DJI Mavic, Air and Mini Drones
Friendly, Helpful & Knowledgeable Community
Join Us Now

Drift to right, IMU calibration?


Well-Known Member
Premium Pilot
Sep 1, 2018
Spring Hill, Fl
Since I had crashed a couple weeks ago, I thought I'd do an IMU calibration.
The manual never says the surface has to be perfectly level.
Ever since doing that, the AC seems to drift to the right, particularly with a cross breeze but not necessarily so.
At one point it particularly went sideways when I was about 700ft out and hardly moving. I thought I had yawed in the direction of the drift but when I looked at my video feed, I was seeing it slip. When I went to make heading corrections it seemed to go nuts, moving every which direction. It was getting rather breezy too do I hit RTH which came back with no apparent issues.
My first flight seemed to be just fine, it was into my second flight that it started being a bit irratic. I thought I went into atti mode, but no.
I'm almost sorry I ran the calibration.
The drift I encountered during tests I can compensate, though without side sensors in most modes, but that one point when it really seemed to drift has me concerned.

Regarding calibration. When setting it on its side or butt, it isn't perfectly flat so there can be deviation from what is intended.
Also, with arms folded, which instructions say to do, I get a mag interfence. This goes away if I unfold the arms and restart.

I did also calibrate the remote, but it isn't sensitive from 0.

I've got the DAT file from the tablet which when played back in the app, shows the AC going nuts with no stick input.
I'm unable to retrieve any flight logs from Assistant 2. Seems many have this problem with M2's even though release notes says it should work.
How do I provide you with the DAT file, since this forum doesn't accept the DAT extension? Also how do I know which DAT file, since there seems to be a few per flight (multiples ending in FLY076.DAT)
I attached the TXT file in the mean time.


  • DJIFlightRecord_2018-09-30_[14-58-27].txt
    539.1 KB · Views: 10
Hi, not sure if I found it in the manual or online (I think online, because I remember an image of a spirit level), but it’s very important to do it on a perfectly flat and level surface. If you just do the calibration again it should be ok. I would suggest to do a camera calibration at the same time, because it all works together. Pretty straight forward, but do remember that the screen has to be at a (or as close to as possible) 90 degree angle.
I would think if it needed to be perfectly level where you should use a spirit level, the manual would be specific enough. It suggests generally level.
In old days, you just put the Phantom on a level surface, run IMU calibration and wait a couple minutes.
Now they have you move it 5 ways, and it automatically advances as soon as you get close to the position for that stage. There's no perfect placement for at least 3 of them.
Based on context, I assume you meant obstacle avoidance calibration. I'm not sure that would be a factor for drift. It would only move off course if using APAS which I was not for my test.
Compass calibration might be the issue though.
I might have an explanation.
I just noticed the rear left arm has a bit of upward play. Right is bit stiffer. I don't suppose anyone else has any play in their rear arms?
Last edited:
  • Like
Reactions: DrorHowley
I was able to get Assistant to show and export flight logs. Used my laptop and disabled Windows Defender firewall.
Only one file even when you select to export all logs.
Can't calibrate vision sensors on my laptop though. I suspect you have to have at least 1080p. My laptop screen is 720.
I might have an explanation.
I just noticed the rear left arm has a bit of upward play. Right is bit stiffer. I don't suppose anyone else has any play in their rear arms?
I’ve noticed that on mine too, one arm doesn’t really close the same, I’ve heard this is normal. I also now take the props off to go in the case. Similar to you I was simply trying to make sure everything was calibrated noticing very slight instability and occasional errors when flying.

At first I was getting magnetic interference from metal and other issues but unfolding the arms seemed fine. Did the calibration, took off and I could barely land it was so off. Tried 2 more times also without success. I took a flat board and a level to an area there was no metal. I leveled the board and went through the calibration props off, turned it off and then did the compass. Finally flew like a champ.

But I still have questions. One side of the craft doesn’t sit quick the same on a flat surface than the other and almost wobbles. Props off helps but do people hold it down to the surface to try and make sure both sides are calibrated in the same position ? Or is it not that critical ?
Lying on its bottom as you would normally have it shouldn't rock, but as we both pointed out, the back legs have some play so they might not sit on a perfectly flat surface. I'm glad you confirmed that's so even without any crash.

Instructions does not say for an IMU calibration you need a perfectly level surface verified with a spirit level. The only orientation that would be reliable if required would be on its bottom. Any other orientation and there are 2 or more ways it could lie solid.

For me, point is moot now that I have to send it in for repair after backing over it with my car while it was in the Fly More case. I'll have a separate post for that momentarily.
I think I figured out why my M2 acted funny out over the water (flight in attached txt file in this topic). I was only 35 feet up which is around the limit of the downward vision sensor. I bet it's input over water along with the moderately high winds confused it. I'm not used to downward sensors being potentially effective that high up, previously having flown P3A which has a 10ft limit.
I got a refresh replacement. Both back legs are solid, so the back left may have gotten a bit loose when it crashed after hitting the wire.
Just looked at my txt log in PhantomHelp. Apparently VPS was not involved.
Not sure how to determine what the corresponding DAT file would be.
You dont need utterly flat .
After calibration you can check it by going into sensor status and checking all fields are green and low numbers (and stable with the drone on the floor and not moving).
I know. My main concern now is why it seemed to toilet bowl. I thought it was VPS tracking shore currents and correcting with GPS, but I now see in the logs, VPS was not being used.

More likely it was moderately high winds. I've found these systems are more stable when moving than when hovering.
Yes you do want the surface to be level,

I set up a box level in both axis to calibrate em on. (I use my phantom's styro case to limit vibrations)
I start with cold drone and with IMU first while working fast to prevent heat build up, I have the dji-go app already running and ready to go just need to refresh the setting page.
Then following up with camera gimbal calibration on that same level surface.
Results in a drone that hovers rock steady and a camera that is level with it and with a level horizon, no adjustment from zero necessary after the fact.

You can get the corresponding DAT file number from the .txt file in CsvView, right panel of the window near the bottom.
It doesn't have to be perfectly level.
Consider this: two of the positions it asks for, there's two ways it can be steady on the surface. IOW you can rock the AC. If it were that critical, they would detail exactly how the AC should sit in those two points.
It doesn't have to be perfectly level.
Consider this: two of the positions it asks for, there's two ways it can be steady on the surface. IOW you can rock the AC. If it were that critical, they would detail exactly how the AC should sit in those two points.

While I don't have a MP2 model yet, my MP sits exact in the positions shown no rock or hard to get it in place.

Think about how it works, the drone has to know where level is and is the reason for moving it following the script in the app.
It gets an "overall" rather than single position by moving it to those different positions, gets it more information to work with during that process about itself.
Yea could account for some small amount of "out of level" to start with but why have it do so. Only takes a second to set up a level surface and only need to do it once unless troubleshooting/reflash firmware etc.

I want my drones to know where true level is, not just what it thinks true level is.

The following is from the link.

IMU calibration, which involves the accelerometer, will help set standards for the aircraft’s attitude and reduce errors caused inaccurate sensor measurements. Make sure to calibrate your IMU on a level surface, and don’t move the aircraft during calibration. If you calibrate on an angled surface or move the aircraft during calibration, the aircraft’s attitude will be angled by default and it won’t hover stability, even without magnetic interference or other environmental issues. If your drone drifts a lot, IMU calibration may help resolve the issue. However, please note that hovering isn’t 100% accurate, it’s normal for a DJI drone to drift a small amount, especially in high winds.

Neither of my two drones waver around or "toilet bowl" as you put it, not even a little bit. They both hover rock steady in place even in pretty stiff wind. What I consider stiff wind and willing to fly in that is.

Has got to be in part giving em the best starting point possible, IMU calibration on a flat and level surface. One that doesn't move and with as little vibration as reasonably possible.
It only has to be reasonably level surface.
Besides my pointing out that it's not possible to know exactly how to position the M2 on two of the indicated positions, if it were that critical, manual would have you check the surface with a bubble level.

No Mavic is rock solid on hover. It's going to drift some. It can't because GPS is not that accurate. VPS is more accurate but even then. If IMU alone was that accurate, it would be able to hover in Atti mode without drift, as it would detect uncommanded wind drift and go back to its original position.
Lycus Tech Mavic Air 3 Case

DJI Drone Deals

New Threads

Members online

Forum statistics

Latest member