Analysis of DAT file info

[Mikemal]

HI all. I had an incident this week that almost resulted in the loss of my I1 V2. Fortunately I was in the hover at about 3ft high and it fell into grass, so no damage.

In short, in P mode it suddenly rolled hard left and pitched, falling the 3ft in about half a second.

I retrieved the DAT file, and have used the CSV view to draw the data into graphical format. After literally hours of work, it is still a mystery what caused the roll and pitch (and subsequent yaw). I thought it may have been a motor failure, but I can't find a smoking gun. What I do find is that one motor lost telemetry of all its data at the same time as the roll and pitch started. I'll include this pic for clarity.

I may be wrong, but I think if that motor failed, it would have rolled the other way, unless the compensation from the other motors resulted in a loss in lift and crash. It seems more likely the ESC had a brain f@rt and it kept the motor at max speed? Hard to determine as all the data from that engine is missing.

Could anyone perhaps help me with an analysis of the data? Perhaps you might have seen this before?

Really appreciate the help.

Mike

 

[BudWalker]

HI all. I had an incident this week that almost resulted in the loss of my I1 V2. Fortunately I was in the hover at about 3ft high and it fell into grass, so no damage.

In short, in P mode it suddenly rolled hard left and pitched, falling the 3ft in about half a second.

I retrieved the DAT file, and have used the CSV view to draw the data into graphical format. After literally hours of work, it is still a mystery what caused the roll and pitch (and subsequent yaw). I thought it may have been a motor failure, but I can't find a smoking gun. What I do find is that one motor lost telemetry of all its data at the same time as the roll and pitch started. I'll include this pic for clarity.

I may be wrong, but I think if that motor failed, it would have rolled the other way, unless the compensation from the other motors resulted in a loss in lift and crash. It seems more likely the ESC had a brain f@rt and it kept the motor at max speed? Hard to determine as all the data from that engine is missing.

Could anyone perhaps help me with an analysis of the data? Perhaps you might have seenView attachment 18116 this before?

Really appreciate the help.

Mike

If you can Dropbox the .DAT file and post a link I’ll take a look.

 

[Mikemal]

Excellent, thanks, Bud. I've snipped the file down to just the last 15 seconds or so, as it looks like everything was dead normal up until then. I'll dropbox it now.

 

[Mikemal]

Bud, here's the link. Thanks for the help!

Dropbox - FLY388 data snip.csv

 

[BudWalker]

Bud, here's the link. Thanks for the help!

Dropbox - FLY388 data snip.csv

Please provide the original .DAT via Dropbox.

 

[Mikemal]

Please provide the original .DAT via Dropbox.

Ok, will do. Home in about 44 mins.

 

[Mikemal]

Hi Bud. Here is the dropbox link for the original dat file.

Dropbox - FLY388.DAT

Mike

 

[BudWalker]

Hi Bud. Here is the dropbox link for the original dat file.

Dropbox - FLY388.DAT

Mike

Looks like you did a pretty good job of looking at the incident. The failure was in the leftFront motor/ESC. Isn't that consistent with a pitch down and roll left?
The eventLog stream shows this
444.328 : 24915 ESC1 link down||
445.402 : 24969 [esc_is_stall] status changed: last(0x00000000) != current(0x00000004)

The Inspire hit the ground before it had time to exhibit the spinning that occurs as a result of lost propulsion. The propulsion issue started around 443.43 and the impact was at 444.26.


There is a bit more to see but CsvView 3.1.0 has some problems that I've been fixing. So I'm going to use version 3.2.0 (the next, unreleased version). 3.2.0 should be released in a couple of days. MotorCtrl: PWM:LFront is the signal that represents what the FC wants the leftFront speed to be. It's not a motor/esc measurement. You can see that the FC wants the leftFront to be at 100% indicating that FC determined that there wasn't enough leftFront thrust.

 

[Mikemal]

Looks like you did a pretty good job of looking at the incident. The failure was in the leftFront motor/ESC. Isn't that consistent with a pitch down and roll left?
The eventLog stream shows this
444.328 : 24915 ESC1 link down||
445.402 : 24969 [esc_is_stall] status changed: last(0x00000000) != current(0x00000004)

The Inspire hit the ground before it had time to exhibit the spinning that occurs as a result of lost propulsion. The propulsion issue started around 443.43 and the impact was at 444.26.
View attachment 18240

There is a bit more to see but CsvView 3.1.0 has some problems that I've been fixing. So I'm going to use version 3.2.0 (the next, unreleased version). 3.2.0 should be released in a couple of days. MotorCtrl: PWM:LFront is the signal that represents what the FC wants the leftFront speed to be. It's not a motor/esc measurement. You can see that the FC wants the leftFront to be at 100% indicating that FC determined that there wasn't enough leftFront thrust.
View attachment 18241

Thanks, Bud, I appreciate your time.

Can you confirm the orientation “Left Front” is as per DJI? As they look at the drone from the front, and determine the LF from that orientation, this confuses me a bit. The machine rolled left (as viewed from the back) if I’m correct here. If the DJI “LF” motor failed would it not have rolled the other way?

 

[BudWalker]

Thanks, Bud, I appreciate your time.

Can you confirm the orientation “Left Front” is as per DJI? As they look at the drone from the front, and determine the LF from that orientation, this confuses me a bit. The machine rolled left (as viewed from the back) if I’m correct here. If the DJI “LF” motor failed would it not have rolled the other way?

Viewing the AC from the back as it is facing away from you the numbering starts with the rightFront and goes CCW. In the eventLog stream they are numbered 0 thru 3, i.e. ESC #0 == rightFront, ESC #1 == leftFront, ESC #2 == leftBack, and ESC #3 == rightBack. But, in the .DAT records they are numbered 1 thru 4. Normally, CsvView translates these numbers to the correct labels. But, if DatDefined signals are used the labels don't appear so you have to know the numbering is 1 thru 4.

I've not seen an instance where the motors are numbered as you described.

 

[Mikemal]

Right! That must drive you nuts when doing your software! So, the LF motor and ESC will need to be replaced. Just to make sure I've the correct culprit, please confirm my selection of the "correct" LF motor. I've included a pic. Sorry for sounding like an idiot!

 

[BudWalker]

Right! That must drive you nuts when doing your software! So, the LF motor and ESC will need to be replaced. Just to make sure I've the correct culprit, please confirm my selection of the "correct" LF motor. I've included a pic. Sorry for sounding like an idiot!

That looks correct to me. Are you sure the motor needs to be replaced. I would think it's just the ESC. But, I'm not a motor guy.

BTW, I had a typo. I meant to say that I've not seen an instance where the motors are numbered as you described.

 

[Mikemal]

Good point about the motor. It runs freely, is not old, has never had any damage, so you're probably right. It's a mission getting just the ESC, though, most vendors provide only a combination. Will investigate. Again, thanks for your time, and thanks for providing the software!