Any thoughts on why on one flight yesterday the altitude of my Mavic 3 was changing without any input on l stick? Firmware of RC-N1 and Mavic3  was updated prior to flight and so was DJI Fly. It was snowing and APAS was on. There was warning of ambiant light issue, and drone was sometimes detecting obstacles when there were none. Primary issue was backward flight always caused increase in altitude sometimes left, right and forward as well. In video below issue occurred in CINE and normal modes, not sport (maybe because obstacle avoidance was odd) Did not have issue in two subsequent flights at different location however so not sure why it happened other than maybe weather/ snow was heavier and was over lake with ice.

Overnight i performed Mavic 3 firmware refresh via DJI Assistant 2, then IMU, RC-N1, and gimbal calibration. Any other steps i should take?

I'm kind of glad I cannot afford one ................yet!

We’re you fling with OA in Bypass or Brake? In bypass the drone will adjust its height to clear obstacles, if you were flying in snow it would be challenging for the OA system.

Weird. Looks like something about those snow covered fields is really confusing the vision system. It basically says as much with its "ambient light varying greatly" warning. It would be handy to get a log to the DJI devs. I dunno if you've opened a support case, might be worthwhile.

Buzzyone Posted at 12-11 09:16
We’re you fling with OA in Bypass or Brake? In bypass the drone will adjust its height to clear obstacles, if you were flying in snow it would be challenging for the OA system.

Was bypass versus brake.

Update : Today I flew in bright sunshine, not snowing, snow on ground. Took with me Mavic 2 Pro and Mavic 3.  Mavic 2 had no issues with elevation increasing. Mavic 3 prior to 12/10/2021 firmware upgrade did not have issues with elevation increasing.  Mavic 3 in bright sunshine does increase altitude. After testing it seemed today the issue occurred only when OA is on and set to bypass s and drone height is under 80 feet. I have nowhere to test where there is not any snow. So I am not sure if snow plays any part in outcome.

Also today I noticed the Mavic 3 lost track of -38.7 feet of altitude!  I went to land in the same spot it took off and Mavic 3 indicated the height was -38.7 feet at that location.

I will be posting another video where I illustrate Mavic 3 is having altitude increase issue in bright sun and Mavic 2 has no issue at same location. Also it will show the loss of elevation.

Montivette Posted at 12-11 14:49
Update : Today I flew in bright sunshine, not snowing, snow on ground. Took with me Mavic 2 Pro and Mavic 3.  Mavic 2 had no issues with elevation increasing. Mavic 3 prior to 12/10/2021 firmware upgrade did not have issues with elevation increasing.  Mavic 3 in bright sunshine does increase altitude. After testing it seemed today the issue occurred only when OA is on and set to bypass s and drone height is under 80 feet. I have nowhere to test where there is not any snow. So I am not sure if snow plays any part in outcome.

Also today I noticed the Mavic 3 lost track of -38.7 feet of altitude!  I went to land in the same spot it took off and Mavic 3 indicated the height was -38.7 feet at that location.

I suspect that this is not your case, but someone else reported that the back usb port cover was being sensed and impacting performance if not properly and completely closed.  In your case, I wonder if it has something to do with the snow dominated landscape.

rookie here so basically i have no clue   But every time it seemed to go up theres the orange or red line,  isn't that the avoidance sensor kicking in

Flying_Amateur Posted at 12-12 06:02
rookie here so basically i have no clue   But every time it seemed to go up theres the orange or red line,  isn't that the avoidance sensor kicking in

Yes, I noticed that too. Could it have been a snow flake hitting the sensor?

kyalami Posted at 12-12 06:40
Yes, I noticed that too. Could it have been a snow flake hitting the sensor?

Sure it was detecting something. However, in my next video posting later today when flying in bright sunshine there are no proximity alerts and the drone altitude still repeatedly increases without any left stick input.

Montivette Posted at 12-12 08:36
Sure it was detecting something. However, in my next video posting later today when flying in bright sunshine there are no proximity alerts and the drone altitude still repeatedly increases without any left stick input.

Sorry, that was sloppy of me, I missed that. Hope that DJI will chip in and assist, or someone else.

I tested earlier today, repeating the same flight pattern as the poster.  I did not experience an inadvertent or unwanted change in elevation, but my weather conditions were superior (beautiful day in Colorado).  I suspect the weather and environment are a factor, but not sure how to accurately test without mother nature's cooperation.

Charles Adams Posted at 12-12 10:52
I tested earlier today, repeating the same flight pattern as the poster.  I did not experience an inadvertent or unwanted change in elevation, but my weather conditions were superior (beautiful day in Colorado).  I suspect the weather and environment are a factor, but not sure how to accurately test without mother nature's cooperation.

Charles,
Perhaps it is not issue that occurs for everyone and maybe due to hardware issue, software anomaly, or perhaps environmental. Happens for me even in the sun. During your test was there snow on the ground or was it dirt, sand, water, or grass? Several people in comments of the video indicate they observe same issue with their Mavic 3 even before December 10, 2021 firmware updates.  Not sure what was below drone at time of their observance of issues.  The temperatures are expected to rise in Minnesota this week so perhaps the snow we have will melt and I can do some more test runs to see if perhaps there is some issue with reflectivity of snow on the ground or something.

Montivette Posted at 12-13 09:15
Charles,
Perhaps it is not issue that occurs for everyone and maybe due to hardware issue, software anomaly, or perhaps environmental. Happens for me even in the sun. During your test was there snow on the ground or was it dirt, sand, water, or grass? Several people in comments of the video indicate they observe same issue with their Mavic 3 even before December 10, 2021 firmware updates.  Not sure what was below drone at time of their observance of issues.  The temperatures are expected to rise in Minnesota this week so perhaps the snow we have will melt and I can do some more test runs to see if perhaps there is some issue with reflectivity of snow on the ground or something.

Light snow on the ground, so it was in patches.  Overcast day, so not sunny but brighter than "dreary" (how's that for unscientific).  I was really testing other features of the drone, and didn't focus on the "altitude change" (though my tests did permit me to observe for it).  If time permits I'll do more focused testing today.

The loss of altitude on take off / landing in the same spot suggests barometric pressure causing issue with sensors. This has come up before where folks had their altitude suddenly go crazy and show them flying at like  -170 feet, despite being up in the air, well above take off point. DJI in one case responded to a user and said it was due to changes in barometric pressure screwing with the sensor in the drone, and that it was harmless. You likely have the same thing going on. It seems you also have some weird issue going on with the vision sensors.

I took my drone out yesterday after seeing your video and tested it just as you did in your videos. I run my drone in Brake mode though. Im still not 100% trusting of the obstacle avoidance until they get it perfected so I prefer the drone to just STOP if it sees something rather than try to navigate it. In brake mode I had no issues in my tests. I am in Florida, USA though, and below the drone was heavy wooded areas, lots of green leaves. Ive had none of the issues though others have reported. Drifting, altitude, etc etc. The drone has flown exactly as you would expect it to everytime Ive went up. Ive always had 16+ satellites locked in and been in good daylight though.
EDIT: One other thought is that since you mention the issues seem to always happen inside of 80 Feet tells me its 100% an issue with the downard facing vision sensors. Something is throwing them off. In the Owners manual it does talk about the vision system could malfunction when flying over certain things like mirrors and things, but Over white snow? I would run a test... Find a spot whre you have lots of snow. as in your video, then perhaps a transition from snow to something other ground cover, like maybe an asphalt parking lot or something. See if, as you exit the snow and enter the change in ground cover, if the drone immediately changes behaviour.

Looking at MN weather graphs, you do see some fair swings of barometric pressure. 40 ft over the course of a flight is not entirely beyond the realms of possibility. That's 0.04 inHG. In the last 24 hours, MN pressure has risen by 0.3 inHg. So if you had a weather front roll across during a flight, you would see significant shift in barometric height.

170ft for Brian, not so much, that would be most likely a fault.

One thing you should try if you have an iPhone or some other phone with a baro sensor, is running a barometer app during a flight, to see if any drift is real or a possible sensor fault.

Below is link to new video in sunshine.  Same issue. I did perform more runs than were indicated in video but cut many of them out to keep video shorter. Video has one run in each direction with OA Bypass, some with OA stop, some with OA off, and a few comparsion runs from Mavic 2 at same location (next day).  The majority of runs on Mavic 3 altitude increased without left stick input. A few times it did not

I just got back from testing.  I could not duplicate the issue when in "stop" and "off", but I did clearly get unwanted elevation when in "bypass".  I did not film it.  But my observations are that it is happening when "bypass" is selected.

I was also in "normal" mode, and I wonder if it occurs in "cine".  I didn't have an opportunity to test that, but the elevation seemed to occur when pitch went past a certain level (I may not be using that term correctly).  The elevation seemed constant until a certain speed was attained, and I wonder if the slower cine will mitigate this.

Tornado12 Posted at 12-13 11:59
The loss of altitude on take off / landing in the same spot suggests barometric pressure causing issue with sensors. This has come up before where folks had their altitude suddenly go crazy and show them flying at like  -170 feet, despite being up in the air, well above take off point. DJI in one case responded to a user and said it was due to changes in barometric pressure screwing with the sensor in the drone, and that it was harmless. You likely have the same thing going on. It seems you also have some weird issue going on with the vision sensors.

I took my drone out yesterday after seeing your video and tested it just as you did in your videos. I run my drone in Brake mode though. Im still not 100% trusting of the obstacle avoidance until they get it perfected so I prefer the drone to just STOP if it sees something rather than try to navigate it. In brake mode I had no issues in my tests. I am in Florida, USA though, and below the drone was heavy wooded areas, lots of green leaves. Ive had none of the issues though others have reported. Drifting, altitude, etc etc. The drone has flown exactly as you would expect it to everytime Ive went up. Ive always had 16+ satellites locked in and been in good daylight though.

After tests on a sunny day it currently seems like the issue is more prevalent at lower altitude than at higher altitudes. Issue has occurred over 80 feet though so not positive whether elevation is contributing or not.   During most test runs altitude increased. In some test runs it did not.   Hopefully snow will melt in next couple of days so I can test at same location without snow to see what occurs.

I decided to put the video of loss of 40 feet of altitude over the course of one flight into a separate video I will post in future.  That flight was 12/11/2021 (same day as other sunny day tests) The test in snow occurred 12/10/2021 around 3pm central time.  

Here is airport data from airport 30 miles away reflecting pressure on the 10th and 11th of December.
https://w1.weather.gov/data/obhistory/KSTC.html

Montivette Posted at 12-13 14:44
After tests on a sunny day it currently seems like the issue is more prevalent at lower altitude than at higher altitudes. Issue has occurred over 80 feet though so not positive whether elevation is contributing or not.   During most test runs altitude increased. In some test runs it did not.   Hopefully snow will melt in next couple of days so I can test at same location without snow to see what occurs.

I decided to put the video of loss of 40 feet of altitude over the course of one flight into a separate video I will post in future.  That flight was 12/11/2021 (same day as other sunny day tests) The test in snow occurred 12/10/2021 around 3pm central time.  

I'm not sure if by "altitude" you mean sea level altitude or AGL.  When I did my test (about 30-45 minutes ago), I did it in Colorado (>5280').  One of my tests was 50' AGL, and another test was 15' AGL.  In both instances, I was experiencing non-user-initiated altitude increase when bypass was turned on.

Just for clarity when you folks are talking about the drone changing height with out being commanded to do so are you talking about watching the drone climb with your own eyes or are you talking about the indicated height increasing.

Separately, if barometers are drifting 'downwards' and the drone is programmed to maintain height based on the barometer reading, would a downwards drifting barometer not cause a climb?

Charles Adams Posted at 12-13 15:00
I'm not sure if by "altitude" you mean sea level altitude or AGL.  When I did my test (about 30-45 minutes ago), I did it in Colorado (>5280').  One of my tests was 50' AGL, and another test was 15' AGL.  In both instances, I was experiencing non-user-initiated altitude increase when bypass was turned on.

Was referring to above ground level AGL. Sorry for any confusion.  

Sean-bumble-bee Posted at 12-13 15:33
Just for clarity when you folks are talking about the drone changing height with out being commanded to do so are you talking about watching the drone climb with your own eyes or are you talking about the indicated height increasing.

Separately, if barometers are drifting 'downwards' and the drone is programmed to maintain height based on the barometer reading, would a downwards drifting barometer not cause a climb?

Talking about when bypass is enabled the Mavic 3 visibly increasing altitude without user input on the stick to move it higher and the app indicating it is increasing in altitude. You can see the increase in altitude AGL occurring in the videos linked in thread.

I think drone ascending without stick input is not likely a barometric pressure issue. If it were then increase above AGL without user input would occur with obstacle avoidance set to bypass, brake, or OFF.  It will only occur in my Mavic 3 when OA is set to Bypass. I can flip back and forth from OA bypass to OA Off or OA brake over the same flight and there is not a problem with OA set to Off or brake. Issue occurs only in bypass.

The secondary observed issue during one flight only where the indicated altitude changed almost  -40 feet over the course of one flight from takeoff to landing on same surface I can get behind it being barometric pressure related.

Montivette Posted at 12-13 23:23
Talking about when bypass is enabled the Mavic 3 visibly increasing altitude without user input on the stick to move it higher and the app indicating it is increasing in altitude. You can see the increase in altitude AGL occurring in the videos linked in thread.

I think drone ascending without stick input is not likely a barometric pressure issue. If it were then increase above AGL without user input would occur with obstacle avoidance set to bypass, brake, or OFF.  It will only occur in my Mavic 3 when OA is set to Bypass. I can flip back and forth from OA bypass to OA Off or OA brake over the same flight and there is not a problem with OA set to Off or brake. Issue occurs only in bypass.

Same observations and experience.  AGL rising is both visible and measured by the app.  I climbed from 50' to >100' without command.  I intend to try today in cine mode, as I have a suspicion the issue is triggered by speed/pitch.

I figured out what we really need to do, and I'm asking for participants who have the time and knowledge to facilitate this request.  We need someone to go out in the field (so to speak), and while in bypass mode flying in "normal" mode, do multiple tests from at rest to forward, then backwards, then each direction left and right.  I'd say 5 times each, but maybe more like 10.  Then take the logs from this flight and examine them, looking to see at what point the climb in AGL occurs.  I suspect that there is a triggering threshold related to speed or pitch or both at which point the rise begins, and I suspect the logs will demonstrate this.

I use to participate in log review with my peers here (Hallmark007 and others), but it's been years, and I'll have to reeducate myself on those skills.  Anybody else with more time can do it faster than I.

This isn't particularly useful information for us consumers, as we now know it happens, we probably don't care about the specific mechanics.  But I think this would be useful info for DJI.

We also don't know what impact this has on other features of the system.  Could this impact RTH?  Other OA circumstances?  We just don't know.

I agree and I suspect the cause of this could be determined with some testing. I think it is very obvious and clear that the climb in altitude has to be related to the vision sensors. The incorrect altitude reading is likely a seperate issue. The climb is happening because the drone thinks it is approaching objects. If you fly at a real object in bypass mode you see the same behavior. The question now is what is the cause of these "ghost objects". What is the sensor seeing? Is it something regarding ground pattern? Is it something on the drone itself it is seeing under certain circumstances? For example, it has already been reported that the little plastic cover door for the Micro SD / USB C port can show up on the sensors if it is not fully closed. These sensors have a fairly wide view angle, so are they seeing the blades of the drone for example? You would think that would not be possible as such a thing would surely have been covered in development stages in the engineering of the drone. Also, if it was the drone seeing parts of itself, then the issue should uniform with most everyone's drone. If it is not uniform then I think that drastically reduces the likelihood that it is something like the drone seeing its own props for example. Could it be ground cover? Montivette indicates that he noticed these issues almost exclusively under 80 feet altitude or so. Does the drone do the same thing at 200 feet? 300 feet? If the symptoms go away at these higher altitudes then I think there again that is a massive clue. It would also rule out the drone seeing some portion of itself in flight.

I will try to do some testing on this myself soon. I have not used bypass at all with this drone. I am actually scared even of the new smart RTH feature. I live out in the country but near my property I have some big high wire powerlines that run for miles, and they are up about 150 feet high. I am very careful everytime I fly in that direction. If I was to lose connection on the other side of those lines though and the drone went into smart RTH it would have a good chance of flying right into those power lines. These sort of thoughts has really put me off the whole bypass and smart obstacle avoidance thing. I use brake mode exclusively. I have some large farm fields and stuff I can test this over here in Florida. If I find anything Ill report back.

I just got back from a round of testing.  My observations are:  It does occur in Cine and Normal flight modes.  It does occur in any direction.  For me it predominantly occurs in reverse.

In Cine, it's an extremely slow rise, but it does rise over time.  The rate of climb does not seem to be linear to speed, but rather seems to be hyperbolic (much higher ascent rate at slightly higher speeds).

There was a strange moment when I went from cine back to normal that the drone was not rising, but that was short lived.

Charles Adams Posted at 12-14 12:45
I just got back from a round of testing.  My observations are:  It does occur in Cine and Normal flight modes.  It does occur in any direction.  For me it predominantly occurs in reverse.

In Cine, it's an extremely slow rise, but it does rise over time.  The rate of climb does not seem to be linear to speed, but rather seems to be hyperbolic (much higher ascent rate at slightly higher speeds).

What about the behavior in different altitudes ? Same exact test, just 100 foot higher, 200 foot higher, 300 foot higher, etc.

Im wonder if once we get to a certain general height (beyond sensor range of the ground) if the behaviour changes or goes away entirely.

I decided to look back over the user manual because I wanted to recall the vision sensing range of the various sensors. I decided to keep reading. Read over this page, as it talks about how the drone may have problems when flying over snow. This would give some plausibility to montivette's situation at least. Still would like to know how it behaves at like 300 foot altitude, where the vision sensors can not read the ground cover.

Tornado12 Posted at 12-14 13:24
I decided to look back over the user manual because I wanted to recall the vision sensing range of the various sensors. I decided to keep reading. Read over this page, as it talks about how the drone may have problems when flying over snow. This would give some plausibility to montivette's situation at least. Still would like to know how it behaves at like 300 foot altitude, where the vision sensors can not read the ground cover.

My flight was with practically no snow on the ground (it's mostly melted).

Tornado12 Posted at 12-14 12:55
What about the behavior in different altitudes ? Same exact test, just 100 foot higher, 200 foot higher, 300 foot higher, etc.

Im wonder if once we get to a certain general height (beyond sensor range of the ground) if the behaviour changes or goes away entirely.

I'll try that out!  My flights were starting at 15' AGL.  Yesterday's were at 50' AGL.  I can try out at 100' and 200', and I'm curious even at 350' and if I'll get warnings when I get towards 400'.

I just got back.  It must be my last flight of the day, as I'm losing daylight.  My methodology was to start at 350' AGL, attempt to trigger the unwanted rise in altitude, and drop in '50 increments until I experienced the issue, and then in small increments climb back up until the issue stopped.

I found 90' and above to be the spot where I no longer was able to trigger the occurrence.  But I was able to trigger it in the low 80's.  Interestingly, if triggered (and I maintained speed and direction) I could not get it to raise above 90'.  In fact my unscientific observation is that the unwanted elevation raise seems to lose "force" the higher the altitude gets, and the lower the altitude, the higher the upward "force" when it is triggered.
P.S.  This is with the new DJI fly app that was pushed today.

Thanks Charles and other for confirming issue and attempting to isolate behavior. Hopefully the thread will result in DJI looking into this anomaly and resolving it if they were not already aware of it and in process of resolving it in software in some manner.  

Had some like this on A2S...pushed forward or back with OA in bypass was just elevating like crazy...
Was resolved by a firmware update.
Still, was not a nice experience as you can:
- break the flight altitude declared at TWR without intentions
- hit something as the increase in height was unpredictable.

Geo_Drone Posted at 12-15 01:42
Had some like this on A2S...pushed forward or back with OA in bypass was just elevating like crazy...
Was resolved by a firmware update.
Still, was not a nice experience as you can:

To add my experiences to this thread:

Mine does it in Normal mode in forward flight, even at 100m off the ground, when in Bypass mode.
It does NOT happen in Sports mode, or if APAS is off.
I have not noticed in in reverse or sideways.

I only give my experiences for bright, dry, clear weather, over land, no snow/ice/lakes/rivers etc.

TheBoy Posted at 12-15 04:27
To add my experiences to this thread:

Mine does it in Normal mode in forward flight, even at 100m off the ground, when in Bypass mode.

I have wondered if the issue is not as much related to altitude, and if it was more related to environment.  By that I mean where I'm flying it's relatively flat.  The tallest "obstacles" are trees and one story buildings.  I've wondered if the visual sensors are "observing" distant objects and interpreting them as closer obstacles.  I've further wondered if I were to fly in an environment where there were taller obstacles (but still some distance away), would I experience the unwanted behavior at a higher altitude.  Like flying near a 5 story building or a really tall tree.  Even if those objects were distant could OA be identifying and reacting to them?  I'm trying to figure out where I have an environment near me that I can test the theory.

Charles Adams Posted at 12-15 12:10
I have wondered if the issue is not as much related to altitude, and if it was more related to environment.  By that I mean where I'm flying it's relatively flat.  The tallest "obstacles" are trees and one story buildings.  I've wondered if the visual sensors are "observing" distant objects and interpreting them as closer obstacles.  I've further wondered if I were to fly in an environment where there were taller obstacles (but still some distance away), would I experience the unwanted behavior at a higher altitude.  Like flying near a 5 story building or a really tall tree.  Even if those objects were distant could OA be identifying and reacting to them?  I'm trying to figure out where I have an environment near me that I can test the theory.

I have experienced it at the full 120m height allowed in the UK, and the only obstacles taller than 50m were literally miles away.

I did wonder if it was detecting things like insects, but ruled that out given the fact it is the start of winter here in the UK.

TheBoy Posted at 12-16 00:17
I have experienced it at the full 120m height allowed in the UK, and the only obstacles taller than 50m were literally miles away.

I did wonder if it was detecting things like insects, but ruled that out given the fact it is the start of winter here in the UK.

Yes, there appears to be a difference in behavior related to the environment, but it's hard to tell what that is.  I'm in Colorado, and we really don't have a large quantity of insects flying around this time of year.  Never-the-less, it is something DJI will need to look into and correct.  I suspect they will.

Charles Adams Posted at 12-14 15:14
I just got back.  It must be my last flight of the day, as I'm losing daylight.  My methodology was to start at 350' AGL, attempt to trigger the unwanted rise in altitude, and drop in '50 increments until I experienced the issue, and then in small increments climb back up until the issue stopped.

I found 90' and above to be the spot where I no longer was able to trigger the occurrence.  But I was able to trigger it in the low 80's.  Interestingly, if triggered (and I maintained speed and direction) I could not get it to raise above 90'.  In fact my unscientific observation is that the unwanted elevation raise seems to lose "force" the higher the altitude gets, and the lower the altitude, the higher the upward "force" when it is triggered.

This sort of confirms my suspicion that it is something with the ground facing sensors. The reason it stops is because the ground get out of range of the sensor. That is my hypothesis anyways. I do see in some previous replies folks sayin it happens much higher, but again only when apas is in bypass mode. It still is obviously a sensor issue. The drone is climbing because it is trying to avoid something based on the sensor system.

I have not been able to go out and test myself yet. Between work and family coming in to stay with us for the holidays I've just been too busy to go out and play. I do plan to test this myself soon though.

Tornado12 Posted at 12-16 08:22
This sort of confirms my suspicion that it is something with the ground facing sensors. The reason it stops is because the ground get out of range of the sensor. That is my hypothesis anyways. I do see in some previous replies folks sayin it happens much higher, but again only when apas is in bypass mode. It still is obviously a sensor issue. The drone is climbing because it is trying to avoid something based on the sensor system.

I have not been able to go out and test myself yet. Between work and family coming in to stay with us for the holidays I've just been too busy to go out and play. I do plan to test this myself soon though.

I concur with you with one addition:  I think that this is a combination of "sensor" and "code".  "Break" seems to work just fine.  I believe that the issue is manifesting when the craft software makes its decision to alter flight path to "bypass".