge on what latitude you are at
than just using the heading. If you know the latitude you know what angle the magnetic vector has.
I don't know if they have an IMU in the gimbal also or if they use the quad IMU, so I'm not sure if the change of camera module would make any difference.
What they try to achieve with the IMU calibration is to get a table with temperatures range from cold to warm and the drift that each gyro (x,y,z) have in steady state at that temperature, and this table should have a temperature range that covers the temperatures it experiences during flight. When the quad is flying they check the IMU temp and looks in the table for the steady state drift and removes that error from the calculation.
I guess that it could be a batch to batch variation for the gyros in the IMU, with more or less drift and also crosstalk (where an y-gyro is affected by a movement in x,z).
There are more types of errors also, don't know all of them.
So it could be an IMU that is below the average IMU in performance?
But as stated above, try a compass calibration and an IMU calibration.
/hakan
You only have one magnetic vector so you can't solve the full position of the quad but it could help more to use the vector info together with knowledge on what latitude you are at
than just using the heading. If you know the latitude you know what angle the magnetic vector has.
I don't know if they have an IMU in the gimbal also or if they use the quad IMU, so I'm not sure if the change of camera module would make any difference.
What they try to achieve with the IMU calibration is to get a table with temperatures range from cold to warm and the drift that each gyro (x,y,z) have in steady state at that temperature, and this table should have a temperature range that covers the temperatures it experiences during flight. When the quad is flying they check the IMU temp and looks in the table for the steady state drift and removes that error from the calculation.
I guess that it could be a batch to batch variation for the gyros in the IMU, with more or less drift and also crosstalk (where an y-gyro is affected by a movement in x,z).
There are more types of errors also, don't know all of them.
So it could be an IMU that is below the average IMU in performance?
But as stated above, try a compass calibration and an IMU calibration.
/hakan
ianwood
Taco Wrangler
A stuck IMU has happened but it is very rare and usually results in the drone racing off in one direction. This type of gradual horizon loss is very commonly caused by a bad compass calibration.
