Atti Mode Drift Crash

[alpinerose]

Have posted files here in hope to understand, more granularly, what happened to me. After thirty flights from this same location and manual take-offs, the A/C dropped out of Opti GPS into Atti Mode and rapidly move away in a no wind situation. Estimate that it was all over in about 12 seconds, from manual Prop Spin-up, establishing home point, hovering at 10 feet to scan read-outs before mission, to commanding an emergency motor stop, with impact and crash. With a previous flight being a dress rehearsal for the failed one, executed flawlessly just thirty minutes earlier. The comments and experience I learn here will not bring my Mavic back any sooner, now just another box with a RMA label inside the cavernous service hall at Carson. And I am not sure how to assess going forward if the possibility of flying a mavic includes such unpredictable maneuvers. Some manual override in the controls may have prevented this crash, but timing and reaction is everything, right?

Here is another view - HealthyDrones.com - Innovative flight data analysis that matters

Thanks for any productive comments.

 

[Mavicsailor]

I'm having problems opening your zip file attachment and the healthy drones info really doesn't show much usefully data. You may want to use Dropbox to allow others to view your dat file.


Sent from my iPhone using MavicPilots

 

[alpinerose]

Good Morning -

Here are the csv files in dropbox -

 

[Mavicsailor]

It would have been best to have the .dat file downloaded directly from the Mavic. This .dat file would provide more information. Anyhow, I haven't seen anyone give this analysis a shot and I'm still learning but here goes.

It looks like the AC dropped into ATTI directly after the GPS Health suddenly dropped from 5 to 1 at 11 seconds into the flight. (purple shading on bottom graph)The number of satellites was 15 and dropped down as low as 3 over the next 3 seconds. During the 3 seconds from the time the GPS health dropped from 5 to 1, the AC moved about 20 feet at speeds up to 12mph. The roll and pitch during this period was all over the place, and only RC rudder and full down throttle appeared to be used, which should only make it rotate and drop, and not move 20 feet away. (Alerion not shown or used). Maybe this was a result of hitting the tree? (Downloaded .dat file would have better been able to determine this) The op stated they did an emergency motor stop, which I assume he meant a CSC. There was no evidence of that. Perhaps they started to do that with the downward throttle and rudder, but the rudder stick was in the wrong direction for a CSC, and the other 2 positions needed for this were not touched. My understanding is CSC will not work on the Mavic using the default settings unless during a critical error. (unless this option is deselected) The timing to drop 10 feet with full down throttle appeared about right. Like the op said, this all happened really quick....a controllable AC to a crash in 3 seconds.

It also looks like the AC started drifting a few seconds prior to the gps drop out. There didn't appear to be substantial stick input that I could see that might cause this. Wondering if all of the trees around could have contributed to loss of gps?

This occurrence may be similar to what others have experienced with the AC very suddenly dropping into Atti for no apparent reason as the GPS Health suddenly drops from 5 to 1. This combined with the inability in some cases to control the AC in ATTi, as has also been reported even without wind, appeared to result in this crash. If the AC was at a higher altitude it may have had the room and time to recover if GPS health increased as could happen within a few seconds.

I really wish DJI would get to the bottom of both of these issues as there have been several reported cases. The AC goes from predictable to chaotic in seconds. Most could handle the AC if it dropped into Atti unless a strong wind, but darting away in no wind is unacceptable and dangerous.

I found it interesting that the pitch was up to -7 degrees while hovering with no reported wind. Maybe just some bouncing around while climbing.

I wonder if the inability of some AC to fly well in Atti has to do with the pilot still using the factory IMU calibration?


For the experts out there, is there anything different or in addition to that listed above?

 

[BudWalker]

It would have been best to have the .dat file downloaded directly from the Mavic. This .dat file would provide more information.
Anyhow, I haven't seen anyone give this analysis a shot and I'm still learning but here goes.

It looks like the AC dropped into ATTI directly after the GPS Health suddenly dropped from 5 to 1 at 11 seconds into the flight. (purple shading on bottom graph)The number of satellites was 15 and dropped down as low as 3 over the next 3 seconds. During the 3 seconds from the time the GPS health dropped from 5 to 1, the AC moved about 20 feet at speeds up to 12mph. The roll and pitch during this period was all over the place, and only RC rudder and full down throttle appeared to be used, which should only make it rotate and drop, and not move 20 feet away. (Alerion not shown or used). The op stated he did an emergency motor stop, which I assume he meant a CSC. There was no evidence of that. Perhaps he started to do that with the downward throttle and rudder, but the rudder stick was in the wrong direction for a CSC, and the other 2 positions needed for this were not touched. My understanding is CSC will not work on the Mavic using the default settings unless during a critical error. (unless this option is deselected) The timing to drop 10 feet with full down throttle appeared about right. Like the op said, this all happened really quick....a controllable AC to a crash in 3 seconds.

This occurrence is similar to what others have experienced with the AC very suddenly dropping into Atti for no apparent reason as the GPS Health suddenly drops from 5 to 1. This combined with the inability in some cases to control the AC in ATTi, as has also been reported even without wind, appeared to result in this crash. If the AC was at a higher altitude it may have had the room and time to recover if GPS health increased as could happen within a few seconds.

I really wish DJI would get to the bottom of both of these issues as there have been several reported cases. The AC goes from predictable to chaotic in seconds. Most could handle the AC if it dropped into Atti unless a strong wind, but darting away in no wind is unacceptable and dangerous.

I found it interesting that the pitch was up to -7 degrees while hovering with no reported wind. Maybe just some bouncing around while climbing.

I wonder if the inability of some AC to fly well in Atti has to do with the pilot still using the factory IMU calibration?
View attachment 5972

For the experts out there, is there anything different or in addition to that listed above?

The only thing I can add here is that, at first glance, this gpsHealth drop seems to be different than several others that I've looked at. In those incidents gpsHealth is independent of numSats; i.e. gpsHealth isn't dropping because of loss of satellites. Here, there seems to be a correlation.

In situations where a numSats drop causes a gpsHealth drop the numSats drop precedes the gpsHealth drop. Here, the opposite occurred. I suspect the numSats drop was caused by the AC losing altitude into what looks like a tree.



@Mavicsailor, just a tip. It's easy to transfer the contents of the SigPlayer (or GeoPlayer) to a post. Just click the "Copy to System Clipboard" button at the top of the Player. Then place the cursor in the post and do a ctrl-V. Takes 3 or 4 seconds.

 

[Mavicsailor]

Thanks Budwalker! And thanks much for creating csvview to help with flight analysis. An incredible tool. Yeah...I saw the tree after going back and looking at the Healthy Drone picture after my post. (And edited that into comments)

I'm still learning. I bow to the experts.

 

[GA_Mavic]

@alpinerose Just out of curiosity, what phone/tablet were you using?

 

[Mavicsailor]

The only thing I can add here is that, at first glance, this gpsHealth drop seems to be different than several others that I've looked at. In those incidents gpsHealth is independent of numSats; i.e. gpsHealth isn't dropping because of loss of satellites. Here, there seems to be a correlation.

In situations where a numSats drop causes a gpsHealth drop the numSats drop precedes the gpsHealth drop. Here, the opposite occurred. I suspect the numSats drop was caused by the AC losing altitude into what looks like a tree.
View attachment 6005
View attachment 6006

@Mavicsailor, just a tip. It's easy to transfer the contents of the SigPlayer (or GeoPlayer) to a post. Just click the "Copy to System Clipboard" button at the top of the Player. Then place the cursor in the post and do a ctrl-V. Takes 3 or 4 seconds.

Budwalker,

In your analysis of many flights, what do you normally see for pitch and roll while hovering at low altitude with no wind? On my Mavic, both pitch and roll are <.5. I've learned to use these measurements as a gauge of how windy it is as altitude increases and the AC attempts to maintain position.

Wondering if pitch and roll are higher when hovering (no wind) would be an indicator of the IMU calibration being off, and ultimately causing difficult handling in Atti? In the case of this accident, the pitch seemed higher when hovering with no wind.

 

[BudWalker]

Budwalker,

In your analysis of many flights, what do you normally see for pitch and roll while hovering at low altitude with no wind? On my Mavic, both pitch and roll are <.5. I've learned to use these measurements as a gauge of how windy it is as altitude increases and the AC attempts to maintain position.

Wondering if pitch and roll are higher when hovering (no wind) would be an indicator of the IMU calibration being off, and ultimately causing difficult handling in Atti? In the case of this accident, the pitch seemed higher when hovering with no wind.

I guess I haven't paid that much attention to pitch and roll at low altitude with no wind. I can't really make that assessment with other pilot's .DATs since I don't know what the wind was. But, here is a .DAT from my P3A I did yesterday when there was no wind.

At the marker the height was about 15 feet, it was hovering, and the pitch and roll were both close to 0.

I get your point about incorrect IMU data. In GPS mode it wouldn't be a problem if the IMU data were biased since the FC will make adjustments to the motors based on GPS data. In ATTI mode all it has is the IMU data.

 

[Mavicsailor]

I guess I haven't paid that much attention to pitch and roll at low altitude with no wind. I can't really make that assessment with other pilot's .DATs since I don't know what the wind was. But, here is a .DAT from my P3A I did yesterday when there was no wind.
View attachment 6013
At the marker the height was about 15 feet, it was hovering, and the pitch and roll were both close to 0.

I get your point about incorrect IMU data. In GPS mode it wouldn't be a problem if the IMU data were biased since the FC will make adjustments to the motors based on GPS data. In ATTI mode all it has is the IMU data.

Thanks,

Will be interesting to look at pitch and roll at hover in future analysis where the AC took off unexpectedly without the influence of wind if dropped into Atti, and whether an IMU calibration had been done since new. Problem is some pilots are so anxious to fly they don't take the time to assure everything appears ok in hover before heading off.

 

[BudWalker]

.... Problem is some pilots are so anxious to fly they don't take the time to assure everything appears ok in hover before heading off.

This won't help much for this problem since, presumably, the AC is in GPS mode prior to launch. I guess launch prior to GPS mode would uncover some problems. But, doing so could be problematical. Not the least of which is facing the disrespect of your fellow pilots

BTW, with regards to wind estimation there is a value that DatCon computes called thrustAngle that is derived from motor speed. But, it's just an angle; i.e. it doesn't include magnitude. Maybe I'll look at that some more and try to include a magnitude.

 

[Mavicsailor]

This won't help much for this problem since, presumably, the AC is in GPS mode prior to launch. I guess launch prior to GPS mode would uncover some problems. But, doing so could be problematical. Not the least of which is facing the disrespect of your fellow pilots

BTW, with regards to wind estimation there is a value that DatCon computes called thrustAngle that is derived from motor speed. But, it's just an angle; i.e. it doesn't include magnitude. Maybe I'll look at that some more and try to include a magnitude.

You are a true champion for our hobby!
Thanks

 

[alpinerose]

@alpinerose Just out of curiosity, what phone/tablet were you using?

iPad Mini 4 WiFi only

 

[erkme73]

Is the fact that it moved horizontally at 12MPH w/o wind or stick input telling at all? Remember @BudWalker, when I had the TBE issues, when ATTI finally kicked in, my MP would dart off in whatever direction the TBE happened to swing it.

 

[Mavicsailor]

iPad Mini 4 WiFi only

OP,

Was the erratic flight after it dropped GPS and it hit a tree, or did it just hit the ground? Appeared to be traveling as this occurred.


Sent from my iPhone using MavicPilots

 

[alpinerose]

@alpinerose Just out of curiosity, what phone/tablet were you using?

OP,

Was the erratic flight after it dropped GPS and it hit a tree, or did it just hit the ground? Appeared to be traveling as this occurred.


Sent from my iPhone using MavicPilots

Yes, the drift occurred after it dropped GPS. It was in clear, still air and stable at the time. Estimating four seconds between recognizing vehicle was drifting to attempting emergency motor shutdown. I found it wedged between my neighbor's garage and adjoining fence, still powered. Feel fortunate it impacted where it did since it was rapidly departing the area. Thanks.

 

[BudWalker]

Is the fact that it moved horizontally at 12MPH w/o wind or stick input telling at all? Remember @BudWalker, when I had the TBE issues, when ATTI finally kicked in, my MP would dart off in whatever direction the TBE happened to swing it.

I took another look at FLY037. It had been toilet bowling. Seems like when the switch to ATTI happened it just started to drift in pretty much a straight line.


I didn't notice this the first time we looked at this. But, gpsHealth had drooped to 1 during the toilet bowling and this is why it switched to ATTI.

 

[ksull72487]

Here is a TIP go indoors cut your signal 100% to GPS. Fly and hover and maybe rotate it. If it drifts or dosent want to stay put something is up with your compass or IMU. Recalibrate. It's a good idea too check. When the aircraft has GPS signal you won't entirely know if it's calibrated correctly, the aircraft automatically utilizes its control to stay put with GPS. So I would do a test in ATTI mode every once in awhile to make sure it hasn't gone wack otherwise if you loose GPS and something's off you'll loose control.

Not a bad idea to recalibrate the controller as well if the problem persists.

The short story is the GPS speeds and slows props down automatically to keep still. In attimode it's mostly manual. So if the drone isn't calibrated correctly and it dosent know it on a switch the handling could go out of whack.

That's my guess as to what happened and why DJI needs to bring ATTI mode back as to where it can be enabled.

When you do this little test make sure you have space the sonar and downward sensors will probably try to keep it still so it helps if you can get a little altitude, be careful as well.

You will get a warning to recalibrate an IMU normally if the aircraft feels like it has to fight something or put to much input in. But it only takes the slightest bit of input to throw it off. I'm sure this will get more reliable with updates as DJI gathers information.


Sent from my iPhone using MavicPilots

 

[Mavicsailor]

Yes, the drift occurred after it dropped GPS. It was in clear, still air and stable at the time. Estimating four seconds between recognizing vehicle was drifting to attempting emergency motor shutdown. I found it wedged between my neighbor's garage and adjoining fence, still powered. Feel fortunate it impacted where it did since it was rapidly departing the area. Thanks.

Curious, had you ever done an IMU calibration? (I know it says you don't need to do one)

Here is a TIP go indoors cut your signal 100% to GPS. Fly and hover and maybe rotate it. If it drifts or dosent want to stay put something is up with your compass or IMU. Recalibrate. It's a good idea too check. When the aircraft has GPS signal you won't entirely know if it's calibrated correctly, the aircraft automatically utilizes its control to stay put with GPS. So I would do a test in ATTI mode every once in awhile to make sure it hasn't gone wack otherwise if you loose GPS and something's off you'll loose control.

Not a bad idea to recalibrate the controller as well if the problem persists.

The short story is the GPS speeds and slows props down automatically to keep still. In attimode it's mostly manual. So if the drone isn't calibrated correctly and it dosent know it on a switch the handling could go out of whack.

That's my guess as to what happened and why DJI needs to bring ATTI mode back as to where it can be enabled.

When you do this little test make sure you have space the sonar and downward sensors will probably try to keep it still so it helps if you can get a little altitude, be careful as well.

You will get a warning to recalibrate an IMU normally if the aircraft feels like it has to fight something or put to much input in. But it only takes the slightest bit of input to throw it off. I'm sure this will get more reliable with updates as DJI gathers information.


Sent from my iPhone using MavicPilots

May be best to order the new prop guard first before trying inside! There are materials that will block the GPS making it drop into Atti. Best to try in a wide open space with no wind.