Mirek6
Second Officer
Flight distance : 609724 ft
Canada
Offline
|
NTQLT,
This will be loooong post.
Your log is thoroughly fascinating. Bare with me in this journey – I will recount what happened and why it happened, with my summary and my findings below.
7.9 sec – GPS mismatch. Altitude 10.5 metres and over the house. Spark drifts NNW even though it is in GPS lock and you command it to fly straight up.
GPS mismatch here means that it is in wrong position – and it is. It changes its physical geo position while his commanded position (the one you give it through sticks) stays the same.
10.5 sec – Another GPS mismatch. Altitude 15.1 metres and still over the house. Drift continues to SE – still no directional input from sticks.
This is sign of trouble to come.
15 sec – still raising. Altitude 24 meters. Breeze from NWW with speed about 2 to 2.5 m/s. Spark cannot hold its position and it is pushed slowly to SEE.
This is trouble. IMU is confused. Spark should be able to hold in one geo spot.
37 sec – still raising. Altitude 69 metres. Wind picks up speed. Still blowing from NWW with speed about 3m/s. Spark continues to drift slightly to SEE even though it should stay put in one spot. IMU is confused. Spark should be able to hold in one geo spot.
1 minute – you stopped raising at 116 metres. Spark is hunting to stay in place. Steady breeze at 3m/s from NWW.
1m 22s – first compass error caused by Spark inability to hold position. It is being pushed to south by few metres while there is no directional input from the sticks and wind is not strong. This means that GPS position change without directional input is confusing Spark. Compass/IMU/GPS combo is not functioning well.
1m 30s – you start raising Spark up again.
2m 32s – altitude 237 m. Compass error again. Spark is confused. IMU and compass are confused. Impending doom!.
Wind from NWW – same speed ~3m/s.
At this moment Spark had enough of a drift and fires up its motors on its own to fly back to its original position.
However, and I have not seen anything like this before, it starts to fly in the toilet bowl effect pattern in wrong direction.
Its pitch / roll and yaw show circular counter clockwise pattern starting in SE direction while, it flies NW in a clockwise pattern.
All is all backwards. It looks like your IMU gyroscope shows South instead of North.
I must see your video (if you were filming) to confirm. Your Spark must have been facing totally opposite direction than your compass believed it was facing.
2m 44s – altitude 260m. You stop raising but clockwise toilet bowl effect spin is in full swing. Spark is already 80 metres north of you.
You are not reacting at all. No movement of your sticks.
By 2m 51s Spark’s systems are so confused that it drops to ATTI. In a massive compass / IMU / GPS failure like this, it is correct reaction.
You are watching and doing nothing to change Spark’s direction.
Spark continues by pure inertia to the south for a few seconds and than it is blown by the NWW wind.
Perhaps it is too high for you to see, but you do not try to change its direction at all. Not even once you pushed on aileron or throttle sticks.
You are just changing the direction the Spark is facing and its gimbal position. Which does nothing to Spark’s trajectory.
At 3m 9s you finally realised that something is totally wrong, and you start landing it.
Spark is at altitude of 260 m – long way to go.
At 4m 13s you give up. Most likely you do not even see the Spark. It is at 188m of altitude and 297 metres from your home.
It drifts with the wind.
At 4m 20s, 316 metres from your home signal breaks. Perhaps you started to move, redirected antennas, stepped behind some obstacle to have clear VLOS with Spark?
When signal is restored 25 seconds later, Spark is already auto-landing.
It started auto-landing with speed between 2.6 and 3m/s 20 seconds after signal was disconnected. This is correct behaviour of Spark even though DJI refuses to admit or confirm it (I asked multiple times) and it is hard to check for sure (hard to force ATTI without messing around with Spark’s firmware and invalidating Spark’s warranty).
But I have seen enough videos when Spark starts to land in ATTI when signal breaks to figure out that it must be the pattern.
Auto-landing sequence started 20 seconds after you lost the signal, which is consistent with what DJI program into Spark when it loses its signal when flying by phone only – after 20 seconds it starts RTH sequence. Since it could not do it here (there is no RTH in ATTI) it started to land.
This failsafe procedure saved your Spark – otherwise it would have drifted into oblivion until battery died down.
UPDATE: (Jul 13, 2018) - My calculations which resulted in a claim that Spark starts autolanding 20 seconds after signal lost when in ATTI are incorrect. Most likely scenario is that it takes over 3 and less than 10 seconds for Spark to start landing but descent is slower for the first few seconds due to inertia. Than Spark descends with speed of 2.6 to 3 m/s. This case showed that Spark dropped 29 metres within 25 seconds of disconnect and than continued autolanding at varying speeds between 2.6 and 3 m/s.
For more explanation see my post: https://forum.dji.com/forum.php?mod=viewthread&tid=156372&page=3#pid1459031
Signal totally breaks over half a kilometer from home.
Summary
1. I am not here to moralize, but you did break the laws of your country. You max altitude should be set to 120 metres. But, hey, maybe you did not know. Now you do :-).
2. Your flight showed abnormalities from the very beginning. My bet is on one or more of the following:
a. You started from the point with a lot of steel and magnetic interference – your home is such a place and if you have interlock or concrete patio, this could be bad. Your compass must have been confused when starting and than “unwinded” as Spark wen up causing confusion.
b. Your IMU and/or compass are not calibrated properly. Please calibrate your IMU on a flat surface and with no vibrations. Calibrate your compass away from any magnetic and electromagnetic interference and away from any steel. Even your mobile, your iWatch, your steel bracelet must be at least a meter from the Spark you are calibrating. Do not calibrate close to rock formations or concrete. Just go outside your house and do it in the middle of the field. When you are calibrating compass, turn the Spark slowly. Do not do it fast.
c. Your Spark IMU is busted. But before you come to this conclusion do your calibration and do not start from places which may interfere with your compass.
3. Post the video of your flight (if you filmed it). I need to check what the hell happened with this wrong direction of toilet bowl effect. Confirmation that Spark was 180 degrees turned to the yaw direction indicated in logs would be a big win for all forum members. This is so unusual that I must see it.
4. After following steps above try to fly. Push the Spark to max but within your VLOS and not too high – just in case. You need to do it to confirm that you do not have a faulty Spark.
Good luck and please keep us informed of your experiments. And don’t forget to post video :-).
Mirek
|
|