Hello everyone! My Phantom 3 Pro crashed unexpectedly at a height of 70m, I'd like anyone who can and is kind enough to see the flight record to try to figure out what happened so he dropped. I was just flying at a height of 70m when he just started falling into free fall.

https://www.phantomhelp.com/LogViewer/92DJ5VM54E794L6WPAUO/


Sorry for your loss.


Did you put it in ATTI yourself?

Did you recover the AC with all four props still on the AC?  

Did you notice anything striking it,  like maybe a bird?

Hello and good day Brunorogi. I'm really sorry to know what happened to your previous flight with your DJI Phnatom 3 Pro. I would recommend you contact our DJI support team to start up a ticket for the said drone at https://www.dji.com/support?site=brandsite&from=nav . We have the professional team who would do their best to fix and find out the reason of the incident and then the corresponding resolution would be provided. Please note that DJI will not perform data analysis for cases occurring after the expiry of the warranty period; otherwise, valid charges will apply. Thank for your support.

Everything looks good up until 6 min 12 sec.  Then it just rolled over and tumbled to the ground.  It doesn't appear to have tried to right itself if it was bumped by a bird, however a birdstrike might have caused even more severe damage than just being bumped off-axis momentarily.  When you recovered it, were the props attached and intact, or missing or broken?  From the error messages it does not look like a motor has shut down, but that is possible.  A thrown propeller would account for the speed error messages -- the aircraft would still be spinning all motors trying to recover but without being able to push air from all four.  That can happen if a prop isn't on tightly and you quickly go from full stick forward to full rear often in the flight, it can work itself off.  I don't see any "acrobatic" flying in your log though, but that's not the only reason for it coming off.  Might need to examine your full log and see if there are more clues in what the individual motors are doing.  You would need to retrieve the DAT file for the flight from the aircraft's internal flight recorder (you will be looking for a file called FLY413.DAT for this flight).  All I can see from the transmitted flight log is that the notEnoughForce flag is set at 372 seconds, which indicates the aircraft detected it was unable to deliver sufficient thrust for flight.

Mark The Droner Posted at 5-20 06:11
https://www.phantomhelp.com/LogViewer/92DJ5VM54E794L6WPAUO/

Yes, it was me who put it in ATTI mode in flight, is not this recommended?
I recovered the well damaged drone because of the fall, but the 4 propellers were there.

ALABAMA Posted at 5-20 06:45
Did you notice anything striking it,  like maybe a bird?

I saw something black very fast, but it may just be my impression, it's very difficult to be sure.

endotherm Posted at 5-20 11:20
Everything looks good up until 6 min 12 sec.  Then it just rolled over and tumbled to the ground.  It doesn't appear to have tried to right itself if it was bumped by a bird, however a birdstrike might have caused even more severe damage than just being bumped off-axis momentarily.  When you recovered it, were the props attached and intact, or missing or broken?  From the error messages it does not look like a motor has shut down, but that is possible.  A thrown propeller would account for the speed error messages -- the aircraft would still be spinning all motors trying to recover but without being able to push air from all four.  That can happen if a prop isn't on tightly and you quickly go from full stick forward to full rear often in the flight, it can work itself off.  I don't see any "acrobatic" flying in your log though, but that's not the only reason for it coming off.  Might need to examine your full log and see if there are more clues in what the individual motors are doing.  You would need to retrieve the DAT file for the flight from the aircraft's internal flight recorder (you will be looking for a file called FLY413.DAT for this flight).  All I can see from the transmitted flight log is that the notEnoughForce flag is set at 372 seconds, which indicates the aircraft detected it was unable to deliver sufficient thrust for flight.

[view_image]

I'll tell you the truth, when I got the drone in my hands I was missing one of the propellers, but this propeller that escaped was at most 2m away from the fallen drone and got stuck in a tree.
I was also thinking that the propeller escaped in flight, but on second thought, what is the chance of the problem happening 70m high and 20m away from the drop site and the propeller falling almost to the side of the drone?
What I thought was this, in the strong impact with the tree the propeller was arracanda of the drone, simple as well.
And from that it was generated the message of "Speed ​​Error"
After also analyzing the flight data a little more I realized that at the time of the failure, there was a lot of variation of the angle of the aircraft in very little time, but much variation even in only 100mS of range.
This led me to believe that there was a failure in some of the sensors, especially in the compass, since physically I did not think that this variation of angle was possible.
Where can I find this DAT file? In the app data of my smartphone?

DJI Stephen Posted at 5-20 08:12
Hello and good day Brunorogi. I'm really sorry to know what happened to your previous flight with your DJI Phnatom 3 Pro. I would recommend you contact our DJI support team to start up a ticket for the said drone at https://www.dji.com/support?site=brandsite&from=nav . We have the professional team who would do their best to fix and find out the reason of the incident and then the corresponding resolution would be provided. Please note that DJI will not perform data analysis for cases occurring after the expiry of the warranty period; otherwise, valid charges will apply. Thank for your support.

Thanks, I'm getting in touch with DJI.

Look at the size of this angle variation in 100mS, is that physically possible? That's why I think the problem was the compass.

Compass has nowt to do with the angle of the drone only the direction its pointing to, you seeing that angle, a lost prop can cause the drone to do this

Bashy Posted at 5-22 19:19
Compass has nowt to do with the angle of the drone only the direction its pointing to, you seeing that angle, a lost prop can cause the drone to do this

And a fault in the gyroscope?
It began to fall very high, the propeller that was loose should be left on the impact with the tree because it fell very close to the drone and what happened was very high and far for this to occur, not that this is impossible, but it is quite unlikely not ?

Brunorogi Posted at 5-22 04:33
I'll tell you the truth, when I got the drone in my hands I was missing one of the propellers, but this propeller that escaped was at most 2m away from the fallen drone and got stuck in a tree.
I was also thinking that the propeller escaped in flight, but on second thought, what is the chance of the problem happening 70m high and 20m away from the drop site and the propeller falling almost to the side of the drone?
What I thought was this, in the strong impact with the tree the propeller was arracanda of the drone, simple as well.

It is improbable, but it could quite easily have come off in flight and ended up landing close to the main wreckage.  The design of the P3 series props is that, as long as the motor is spinning, it tends to thread itself ON to the prop hub.  The prop motor won't run backwards, so it shouldn't be able to unthread itself.  However if it wasn't on tight, the normal acceleration could be enough to turn the prop in the correct direction, but not aggressively enough that it "self-tightens" itself into the hub.  Therefore, normal flight can be achieved, but it is a delicate balance.  If you slow suddenly or abruptly return the sticks to centre, the momentum of the prop is enough to keep it in rotation.  The motor is slowed or stopped in relation to this momentum.  That can be enough to have a counter rotational effect on the prop and it unscrews itself.  Now you have two separate objects tumbling in the air, without propulsion.  They will continue on a trajectory consistent with their previous momentum and fall to earth at the same speed.  Assuming they weren't acted upon by a lateral wind, the lighter prop could conceivably end up very close to the heavier body.  Even when the space shuttle broke up on re-entry, bits that came off the main craft at the same time ended up very close to each other on the ground.  It has more to do with momentum than weight.  Had it been intact when it hit the trees, you would then have to come up with a reason for the props, which are spinning "forward" to spin backwards in relation to the motors which are also spinning "forwards".   Obstructing the leading edge (hitting a branch or the ground etc.) will serve to drive the prop on tighter.  Obstructing the trailing edge will have no effect as the rotation will quickly move away from the obstruction and strike the leading edge instead.  The only way a prop should come off on the ground is if it breaks or the hub is stripped.  It is unlikely to just come off if it impacts a tree/branch or the ground, it is more likely to snap the prop blade.  Are they genuine DJI props, and for the correct model?  (P3 props have a white plastic centre, not a metal hub).  Is the thread cut into this hub stripped out at all?

If you look at the flight data, all the speed error messages were after 6min 12 sec.  Anything after that time is when you were crashing and the aircraft was tumbling out of the sky.  You can see the X, Y, and Z axes are changing quickly and wildly.  This is typical of a tumble, I have reviewed many crash data and this is very normal.  It suddenly rotates forward nose over tail, while rolling side over side and spinning to face every compass heading.  

This is an example of another crash investigation, see the directions and angles change as it falls from the sky?


This will happen every time you only have 3 props spinning -- it is not enough for flight, but enough to flip direction.  Any angle beyond about 30° pitch or roll is the aircraft crashing, it is out of control.  With 4 intact props it would have a chance of recovering level flight all by itself, especially at your height of 70m or more.  If you notice the speed errors, they all occur as the aircraft is crashing and tumbling with the pitch/roll values changing wildly at that time.  Those values are physically possible.  The compass and gyroscope are working properly.  It is showing that your aircraft is on its side or upside down at the instant that the data is sent.  The telemetry will continue sending these sideways and inverted artificial horizon values for all the time until it hits the ground and the signal is obstructed or the electronics are damaged and it is incapable of transmission.  The speed error message isn't particularly verbose to tell us what was going on, but I'd suggest one of the motors was spinning overspeed with no resistance.  The aircraft was trying to regain control and probably spinning a motor at 100%  (depending on its angle).  However if the prop was missing, it is possible that the motor is spinning faster than that, and the system reported an overspeed error.  Depending on the firmware a different error could be returned.  Sometimes you get a missing or obstructed prop or a loose prop warning.  It might not have had enough time to make a determination that a prop was missing and that is the best it could determine considering all the rapid X-Y-Z axis changes.  The last data entry of the record is when it hit the ground (the battery possibly popped out?) or the aircraft was so low that you lost line-of-sight between the aircraft transmitter and the remote control receiver, and the data stream cut out.

The DAT file is a LARGE file recorded inside the aircraft's flight recorder.  It records ALL the sensors and performance and movement commands.  Hundreds of them, dozens of times per second.  It will keep recording even when stationary on the ground until the power is interrupted.  I haven't had to retrieve one for a while, but if I remember correctly, you set the aircraft to Enter Flight Data Mode in Advanced Settings through the Go app, then connect the aircraft to your computer like an external USB hard disk.  The internal storage will show up and you will need to search for that file and copy it off.  You need to plug into the micro USB port on the body of the aircraft beneath the nameplate, not the camera/gimbal USB port.  The data transmitted and recorded for your flight is much smaller in size but it does contain 99% of what we would need to analyse the flight in a spreadsheet or dedicated analysis software.  The PhantomHelp web page is good visually, but does not show you all the parameters.  It is much better than it used to be and now shows horizon and stick movements for that instant, for example.  I am hoping that there is definitive evidence that a particular motor lost a prop or stopped completely in flight.  It will show a sudden loss of air resistance and thrust, motor speeds, current draw, and if it was overheating etc.

At the moment it is looking highly likely that there was a birdstrike and you lost a prop, but I'd be more confident after viewing the DAT.

What do the threads of the prop look like?  It's rare that the threads would strip, but it's possible - especially if you tightened them with a wrench.  That might explain the speed error coupled with the prop landing nearby to the drone.

endotherm Posted at 5-23 04:38
It is improbable, but it could quite easily have come off in flight and ended up landing close to the main wreckage.  The design of the P3 series props is that, as long as the motor is spinning, it tends to thread itself ON to the prop hub.  The prop motor won't run backwards, so it shouldn't be able to unthread itself.  However if it wasn't on tight, the normal acceleration could be enough to turn the prop in the correct direction, but not aggressively enough that it "self-tightens" itself into the hub.  Therefore, normal flight can be achieved, but it is a delicate balance.  If you slow suddenly or abruptly return the sticks to centre, the momentum of the prop is enough to keep it in rotation.  The motor is slowed or stopped in relation to this momentum.  That can be enough to have a counter rotational effect on the prop and it unscrews itself.  Now you have two separate objects tumbling in the air, without propulsion.  They will continue on a trajectory consistent with their previous momentum and fall to earth at the same speed.  Assuming they weren't acted upon by a lateral wind, the lighter prop could conceivably end up very close to the heavier body.  Even when the space shuttle broke up on re-entry, bits that came off the main craft at the same time ended up very close to each other on the ground.  It has more to do with momentum than weight.  Had it been intact when it hit the trees, you would then have to come up with a reason for the props, which are spinning "forward" to spin backwards in relation to the motors which are also spinning "forwards".   Obstructing the leading edge (hitting a branch or the ground etc.) will serve to drive the prop on tighter.  Obstructing the trailing edge will have no effect as the rotation will quickly move away from the obstruction and strike the leading edge instead.  The only way a prop should come off on the ground is if it breaks or the hub is stripped.  It is unlikely to just come off if it impacts a tree/branch or the ground, it is more likely to snap the prop blade.  Are they genuine DJI props, and for the correct model?  (P3 props have a white plastic centre, not a metal hub).  Is the thread cut into this hub stripped out at all?

If you look at the flight data, all the speed error messages were after 6min 12 sec.  Anything after that time is when you were crashing and the aircraft was tumbling out of the sky.  You can see the X, Y, and Z axes are changing quickly and wildly.  This is typical of a tumble, I have reviewed many crash data and this is very normal.  It suddenly rotates forward nose over tail, while rolling side over side and spinning to face every compass heading.  

That is, we can not rule out the probability that a propeller escaped in flight, I thank you very much for your attention and explanation!
But now let's go the part that interests, I was able to get the DAT file, I do not know to analyze this type of file but I am very anxious for your answer to heal my doubt of what happened once so that I can prevent it from occurring more often, can be downloaded from the link below:

https://drive.google.com/open?id=11YASGFQ4V5o7DxW_jzroIOWb5XlJFaC_

Mark The Droner Posted at 5-23 05:29
What do the threads of the prop look like?  It's rare that the threads would strip, but it's possible - especially if you tightened them with a wrench.  That might explain the speed error coupled with the prop landing nearby to the drone.

Only the one of the propeller that escaped was damaged, but this may have been either by the fall or already this way, we will not know until the analysis of our friend.

Brunorogi Posted at 5-23 17:30
Only the one of the propeller that escaped was damaged, but this may have been either by the fall or already this way, we will not know until the analysis of our friend.

I've had a look at the DAT file and extracted the motor parameters.   I can see that at 372 seconds into the flight an event happened.  The problem seems to be at the left back motor.   The motor continued to spin and function as expected for the entire time.  At the incident, I can see the LBack motor went suddenly from ~6000rpm to over 10000rpm (column JZ) and drew over 100% of its rated voltage as it was commanded to spin at maximum speed (columns ARR, ARY, and ASG).  The other motors momentarily slowed in response, as expected, but shortly thereafter returned to unsuccessfully trying to maintain stable flight.  The current draw drops off almost immediately during the fall, suggesting it wasn't pushing any air.  The motor was able to "free-wheel" and spin freely with the lack of resistance, a task it is able to perform while drawing minimal current.  There seems to be an issue for a few milliseconds with two other motors (columns KJ-KQ) as voltages fluctuate.  At first I thought this might indicate a birdstrike touching three motors but damaging one, but it seems to be too short and there is no other data confirming this.  It is probably just a voltage spike from a sudden event.  The motor controllers were not damaged and showed nothing abnormal, so we know the motors didn't stop.  Strangely, the motor seems to be drawing higher than normal current for some time prior to the incident as well, suggesting some sort of obstruction or drag on the motor.  I would have expected an accompanying Motor Obstructed error to be displayed as well, but it seems that the current draw was not excessive and within normal parameters, not qualifying for its own error message.









I now think the data points to either the prop being thrown, or the hub was damaged and the thread was chopped out.  The second circumstance would explain the motor spinning at full speed, but not providing any lift, at the same time keeping the prop in place together with the aircraft but not spinning.  It would then be easily separated from the aircraft upon striking the trees when it crashed.

I would very much like to see some photos of the prop in question, showing the top and leading edge (there will be nicks and damage if it impacted while spinning), and the underside, concentrating on the condition of the threaded hub.  Can you upload them here?  You might as well include the extent of the damage to the aircraft while you are at it, we are curious to know how it held up.  Could you also try to reattach the prop and see if you can pull it off easily without unscrewing it, or lifting the weight of the airframe while holding the prop.  If it pops right off, I'd say we have located the reason for the failure.

As to preventing this from happening again, I can only suggest a more thorough pre-flight inspection, making sure the props are attached properly and tight, but not too tight and thereby stripping the threads from the plastic hub.  It may have been slow incremental damage over time and went unnoticed until it failed.  Props are cheap and should be replaced when any damage is noticed -- it is cheap insurance for the loss of an expensive aircraft.  I'd suggest a close examination of all props as they might be on the way out too.  It is not necessarily your fault for over-tightening them, it might just be accumulated wear.   Unfortunately in this situation, I would put this failure down to bad luck.

endotherm Posted at 5-24 02:26
I've had a look at the DAT file and extracted the motor parameters.   I can see that at 372 seconds into the flight an event happened.  The problem seems to be at the left back motor.   The motor continued to spin and function as expected for the entire time.  At the incident, I can see the LBack motor went suddenly from ~6000rpm to over 10000rpm (column JZ) and drew over 100% of its rated voltage as it was commanded to spin at maximum speed (columns ARR, ARY, and ASG).  The other motors momentarily slowed in response, as expected, but shortly thereafter returned to unsuccessfully trying to maintain stable flight.  The current draw drops off almost immediately during the fall, suggesting it wasn't pushing any air.  The motor was able to "free-wheel" and spin freely with the lack of resistance, a task it is able to perform while drawing minimal current.  There seems to be an issue for a few milliseconds with two other motors (columns KJ-KQ) as voltages fluctuate.  At first I thought this might indicate a birdstrike touching three motors but damaging one, but it seems to be too short and there is no other data confirming this.  It is probably just a voltage spike from a sudden event.  The motor controllers were not damaged and showed nothing abnormal, so we know the motors didn't stop.  Strangely, the motor seems to be drawing higher than normal current for some time prior to the incident as well, suggesting some sort of obstruction or drag on the motor.  I would have expected an accompanying Motor Obstructed error to be displayed as well, but it seems that the current draw was not excessive and within normal parameters, not qualifying for its own error message.

[view_image]

Impressive! Can we say now with 100% certainty that the propeller has come loose?
Because after the fall I already repaired it, fortunately there was not much damage from falling on the tree, I did not make any repairs on the central plate, it does not replace any electric wire and even then after all 4 motors started normally, I even made a small flight and he behaved normal.
I honestly had lost confidence in him because I thought it was an internal problem on the electronics side, but if it really was the propeller I'm relieved to know that! Well, it's an easy problem to solve.
I even had the propeller that had come loose, I kept it, but now I do not have the photo of it, but I did a test after it fell and it was not possible to pull the propeller just by pulling (Of ​​course I did not do much force either, but enough to hold his own weight).
But when I noticed her wire was really broken, so I did not use it anymore.
Could it be that with all this we can say that it was the propeller that was released?
Note: Unfortunately I only have these two photos after the fall, should have taken much more

Brunorogi Posted at 5-24 04:41
[view_image][view_image]

Impressive! Can we say now with 100% certainty that the propeller has come loose?

I'd be comfortable with 99.9% certainty that it was a propeller .  Whether the hub was chopped out or it released completely is unknown, but either way, it was the cause of the flight failure.   Based on your test that it could hold its own weight, it is more likely it spun off in flight.  I'd definitely examine the remaining props for excessive wear on the threads and if there is any looseness or sloppiness when it is screwed on.

As to it being an easy problem to solve... not really, there is lots of data to review and responses to write, which takes many hours.  Fortunately I enjoy doing that and crashes that require forensic investigation aren't that common these days (for Phantom at least, that's the forum I hang out in).  Fortunately it's an easier problem to solve for you, there is no complex electronic repair to address, mainly structural issues.   Good to see the damage wasn't that extensive and you were able to salvage and repair it.

endotherm Posted at 5-24 09:17
I'd be comfortable with 99.9% certainty that it was a propeller .  Whether the hub was chopped out or it released completely is unknown, but either way, it was the cause of the flight failure.   Based on your test that it could hold its own weight, it is more likely it spun off in flight.  I'd definitely examine the remaining props for excessive wear on the threads and if there is any looseness or sloppiness when it is screwed on.

As to it being an easy problem to solve... not really, there is lots of data to review and responses to write, which takes many hours.  Fortunately I enjoy doing that and crashes that require forensic investigation aren't that common these days (for Phantom at least, that's the forum I hang out in).  Fortunately it's an easier problem to solve for you, there is no complex electronic repair to address, mainly structural issues.   Good to see the damage wasn't that extensive and you were able to salvage and repair it.

I am extremely grateful for all your help!
I had at first discarded the possibility that the propeller had come loose because it had fallen too close to the drone, but it seems the world conspired to make the investigation more difficult.
I am also 99.9% sure that we have solved the mystery! Now is to put it to work in flight and check if everything is normal, and of course, would pay close attention to the propellers!
Again, thank you very much for the help, how nice that there are people like you who take precious time to help others, thank you!

endotherm what program are you using to analyze the flight?

Tree Wolf Posted at 5-24 13:46
endotherm what program are you using to analyze the flight?

There are a combination of things.  Firstly I'll view the flight record on Phantom Help to see if anything obvious really stands out, and to get an overview of what happened.  I'll download the kml file from the website and open it in Google Earth and get a 3D display of the flight path, sometimes that will reveal things like a power line running overhead (an unexplained fall from the sky) , or metal buried underground (compass errors reported in flight) by reviewing historical imagery from previous satellite passes.  Then I will download the csv (comma separated value) file that contains much more information than the obvious ones displayed on Phantom Help or the Go app.  I use Excel to load it into a spreadsheet program and apply a bunch of formatting, colouring and data bar rules that I created.  That way I can visualise everything that is happening and shapes and colours tend to jump off the screen easier than a bunch of numbers.  However even that csv doesn't contain everything.  For the particularly difficult investigations I load it into a program called CsvView.  This will show up a number of hidden parameters and "engineered" parameters, calculated and derived from analysis of other sensor data.  This data can be selected individually or overlayed with other data and a graph drawn to visualise the data.  It allows you to quickly select a point in time and return the graphical data, or zoom in to an area to concentrate on, etc.  e.g. display all the motor control pulse width modulation signals simultaneously.  Then there is the mother of all analysis programs, the big brother to CsvView called DatCon.  This is able to export a new csv file with (currently) 1198 columns of parameters (up to column ATB) and the ability to define the frequency of the sample rate (by default, 10 records per second is recorded to the standard flight record, or every 100ms)  but if necessary you can extract as much as 200 times per second (5ms).  These are necessary to view the very large DAT file that is recorded internally on the "flight" recorder, even though it records everything while the aircraft is powered on.  Much like a real aircraft flight recorder or your car crash cam, it will record over old data once the internal storage fills up.  It usually has enough to retain the past few flights for examination.  This is the main reason DJI require you to send the aircraft to them for analysis for warranty repair, for example.  It contains everything they could possibly want to know and it is reliable data, unlike an extracted flight record which might be tampered with and falsified.

endotherm Posted at 5-24 22:03
There are a combination of things.  Firstly I'll view the flight record on Phantom Help to see if anything obvious really stands out, and to get an overview of what happened.  I'll download the kml file from the website and open it in Google Earth and get a 3D display of the flight path, sometimes that will reveal things like a power line running overhead (an unexplained fall from the sky) , or metal buried underground (compass errors reported in flight) by reviewing historical imagery from previous satellite passes.  Then I will download the csv (comma separated value) file that contains much more information than the obvious ones displayed on Phantom Help or the Go app.  I use Excel to load it into a spreadsheet program and apply a bunch of formatting, colouring and data bar rules that I created.  That way I can visualise everything that is happening and shapes and colours tend to jump off the screen easier than a bunch of numbers.  However even that csv doesn't contain everything.  For the particularly difficult investigations I load it into a program called CsvView.  This will show up a number of hidden parameters and "engineered" parameters, calculated and derived from analysis of other sensor data.  This data can be selected individually or overlayed with other data and a graph drawn to visualise the data.  It allows you to quickly select a point in time and return the graphical data, or zoom in to an area to concentrate on, etc.  e.g. display all the motor control pulse width modulation signals simultaneously.  Then there is the mother of all analysis programs, the big brother to CsvView called DatCon.  This is able to export a new csv file with (currently) 1198 columns of parameters (up to column ATB) and the ability to define the frequency of the sample rate (by default, 10 records per second is recorded to the standard flight record, or every 100ms)  but if necessary you can extract as much as 200 times per second (5ms).  These are necessary to view the very large DAT file that is recorded internally on the "flight" recorder, even though it records everything while the aircraft is powered on.  Much like a real aircraft flight recorder or your car crash cam, it will record over old data once the internal storage fills up.  It usually has enough to retain the past few flights for examination.  This is the main reason DJI require you to send the aircraft to them for analysis for warranty repair, for example.  It contains everything they could possibly want to know and it is reliable data, unlike an extracted flight record which might be tampered with and falsified.

You must have studied and researched a lot to get to this level of research, congratulations!

I'm still thinking about the case, I also managed to open the flight data and I came up with a doubt, the engine on the opposite side had a much more coherent attitude, if the drone is spinning madly the engine has to vary its speed (I imagine myself) , but it was not what happened to the Lback engine, it simply locked in 100%.
Is this correct? Can not be a broken wire in the Lback engine that did it? Cold welding? Although I tested the drone after the fall, I flew only 4min, sometimes the problem did not repeat itself but it is there.
Is there no chance that the engine sent a wrong signal to the ESC because of an electrical problem? Broken wire, cold welding, etc?

Brunorogi Posted at 5-28 09:53
I'm still thinking about the case, I also managed to open the flight data and I came up with a doubt, the engine on the opposite side had a much more coherent attitude, if the drone is spinning madly the engine has to vary its speed (I imagine myself) , but it was not what happened to the Lback engine, it simply locked in 100%.
Is this correct? Can not be a broken wire in the Lback engine that did it? Cold welding? Although I tested the drone after the fall, I flew only 4min, sometimes the problem did not repeat itself but it is there.
Is there no chance that the engine sent a wrong signal to the ESC because of an electrical problem? Broken wire, cold welding, etc?

The thing is, the motor does not know that the prop is not there.  The flight system analyses the flight and realizes it is not in level flight, or the expected pitch or tilt if it is in forward or lateral acceleration and tries to correct it.  It is the equivalent of applying another joystick control signal on top of the one it is receiving from the pilot.  It does this continuously.  If there is a lateral influence (a bump or the wind blowing against it), it will react by increasing the opposite side and decreasing the facing side to counteract the force and level out the craft again.  In this case, one motor is not providing ANY thrust at all.  As a result that arm goes downward due to gravity.  The response is to spin up that motor until it provides enough thrust.  Even spinning at 100% isn't enough, and it would spin faster if it could, hence the flat line at maximum we see here.  The system does not know that a prop is missing, all it could do is spin faster to address this.  Even if it did know the prop wasn't there, there is nothing it could do about that.  The thing that normally happens in flight is that one motor will spin up and the diagonally opposite will slow down, in order to rotate it back into position and provide consistent power with the remaining motors.  If both diagonal motors remained higher, the aircraft would try to climb, which is not what was commanded.  The response is different if it is mainly upright, to what it will do if it becomes inverted.  If it is mainly upright, you will see pulsing various motors as you described and expected.  However as we have determined a prop was missing, this mode won't apply.  At the point where it realizes it is tumbling and inverted, the response is to spin up two adjacent motors to flip it upright again, while slowing the other two.   This critical operation is so important it also ignores any other remote control commands until completed, including CSC.  A simple formula, but it is the only thing that works.   With only three props it will be unsuccessful in righting itself and it will continue to tumble with little observable change in the motor outputs.  I'm not sure what your graph is plotting as you haven't provided a legend.  Your graph is different to the one I show in post #18, which shows the four command signals to the motors.  It is easy to select the wrong motor data when so many different sets are provided in CsvView, which makes it inadvisable to try and make a determination from the data if you don't know what the difference is, or what each set means.  Your red line seems to represent something else -- it's not power command signals because you would have a corresponding dip in the graph like the purple line in mine.   I assume you are mixing different signals here, and your red line shows the actual speed of the motor, with fluttering speeds of a couple of percent.  The command from the aircraft flight controller is instructing it to spin at 100% the whole time though (blue line).

It's not a broken wire because it isn't wired up that way inside to return these results.  If the power wire to the motor were to break, the power output would drop to zero and the command signal would still be at 100%.  If it were to short out to the power rail causing it to spin at 100%, the command signal would show something else other than 100%.  One of the wires returns a signal for motor performance, so it tells us what it was doing in reality, not what we were expecting it to do.  So you need to look at a number of things simultaneously and that takes experience and a trained eye.

Any electrical problem you might experience on subsequent tests is probably due to the crash and impact.  It is quite possible a wire could crack or form an intermittent connection like a cold solder joint.  Your arms appear to have twisted a bit and there is not a lot of spare wire in there, so it is possible they were stressed and damaged.  Any repair would involve popping the top of the body off and examining the heavy wires from the ESC/motor controllers to the motors.  Inspect the wires for continuity and that the joints are sound.

endotherm Posted at 5-29 05:37
The thing is, the motor does not know that the prop is not there.  The flight system analyses the flight and realizes it is not in level flight, or the expected pitch or tilt if it is in forward or lateral acceleration and tries to correct it.  It is the equivalent of applying another joystick control signal on top of the one it is receiving from the pilot.  It does this continuously.  If there is a lateral influence (a bump or the wind blowing against it), it will react by increasing the opposite side and decreasing the facing side to counteract the force and level out the craft again.  In this case, one motor is not providing ANY thrust at all.  As a result that arm goes downward due to gravity.  The response is to spin up that motor until it provides enough thrust.  Even spinning at 100% isn't enough, and it would spin faster if it could, hence the flat line at maximum we see here.  The system does not know that a prop is missing, all it could do is spin faster to address this.  Even if it did know the prop wasn't there, there is nothing it could do about that.  The thing that normally happens in flight is that one motor will spin up and the diagonally opposite will slow down, in order to rotate it back into position and provide consistent power with the remaining motors.  If both diagonal motors remained higher, the aircraft would try to climb, which is not what was commanded.  The response is different if it is mainly upright, to what it will do if it becomes inverted.  If it is mainly upright, you will see pulsing various motors as you described and expected.  However as we have determined a prop was missing, this mode won't apply.  At the point where it realizes it is tumbling and inverted, the response is to spin up two adjacent motors to flip it upright again, while slowing the other two.   This critical operation is so important it also ignores any other remote control commands until completed, including CSC.  A simple formula, but it is the only thing that works.   With only three props it will be unsuccessful in righting itself and it will continue to tumble with little observable change in the motor outputs.  I'm not sure what your graph is plotting as you haven't provided a legend.  Your graph is different to the one I show in post #18, which shows the four command signals to the motors.  It is easy to select the wrong motor data when so many different sets are provided in CsvView, which makes it inadvisable to try and make a determination from the data if you don't know what the difference is, or what each set means.  Your red line seems to represent something else -- it's not power command signals because you would have a corresponding dip in the graph like the purple line in mine.   I assume you are mixing different signals here, and your red line shows the actual speed of the motor, with fluttering speeds of a couple of percent.  The command from the aircraft flight controller is instructing it to spin at 100% the whole time though (blue line).

It's not a broken wire because it isn't wired up that way inside to return these results.  If the power wire to the motor were to break, the power output would drop to zero and the command signal would still be at 100%.  If it were to short out to the power rail causing it to spin at 100%, the command signal would show something else other than 100%.  One of the wires returns a signal for motor performance, so it tells us what it was doing in reality, not what we were expecting it to do.  So you need to look at a number of things simultaneously and that takes experience and a trained eye.

I understand perfectly.
But let's get back to the theory, if the drone is spinning madly in the air, is the Lback side somewhere higher than the correct Rfront? Because in this case it is not commanded for the Lback engine to slow down a bit and accelerate Rfront (Rfront accelerated and decelerated)?
The graph I put shows just that, the Rfront speeding and decelerating depending on the current position of the aircraft but Lback is at 100%.
It's okay that the propeller is not there, but like you said the engine does not know it! Why did not he slow down when his side was higher than Rfront?

Atti mode?  Speed error?

Maybe an ESC failure.  I'm not even sure why they've even bothered to put Atti mode in the first place.  I think that's for indoor flying?

AntDX316 Posted at 5-29 10:06
Atti mode?  Speed error?

Maybe an ESC failure.  I'm not even sure why they've even bothered to put Atti mode in the first place.  I think that's for indoor flying?

No, it was me who selected the ATTI mode, but I do not think it has anything to do with the fall, I just think ...

Brunorogi Posted at 5-29 06:47
I understand perfectly.
But let's get back to the theory, if the drone is spinning madly in the air, is the Lback side somewhere higher than the correct Rfront? Because in this case it is not commanded for the Lback engine to slow down a bit and accelerate Rfront (Rfront accelerated and decelerated)?
The graph I put shows just that, the Rfront speeding and decelerating depending on the current position of the aircraft but Lback is at 100%.

The reason for that is that there is a complex comparison of the data going on in the flight computer -- the commands in to the motor, the power output it is generating, the current draw etc.  Lots of different parameters from the dozen of so "motor" groups of data shown in post #17.  There is an expected value range from the calculations, and if it is outside of that tolerance, it throws up an error message.   From these calculations it knows that the motor is not generating sufficient lift (or any at all), so it tries to spin it faster.  Even when it is tumbling, it is trying to "lift" on that motor (even if that is thrusting sideways or upside down), but not able to achieve the amount that is commanded.  All you see is that the motor is pegged at 100%.  It makes no sense to slow it down if it isn't providing enough lift.  The other three are at least occasionally producing sufficient thrust for their corners, and we see their thrust values pulsing up and down.

AntDX316 Posted at 5-29 10:06
Atti mode?  Speed error?

Maybe an ESC failure.  I'm not even sure why they've even bothered to put Atti mode in the first place.  I think that's for indoor flying?

There is nothing wrong with flying in ATTI mode, all that happens is that you lose help from the GPS satellites in order to correct its orientation/tilt/roll/yaw and keep it on track.  You are under the influence of the wind, but if you are skilled and have practiced, or there is little wind, you can fly it quite normally.  Getting pushed around a bit isn't a huge problem in the wide open sky.  The problem is when an inexperienced pilot comes up against it and has never experienced full manual control or is tackling a stronger wind than he can deal with, he will get into trouble.  It is a bit like driving a car yourself "manually", versus going for a ride in a driverless computer-guided/robot controlled car -- the tech is doing lots of the navigation for you.  It's still pretty safe having an unpredictable human driving instead though.

endotherm Posted at 5-31 11:58
The reason for that is that there is a complex comparison of the commands going in to the motor, the power output it is generating, the current draw etc.  Lots of different parameters from the dozen of so "motor" groups of data shown in post #17.  There is an expected value range from the calculations, and if it is outside of that tolerance, it throws up an error message.   From these calculations it knows that the motor is not generating sufficient lift (or any at all), so it tries to spin it faster.  Even when it is tumbling, it is trying to "lift" on that motor (even if that is thrusting sideways or upside down), but not able to achieve the amount that is commanded.  All you see is that the motor is pegged at 100%.  It makes no sense to slow it down if it isn't providing enough lift.  The other three are at least occasionally producing sufficient thrust for their corners, and we see their thrust values pulsing up and down.

So the prop probably came off in flight?

endotherm Posted at 5-31 11:58
The reason for that is that there is a complex comparison of the commands going in to the motor, the power output it is generating, the current draw etc.  Lots of different parameters from the dozen of so "motor" groups of data shown in post #17.  There is an expected value range from the calculations, and if it is outside of that tolerance, it throws up an error message.   From these calculations it knows that the motor is not generating sufficient lift (or any at all), so it tries to spin it faster.  Even when it is tumbling, it is trying to "lift" on that motor (even if that is thrusting sideways or upside down), but not able to achieve the amount that is commanded.  All you see is that the motor is pegged at 100%.  It makes no sense to slow it down if it isn't providing enough lift.  The other three are at least occasionally producing sufficient thrust for their corners, and we see their thrust values pulsing up and down.

Also, I assume the esc wires or esc has failed? They've revised the ESCs with V2.0.  I've lost my hexacopter due to ESC failure.  This is why getting the lastest is the best way to be safe as the latest usually addresses all the known issues of previous generations.  Crashing out of the sky may result in many of the latest drones becoming free.  We can only do so much as an operator that if anything goes wrong beyond our control there is nothing we can do.

AntDX316 Posted at 5-31 12:24
Also, I assume the esc wires or esc has failed? They've revised the ESCs with V2.0.  I've lost my hexacopter due to ESC failure.  This is why getting the lastest is the best way to be safe as the latest usually addresses all the known issues of previous generations.  Crashing out of the sky may result in many of the latest drones becoming free.  We can only do so much as an operator that if anything goes wrong beyond our control there is nothing we can do.

I was also in doubt whether the propeller was the cause, but sure now! I made a 15 minute flight today without problems, with the propellers well attached, the flight was a success!

Brunorogi Posted at 5-31 15:18
I was also in doubt whether the propeller was the cause, but sure now! I made a 15 minute flight today without problems, with the propellers well attached, the flight was a success!

Your drone crashed from 70m and you are using it now?

AntDX316 Posted at 5-31 12:15
So the prop probably came off in flight?

That is what the substantial analysis above concluded.

AntDX316 Posted at 5-31 12:24
Also, I assume the esc wires or esc has failed? They've revised the ESCs with V2.0.  I've lost my hexacopter due to ESC failure.  This is why getting the lastest is the best way to be safe as the latest usually addresses all the known issues of previous generations.  Crashing out of the sky may result in many of the latest drones becoming free.  We can only do so much as an operator that if anything goes wrong beyond our control there is nothing we can do.

ESC didn't fail, they were providing the expected signals to the motors, the motors were performing as expected.  The only thing missing was the anticipated lift from a prop.

Brunorogi Posted at 5-31 15:18
I was also in doubt whether the propeller was the cause, but sure now! I made a 15 minute flight today without problems, with the propellers well attached, the flight was a success!

Great to hear!

AntDX316 Posted at 5-31 17:33
Your drone crashed from 70m and you are using it now?

Yes. Crashing into trees definitely saved it.  Despite thinking damage would be catastrophic from hitting branches on the way down, hitting lighter foliage can actually cushion the fall.  There are even cases in WWII of paratroops jumping from planes and landing in trees due to their parachutes not opening.  They survived.

endotherm Posted at 6-1 02:23
Yes. Crashing into trees definitely saved it.  Despite thinking damage would be catastrophic from hitting branches on the way down, hitting lighter foliage can actually cushion the fall.  There are even cases in WWII of paratroops jumping from planes and landing in trees due to their parachutes not opening.  They survived.

wow, the power of trees

AntDX316 Posted at 6-1 03:09
wow, the power of trees

I had a hard time escaping the propeller but great luck falling on the tree! There was concrete floor near the fall, if it falls on that floor would not even see the size of the damage ... Maybe I would have lost the drone almost 100%, as it fell from the tree a simple exchange of carcass has already left it as new, except for the flat cable that I'll have to change from gimbal, but that's all!
I would like to thank everyone, and especially Endotherm, for making your valuable time available to help me! My drone is in the heavens again!