Mavic Pro - Vision sensors issue & unpredictable behavior

ReBreaker · Jun 14, 2019

Hi,

I have something to say…

Background:
I have Mavic Pro for some time. I have bought it on 12th of December, 2017 on GB. Initial check was ok, I did a couple flights with beginner mode enabled, really nice piece of hardware I thought that time. Then my live situation changed rapidly so I had to sacrifice my hobbies and left the Mavic to rest for a couple months. After a while I came to an idea that I can try the Mavic again and do some video footage of my backyard. Around that time a got force update through the dji go4app. After the successful update I got it in the air and at the altitude close to 30m I have noticed that it has yaw drift to the left side in hover. I didn’t care much that time, I have kept it in the beginner mode and flew just a battery. Many months later I was confident enough that I left the beginner mode and did a couple flights on a wide field. During that flight I have noticed for a first time, that the Mavic losses its course and tends heavily to the left. Like it was not enough, after braking it has slowly drifted to the side about a meter or so until it has finally stopped in hover. On the way back to me I got notice about magnetic interference and it switched to the ATTI mode. I got lucky enough and brought it back to me. Then I have noticed that the vision sensors aren’t work during landing. I have messed up the compass calibration or it got magnetics interference from the towers nearby at least I thought that time. I left the spot and again the Mavic took rest for a couple weeks. A couple flights later the problems escalated to the level that I couldn’t use it for its purpose. Yawdrift, after-brake drift about a meter and a bow line instead of a straight line, you name it. I made a decision that I need to downgrade the firmware. I brought it back to 01.03.0400 and got some semi-successful results like less yaw drift, no signal breakups, no magnetic interference, no compass error and a sudden switch to the ATTI mode, no obstacle avoidance off without warning, but a straight line flight was more a bow line flight again. I said something like “Oh, I´m sure I can live with that” and moved towards to another things in my life. Time changes and I slowly get more room to my hobbies again and to fly the Mavic is a part of the list, so I foolishly upgraded the Mavic to the latest firmware through the Assistant 2 to the version 01.04.0500. What a mistake. I did the test flight on my backyard, no big issue aside from yawdrift and inability to flight straight on longer distance at more substantial speed.

Incident I:
Note: Pre-flight check list includes cold IMU calibration & gimbal calibration at home and compass calibration on a place.
I have done all the pre-flight checks and waited for a good number of satellites and strength of the GPS. It was ready to takeoff so I brought it to the air. Then it was ready to higher attitude just right after obstacle avoidance check. I ascended to 30 meters, turned on the course lock and I pushed elevator forwards. Just a second after that the Mavic switched into ATTI mode and have done unpredictable dance 30 meters above my head close to the trees all around me. Luckily it didn’t crashed. I have descended and turned it off with compass error and check redundancy state notice. I was afraid to give another try so I wrapped it up. Whole incident is attached in the .dat file extracted from the Mavic.

Airdata analysis:
Go App notice:

.dat files:
Incident I

Incident II:
I have proceeded refresh of firmware 01.04.0500 in the Assistant 2 at home and did some research about check redundancy state notice. My conclusion was to do a proper compass calibration on a wide field with no interference, so I did it another day. No avail. A brief notice about ATTI mode right after takeoff with TBE followed. And what is really interesting is that the vision sensors issue came back again – false alarms, no landing protection, radar and RC obstacle warning stuck. If you are interested in .dat files here you go.

.dat files:
Incident II

Many calibrations later
Many firmware refreshes later
Many downgrades and upgrades later

Current state:
Firmware 01.03.0700. Proper cold IMU calibration on level surface, gimbal calibration on level surface, successful vision sensors calibration in Assistant 2 many and many times again, compass calibration on my backyard. Results are in the video below.

link to the video on YT:
.dat files copying events in the video are attached below:
Current state

In general:

Problem? Vision sensors aren’t working

Was unit in a crash? No, never.

What have you tried so far? Calibrations of everything, Downgrade, Upgrade, Reflesh, Reinstalled the Go App, Different version of Go App

What device are you using? LG V10

What firmware are you running? 01.03.0700

What Go app version are you using? 4.3.20

Any modification? None

Did you change anything or install any apps? Reinstalled the Go app, no change whatsoever

Do you have a video or pictures of the problem? In the text above

I’m hopeless guys. Tired of all the calibrations. The last thing I can do is to put together the documentation of it all. Your amazing knowledge in analyzing the .dat files is the only thing what keeps my hope alive. Your help is widely welcomed. If anyone can bring some light to the culprit of the issue, it will save my Mavic days and I will buy you a beer. Thanks

brett8883 · Jun 15, 2019

Alright there the biggest thing here is a somewhat common issue though not as common with the MP but we call it Yaw Error. Yaw Error happens when your Mavic starts up and calibrates itself and locks in its reference heading direction incorrectly. It often happens due to compass issues on start up relating to magnetic interference. I notice your yard is elevated, any metal under there? Then when the mavic lifts off the compass goes back to normal but its reference heading is off. I notice in your video you turn the mavic on and then walk over to put it on the launch pad. DO you do that normally or was that just for the video?

This graph is interactive and can be made full screen by clicking the expand button in the lower right corner. You can slide the slider at the top to hone in on a certain section

In this first graph we see the yaw-mag to yaw. Your yaw-mag should be roughly the same as your yaw. But there is a point right after you take off things go haywire for a sec and then go back to normal as you get further from the ground. then when you come in to land that when the mavic give a code red warning and you are on your own. Notice how in the next graph that shows your rc inputs to yaw actual and notice where they split from eachother in relation to the first graph.

When you start up the mavic make sure its on the ground in a fixed position and let it start all the way up before moving it. Orientate it to a know road you can see on the map and when it get gps lock make sure the little arrow on the map is pointing in the actual direction the mavic is facing.

Now when you took off in opti mode in that video and it seemed to have a mind of its own thats expected behavior and the front sersors are not active in that mode. Opti refers to the VPS sensors on the bottom. Landing protection refers the the sonar sensors and means the mavic will stop and wait for confirmation to land. It doesnt have anything to do with activation of the VPS sensors or not.

ReBreaker · Jun 15, 2019

Thanks for your quick response. Now to your questions. I’m not aware of any metal under the yard. I have tried a different location on a wide field near the city where I live (as seen in the Incident II). No metal objects around, high voltage power lines far to the horizon and no towers nearby. I did the powering up in a hand just for the video. What I always do is to take place for the take-off, then put the Mavic to the place and switch it on a ground with no movement involved.

I have carefully looked at the graphs. It is still unclear to me what may have been the cause of the magYaw deviation from yaw.

I took another video with a closer look at the observed problems with obstacle avoidance and landing protection. It also includes a drift on the yaw axis. It may show you what I’m dealing with. It shows that the Mavic goes to the air with the front and the down sensors already disabled. I can replicate that even when I sit next to the Mavic at home. I turn it on, there is no GPS signal so it stays in OPTI (2 quick green flashes), radar chart shows itself as well as clearance shows values and after a minute or so the radar chart hangs, clearance values disappear and it switches to ATTI mode (one slow yellow flash at time) without reason. Ambient conditions are good with enough light, so the Mavic should stay in OPTI mode if there is no GPS signal if I’m correct. What I want to add is that spin the motors doesn't change that state as well as warm power cycle. What changes that is a cold power on, but OPTI state won’t last long again. What I have tried also is to take-off in my garage, OPTI mode keeps the Mavic steady and in a minute it switched into ATTI by itself and I had to land immediately (no landing protection, no obstacle notice). I’m really confused from that.

ReBreaker · Jun 21, 2019

In addition, there is my indoor test of how the Mavic switches back and forth between OPTI mode and ATTI mode without any intervention on my part. Ambient lightning is normal daylight. That whole thing is a real mystery if you ask me.

I can no longer rely on obstacle avoidance system anymore as if the massive yaw drift wasn’t enough. Can it be a compass unit issue? I mean if the compass unit is defected in some way, can it force to switch from OPTI mode to ATTI mode even when the ambient conditions are sufficient for VPS?

It drives me nuts that it behaves so unpredictably. I’m ready to rip the entire device apart into the last screw. Go for it to the end and replace everything that can cause these troubles. The easiest replacement is the front compass board and integrated GPS & compass unit in the back, but what keeps me back is the question: Will it help to resolve the issues?

brett8883 · Jun 21, 2019

ReBreaker said:
In addition, there is my indoor test of how the Mavic switches back and forth between OPTI mode and ATTI mode without any intervention on my part. Ambient lightning is normal daylight. That whole thing is a real mystery if you ask me.

I can no longer rely on obstacle avoidance system anymore as if the massive yaw drift wasn’t enough. Can it be a compass unit issue? I mean if the compass unit is defected in some way, can it force to switch from OPTI mode to ATTI mode even when the ambient conditions are sufficient for VPS?

It drives me nuts that it behaves so unpredictably. I’m ready to rip the entire device apart into the last screw. Go for it to the end and replace everything that can cause these troubles. The easiest replacement is the front compass board and integrated GPS & compass unit in the back, but what keeps me back is the question: Will it help to resolve the issues?

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

The switching from Atti to Opti mode isn’t all that odd I find that happens indoors for me too but I get what you are saying. Let’s not lose sight of the bigger issue which is the yaw drift (yaw error) for now.

Given everything you’ve already tried it would seem to me it’s very likely a hardware issue but try a couple of things for me.

#1)Take the drone outside, turn on the remote with a device attached and start recording your screen.

Change the mini map to the “compass/attitude radar” and then put the remote down on the ground directly in front of you then stand up and point the drone in exactly the same direction the remote and device are facing and turn the drone on while you are holding it.

Keep the drone facing the exact same direction as remote and hold it as still as possible while it starts up. Continue holding it still until GPS is acquired and continue holding it there for a few more seconds.

After a few seconds slowly rotate the drone 360 degrees along the yaw axis as if it were hovering and yawing. When you get it back to facing the same direction as the remote note if the icon for the drone on the radar in the map is facing the same direction as The remote as noted by the triangle at the top of the radar thingy.

Then hand launch the drone while it is facing the original direction the remote was facing and yaw 360 degrees in place and land. Don’t do anything else in the flight, keep the drone on and remote connected and take off from the ground do the same thing yaw 360 and then land do nothing else. Take off again and see if it is flying normally.

Upload the screen recording video ad the flight log from this experimentZ

This is to rule out magnetic forces in your ground and to verify if the compass in the drone is able to consistently notice the direction the drone is facing and get a controlled h

#2. Turn off the drone and the remote shut it all down restart everything and then take off hover momentarily, land and then without turning anything off rotate the drone as if you were doing a compass calibration (you know how to do that without being prompted right?) then upload that flight log as well.

This will give us a full mag mod picture to compare to the mag mod in the first flight. You needed to take off just to ensure the drone records and flight log.

It very well and probably is a hardware issue but let’s completely rule out everything else before making that conclusion

ReBreaker · Jun 22, 2019

ok @brett8883 , I will do all the described procedures and I will post the results as soon as possible.

ReBreaker · Jul 25, 2019

So @brett8883 I am finally here with the results. I apologize for the delay, but after 1860 miles driven by car I was exhausted. Holiday season can be tough sometimes…

I had instructions printed so hopefully everything is done properly. I have noticed during the part #1 that even the drone and the remote are facing the same direction, the compass thing shows a slight offset before any movement and a bit larger offset after 360° rotation. Just as a note I tried to follow the instructions off the record after arrival at my vacation’s location miles away and I noticed even larger offset there.

I was so focused to hold the mavic still during startup and noticed late that the compass thing is covered by GUI, so I brighten up the video a bit at that point. It should be visible enough. I noticed also, that after unintentional phone display flip during the test the compass thing flips too. I hope it did not affect the test.

Flight log attached here: Flightlog ExperimentZ