Last December, high altitude flight at 2200m. S mode about 4m fast-flying above the river. Slight wind and just above zero C. At the endpoint, I did yaw to the left and suddenly the drone dipped fast almost touching the water. Stick was not responsive momentarily! .300 firmware




Come on DJI this is not cool. That was so scary... Please find out wat is going on. I will PM you the email of the account so you can check the uploaded log.

See phantomhelper log here, the yaw happened after 2m13s.

UPDATE: Guys I may have uncovered the issue. Try flying at eye level in sport mode without touching the throttle stick. I was doing this repeatedly today over a shallow pool. The result is if there is a suddenly gust of wind my mini may find it difficult to hold altitude and dip. My theory is that higher airflow resulted in the barometer detecting a lowered pressure. When mini senses that and no manual inputs from the throttle stick it attempts to correct its own altitude. In my case, it dipped in an attempt to keep altitude without knowing that the pressure change is caused by a gust not an unexpected change in altitude from other factors rather than the pilot's intention. I hope I am making sense. My issue now is is that every unit's behaviour or just mine? You are welcome to repeat my test over grass or shallow water to test it.

Some after thoughts are in post #119 on the third page. I suggest you read that for more details on pressure deviation and height keeping. And certainly take home message.

Can you post your flight log - https://www.phantomhelp.com/LogViewer/upload/

Hello and good day Guorium. I am sorry to read and to know what happened to your previous flight with your DJI Mavic Mini. Since this unfortunate issue happened. Kindly please contact out DJI Support Team for further assistance at https://www.dji.com/support?site=brandsite&from=nav. We would do their best to find out the reason of the incident and then the corresponding resolution would be provided. Again I am sorry for the trouble and thank for your support.

Scary moment for sure!! Have you looked at the flight logs? Perhaps post them here if you wouldn't mind?

uuuufffffff, that was close!
Thats the problem I noticed when the drops happen, it has some delay to regain altitude, not sure if delay or the loose of power in motors...
I do the tests at 40-50m to have enough time.

DJI Stephen Posted at 1-7 07:41
Hello and good day Guorium. I am sorry to read and to know what happened to your previous flight with your DJI Mavic Mini. Since this unfortunate issue happened. Kindly please contact out DJI Support Team for further assistance at https://www.dji.com/support?site=brandsite&from=nav. We would do their best to find out the reason of the incident and then the corresponding resolution would be provided. Again I am sorry for the trouble and thank for your support.

I contacted support for the engineers to look at. Others can look at the log I uploaded to phatomhelp

Hiya,

Had a look at your log, see chart.
Cannot find the situation you decribed. (searching for left yaw, dropping heigt and no up/down stick input)
Wich part in the log is the short video ?

cheers
JJB

JJB* Posted at 1-8 00:13
Hiya,

Had a look at your log, see chart.
[Image]

Right after 2m 13s

Guorium Posted at 1-8 01:48
Right after 2m 13s

thanks,

Cannot find errors in the log.
Loosing height after yaw turn, craft doesn`t look stable, cam should be stabilized in normal conditions.
VPS just started to indicate values....

Flying low over moving water with VPS yes/no measuring is always tricky.

See the data in the chart what happend in numbers  ;-)

Cheers
JJB

Guorium Posted at 1-8 01:48
Right after 2m 13s

It’s all explained in your manual, Mavic mini is not great flying low over moving water .

JJB* Posted at 1-8 04:38
thanks,

Cannot find errors in the log.

Thanks for the analysis! Does appear that my throttle up coincided with values in ALT VPS column. What is ALT VPS? From my understanding, VPS should not mess with altitude though for altitude is determined by the barometer in the drone.

Guorium Posted at 1-8 08:59
Thanks for the analysis! Does appear that my throttle up coincided with values in ALT VPS column. What is ALT VPS? From my understanding, VPS should not mess with altitude though for altitude is determined by the barometer in the drone.

Your bottom sensor measurs height above the ground at low heights, between 0 to 6-8 * meters but the actual upper range depends on the surface structure.
* find the exact value in the specs/manual.
And yes, it does do something flying low with or without correct measuring height.  see the manual.

Sensor also used for AutoLanding protection, try this:

Hover at 5 meter
Apply quickly some RC stick down (50%) - craft drops in height
Do it again and again, see craft dropping height is steps.
At 0.5 meter craft does not go down!
Only after 100% down for some moments craft goes into AutoLanding.

DON`T do this flying low above water (AutoLanding does works too with forward speed!) .....if you not aware of this craft will go swimming..

ONLY way to stop this landing is 100% UP.

cheers
JJB

JJB* Posted at 1-8 09:40
Your bottom sensor measurs height above the ground at low heights, between 0 to 6-8 * meters but the actual upper range depends on the surface structure.
* find the exact value in the specs/manual.
And yes, it does do something flying low with or without correct measuring height.  see the manual.

I flew tello which had identical vps module. The two black spheres around the center camera are infrared transmitter and receiver. The pair works as a proximity sensor for distance less than about 6m (according to similar sensors sold online). I am aware of the landing behavior and cancellation by stick up, as I take advantage of this behavior when palm landing mavic mini.

I do not know which part of the manual you are referring to. I have read every page of it and the part describing bottom sensor function isn’t specific about interfering with altitude. It just says the sensors are not working at their best over water, moving fast and low over terrain. The bottom camera solely locks horizontal position. My drone did not go into landing mode (the app would warn me about it). It lost altitude too fast mid yaw and the infrared sensor stopped further drop by halting it over shallow water as infrared can penetrate shallow water. Should the river be deep it would have gone swimming.

Guorium Posted at 1-8 14:29
I flew tello which had identical vps module. The two black spheres around the center camera are infrared transmitter and receiver. The pair works as a proximity sensor for distance less than about 6m (according to similar sensors sold online). I am aware of the landing behavior and cancellation by stick up, as I take advantage of this behavior when palm landing mavic mini.

I do not know which part of the manual you are referring to. I have read every page of it and the part describing bottom sensor function isn’t specific about interfering with altitude. It just says the sensors are not working at their best over water, moving fast and low over terrain. The bottom camera solely locks horizontal position. My drone did not go into landing mode (the app would warn me about it). It lost altitude too fast mid yaw and the infrared sensor stopped further drop by halting it over shallow water as infrared can penetrate shallow water. Should the river be deep it would have gone swimming.

Hi Guorium,

Indeed the manual does not say anthing about bottom sensors interfere with holding height.
But somehow it does have efffect on low flying, loosing and or getting VPS height indicating does have effect on the height behavior of the craft. Just my experience.

Reading your file i think the unwanted descend was stopped by pulling up, not because of the bottom sensor. Your began the UP stick when VPS height indicates 5.1 meters, sensor acts at 0.5 meter.
Your camera wasn`t stable, wonder why ?  windy out there?  mayby the turning and pushing down the craft got the craft into false wind or so.
I cannot find a reason why it decended as it did.

BTW not many people know about that Autolanding 'thing', some crafts were lost (forward flying) because entering an autolanding without knowing it.

Happy many landings,
cheers
JJB

hallmark007 Posted at 1-8 08:29
It’s all explained in your manual, Mavic mini is not great flying low over moving water .

I would expect (and from what I remember, that's what the manual says) it would (at worst) affect the hover position. Not cause the drone to go down! Also, I don't understand why anything beneath it would affect it while having GPS lock.

Ice_2k Posted at 1-9 01:53
I would expect (and from what I remember, that's what the manual says) it would (at worst) affect the hover position. Not cause the drone to go down! Also, I don't understand why anything beneath it would affect it while having GPS lock.

Exactly my thoughts, with good GPS signal and light, a drop of more than 20-30cm is unacceptable, I have some more logs to post regarding drops. They do happen many times.

JJB* Posted at 1-9 00:32
Hi Guorium,

Indeed the manual does not say anthing about bottom sensors interfere with holding height.

I was doing yaw left and descent at a constant rate. When I reached the height I like I stopped descend and kept yaw left. Then the drone spontaneously dropped fast. I immediately stick up but it did nothing. It kept dropping at that rate until the lowest point and I simultaneously released the stick up there. You can see this in the log record.

Camera isn’t stable because I tried to fly low and fast with FPV mode. That is not an issue.

InspektorGadjet Posted at 1-9 02:06
Exactly my thoughts, with good GPS signal and light, a drop of more than 20-30cm is unacceptable, I have some more logs to post regarding drops. They do happen many times.

But GPS is not used to maintain height, even in ATTI mode the height is stabilized by the baro sensor.

if this sensor, this measuremnt is disturbed by airflow for instance, it will believe its flying higher or at a lower altitide, thus correcting.

When i fly at 6 meters, from hover to full speed in Sport mode my MM starts to climb a little.

cheers
JJB

JJB* Posted at 1-9 03:58
But GPS is not used to maintain height, even in ATTI mode the height is stabilized by the baro sensor.

if this sensor, this measuremnt is disturbed by airflow for instance, it will believe its flying higher or at a lower altitide, thus correcting.

Absolutely, GPS should lock the position and barometric sensor the height however baro height is being affected by something else.

Guorium Posted at 1-9 02:32
I was doing yaw left and descent at a constant rate. When I reached the height I like I stopped descend and kept yaw left. Then the drone spontaneously dropped fast. I immediately stick up but it did nothing. It kept dropping at that rate until the lowest point and I simultaneously released the stick up there. You can see this in the log record.

Camera isn’t stable because I tried to fly low and fast with FPV mode. That is not an issue.

oke, guess i see other thing in logs  ;-)

But had a further look at this log.
See my chart, wich is just a graphic representation of your log.
See the blue lines ; the upper blue line is the Altitude, where it is red craft is ascending/descending without RC input on up/down.
At ID 1750 it started to climb, up to 30 meters!
But in the middle blue line ; no UP rc input.
Lower blue line the Vertical speed, wich is zero....

So did it climb to 30 meters??  guess not, wich means that the baro info is not correct, thus explaining the drop before.

cheers
JJB

I wonder if all the hard investigation work from JJB is getting to the DJI technical department or if we’re just wasting our time

Ice_2k Posted at 1-9 01:53
I would expect (and from what I remember, that's what the manual says) it would (at worst) affect the hover position. Not cause the drone to go down! Also, I don't understand why anything beneath it would affect it while having GPS lock.

That’s not correct it’s the readjustment because of bad reading of vps, there have been cases many of flying low over flowing water vps reports sudden landing because it thinks it’s below 0.5m so auto landing is initiated. It is recommended in some dji manuals not to fly below 15m over fast flowing water .

Ice_2k Posted at 1-9 04:33
I wonder if all the hard investigation work from JJB is getting to the DJI technical department or if we’re just wasting our time

Most of what’s said on the forum will be noted, but it maybe slow for dji to act on and in some cases nothing will be admitted but you might see bug fix sort out the problem.

Ice_2k Posted at 1-9 04:33
I wonder if all the hard investigation work from JJB is getting to the DJI technical department or if we’re just wasting our time

Lets hope that it is getting "cascaded" to the right department within DJI.  If not, they are missing out on some great investigative work by JJB.

hallmark007 Posted at 1-9 04:59
Most of what’s said on the forum will be noted, but it maybe slow for dji to act on and in some cases nothing will be admitted but you might see bug fix sort out the problem.

But still, Landing is prompted via both audio and visual cue. I had none of those.

DowntownRDB Posted at 1-9 05:03
Lets hope that it is getting "cascaded" to the right department within DJI.  If not, they are missing out on some great investigative work by JJB.

They will see the post. It takes time to get to the right place in DJI. Also I have a case number from DJI on this issue so they are looking at my flight log.

JJB* Posted at 1-9 04:16
oke, guess i see other thing in logs  ;-)

But had a further look at this log.

Very impressive analysis. It is interesting to see vps altitude functions only less than 10m. In the untrimmed video I see it indeed went up to 30m near the end of the flight. I was there to check out a waterfall above the treeline.  So the barometer was ok. I can see stick up in the dji fly app during that ascend so I am not sure why your analysis shows no input. Additionally, I wonder if you analyzed the entire log? Because at the end the altitude is 20m. I cannot land 20m above my takeoff point nor start 10m above.

Guorium Posted at 1-9 19:25
Very impressive analysis. It is interesting to see vps altitude functions only less than 10m. In the untrimmed video I see it indeed went up to 30m near the end of the flight. I was there to check out a waterfall above the treeline.  So the barometer was ok. I can see stick up in the dji fly app during that ascend so I am not sure why your analysis shows no input. Additionally, I wonder if you analyzed the entire log? Because at the end the altitude is 20m. I cannot land 20m above my takeoff point nor start 10m above.

Hi,

Yes, look at the whole flight.

This chart is now the whole flight, so you landed at zero baro height.
Chart in post #20 only part of your flight, to make that unwanted descend more clear.

Weird, in the converted flightlog no up steering written during the climb to 30 meters.

cheers
JJB

Guorium Posted at 1-9 19:16
But still, Landing is prompted via both audio and visual cue. I had none of those.

I was merely making the point what can happen with vps over moving water, I’m sure your now aware that there is risk with vps of flying within range of moving water and it’s pretty similar at times with grass.

JJB* Posted at 1-8 04:38
thanks,

Cannot find errors in the log.

Because when the Mini was made, the actions it does is thinking there is a normal situation w/ no error because some of the parts weren't built right.  Like the Boeing MCAS error where the plane thought it was stalling but it wasn't eventually bringing down the aircraft to the ground completely.  Of course, if the MCAS detected an error, it would not have brought the aircraft down into the ground.

Nobody most likely here knows what is truly going wrong.  DJI probably has no idea either.  This is why when you put stuff together you need to know what everything does and their reliability because when you start putting stuff together w/o proper analysis for whatever reason, you will get situations like this in mass with no idea what is wrong.

Probably some components made during prototype were done right but at production, it was built incorrectly but passed QA for the sake of profits and bonuses.  People can tell you things are ok and great but when the science shows something is wrong and you don't know science, you will think its ok and great until things go totally wrong.

The higher-end DJI drones are reliable and built right.  It's like another planet compared to bad drones.

hallmark007 Posted at 1-10 01:06
I was merely making the point what can happen with vps over moving water, I’m sure your now aware that there is risk with vps of flying within range of moving water and it’s pretty similar at times with grass.

I cannot say you are wrong in that regard. I hovered a tello close to some treetop before and the vps got confused. As a result, the tello started bouncing up and down and yawing randomly. No issue when I am further above that tree. Confused vps could cause erratic behavior in all directions. That is not good, vps should not cause vertical deviation. That is the sole job of barometer. The fact that my stick up could not correct the fall was also very dangerous. Tello was at least able to be corrected by stick up. Mini was a good 4m above that water with clear pattern below and stationary.  I have flown fast and low over the same river and altitude held fine. So I am still 100% sure it was caused by confused VPS.

What I was disagreeing with is that it could not have been the autolanding causing the drop or the app would prompt that.

JJB* Posted at 1-10 00:46
Hi,

Yes, look at the whole flight.
[Image]

DJI replied me pretty fast. Their conclusion is that the barometer malfunctioned and they will offer a free checkup/repair for me.

You think the barometer data acted strange during that incident? I really doubt it was misbehaving. The barometer altitude correctly reflected the abnormally fast descent speed (steeper blue trace gradient) shortly before ID 1400. The vps height was spotty in that window of time. That could mean vps was not functioning well? Can see before ID1400 my full stick up was not able to correct the fast fall.

I think that the baro sensor is correct, if it was correct during that unwanted descend i don`t know. Airflow around the sensor can disturb correct measurements.

In my chart see ID 1372, 2m17.1s : no more down input, but craft is still descending (see V-z)
V-z zero but when VPS height starts to indicate ; V-Z in minus values again.
Pulling up and time needed to stop the descend 0.8 second. I this too  much?  normally not noticed when flying at higher altitudes.

Ofcourse i don`t knwo the reason for dropping, but imo has something to do with suddenly and not correct measurement of VPS  bottom sensor ; not a fault, moving water is killing for accurate measuring.

So i think that your fully UP stick did the job, drone stopped down move and went up again. It did react OK imo, but for you to slow seeing the craft almost going down!

Its up to you to have a check done by DJI or not.

DJI text about Vision is, for me, confusing.
Vision is added to the GPS to maintain better hover performace, ofcourse withing vision range.
In their text they write about 'enabled', i think they mean to say when the system is used as its always ON.

I do believe that Vision is also used for height hold, in addition to baro sensor. Vision as fine tuning for baro info??
If not   why all the warnings and exceptions for the vision in the manual?  If i fly with good GPS lock my position flying low altitude is no problem, height is kept steady by baro sensor. But when VPS start to measure (thus for positioning and height) things can go wrong....

Hope that DJI will explain better in their manuals, how the systems interact wich each other.

cheers
JJB

I hope there will be an update that will allow us to disable the vps system.

JJB* Posted at 1-10 04:29
I think that the baro sensor is correct, if it was correct during that unwanted descend i don`t know. Airflow around the sensor can disturb correct measurements.

In my chart see ID 1372, 2m17.1s : no more down input, but craft is still descending (see V-z)

DJI definitely needs to outline better how GPS and VPS work together...

After the blue cell Vz turned positive yet altitude kept going down. That is not very consistent. VPS alt looks ok though. I cannot for sure say if the sitck up did nothing or something now that I carefully think about all the evidence I have. What I can be sure of is, if the vps is to be trusted my min atl is still 1.1m which isn't low enough to automatically halt the drone for landing. I cannot say from the video if it was indeed 1m above the river bed.  Maybe it was 1m above. But still there is no solid evidence leading to the cause of the altitude drop. I flew it 30cm above faux grass pretty fast in P mode and observed no altutide dip. Will try again with sport mode.

Seems like covering VPS sensor might help. Just we should be careful while landing and flying low

Eka Posted at 1-10 05:38
Seems like covering VPS sensor might help. Just we should be careful while landing and flying low

Can never be too careful haha

Guorium Posted at 1-10 06:44
Can never be too careful haha

Reminder for myself.

Covering the sensors will not deactivate them it will be interpreted as if it were dark , I think.

minilou Posted at 1-10 07:57
Covering the sensors will not deactivate them it will be interpreted as if it were dark , I think.

Yes, but after you overide it and start flying, it won't be able to use them anyway. Looking at all those posts about landing in water and general problems when flying over moving water, to me, seems like they  bring more bad than good. Maybe they are useful to beginners, but more advanced pilots might benefit more from disabling them. When that sensor feels ground and you input longer left stick down, AC goes into landing procedure. Then when going further down, more chance to lose signal and puting left stick up might not help. Now combine that with posible wrong reading from sensor above water and there you have a disaster.