Hello,
My teenage son was out flying the drone yesterday (as he has done many times) when he says it 'disconnected' and flew away.
I have tried uploading the flight log to phantomhelp.com but keep getting an internal server error, i have managed to upload it to healthydrones
https://healthydrones.com/main?f ... _id=GENERALOverview
Can anybody give me an idea on where to start searching? Please bare with me, im not very clued up on these things, any help would be much appreciated!
Thanks

I'm looking into it but Healthydrones isn't great for analysis and this one looks a bit complicated.
If you can't get Phantomhelp to work, can you put the .txt flight record into Dropbox and post a link?

Just quickly, the flight data is peppered with compass errors and atti mode so it looks like he's let it blow away in atti mode.
Can you give any information about where and when the compass was last calibrated?

Drive to Bancombe rd and Somerton rd intersection and look fot the drone in the field close to the auto tire business. Also check in the pond that is close in that area.

MJLSTUDIOS Posted at 2017-4-29 00:07
Drive to Bancombe rd and Somerton rd intersection and look fot the drone in the field close to the auto tire business. Also check in the pond that is close in that area.

It's not going to be there.
That's simply where the last position that was recorded before signal was lost.
At the time, the Phantom was still 450 feet up in the air and being piloted even further away at 37 mph.
It's going to be even further north east.
There are indications that the wind was assisting and will have taken it further once signal was lost before it reached critical low battery and autolanded somewhere.
There's plenty of evidence that the pilot was not fit to be flying the Phantom and was responsible for the loss.

Do you have any idea what the failsafes in the App were set to?  Without some more data the search area will be impossibly large...

Wow, that log is a disaster. I wouldn't trust most of the coordinates or flight path showing there, the compass was so out of whack the phantom didn't know if it was coming or going.

Labroides Posted at 2017-4-29 00:04
I'm looking into it but Healthydrones isn't great for analysis and this one looks a bit complicated.
If you can't get Phantomhelp to work, can you put the .txt flight record into Dropbox and post a link?

https://www.dropbox.com/s/bv8yzw ... 6-30-11%5D.txt?dl=0
hopefully that works.
the compass was calibrated on the previous fight 2 days prior to this one. Both took off from the same location.

MJLSTUDIOS Posted at 2017-4-29 00:07
Drive to Bancombe rd and Somerton rd intersection and look fot the drone in the field close to the auto tire business. Also check in the pond that is close in that area.

Thank you, we did go there last night. the owner of the farm with the pond said they saw it flying over at speed.

Did you check back at the launch point to see if it ever made it back by itself?  

I see the pilot walked away from the launch point when the signal was lost and then regained signal for a while when he reached the other end of the field, at that point, as he went through the gate/hedge, it was on the way back to the launch point and apparently flying OK.  It may have taken some time to get back though so may have arrived after he gave up...

But I suspect it may have had another compass/magnetic interference error and drifted away with the wind, predicting where it ended up is going to be very difficult and finding it even more difficult.

If you get a compass error then you should always land as soon as is safe and then sort the problem, this has to be considered pilot error.

Sorry to hear the lost, may I know how long you've owned the drone, did you send an e-mail to support.eu@dji.com to report the issue?

(Oops - I see some others posted while I wrote this. I leave it anyway)

I'm afraid the search area is so huge that you hardly have a chance to find it.
The drone was 5km away when total signal loss ocurred - over a cell tower in Somerton Business Park.
At that moment is was at a speed of 40mph and a height of 400ft. The battery was at 22% so it will have travelled quite some more distance.

Over the "Pibsbury Pigs" farm is when the problems became severe. That was 2km from home point and that is ok for running into signal issues, even if it is countryside.
One severe warning was: "Warning:Compass Error. Exit P-GPS Mode".

38 sec later it switched to Go_Home. From there on it flew in a straight line going northeast.
Maybe some others here can explain this. If the drone switches to RTH even while being in Atti mode it should have gone home.
It always had good GPS signal...

That's a long way to fly on 12% battery (if the P3A autolands at 10%?). I guess it auto landed somewhere about half way back in a straight line to the launch point.

Having seen where you were flying, can I suggest you read and comply with the UK drone code: http://dronesafe.uk/drone-code/
Flying out of sight and hearing that close to a busy military airport where new pilots train on low flying helicopters is stupid, not just illegal!

Nigel_ Posted at 2017-4-29 02:31
That's a long way to fly on 12% battery (if the P3A autolands at 10%?). I guess it auto landed somewhere about half way back in a straight line to the launch point.

Having seen where you were flying, can I suggest you read and comply with the UK drone code: http://dronesafe.uk/drone-code/

Could the drone have been forced into auto land when entering restricted airspace?

Nigel_ Posted at 2017-4-29 02:31
That's a long way to fly on 12% battery (if the P3A autolands at 10%?). I guess it auto landed somewhere about half way back in a straight line to the launch point.

Having seen where you were flying, can I suggest you read and comply with the UK drone code: http://dronesafe.uk/drone-code/

thanks, but the drone is normally  kept in sight, it wasn't intentional that it went out of sight. The airbase you mention is some 12 miles away and we are not in a no fly zone, again it wasn't meant to go out of sight.

Nigel_ Posted at 2017-4-29 01:52
Did you check back at the launch point to see if it ever made it back by itself?  

I see the pilot walked away from the launch point when the signal was lost and then regained signal for a while when he reached the other end of the field, at that point, as he went through the gate/hedge, it was on the way back to the launch point and apparently flying OK.  It may have taken some time to get back though so may have arrived after he gave up...

yes, have checked launch point. Not disputing its pilot error, just trying to get a rough idea of where it might of gone

blackcrusader Posted at 2017-4-29 02:46
Could the drone have been forced into auto land when entering restricted airspace?

Don't think so, I think it was over the 5 mile limit at all times, although not by much, so I don't think there would even have been a warning.  They just have to give way to all manned aircraft, which isn't possible if you are so far away that you cant see or hear them.

I remember standing on the runway of that airport watching new pilots practicing take offs and landings in Harrier Jump Jets - one of the most amazing sights I have ever seen, aeroplanes should not just sit 100ft above the runway with no movement whatsoever!

mumofboys Posted at 2017-4-29 02:55
thanks, but the drone is normally  kept in sight, it wasn't intentional that it went out of sight. The airbase you mention is some 12 miles away and we are not in a no fly zone, again it wasn't meant to go out of sight.

You are also not allowed to fly over residential areas, I'm not going to look at the logs too closely, but I think a careful read of the Drone Code might be sensible!  Remember that the pilot is always responsible for a safe flight and for the safety of all other aircraft and the people and traffic on the ground below.  Helicopters don't have to stay above the 400ft Drone limit at all times.

mumofboys Posted at 2017-4-29 02:58
yes, have checked launch point. Not disputing its pilot error, just trying to get a rough idea of where it might of gone

Phantomhelp.com seems to be down so the log cannot be uploaded currently.

Something I noticed:
Neither Airdata or any entry in the csv file suggest that a home point was recorded at the beginning.
The "H" on the map is just an indicator for where flight recording started.

At Airdata an entry like this first screenshot below should be seen.
You don't have such an entry.

A csv file always shows in the early stage an entry:
Home Point Recorded. RTH Altitude: ...m.
There is no such entry in your csv file.

The drone switched into RTH mode starting at the Piggy farm and went straight Northeast.
Have you been with the drone prior to this flight somewhere northeast of the business park where signal was lost?
Maybe it had in its system a previous home point?
I wouldn't know how that can happen.

Eric13 Posted at 2017-4-29 04:05
Phantomhelp.com seems to be down so the log cannot be uploaded currently.

Something I noticed:

"A csv file always shows in the early stage an entry:
Home Point Recorded. RTH Altitude: ...m.
There is no such entry in your csv file."

Could this be because it is a P3, not a P4?  I don't normally look at P3 logs so don't know the differences.

I can however see that it knew the correct home point, that is recorded in the log and was never changed.
I can also see that by the time of the last log entry the pilot was in the gateway the other side of the field  and may not have see the P3 return even if it had, I assume he gave up and turned off the control unit at that time. It would have arrived a lot later if it had not run out of battery which it probably did.  Had he left he controller on then he may have seen where it auto landed since the signal would have been a lot better by that time.

At the end of the log, even though it is in go home mode and has turned around and claims descent speed, the distance is still increasing - have to suspect the compass wasn't working correctly, some of the data is not correct but I can't tell which.  There doesn't seem to have been much wind so it wasn't being blown away.

Unfortunately this is a pretty typical example of flying too far and high, and not being able to see the aircraft.  It seems it got into difficulty fighting some wind and was blown away out of control of the pilot.  The critical thing here is that the pilot was relying on return to home which is limited to 10m/s flight speed. In this case he was trying to retrieve it against a stronger wind and lost.  Had he returned to manual flight, he would have had the necessary power to overcome the wind and could have flown home safely.  Also, he forgot the cardinal rule in these situations -- descend to where the wind is likely to be weaker.

I closely examined the logs from the csv file.  Although there were some compass errors reported, they were short blips and did not corrupt the other data, they were therefore inconsequential.  GPS coordinates and control inputs were accurately recorded and reliable.  The aircraft tried to RTH a few times but was unsuccessful at that altitude (not above 400ft, take-off was 64ft).  There were large chunks of loss of signal to the aircraft reflected in the logs, perhaps due to trees or other obstructions, or poor antenna orientation.  These dropped out and restored periodically during the flight.

If we examine the last long straight leg of the flight, the aircraft was travelling around 40mph (18m/s) on a heading of around 70° while facing backwards at 290°, having turned to face home point due to the RTH command.  Unfortunately the P3 has a top speed of around 15m/s in manual flight and only 10m/s in RTH mode, so either were unable to overcome the speeds reported (18m/s).

We can assume the aircraft was drifting in a 40mph wind from the SW and was last tracked heading on a course of 70° at 400ft.  The last reported battery level was 22%.  The aircraft would continue on this last direction until critical battery kicks in (assumed at 10%), and autolanding occurs.  As long as wind speed and direction did not vary considerably in the next few minutes, we can extrapolate where it was heading.

Based on the previous 12% of battery consumption with similar load, I estimate the aircraft travelled 2.36km in 132 seconds.  This is represented on the map by the red line.  At that point the 10% critical battery autolanding occurred.  The P3 is limited to 3m/s descent to avoid a vortex ring state, and from 400ft it takes 2 minutes to descend to the ground, while still being influenced by the 40mph wind.  This could take it a further 2km if the windspeed remained constant at 40mph to the ground.  This is unlikely and the windspeed is expected to drop off.  Therefore the aircraft would touch down somewhere in the second half of the green line on the map.  This places it in the heavily treed area SW of the Kingweston Golf Club in Somerton.  Your likely search area is within the yellow circle.  Around coordinates  51° 4'6.00"N   2°42'7.00"W.

endotherm Posted at 2017-4-29 05:53
Unfortunately this is a pretty typical example of flying too far and high, and not being able to see the aircraft.  It seems it got into difficulty fighting some wind and was blown away out of control of the pilot.  The critical thing here is that the pilot was relying on return to home which is limited to 10m/s flight speed. In this case he was trying to retrieve it against a stronger wind and lost.  Had he returned to manual flight, he would have had the necessary power to overcome the wind and could have flown home safely.  Also, he forgot the cardinal rule in these situations -- descend to where the wind is likely to be weaker.

I closely examined the logs from the csv file.  Although there were some compass errors reported, they were short blips and did not corrupt the other data, they were therefore inconsequential.  GPS coordinates and control inputs were accurately recorded and reliable.  The aircraft tried to RTH a few times but was unsuccessful at that altitude (not above 400ft, take-off was 64ft).  There were large chunks of loss of signal to the aircraft reflected in the logs, perhaps due to trees or other obstructions, or poor antenna orientation.  These dropped out and restored periodically during the flight.

Weather records give the wind speed at 6mph - 10mph from the NW.  Yes, they would have been higher at 400ft but a constant 40mph seems unlikely.
Plus when it first went to 400ft, it was able to hold position with little effort.

Endoterms theory doesn't make sense to me:
The AC was travelling 40mph northeast/later east.
See below what the wind speed calculation was for that flight.
The AC can't have been drifting, it actively flew at high speed.

I don't see any home point being recorded. An "H" on the map only shows the beginning of a flight record.
Usually it goes along with home point recording but in this case that never happened.

The drone had GPS signal from the beginning and it should have recorded a HP.
But it never did. Phantomhelp is back up. Check it out there.

He had lots of compass issues over the town. Later on countless yaw errors.
The system must have got messed up somehow.

Nigel_ Posted at 2017-4-29 06:20
Weather records give the wind speed at 6mph - 10mph from the NW.  Yes, they would have been higher at 400ft but a constant 40mph seems unlikely.
Plus when it first went to 400ft, it was able to hold position with little effort.

Wind is variable.  It changes speed and direction often.  It is frequently stronger at higher altitudes.  A constant 40mph is quite plausible.  A recorded windspeed at a particular time and place may be completely different at only 100ft difference in altitude.

It's initial ascent to 400 feet appeared to have been under controlled flight.  The P3 is quite able to hold station up to around 35mph.  It has sufficient power to do so in ATTI or GPS modes.  In GPS mode it can hold quite still.  In ATTI you would be working very hard.  If you were not providing any stick inputs, the aircraft will drift away.  In this case, the aircraft was primarily in RTH mode with little to no manual stick inputs.  Being caught in a 40mph wind stream is beyond the capability of this aircraft to overcome.

Eric13 Posted at 2017-4-29 06:36
Endoterms theory doesn't make sense to me:
The AC was travelling 40mph northeast/later east.
See below what the wind speed calculation was for that flight.

At the end of the record, the aircraft rotated right and changed heading by about 30°.  This could be a momentary change of a couple of seconds or it could be a more lasting change.  The last leg is too short to make any certain prediction, however it is plausible to assume the wind had changed direction to 70° as the previous leg wasn't choppy and was quite stable for some time.  I don't have a crystal ball to say that the predicted search area would be fruitful, but I have examined quite a few cases such as this one and the data indicates a likely solution.

You can't say "it can't have been drifting" and "it was actively flying at high speed".  The data proves otherwise.  The P3 is incapable of actively flying at that speed by itself.  The only explanation for it to be travelling at that speed is for it to be carried by faster wind.

Having used Healthy Drones/AirData for some time and analysing the csv data from flight records, I can tell you that they do not report a home point being recorded in their data.  This does not indicate a fault, it is just something that that site does not report.  Phantom Help does things differently, and also reports some different parameters, and describes things differently.

A compass error is reported when the heading data does not coincide properly with GPS data.  These faults were numerous, but short lived.   It may have been due to the RC connection being unreliable and cutting in and out.  Although the compass may have suffered from the equivalent of "noisy static" it was still sufficient to plot a flight track and report headings.  There was not a constant string of "compass error" faults reported.

The wind report generated by this site is not particularly reliable in all circumstances.  It is an analysis of the aggregate of forces applied against the aircraft, and the motor power and tilt of the body required to overcome those forces.  Depending on a number of factors including gaps in the flight log data, this calculated value could be quite dissimilar to the actual air currents.  It should be particularly noted that there are big chunks of data missing when the aircraft is travelling on that long last leg; 2 minutes, 45 seconds, 20 seconds...etc.  That's lots of vector values of 40° missing when added to the AirData estimate of 125° results in a true wind vector towards 70°, as I indicate. The apparent straight line is actually an illusion, joining dots between missing parts of the log with the shortest path between known points.  The true path is unknown but the approximate heading is able to be determined.

Keep in mind a proper analysis is performed (by me and others) by actually analysing the kml and csv data on a spreadsheet, with additional data not visualised on the website.  HD/AD does provide a number of conclusions from the data, but much of this isn't a lot of help in crash/flyoff investigations.  It is good for tracking your battery cell health and temperatures etc, but it does not report what happened in 100 millisecond increments, which is the data that I was examining, not just a line on a map.  I am able to see the tilt of the body, the direction it is facing and direction of travel.  It is from this data that I can confirm the aircraft was recording data correctly for those times it was recording in the flight log.   This can be confirmed if we can establish where the farmer who saw it fly overhead was located.

Edit:  examining the EXIF data in the photos taken during the flight and comparing it to the images taken, it confirms the GPS positions, compass headings, gimbal/camera tilt etc. all correspond exactly with what we see in the photo and what is on the map.  This confirms that the telemetry data is accurate.

endotherm Posted at 2017-4-29 07:17
At the end of the record, the aircraft rotated right and changed heading by about 30°.  This could be a momentary change of a couple of seconds or it could be a more lasting change.  The last leg is too short to make any certain prediction, however it is plausible to assume the wind had changed direction to 70° as the previous leg wasn't choppy and was quite stable for some time.  I don't have a crystal ball to say that the predicted search area would be fruitful, but I have examined quite a few cases such as this one and the data indicates a likely solution.

You can't say "it can't have been drifting" and "it was actively flying at high speed".  The data proves otherwise.  The P3 is incapable of actively flying at that speed by itself.  The only explanation for it to be travelling at that speed is for it to be carried by faster wind.

I'm sure it was assisted by the wind, however when it first went to 400ft it could apparently cope easily with the wind, I wouldn't expect huge and sudden changes in that area yesterday, a 40mph constant wind would be unusual at any time.

How do you explain that when it turned around to fly home under RTH, it pitched forward to fly fast back home, but it started moving away faster?  It seems like it turned to the wrong direction as though it had a compass calibration problem.  You can't trust the yaw, pitch and roll data from the logs if the compass was badly calibrated, it is not accurate data, and so you can't tell if it was flying against the wind or with the wind which means you can't calculate the wind speed.

endotherm Posted at 2017-4-29 07:17
At the end of the record, the aircraft rotated right and changed heading by about 30°.  This could be a momentary change of a couple of seconds or it could be a more lasting change.  The last leg is too short to make any certain prediction, however it is plausible to assume the wind had changed direction to 70° as the previous leg wasn't choppy and was quite stable for some time.  I don't have a crystal ball to say that the predicted search area would be fruitful, but I have examined quite a few cases such as this one and the data indicates a likely solution.

You can't say "it can't have been drifting" and "it was actively flying at high speed".  The data proves otherwise.  The P3 is incapable of actively flying at that speed by itself.  The only explanation for it to be travelling at that speed is for it to be carried by faster wind.

I'm sure you guys have much more knowledge than me.
Esp. when it comes down to AC attitude, reading the spreadsheet...

Didn't know the P3A can't travel that fast by itself.
The kid put probably some rocket boosters on it and never told his father.


For me the irritating point is that I don't see anywhere a home point recording.
So I figured the drone was heading towards some previous HP.

BTW:
In each of my 100+ flight records at Airdata I have the HP confirmation shown.
The same applies to the csv files, containing this text entry: "Home Point Recorded. RTH Altitude: ..."

Eric13 Posted at 2017-4-29 09:49
I'm sure you guys have much more knowledge than me.
Esp. when it comes down to AC attitude, reading the spreadsheet...

It may have something to do with whether it is a CSC/manual takeoff or an autotakeoff swipe from the app.  It is quite common not to see it in the logs, it is nothing to worry about, whether or not there is a marker indicated on the map for a home point.  There was not any indication that a new home point was set, if that is where this line of investigation is going.  Each time an RTH was initiated by this pilot, the aircraft turned to face the correct indicated home point, even when it was flying "backwards", assisted by the wind.  This is exactly what a "blown away" looks like.  This supports the data showing that the compass was not completely useless.  Despite some error messages popping up, there is no evidence of the headings being incorrect.

Nigel_ Posted at 2017-4-29 07:56
I'm sure it was assisted by the wind, however when it first went to 400ft it could apparently cope easily with the wind, I wouldn't expect huge and sudden changes in that area yesterday, a 40mph constant wind would be unusual at any time.

How do you explain that when it turned around to fly home under RTH, it pitched forward to fly fast back home, but it started moving away faster?  It seems like it turned to the wrong direction as though it had a compass calibration problem.  You can't trust the yaw, pitch and roll data from the logs if the compass was badly calibrated, it is not accurate data, and so you can't tell if it was flying against the wind or with the wind which means you can't calculate the wind speed.

Of course it can be dead calm at the surface, yet be quite windy a few dozen meters higher.  You have no way of knowing what is going on at altitude as it is generally quite invisible.  Sure, when it took off and went to 400ft it was in control, I previously mentioned this.  The wind may have been 30mph gusting to 34mph.  Once again, the p3 can handle that speed. It has enough power to correct any lateral forces and stay in control, holding position.  Yet if the wind picks up by 5-10 mph you are going to have problems.  That's nothing drastic/abnormal/unusual.  It's not a huge change, but it's enough.  There is no proof that it was coping with ease in these conditions, it was either stable, or it was out of control.  When we notice it moved off, it lost GPS mode and went to ATTI mode.  There were no pilot inputs holding it in position any more, and GPS mode wasn't contributing anything, so the only forces left is wind.  This is what the data shows.  The logical explanation is that it is wind.  The most obvious solution is probably the right solution.  The counter argument is to ignore all the data as unreliable on the basis of a few error messages, and assume the aircraft developed its own intelligence and decided to fly off on its own, or was so confused it had no idea what it was doing.  Somehow that means it can fly faster than physically possible while maintaining a constant altitude, instead of just corkscrewing into the ground.

When they fly back in RTH the "nose" will turn to point towards home, the front motors slow in relation to the rear, This pitches the aircraft into negative pitch.  The combined motor output spins up to 10m/s, leaving some capacity in reserve to maintain GPS mode stabilisation.   Despite trying, it is unable to overcome the headwind, and it will pitch forward in the correct orientation into the wind, somewhere around -10° to -30° depending on how strong a wind it encounters.  However it will travel backwards as it is pushed away.  RTH will not come home "fast", it's more like medium speed.  If you want full power you need to come home with full forward stick input which did not happen.  I analysed the stick movements.  They weren't present significantly when the aircraft flew off.  They were activated in bursts to rotate the aircraft port and starboard.  The compass headings changed in unison and consistently.  The stick movement data is not dependent on compass calibrations,there is no reason to distrust them.  Therefore there is no evidence that the compass readings are wrong.

For hundreds of entries there were compass readings recorded.  There were no "compass error" warnings associated with them.  Why do we need to discard all of them as untrustworthy?

There is no evidence of a faulty compass calibration, merely a speculation.

All a compass error is going to do is cause confusion regarding its heading and which direction it is facing.  It won't affect the GPS coordinates being logged from a different subsystem.  Even in ATTI, it is recording GPS location, just not relying on GPS for stabilisation.

I'm not privy to DJI's system firmware code, but it is quite logical to assume that the "compass error" message is a sanity check code.  A value that is calculated or expected isn't making sense... "it does not compute".  I haven't looked at them all, nor am I saying that this is the reason, but I saw a couple of examples where there was a stream of data, then there was data loss (broken connection with the RC).  The data resumed for 100-200ms, then there was another gap, then the data resumed.  During the one or two stranded lines of data, it showed a large change in orientation or speed from what was there previously.  It looks like inconsistent data, but it could just as easily be correct in that the aircraft legitimately changed orientation and speed during the blackout.  Curiously, there was a corresponding compass error message reported, but it disappeared once the data stream resumed.  There were numerous dropouts and absent data periods during the flight.

With regards to the speed, it is calculated by the firmware, not me, It's a product of the change in GPS coordinates.  The heading has no bearing on its speed.  It can be easily verified by examining GPS coordinate pairs over time and calculating the difference in distance.  Similarly the roll/yaw/pitch data is derived from an accelerometer which is calibrated independently of the compass.  If there were this many errors in place, there would be a slew of IMU errors reported and the aircraft would barely fly.

I'm sorry to hear. Have you gotten in contact with the EU Support team as Natalia mentioned to have the Go app flight records evaluated?

endotherm Posted at 2017-4-29 10:53
Of course it can be dead calm at the surface, yet be quite windy a few dozen meters higher.  You have no way of knowing what is going on at altitude as it is generally quite invisible.  Sure, when it took off and went to 400ft it was in control, I previously mentioned this.  The wind may have been 30mph gusting to 34mph.  Once again, the p3 can handle that speed. It has enough power to correct any lateral forces and stay in control, holding position.  Yet if the wind picks up by 5-10 mph you are going to have problems.  That's nothing drastic/abnormal/unusual.  It's not a huge change, but it's enough.  There is no proof that it was coping with ease in these conditions, it was either stable or out of control.  When we notice it moved off, it lost GPS mode and went to ATTI mode.  There were no pilot inputs holding it in position any more, and GPS mode wasn't contributing anything, so the only forces left is wind.  This is what the data shows.  The logical explanation is that it is wind.  The most obvious solution is probably the right solution.  The counter argument is to ignore all the data as unreliable on the basis of a few error messages, and assume the aircraft developed its own intelligence and decided to fly off on its own, or was so confused it had no idea what it was doing.  Somehow that means it can fly faster than physically possible while maintaining a constant altitude, instead of just corkscrewing into the ground.

When they fly back in RTH the "nose" will turn to point towards home, the front motors slow in relation to the rear, This pitches the aircraft into negative pitch.  The combined motor output spins up to 10m/s, leaving some capacity in reserve to maintain GPS mode stabilisation.   Despite trying, it is unable to overcome the headwind, and it will pitch forward in the correct orientation into the wind, somewhere around -10° to -30° depending on how strong a wind it encounters.  However it will travel backwards as it is pushed away.  RTH will not come home "fast", it's more like medium speed.  If you want full power you need to come home with full forward stick input which did not happen.  I analysed the stick movements.  They weren't present significantly when the aircraft flew off.  They were activated in bursts to rotate the aircraft port and starboard.  The compass headings changed in unison and consistently.  The stick movement data is not dependent on compass calibrations,there is no reason to distrust them.  Therefore there is no evidence that the compass readings are wrong.

That's a top class investigation by Endotherm and he's put a lot of effort into it.

Labroides Posted at 2017-4-29 17:09
That's a top class investigation by Endotherm and he's put a lot of effort into it.

Yup if you have another drone or a friend who has one maybe he could go to the area and fly over and see if can locate your drone.  

blackcrusader Posted at 2017-4-29 23:37
Yup if you have another drone or a friend who has one maybe he could go to the area and fly over and see if can locate your drone.

I've never tried this.  If a Phantom is sitting on the ground after a successful low battery critical landing, how low do you have to fly with another Phantom to be able to reliably spot it as you fly over at 20mph?


I'm still not happy with the wind explanation.  4 minutes before the end of the log, it was in atti mode, drifting with the wind with zero roll and zero pitch, that should show us the wind speed - 5.81m/s.   At the end of the log, after it has turned to fly towards home and is pitched forward at 27 degrees flying hard forwards, it is now doing a speed of 17.69m/s, which would be fine except that it is doing that speed in the opposite direction to home.  That either requires that there was a huge and consistent gust of wind immediately it tried to fly home which is extremely unlikely, or far more likely, when it turned towards home it actually positioned itself facing away from home due to a bad compass calibration.

Nigel_ Posted at 2017-4-29 23:58
I've never tried this.  If a Phantom is sitting on the ground after a successful low battery critical landing, how low do you have to fly with another Phantom to be able to reliably spot it as you fly over at 20mph?

Well you don't need to be doing 20 mph you can fly slower and do a grid pattern.

Nigel_ Posted at 2017-4-29 23:58
I've never tried this.  If a Phantom is sitting on the ground after a successful low battery critical landing, how low do you have to fly with another Phantom to be able to reliably spot it as you fly over at 20mph?

I'm still not happy with the wind explanation.

Prior to that -4 minute RTH command, you are correct, it was drifting in the wind at about 13mph at 82°.  Prior to that, it was travelling at close to 50mph on the same course.  At that time, the pilot was inputting maximum right stick forward with the wind, which correctly accounts for 35mph.

As we examine the long leg, we see the speed gradually increasing towards 40mph without any input from the aircraft. This indicates the wind was picking up continuously and it sustained a constant speed.  It isn't a gust, which is a sudden increase of the wind's speed that lasts no more than 20 seconds, usually occurring when wind peaks around 20mph, and typically cyclic in 2 minute intervals.  The wind direction was varying around 70° ± 30°.  

At the end of the log when the windspeed picked up to 40mph at 70° the pilot issued a RTH "go home" command with no forward stick input.  It continued on the previous course, but rotated 180° to face home.  It had to pitch forwards as far as it could (no more than about 30° nose-down, or it fails to remain aloft) to combat the headwind.  Because it is in RTH mode, the maximum unassisted speed it supplied is 10m/s.    It is fighting an opposing wind that we know is nearly twice as fast.  So in this configuration it is pitched down, facing home, but flying backwards as the superior wind is overcoming the aircraft speed.  This perfectly explains the scenario, the telemetry and observations.


At  the end of the record, we can see the aircraft is being carried on a  course of 70° while facing towards home at 254° (roughly WSW) at a speed  of 40.8 mph.  It is tilted down almost at maximum at 27.2°, indicating a  maximum headwind or beyond.  It is also slightly rolled, with "left  wing" down 2.9°, suggesting it is also tackling a wind vector from the  left side.  This indicates the wind is coming more from the SW than the  WSW course it is attempting.

Sustained winds at certain speeds, altitude and direction are quite common and likely to occur.  Some are in fact quite predictable, for example around mountains and buildings.  Windspeed and direction can shift due to friction, wind shear and solar ground effects which are also possible factors in this instance.   There is no proof that any of these accepted meteorological facts are "unlikely", aside from your assertion that you believe it to be extremely unlikely.  Similarly, there is no proof that it turned away from home due to a bad compass calibration, aside from your assertion that it is a far more likely scenario.  The telemetry does not support this contention at all.  There is no proof that the data is unreliable.  There are confirmed verifiable occurrences of the data coinciding exactly with physical features, e.g. the EXIF data in the photos taken in flight correspond exactly with the position reported, and the images obtained that could only be obtained if the camera was at the declared height and gimbal tilt. Why did the first half of the flight contain NO compass errors if there was a bad compass calibration?  The reported errors later in the flight seem to be related to signal dropouts with the RC.

If you want to offer some facts or evidence to support your contention, I'd be happy to examine and debate it.  However all we have at the moment is "I reckon.." or "I believe this is more likely", with no supporting evidence.  I'm not sure why you would have difficulty accepting my explanation which is supported by facts.