My P3P crashed, DJI not cooperating

[msinger]

Others have suggested the tree scenario but after viewing the CSV logs they have also said that they agree it flew directly into the rocks. Looking at the location of the trees, any reasonable person would realize that flying at the trajectory outlined in the logs, hitting those trees, which are probably 40' from the shoreline, the drone would have fallen directly down and would have crashed on land, not in the water.

To be perfectly honest with you, I cannot tell how close the trees are in real life by looking at the poor imagery available in Google maps. Do you have a GPS receiver? If so, perhaps you could go to the location where the log says the Phantom crashed and take a photo at that spot (showing the coordinates on the GPS receiver).

 

[Chris_P]

Thats a great idea. Thank you. If that will help my case with DJI, its certainly worth a shot. I boat in that area every weekend.

 

[msinger]

That of course would only help prove your point that the altitude is incorrect in the logs.

You'd also need to show proof that there are no obstacles higher than ~50 feet between your boat and the takeoff location. Had the Phantom actually climbed to 50 feet, it would have flown back to the home point at that altitude. If there were any obstacles higher than ~50 feet, your Phantom would have crashed there instead.

 

[MikeyOnline]

The TXT and DAT logs both indicate you hit a tree. I don't think they'd come to a different conclusion with the DAT log in hand.

You may remember that I came to the same conclusion initially. To me, both the baro altitude and the Z axis sensor seemed to indicate a ~40 foot fall near the end of the log. I then see some anomalies in the GPS height where GPS height jumps up and down and I think, maybe it was already in the water when those readings show up. I also took the lat/long coordinates out of the DAT file and plugged them into Google Maps and they show the same location as indicated in the log: it never went beyond that rock and was probably 30-40 feet from the nearest tree. At first I thought, maybe you can't depend on that level of accuracy plugging coordinates into Google Maps but I remember every time I've done that, the spot has been accurate within a few feet. There are also other anomalies in the log that I can't explain. At 981.613 seconds in the log (row 29348), the front left motor gets commanded to 10,000 RPM without any stick input from the pilot. At that same instant, the AC registered a -12G acceleration from the X axis sensor. In that same row, the Z-axis sensor registered -4.7G, indicating that it hit something and abruptly stopped and bounced UP. If you plug in the coordinates of that spot, it was right where the water and rock meet. If he hit something at that point, I don't know how it could have been a tree unless the GPS coordinates can't be relied upon. The stop of forward motion and bounce upward would also lend credence to hitting the leading edge of that rock.

Edit: I do try to factor in that, when reading any type of log like this, there can be some lag in parameters where each column is not always sampled at the same interval. Given that, it's possible that the motors have already stopped for example, when looking at the row with the 12G hit, but the motor RPM doesn't update for a fraction of a second. But GPS coordinates, especially with 16 sats, I would think would be accurate to within a few meters.

Mike

 

[Chris_P]

Bing Maps have a much more detailed view of the area ....

 

[msinger]

The stop of forward motion and bounce upward would also lend credence to hitting the leading edge of that rock.

The OP said that's what happened, so let's just assume that actually happened for a minute.

If the Phantom was actually at 50 feet over top of those rocks, what are the chances that there were no obstacles higher than that between that point and the takeoff location? If the answer is no chance, then a crash was inevitable either way.

 

[Chris_P]

I will certainly explore that possibility and take a look at any possible obstacles in the RTH path. You were kind enough to confirm that my RTH altitude was 30M (98 '). Isn't that the measurement that I should use? Another point to consider is I've read of many crashes in trees where the drone survived. This collision with a rock took out my camera and plunged it into the lake resulting in water damage as well.

Due to the incorrect altitude calculations, the aircraft flew directly forward and crashed within seconds. Had it started to rise to the proper RTH altitude, that may have given me enough time to clue in to what was happening and have a chance to stop it.

If I'm reading the logs correctly, I had 9 seconds between the auto RTH function engaging and the impact with the rock.

 

[Rypper1]

With all due respect, there are others who have read the logs and think differently. The Google earth photo is proof that there are no trees nearby. The flight path is plotted from the boat where I was attempting to land directly to the rocks. I understand DJI and their affiliates may suggest otherwise for obvious reasons but the Google earth photo speaks for itself. The only info that suggests any contact with a tree is the unexplainable sudden altitude drop which is pure speculation.

@Chris_P Please remember that msinger is simply giving you his interpretation of your logs. I have never seen him side with DJI to try to help DJI. He has always given his honest opinion from what I have seen.
Even though DJI said they didn't need these files I would send them in. It can't hurt anything.

 

[ferraript]

I have converted the DAT files from the drone to CSV and uploaded here ... maybe someone can make sense of that.

thanks
as I said before, it confirms that barometer's readings were off
and it also confirms, that moment of crash is at offsetTime 981.613
I counted that bird made two or three 360 rollovers in first second after the impact
after another second all sensors went crazy
and then the bird is just rolling down into water
it must have been scary to see something like that
fortunately nobody was hurt

 

[MikeyOnline]

The OP said that's what happened, so let's just assume that actually happened for a minute.

If the Phantom was actually at 50 feet over top of those rocks, what are the chances that there were no obstacles higher than that between that point and the takeoff location? If the answer is no chance, then a crash was inevitable either way.

True. Although a little more time might have been enough to figure out that RTH was on and cancel it. Also, hitting a pine tree would be preferable to hitting a rock. You can sometimes get away with no damage hitting a pine. But the more I look at this, the more questions I have. It's become a bit of a mystery investigation for me. The picture that was posted above the boat appears to show the Phantom well above the level of said rock. So I still think it had to lose quite a bit of altitude before hitting that rock but given where the photo was taken, both the GPS altitude and baro altitude say that the 12G deceleration (our assumed crash) occurred at the same altitude as the photo was taken (within 10 feet). It shows you were within about 10 feet of the same altitude for most of the last part of the flight. Given that photo, plus your trajectory and altitude, and assuming some "slop" in Google Maps/Earth, I assume DJI concluded that the only thing you could have possibly hit was a tree. I actually looked up Google Maps and Google Earth accuracy and found people reporting accuracy as good as 2.4m in some locations but also some studies that showed a 24m average variation. So whatever happened, my guess is that DJI considers positioning on the map as only a general indication of where you actually were. Keep in mind of course, that all of this is irrespective of what actually happened or what was observed because I assume DJI can only go by what the log shows.

Mike

 

[Chris_P]

@Chris_P Please remember that msinger is simply giving you his interpretation of your logs. I have never seen him side with DJI to try to help DJI. He has always given his honest opinion from what I have seen.
Even though DJI said they didn't need these files I would send them in. It can't hurt anything.

I absolutely appreciate msinger's input. He was one of the first people to respond with a very detailed analysis of what happened. I just don't understand how/why his interpretation of the logs differ from my recollection of what happened. If I hit a tree I would own it and happily accept DJI's offer. That wasn't the case and I hope a thorough review by DJI confirms that.

 

[Chris_P]

True. Although a little more time might have been enough to figure out that RTH was on and cancel it. Also, hitting a pine tree would be preferable to hitting a rock. You can sometimes get away with no damage hitting a pine. But the more I look at this, the more questions I have. It's become a bit of a mystery investigation for me. The picture that was posted above the boat appears to show the Phantom well above the level of said rock. So I still think it had to lose quite a bit of altitude before hitting that rock but given where the photo was taken, both the GPS altitude and baro altitude say that the 12G deceleration (our assumed crash) occurred at the same altitude as the photo was taken (within 10 feet). It shows you were within about 10 feet of the same altitude for most of the last part of the flight. Given that photo, plus your trajectory and altitude, and assuming some "slop" in Google Maps/Earth, I assume DJI concluded that the only thing you could have possibly hit was a tree. I actually looked up Google Maps and Google Earth accuracy and found people reporting accuracy as good as 2.4m in some locations but also some studies that showed a 24m average variation. So whatever happened, my guess is that DJI considers positioning on the map as only a general indication of where you actually were. Keep in mind of course, that all of this is irrespective of what actually happened or what was observed because I assume DJI can only go by what the log shows.

Mike

Thanks Mike, bear in mind though that when that picture above the boat was taken, the Phantom was at a much higher altitude. After that picture was taken, I flew it around behind the boat and I was bringing it in to land so it wasn't that high above the boat when the auto RTH engaged.

 

[ferraript]

both the GPS altitude and baro altitude say that the 12G deceleration (our assumed crash) occurred at the same altitude as the photo was taken (within 10 feet).

not true
photo was taken at 16:51:12.230 (flightTime/offsetTime 925.3)
at that time bird was 14 m higher than in the moment of crash

 

[msinger]

Bing Maps have a much more detailed view of the area ....

Yes -- I took a look at Bing's map too. It's still a poor image and I'm not sure when it was last updated. Here's an interesting look at it though:

 

[msinger]

I just don't understand how/why his interpretation of the logs differ from my recollection of what happened.

Like I said a few times now, I'm simply telling you what the logs show.

 

[MikeyOnline]

not true
photo was taken at 16:51:12.230 (flightTime/offsetTime 925.3)
at that time bird was 14 m higher than in the moment of crash

You're right of course. I show 13m... but I was thinking feet. Forgot it was meters.

Mike

 

[Chris_P]

Yes -- I took a look at Bing's map too. It's still a poor image and I'm not sure when it was last updated. Here's an interesting look at it though:

View attachment 56947

The location of your line in the photograph is quite a bit off though. If you look closely you will see three people in the photo. Two adults and a child. Those are my friends who fished it from the water seconds after impact before it sank into the bay. The rock that it hit is directly in front of where they are standing.

The tree theory is also based on the 40' fall in altitude. There are no 40+' trees anywhere near the crash site. Hitting one of those shrubs near the water would have been much better !

 

[msinger]

The location of your line in the photograph is quite a bit off though.

That line was just my estimation based on your photo. The Bing photo is pretty lousy, so I did my best to visually line it up based on the shape of the land.

 

[Meta4]

as Oso wrote, you can use the DAT converter to get CSV file
it contains barometric altitude but also GPS altitude, so we can see the difference

GPS altitude data is completely unreliable and cannot be relied on for anything.
As well as having a large variable error, they shift substantially over short time frames.
The barometer is much, much more reliable as a source of altitude data.

Here's an explanation from Garmin
How accurate is the GPS elevation reading?

GPS heights are based on an ellipsoid (a mathematical representation of the earth's shape), while USGS map elevations are based on a vertical datum tied to the geoid (or what is commonly called mean sea level). Basically, these are two different systems, although they have a relationship that has been modeled.

The main source of error has to do with the arrangement of the satellite configurations during fix determinations. The earth blocks out satellites needed to get a good quality vertical measurement. Once the vertical datum is taken into account, the accuracy permitted by geometry considerations remains less than that of horizontal positions. It is not uncommon for satellite heights to be off from map elevations by +/- 400 ft. Use these values with caution when navigating.

 

[Meta4]

I will certainly explore that possibility and take a look at any possible obstacles in the RTH path. You were kind enough to confirm that my RTH altitude was 30M (98 '). Isn't that the measurement that I should use? Another point to consider is I've read of many crashes in trees where the drone survived. This collision with a rock took out my camera and plunged it into the lake resulting in water damage as well.

Due to the incorrect altitude calculations, the aircraft flew directly forward and crashed within seconds. Had it started to rise to the proper RTH altitude, that may have given me enough time to clue in to what was happening and have a chance to stop it.

If I'm reading the logs correctly, I had 9 seconds between the auto RTH function engaging and the impact with the rock.

The Phantom initiated RTH at 16:09 at (indicated) altitude of 62 feet and commenced RTH.
It ascended to 65 feet but joystick input from 16:16 cancelled the ascent to the set RTH height.
In this 7 second period, the Phantom had flown 25 feet and reached a velocity of 11 mph.
Further joystick input continued the descent down to (indicated) 54 feet as RTH continued for 6 seconds until impact at 16:22.

The difference between the altitude you observed might be explained by a pressure drop Ferraript mentioned in post #42.
If this was the case, the disparity between indicated and observed altitudes could have been observed in the app.
The Phantom did not ascend to the set RTH height because of joystick input during the ascent phase, as described in the manual.
The home point had not been reset during the flight.
This would also have been indicated in the app.
Once RTH commenced, there was 13 seconds of flight covering 200 feet.
During this time the Phantom could have been flown to a safe altitude or RTH could have been cancelled.

The Phantom was in RTH with the operator nearby and able to clearly see what was happening.
In the end it doesn't matter whether it crashed into a rock or a tree.
Without appropriate operator intervention, the Phantom was going to crash into one of them.