Nearly Crashed: Don’t Know Why, and Don’t Know How It Was Saved

General Drone Discussion
1

Basic Story
Planned a Litchi flight for my Mavic Air along a river: good line of sight available all the way, with a short cut back to the starting point over two car parks.

image
image997×457 237 KB

Waypoint height: 90-110m AGL, cruising speed 25 km/hr, 9 mins estimated, 5’30" actual

Yesterday
Flew at ~3pm, and the mission went exactly to plan. The weather was dull, and the pictures weren’t spectacular:
image

The sun popped out for a minute, so I used my second battery for another go at the same mission. However, the sun disappeared quickly, so disappointing results again:
Screenshot 2021-04-29 at 16.20.23

Today
Spotted the sun from my hotel room, and still had one battery left, so ran the same mission again.
This time things went badly wrong between Waypoints 14 and 15 - see 4K YouTube video:

(I will leave this video up for a few days)

This was the scariest point:

image
image1104×622 247 KB

When I was supposed to be 90m in the air over a car park, I was 2-5m above the flat roof of a two-storey building.

Analysis
I have looked at the flight record:

I may accidentally broken the Mavic Air speed record:

image
image935×685 271 KB

Other than that, the data tells me very little. The batteries seem fine, there are 13 satellites, and even when the flight was 70m lower than intended, it’s got 68% radio signal…

Lastly…

I have no idea how the death spiral was averted. To the best of my knowledge, I didn’t touch the sticks. I think a message came up on my screen, and I remember thinking from the view on the screen that I was far too low, but soon the familiar Mavic Air noise was heard above me.

So

Any idea what caused this?
Is it going to happen again?

2

Hit by a bird?

1 Like
3

I would say possibly interference from the aerials and sat dishes on the BBC building maybe

1 Like
4

I had similar with my Mavic Air back in 2918. It was happily following a Litchi mission when it suddenly spun and recovered with the speed apparently reaching just over 100 mph. I have a suspicion that the local Royal Artillery base, who both operate drones and are involved in anti-drone measures were involved.

So my immediate thoughts in your case are some form of interference and I notice a couple of dishes on top of a building just before things kick-off. They may be receiving but the building is the Glasgow BBC studios so I suspect they may also be transmitters - in which case …

1 Like
5

Wow, that’s a proper fill your pants moment right there :scream: can’t imagine what was going through your head watching that happen.

6

Wow that was a close one! I would agree, some interference of some kind. Happy it recovered ok :+1:t2:

7

Thanks everyone for your thoughts…

@Avianskipdiver : there’s no sign of birds at any point, and given the drone swings wildly, you would think that a wing or something would be visible. There certainly are birds seen in the film, however, so it’s possible…

@scottydog, @macspite and @SirGunner : the tricky things with the interference theory are:

  1. it didn’t happen on the two identical missions before this one
  2. the disruption in the flight starts about here:
    image
    This is nowhere near those nasty large dishes on the top of the BBC roof - they’re visible in the picture above. In fact, it recovers when it is in front of those beams…
    The thing that cheers me up about the interference theory is that it means there’s little wrong with the drone, and it will fly again (in a different place!)

@McSteamy2010 : strangely, there was no “pants filling”. I think I saw a message on my phone screen (possibly about a motor error), but by the time I could start worrying, I could hear the Third Twin was on its way home to me. However, watching the film has me more worried - too close to cars, people and buildings.

I was rather hoping that someone could give the log a look to find out if there is anything extra tucked in there that I don’t understand…

8

Bit off topic, isn’t your horizon tilted?

9

Close call that. Glad it got back to you ok.

2 Likes
10

Brown underwear moment indeed.
Glad to see that landed safely after this drama.

11

I offer up another hypothesis.

If we dismiss the RF interference, the dishes on the BBC operate in the Ku Band (12GHz to 20GHz) and are mainly used for receiving files for in-studio editing. Dismiss operator error as the drone was operating autonomously, then another possibility arises that the data from Litchi was somehow corrupted causing the drone to spazz out and later correct itself. This may have credence as Rob also experienced a very similar issue using Litchi for a Waypoint mission.

RF interference needs to be caused by one or more of the following.

  1. Co-channel interference. Interference from a signal on the same frequency.
  2. Adjacent Channel Interference. The interfering signal needs to be suitably close in frequency as to break through the filtering of the receiver.
  3. Harmonic Interference. The interfering signal is on a sub integer frequency of the receiver’s tuned frequency.
  4. The interfering signal is so harmonically rich (filthy) it becomes extremely broadband. This pretty much went away when Spark Gap Transmitters stopped being used

There is another form of interference whereby the interfering signal induces unwanted current flow in the drones circuitry through close coupling, but in such a case the drone would need to be very, very close to the source of interference, e.g. within a few feet of a very high power transmitter.

3 Likes
12

Looking at the Airdata log there is a big current draw right around the point it went crazy. Could it be a physical issue with one of the motors maybe?

Screenshot_20210430-095523_Chrome
Screenshot_20210430-095523_Chrome1440×3088 526 KB

Screenshot_20210430-095629_Chrome
Screenshot_20210430-095629_Chrome1440×3088 655 KB

13

have you gone back to Litchi to verify it wasnt a bug their end? Thats where I would start…

14

Just wondering, if you put the Litchi flight into Google Earth as a Virtual Litchi Mission would it display any strange behaviour.

I’m not sure what the output of a Litchi mission is when uploaded to an aircraft but it loads in a couple of seconds, I doubt that it is that complex - probably a *.csv type list of waypoint values - speed, height, bearing, gimbal attitude. I don’t see that causing a mechanical malfunction.

Maybe someone could enlighten me as to what is sent from the Litchi program to the aircraft?

15

scary stuff right enough …plus you went over the main road just after its tumble … hope you are insured …that could have been an interesting one had it crashed…

Whats the steamer there? Obv not the waverly as its off to get new boilers i think … looks like the old CalMac QM2 from the 70’s …

16

Again, thanks for the consideration that you have all given to my dilemma.

@TassyWass : Certainly by the time of the incident, the horizon was significantly askew. However, it looks normal enough at the beginning of the flight:
image

@Nidge : Agreed that radio interference is an unlikely cause. Thank you for the detailed technical analysis.

@SirGunner : Thank-you for the observation of the massive current draw - hadn’t seen that before - and for the suggestion of a possible motor error. I have spent this afternoon (i.e. three batteries worth) trying to get the motor to fail at 2.5m in the air, and it hasn’t. I eventually took it up to 100m in the local park and it just wasn’t failing.

@Gkinghrn : I”m not sure about going back to Litchi on this. The mission worked twice, and the third time I would have needed to deliberately unlock the mission before flying it, and then accidentally set an impossible waypoint. I don’t remember either of these things happening.

You raise an excellent point about insurance, and it’s about time I got some.

I am afraid that I am not certain of the name of the steamer, but agreed that it is not the Waverley, that features in the GoogleEarth simulation:

image
image820×475 115 KB

It may be the TS Queen Mary - see Historic steamer TS Queen Mary finds permanent home in Glasgow - BBC News.

@macspite : I created the mission at home, and tested it on GoogleEarth using VLM before leaving for Glasgow. The film generated by the first two flights closely matches the virtual film created by GoogleEarth. The data “sent from the Litchi program to the aircraft” is why I am (was) still happy with the Mavic Air - no requirement for “virtual sticks”, and the aircraft always comes back even if the signal is dodgy. (To be fair, it came back this time as well… …but only just!)

Unless some other explanation comes up, I am going to build my confidence in this machine gradually, keeping it closer than usual. And I’m going to get that public liability insurance!

2 Likes
17

I had the major twitch previously described back in September 2019 and no recurrence since. I had tested the flight in VLM and most of my subsequent, unevent full flights have beenLitchi controlled

I can recommend the FPVUK membership as a cheap way to get 5 millon cover/

1 Like
18

+1 for FPV UK :+1:

19

What about compass interference, or something affected the Compass and or the IMU

You are approaching the BBC building when it starts to go wrong and once you clear it you regain control

image
image947×457 130 KB

20

BMFA gives you a much higher insurance coverage … £35m ? ? Can’t recall

That indeed is the Queen Mary - I went on it back in 1968 I think! Never forgot it…

Still think there is merit in just asking Litchi to look at it since it was their s/w that controlled it in effect …

Nothing to lose…