Help me decipher this P4PV2.0 flyaway. Drone op with 100+ hrs of flight needs your assistance!

Joined
Feb 6, 2018
Messages
13
Reaction score
10
Age
41
Hey Everyone! I know, it's another post asking for your help to figure out why a drone crashed. Maybe the slight difference is I have a decent amount of experience under my belt so this isn't your typical first flight gone wrong type of post. The short version of the story is I took off, flew straight up to an altitude of 30 ft and a few seconds later, without touching the sticks, the drone flew forward on its own, crashing into a building (and then to the ground). I've attached a few files that may help decipher what happened. The "flight midpoint" screenshot was from right before the drone took off on its own. The "flight end" is the last recorded info. The "GPS Signal Weak" notification only appeared on that very last frame. The weird thing is the flightpath shows a very wonky path whereas in real life, the drone only flew up (under my power) and then forward (on its own).

Now for the longer version of the story:
I was hired to shoot some new civic center buildings which consists of two long buildings with a courtyard between them. I started with a homepoint across the street and got several passes through the courtyard (where my drone would later crash). A battery and a half later, I brought the drone inside the courtyard and powered it up. I got a compass error for a brief second and was about to calibrate it but the warning went away and all the criteria looked good for taking off. My new homepoint was recorded, I had enough satellites and took off, flying up 30 ft. A few seconds later the drone flew forward on its own. In my panic, I pulled down on my left stick and in doing so, caused the drone to rotate 150 degrees which I saw it do before hitting the building. According to my flight logs, I pulled up on my left stick (not down) which is why it climbed to 60 ft before crashing. My Visual Observer also said it looked like the drone crashed from 60 ft up.

Replaying the crash in my head, I was positive that I pulled down on my left stick and realized what I should have done was pull down on my right stick and switched the remote from P to A mode. Unfortunately, I only had a couple seconds to react.

Even if I'm completely wrong and I really did pull up on my left stick, it doesn't explain why the drone flew forward on its own (you can see in the logs the sticks aren't being touched when it flies away from me) and why the recorded flightpath is all over the place when it only flew up and then forward.

A couple hours later I had my back-up drone in the air (launching from homepoints away from the crash site) and ran into several compass and lack of satellite errors. I know that there is a lot of RF in dense downtowns (and who knows what underground) plus blockage from all the tall buildings. I kept picking up and trying to find new places I could launch from (and recalibrated my compass often).

As mentioned, I've flown well over 100 hours and I've never had a drone take off on its own before. I've also flown in similar type areas where interference is occasionally a problem but never caused a flyaway. This is scary stuff if things like this still happens with the newer drones. Thank goodness I insisted on doing this flight early on a Sunday morning when no one would be around.

Some other side notes: One block away was a prison which the DJI app wouldn't let me fly over when I was at a higher altitude. Yet this was the direction my drone headed on its own.

And if anyone was wondering, yes, I'm Part 107 certified (since the first day it was offered), I have liability (but not hull) insurance, I had airspace authorization and even the Chief of Police was aware of my flight that happening that morning. The drone was less than 6 months old. I've flown with a P3P the previous 3+ years.

My biggest question is, "Why did the drone fly away on its own?" with the follow up, "What could I have done differently?"
Thank you in advance for your help!

Mark
 

Attachments

Fly Dawg

"Electricity is really just organized lightning"
Premium Pilot
Joined
Apr 29, 2016
Messages
4,362
Reaction score
1,919
Location
Watkinsville, Georgia
@makneyse Can you confirm this is the correct log? This log begins at 15min in.....something is definately not right here. The .DAT file shows nothing. See here: Flight View
 
  • Like
Reactions: makneyse
Joined
Feb 6, 2018
Messages
13
Reaction score
10
Age
41
@makneyse Can you confirm this is the correct log? This log begins at 15min in.....something is definately not right here. The .DAT file shows nothing. See here: Flight View
Thank you for the reply. Yes, that is the correct log. I had flown half a battery's worth a block away and then picked up the drone and walked it to this new launch point. I thought I had powered down in-between. So that's probably why it's 15 minutes in. I'm not sure why the .dat file isn't working...
 

Fly Dawg

"Electricity is really just organized lightning"
Premium Pilot
Joined
Apr 29, 2016
Messages
4,362
Reaction score
1,919
Location
Watkinsville, Georgia
  • Like
Reactions: makneyse
Joined
Jun 24, 2015
Messages
1,390
Reaction score
520
Age
71
Location
Arroyo Grande, California
Hey Everyone! I know, it's another post asking for your help to figure out why a drone crashed. Maybe the slight difference is I have a decent amount of experience under my belt so this isn't your typical first flight gone wrong type of post. The short version of the story is I took off, flew straight up to an altitude of 30 ft and a few seconds later, without touching the sticks, the drone flew forward on its own, crashing into a building (and then to the ground). I've attached a few files that may help decipher what happened. The "flight midpoint" screenshot was from right before the drone took off on its own. The "flight end" is the last recorded info. The "GPS Signal Weak" notification only appeared on that very last frame. The weird thing is the flightpath shows a very wonky path whereas in real life, the drone only flew up (under my power) and then forward (on its own).

Now for the longer version of the story:
I was hired to shoot some new civic center buildings which consists of two long buildings with a courtyard between them. I started with a homepoint across the street and got several passes through the courtyard (where my drone would later crash). A battery and a half later, I brought the drone inside the courtyard and powered it up. I got a compass error for a brief second and was about to calibrate it but the warning went away and all the criteria looked good for taking off. My new homepoint was recorded, I had enough satellites and took off, flying up 30 ft. A few seconds later the drone flew forward on its own. In my panic, I pulled down on my left stick and in doing so, caused the drone to rotate 150 degrees which I saw it do before hitting the building. According to my flight logs, I pulled up on my left stick (not down) which is why it climbed to 60 ft before crashing. My Visual Observer also said it looked like the drone crashed from 60 ft up.

Replaying the crash in my head, I was positive that I pulled down on my left stick and realized what I should have done was pull down on my right stick and switched the remote from P to A mode. Unfortunately, I only had a couple seconds to react.

Even if I'm completely wrong and I really did pull up on my left stick, it doesn't explain why the drone flew forward on its own (you can see in the logs the sticks aren't being touched when it flies away from me) and why the recorded flightpath is all over the place when it only flew up and then forward.

A couple hours later I had my back-up drone in the air (launching from homepoints away from the crash site) and ran into several compass and lack of satellite errors. I know that there is a lot of RF in dense downtowns (and who knows what underground) plus blockage from all the tall buildings. I kept picking up and trying to find new places I could launch from (and recalibrated my compass often).

As mentioned, I've flown well over 100 hours and I've never had a drone take off on its own before. I've also flown in similar type areas where interference is occasionally a problem but never caused a flyaway. This is scary stuff if things like this still happens with the newer drones. Thank goodness I insisted on doing this flight early on a Sunday morning when no one would be around.

Some other side notes: One block away was a prison which the DJI app wouldn't let me fly over when I was at a higher altitude. Yet this was the direction my drone headed on its own.

And if anyone was wondering, yes, I'm Part 107 certified (since the first day it was offered), I have liability (but not hull) insurance, I had airspace authorization and even the Chief of Police was aware of my flight that happening that morning. The drone was less than 6 months old. I've flown with a P3P the previous 3+ years.

My biggest question is, "Why did the drone fly away on its own?" with the follow up, "What could I have done differently?"
Thank you in advance for your help!

Mark
The EXPORT_FILE_2019-08-13_11-11-28.DAT contains FLY105.DAT but we need FLY104.DAT. Can you try again with DJI Assisant 2?

From your discourse I suspect that the Yaw value somehow got set to the wrong value. Possibly the courtyard you referred to was geomagnetically distorted. Hard to tell without the FLY104.DAT.

Which direction was the P4PV2 pointed? The .txt has it pointing north. I suspect that's not right.
1565735408489.png

The green line emanating from the red A depicts the orientation of the P4PV2. The light green curve is the trajectory derived from the IMU sensor fusion algorithm - quite different than the actual trajectory shown in red.

The numSats was 9 or 10 which is marginal. But, navHealth was 4, which is adequate, until the FC lowered it to 3 when it realized the position computed using the IMU data didn't match the GPS data.

Good luck with DJI Assistant 2 :) If that hurts too much you could try retrieving the .DAT from the tablet.
The one you want ends with FLY104.DAT. There may be more than one.
 

Meta4

Moderator
Staff Member
Premium Pilot
Joined
Jan 8, 2014
Messages
13,776
Reaction score
7,799
I got a compass error for a brief second and was about to calibrate it but the warning went away and all the criteria looked good for taking off.
...
I know that there is a lot of RF in dense downtowns (and who knows what underground) plus blockage from all the tall buildings. I kept picking up and trying to find new places I could launch from (and recalibrated my compass often).
My biggest question is, "Why did the drone fly away on its own?" with the follow up, "What could I have done differently?"
You could get a better understanding of what a compass error is and what compass calibration actually does (and what it doesn't do).
A compass error is not telling you about a fault in the compass that needs to be fixed.
It's a good compass warning you of a magnetic problem in your launch area - there's too much steel too close to the compass.
No amount of compass recalibrating will ever fix that problem.
The only correct action is to move the compass further away from the problem.

In your downtown environment, the biggest worry is all the steel reinforcing in concrete and launching from reinforced concrete surfaces is asking for trouble.
If you have to launch from concrete on a regular basis, you should get a a small launch pad - an upturned plastic bucket or a wooden table to get your compass further from the steel.

If you tried to launch from bare steel, the compass would probably prevent you but the problem comes when the steel is close enough to cause local magnetic distortion but not big enough or close enough to cause a compass error.
You can confirm that the drone has properly initialised by checking the direction your drone icon on the map or radar display is facing the same compass direction as the actual drone is facing.
 
Joined
Feb 6, 2018
Messages
13
Reaction score
10
Age
41
Thank you for taking the time to help. I was able to retrieve FLY104.DAT but according to the time stamp of the file, it is associated with the flight before my crash. I tried to upload it here but it didn't work. Maybe because it's too large (it's 10.185 Kilobytes)? Is there another way to get it to you? It wouldn't let me send it in a private message either.

The drone was turned on facing north and then flew forward (north) on its own. When I hit down on the left stick, I caused it to yaw to its right 150 degrees so it was almost facing me (towards the south) when it hit the building.

Yeah, Meta4 what you're saying makes sense. I flew over the same area earlier (like 100 ft up) and didn't have any problems. But when I took off from that part of the courtyard (with who knows what under the concrete) it affected my drone a few seconds into my flight. I've heard of similar issues launching from a metal boat. I didn't realize that just a few feet between the drone and the ground might help avoid a repeat of my situation. Thank you for the advice.

Watching the flight record/data in the DJI GO 4 app, the direction of the drone looks mostly correct but the direction it travels is the opposite of what really happened.
 

Attachments

Joined
Jun 24, 2015
Messages
1,390
Reaction score
520
Age
71
Location
Arroyo Grande, California
Thank you for taking the time to help. I was able to retrieve FLY104.DAT but according to the time stamp of the file, it is associated with the flight before my crash. I tried to upload it here but it didn't work. Maybe because it's too large (it's 10.185 Kilobytes)? Is there another way to get it to you? It wouldn't let me send it in a private message either.

The drone was turned on facing north and then flew forward (north) on its own. When I hit down on the left stick, I caused it to yaw to its right 150 degrees so it was almost facing me (towards the south) when it hit the building.

Yeah, Meta4 what you're saying makes sense. I flew over the same area earlier (like 100 ft up) and didn't have any problems. But when I took off from that part of the courtyard (with who knows what under the concrete) it affected my drone a few seconds into my flight. I've heard of similar issues launching from a metal boat. I didn't realize that just a few feet between the drone and the ground might help avoid a repeat of my situation. Thank you for the advice.

Watching the flight record/data in the DJI GO 4 app, the direction of the drone looks mostly correct but the direction it travels is the opposite of what really happened.
FLY104.DAT is the one we need. I expect it will be much larger than 10,185 KB. You'll need to use a public sharing site like Dropbox and then provide a link to the uploaded FLY104.DAT. FLY104.DAT will have started recording when the battery was turned on coincident with the beginning of the flight before the crash.

According to the .txt the throttle was increased (left stick up not down). The CW Yaw was caused by rudder input.
1565769448733.png


Can you confirm you were south of the AC? The tablet was reporting that it was north of the AC.
1565769536966.png
 
Last edited:
  • Like
Reactions: makneyse
Joined
Mar 24, 2019
Messages
19
Reaction score
9
Age
59
I too have had my P4p take off wildly twice now but was able to recover somehow manually and land not crash. I explained to DJI but they weren’t concerned and I never knew the forum members had such knowledge or details on tracking it. It happened once at the beach and again just two days ago in a remote mountain area.
 

Meta4

Moderator
Staff Member
Premium Pilot
Joined
Jan 8, 2014
Messages
13,776
Reaction score
7,799
It happened once at the beach and again just two days ago in a remote mountain area.
If you want any help working out what happened ...
Go to DJI Flight Log Viewer | Phantom Help
Follow the instructions there to upload your flight record from your phone or tablet.
That will give you a detailed report on the flight data.
Come back and post a link to the report it provides and someone might be able to analyse it and give you an understanding of the cause of the incident.

ps ... post your data in a new thread.
 
Joined
Feb 6, 2018
Messages
13
Reaction score
10
Age
41
Here's the link to FLY104.DAT. I was mistaken and the 10,000 KB file was FLY105.DAT (which was a very short flight - and crash).

I (and my RC) were south of the AC just a few feet. Then the AC flew north, away from me (and the RC).

Could the interference have messed up the RC communications with the drone?
 
Last edited:

Fly Dawg

"Electricity is really just organized lightning"
Premium Pilot
Joined
Apr 29, 2016
Messages
4,362
Reaction score
1,919
Location
Watkinsville, Georgia
It appears you had a compass issue. @BudWalker can give you the detailed analysis.

1565804829180.png
 
Joined
Jun 24, 2015
Messages
1,390
Reaction score
520
Age
71
Location
Arroyo Grande, California
It appears you had a compass issue. @BudWalker can give you the detailed analysis.

View attachment 113774
I can see why you're thinking compass issue. But, I'm not so sure.

The Yaw/magYaw separation was about 12° for most of the .DAT which is correct since the geoDeclination was 12°. This included the first flight followed by the pilot carrying the P4PV2 to the next launch site. From the accelerometer (green) and gyro (red) data we can see when that when the P4PV2 was placed on the ground at 1053 secs the Yaw/magYaw separation (blue) became 65°.
1565827317952.png


Normally, the FC would try to reconcile this Yaw/magYaw separation by changing the Yaw to align with magYaw. Had this happened the Yaw would have been compromised in the following 40 secs when launch occurred. This would have lead to your usual launch-from-geomagnetically-distorted-launch-site-causes-flyaway incident. But, the reconciliation didn't take place and as soon as the P4PV2 gained about 0.5 meters altitude the P4PV2 was clear of the geomagnetic distortion and the Yaw/magYaw separation returned to the correct 12°.
1565827734502.png


Why didn't the FC try to reconcile the Yaw/magYaw separation? This is the best part. I think it's because the distortion was so strong that the FC decided to ignore it and not attempt a reconciliation. This from the eventLog stream
1055.122 : 52424 [L-FDI]NS(0) COMPASS(1): fault on , noise
1055.561 : 52445 [L-GEO][record_flysafe_info]drone relation NFZ_UNKNOWN => OUT_NFZ|
1057.162 : 52522 [L-FDI]NS(0) COMPASS(1): fault on , interfere
1058.791 : 52600 [L-FMU/MOTOR]mag_need_action: cali mag

The question remains - why did the P4PV2 move north but the data is saying it moved south. I suspect it's because the GPS data was compromised. The DOP was OK and the FC thought the solution it had for position was good (navHealth ==4). But, this I'm not sure about, the reflections off the buildings could have introduced position error without increasing the DOP. Maybe @sar104 can take a look and see if there is a way to tell if this happened.
 
Joined
Feb 6, 2018
Messages
13
Reaction score
10
Age
41
Wow. Thank you for the detailed breakdown. I didn´t realize the yaw was affected.

Side note: I´m going to amend my original post a little even though it is probably just going to muck things up.
What I remember happening is I took off, hovered, and then the drone flew a straight line forward into the building. I thought without a doubt I pulled down on my left stick (and possibly slightly to the side), causing the drone to yaw but I didn´t see it change in altitude any.

I slightly changed my story (the one you see in my original post) after reading the flight logs. They said I hovered at 30 ft and when the drone started to fly forward, I pulled my left stick up rising the drone to 60 ft before it hit the building.

I thought for sure I didn´t pull the left stick up but since that´s what the logs said happened I figured maybe in my panic I didn´t remember correctly. But after talking with my visual observer, he said the drone was already at 60 ft and flew a straight line into the building (not changing altitude).

So, I don´t know what that means about my story vs. the logs but obviously something was not right courtesy of some major interference.

Again, thank you everyone for sorting through the files to try to figure out what the heck happened. I don´t know if the interference compromised the logs saying I pulled the sticks one way and that the drone was at a different altitude but at least my V.O.´s story was the same as what I remembered in relation to the drone´s altitude and its flight path.
 

Fly Dawg

"Electricity is really just organized lightning"
Premium Pilot
Joined
Apr 29, 2016
Messages
4,362
Reaction score
1,919
Location
Watkinsville, Georgia
But after talking with my visual observer, he said the drone was already at 60 ft and flew a straight line into the building (not changing altitude).

So, I don´t know what that means about my story vs. the logs but obviously something was not right courtesy of some major interference.
It is true that there was indeed some interference, however, for what it is worth, both the .txt and .dat logs match up, as far as your control inputs and altitude goes. You can see in both of these where the aircraft was at approx 30ft in altitude and the throttle was moved, ( Actually all controls were active ) climbing to 60ft at the end of both logs. This, I think confirms the original climb to 60 ft. Not that it answers any questions other than the re-call of the incident.

TEXT LOG PLOT


Capture2_txt log.PNG


DAT LOG PLOT



Capture.PNG
 
  • Like
Reactions: BudWalker
Joined
Jun 24, 2015
Messages
1,390
Reaction score
520
Age
71
Location
Arroyo Grande, California
Wow. Thank you for the detailed breakdown. I didn´t realize the yaw was affected.

Side note: I´m going to amend my original post a little even though it is probably just going to muck things up.
What I remember happening is I took off, hovered, and then the drone flew a straight line forward into the building. I thought without a doubt I pulled down on my left stick (and possibly slightly to the side), causing the drone to yaw but I didn´t see it change in altitude any.

I slightly changed my story (the one you see in my original post) after reading the flight logs. They said I hovered at 30 ft and when the drone started to fly forward, I pulled my left stick up rising the drone to 60 ft before it hit the building.

I thought for sure I didn´t pull the left stick up but since that´s what the logs said happened I figured maybe in my panic I didn´t remember correctly. But after talking with my visual observer, he said the drone was already at 60 ft and flew a straight line into the building (not changing altitude).

So, I don´t know what that means about my story vs. the logs but obviously something was not right courtesy of some major interference.

Again, thank you everyone for sorting through the files to try to figure out what the heck happened. I don´t know if the interference compromised the logs saying I pulled the sticks one way and that the drone was at a different altitude but at least my V.O.´s story was the same as what I remembered in relation to the drone´s altitude and its flight path.
Some clarification.

Yaw was never affected but magYaw was. Yaw is a value computed by the Flight Controller that represents what the FC thinks is the heading. MagYaw is computed from magnetometer and attitude data - it's independent of the Yaw value. Yaw is used by the FC for navigational purposes. Yaw is subject to small random errors and left uncorrected will slowly drift from the correct value. MagYaw is used by the FC to make the necessary minor corrections to the Yaw value.

If the P4PV2 had been powered up at the launch site Yaw would have been initialized to the incorrect magYaw value. This would have lead to erratic flight and probably a fly away. But, it wasn't powered up at the launch site and Yaw wasn't then set to the incorrect magYaw value.

Normally what happens is that the FC tries to fix a Yaw/magYaw separation occurring before flight by changing the Yaw value to agree with the magYaw value. In this instance the Yaw value would have been set to an incorrect magYaw value. Fortunately, that didn't happen here and the Yaw value wasn't compromised. At launch Yaw was correct. The erratic flight was not caused by the fact that the launch site was geomagnetically distorted.

The reason I suspect that the cause was due to poor GPS data can be seen here.
1565877806502.png


The blue plot is the Latitude computed by the FC and shows the P4PV2 moving south. This in incorrect since the P4PV2 was moving north. But this isn't totally conclusive since the FC uses other data besides actual GPS data to compute this value for Latitude. More significantly is that the raw GPS data (red) also shows the P4PV2 moving south.
 

New Threads

Forum statistics

Threads
135,374
Messages
1,400,694
Members
98,606
Latest member
AceDoodle