I do not understand why a compass (or other system / subsystem) error should trigger a flyaway.

  

  

There are five different systems on Spark, and they do not seem to work in cooperation.

  

  

1 – Compass

  

2 – Accelerometers

  

3 – GPS

  

4 – Vision System

  

5 - Barometer

  

  

The Aircraft should work well with three, or even two. I really don´t understand why error on just one triggers a fail.

  

  

Let´s say:

  

  

1 -  you lose compass: You still have GPS (from which you can get heading). The AC can even RTH very, very, safely. I DON’T UNDERSTAND A COMPASS ERROR COMPROMISES THE FLIGHT

  

  

2 – you lose GPS: You still have Accelerometers, compass, and if not too high, vision system. The AC should not RTH, but it can use those subsystems to hover in place (not flying away), AND, yes, WITH ACELEROMETESR YOU CAN CONTERACT THE WIND, making it simple to control it back (as you still have connection)

  

  

With Accelerometers, it would even be possible to make what in aviation is called INS (Inertial navigation system). You know where you were when you lost GPS, and you know where you shoud RTH. Using accelerometers and compass, you can get really near launch place.

  

  

https://en.wikipedia.org/wiki/Inertial_navigation_system

  

  

Just trying to figure it out. Bad software design? Not getting full potential of the system?

the "fly aways" have been typically having the Compass, GPS, and transmission fail and with no transmission you have no Accelerator. this is why it has been flying away. they were too high for the VPS to work(at least mine was too high) so the barometer actually made it worse. since the barometer did not fail the drone stayed at the same altitude thus flying away. if the barometer had also failed the drone would start to fly away but would probably hit the ground before going too far.

This is only my speculation and opinion, if anyone can prove me wrong then please educate us!

I agree with you 100% !

OP: Just a slight correction to your point 1: GPS alone will not provide a heading. That is what the compass is for. Yes, you can derive heading from looking at two GPS points, however the drone will not know which way it is physically pointed relative to that direction of travel. It is the compass which corrects this, which is why when you lose compass, it disables GPS as well.

The rest I agree with. The Spark should be able to, at a minimum, fly flat and level with the accelerometers/gyros only. In conjunction with VPS, should be able to hold a steady hover, too.

Some have suggested that when entering atti, if the drone is unaware of what level is, it triggers the fly away. But as I have posted before, I have used much more primitive tech which never had such issues. The Spark should overcome this.

OneMatt Posted at 2017-8-17 05:23
OP: Just a slight correction to your point 1: GPS alone will not provide a heading. That is what the compass is for. Yes, you can derive heading from looking at two GPS points, however the drone will not know which way it is physically pointed relative to that direction of travel. It is the compass which corrects this, which is why when you lose compass, it disables GPS as well.

The rest I agree with. The Spark should be able to, at a minimum, fly flat and level with the accelerometers/gyros only. In conjunction with VPS, should be able to hold a steady hover, too.

Yes, but this is an aircraft that can fly in all four directions. It does not need to face forward to fly. It could calculate direction from movement. I really don't care if it returns to the relative position flying sideways, backwards, or forward.

OneMatt Posted at 2017-8-17 05:23
OP: Just a slight correction to your point 1: GPS alone will not provide a heading. That is what the compass is for. Yes, you can derive heading from looking at two GPS points, however the drone will not know which way it is physically pointed relative to that direction of travel. It is the compass which corrects this, which is why when you lose compass, it disables GPS as well.

The rest I agree with. The Spark should be able to, at a minimum, fly flat and level with the accelerometers/gyros only. In conjunction with VPS, should be able to hold a steady hover, too.

“ which is why when you lose compass, it disables GPS as well. ”

In my point of view, design flaw.

The best would be: Maintain GPS, hover steady until regain compass or receive remote command..

fans99d24711 Posted at 2017-8-17 05:45
Yes, but this is an aircraft that can fly in all four directions. It does not need to face forward to fly. It could calculate direction from movement. I really don't care if it returns to the relative position flying sideways, backwards, or forward.

And Adriano Araujo....

You both don't understand what each function does, and how they tie in.

The compass tells the drone which way is forward. If the aircraft doesn't know which way is forward, it also does not know which way is left, right, or back.

GPS tells the drone what its specific location is at this moment. That is all. If the GPS calculates a different location, it will know it moved from A to B, but not HOW it got there.

However, without a compass, it will know where A was and where B is, but wont know if the new location B was to the front, back, left, or right of A, relative to the direction the drone was pointed originally. Therefore, it cannot counter the movement in the correct direction.

TRY THIS: Go sit on a swivel chair. Blindfold yourself. Now get a friend to spin you around a bunch of times, then while spinning, slide you to a new location. Now, still blindfolded, walk back to the original location. You can't, because you are dizzy (no compass) you dont know which way you moved, so you can't get back there. So, you fall over (fly away). Your best solution at this time so you don't crash is to sit down until your dizziness stops and you take off the blindfold (return compass).

So, the drone  says "i have no compass bearing, so I dont know which way I went, so I should also ignore GPS input, because I can't do anything with it". The best solution would be to go into self-level hover (sit down) and wait for control input. This means it may drift with the wind, but that is a small price to pay. Of course, if you are in VPS range, it should know the orientation of the camera, and therefore determine movement direction based on that, as it does indoors.

Again, the real issue with flyaways is "why doesn't it actually go level?". When my Spark flew away, it was at a sharp bank, causing it to accelerate. There was no wind that day (or at least not 32km/h of wind, the last speed recorded). The flight log still recorded GPS location, so it knew it was drifting, it knew it was accelerating, and this is probably why DJI granted the warranty claim. Why at a mere 1.6m height the VPS didnt intevene is another question I have...

OneMatt Posted at 2017-8-17 10:58
And Adriano Araujo....

You both don't understand what each function does, and how they tie in.

I love your suggestion to help understand.  Gonna try it, just for fun and to be a flyaway Spark for a just one minute of my life.   

for a TL;DR

if IMU goes down, it cant figure out where it is in 3D space very well or how to compensate...  

if GPS goes down its lost to where it is on plant earth...

if compass goes down is doesnt know where it is facing

if VPS isnt available (too high) then it doesnt know where it is in relation to surroundings

if barometer does down it doesnt know how high up it is once VPS is unavailble

only 1 stack of each thing here... if one thing does down its heavily relaying on other single points of failure as well..

OneMatt Posted at 2017-8-17 10:58
And Adriano Araujo....

You both don't understand what each function does, and how they tie in.

“So, the drone  says "i have no compass bearing, so I don’t know which way I went, so I should also ignore GPS input, because I can't do anything with it".”

I respectfully desagree.  ;-)
It would use the 10...12..14 satellites lock to maintain it´s position without drifiting to the wind.
The flyaway, as stated, shoul be software/hardware failure, since it happens way faster than the wind speed.

What I am saying is that with the available variables/sensor even with the failure of one system, the others should be able to maintain AC manageable.

Lot´s o time ago (90´s) I used to fly remote controller helicopters. At that time, no GPS, and Gyros were just starting popping out.

So, with no GPS, no Gyro, if you loolse radio, the AC would maintain it status-quo ´till got connection again. And it, back in the 90´s.

In my opnion.. DJI cropped the firmware to make Spark a simple drone and ends making a big mess.. DJI uses the same firmware since P3 launch (maybe since p1,p2), over last years are improving functions and stability, adding functions on new models,  making her drones much more reliable and.. decided to launch a drone with a cropped firmware that doesnt have some ESSENTIAL functions, like turn off VPS and others, and thats just the funcions whe can see directly on APP, and the ones on flight controll part of firmware that whe cant see and are missing or modified?

Years of improvement doing almost everything in the right way to.. cut everything off and make a big mess.

Who knows? If DJI modify or cut out a important ( and working like a charm since p1-p2-p3) part of firmware to make Spark a more "user friend" "smart" "simple" "idiot proof" drone, and end making a much less reliable drone and full of bugs.

Adriano Araujo Posted at 2017-8-17 12:22
“So, the drone  says "i have no compass bearing, so I don’t know which way I went, so I should also ignore GPS input, because I can't do anything with it".”

I respectfully desagree.  ;-)

Disagree all you want, you are wrong. You csnt navigate by GPS without knowing where North is!

Adriano Araujo Posted at 2017-8-17 12:30
What I am saying is that with the available variables/sensor even with the failure of one system, the others should be able to maintain AC manageable.

Lot´s o time ago (90´s) I used to fly remote controller helicopters. At that time, no GPS, and Gyros were just starting popping out.

I do agree with this.

When you consider that this is designed to work without any connected controller, there is no excuse for this thing, with its multitude of sensors, to fly away..

Adriano Araujo Posted at 2017-8-17 12:30
What I am saying is that with the available variables/sensor even with the failure of one system, the others should be able to maintain AC manageable.

Lot´s o time ago (90´s) I used to fly remote controller helicopters. At that time, no GPS, and Gyros were just starting popping out.

your reasoning and theory is solid. however your failing to understand a few fundamentals..  

  first of all, GPS sats dont know which way you are facing, that requires a compass..  GPS software can roughly figure out your direction using math, yes.. but thats alot of wasted time since it needs distance over cords and time to figure out heading/bearing.. it will take multiple seconds every time you slidely change course to figure out new rough bearing/heading..  this is why compasses are used in GPS navigation in your car, its faster and requires less wasted processing time.. just like with car Dji uses compass, its faster, it know which way your facing without even moving or tracking.the spark only has 1 Compass stack.. so single point of failure on compass.. and it should be fine as long as no flight automation is used..  like RTH

IMU system is like an indoor GPS yes sort of.., mostly used to virtualize where it is in 3D space and maintain a specific value... preventing external forces or its own forces from impacting it...  but its not precise, only single unit in the spark also...  but again, without a compass, it can only know where it is in relation to its own info... with GPS info and IMU info the aircraft can likely navigate very safely without compass,  as long as you do not use any automation flights like RTH.... without GPS or compass the IMU is virtualizing where it is in relation to its own info and movements its tracking its a bit sloppy..    if IMU goes down, it doesn't know where it is in 3D space, or how to compensate for its own forces or external forces acting upon it.. it will still maintain its self if IMU is only thing down, dont dont expect flight automation to work well...

barometer is another single stack, but its considered secondary info, it is important, but its only relative and not absolute, if barometer fails i expect the drone to maintain itself otherwise.   if barro and IMU fails then i expect problems..

back to the issue at hand... why does a compass failure cause flyaways or crashes and problems.. because flight automation...  if RC connection fails, RTH initiated, or panicky flyer... drone is confused, it has its directions mixed up, the IMU is only maintaining the aircraft on a predictable path, GPS is only a data stream, not enforced.. says return to these cords marked and it tried but it fails to because imagine your car's GPS mixed up west and south and east and noth in compass... sure it give you directions, but ill promise you they will be horribly inaccurate with a ton of recalculations...  dont beleive me or dont have a car with build in GPS... put a magnet on your iPhone or andriod until compass flips... use google maps with that magnet right there... take a drive somewhere across town.. 10 miles to give it a change to push you around random streets.... it be mixed up it be really confused..it probably get you there, eventually, but it wont be smooth

Adriano Araujo Posted at 2017-8-17 12:22
“So, the drone  says "i have no compass bearing, so I don’t know which way I went, so I should also ignore GPS input, because I can't do anything with it".”

I respectfully desagree.  ;-)

As OneMatt mentioned, the GPS need compass to indicate the direction. When compass is interfered and does not work, the drone still wants to go home, but it doesn't know the correct direction, and it may drift. Switching to ATTI mode, the drone will not want to go home any more, you can control it manually.

It seems to me that the fail safe protocols have not been thought through correctly, most flyaways I have read about have had a system failure and a loss of connection rendering attitude mode in-operable. At this point, no position hold and loss of user input the AC is going to drift as it waits to regain signal from controller. If there is an imu failure in the mix too then it will no longer know where level is and will therefore fly-off (The IMU contains two sensors, gyroscope and accelerometer. these calculate attitude, level) from what I understand, if connection to controller is lost the drone will wait 20 seconds then start a RTH, If there is also a sensor failure then it will fly off. The fail safe needs to be programed to react differently when it has detected a sensor fault. I would suggest that it should cut power and land (this is how we fail safe our racing drones)
Maybe someone from DJI could confirm what the fail safe protocols are ?
The correct fail safe for sensor failure and loss of connection should happen much faster than 20 seconds. I would recommend just a few seconds.
As for needing a compass to calculate direction with GPS, If you look at Inav for fixed wing aircraft you will see that this system dose not use a compass. It still can perform position hold, RTH, Land and do waypoint missions. (with a flight controller and GPS that can be purchased for under £50)  

“As for needing a compass to calculate direction with GPS, If you look at Inav for fixed wing aircraft you will see that this system dose not use a compass. It still can perform position hold, RTH, Land and do waypoint missions. (with a flight controller and GPS that can be purchased for under £50)”

That is the point. I don´t understand why people says that the compass is a NEEDED info.

Of course, compass can speed and smooth up things, BUT in critical situations it is unnecessary. IT IS UNECESSAYY. #Fact

You can derive heading from two points very easily.

My first GPS (back on 90´s) did that:
https://www.amazon.com/Garmin-eT ... g-GPS/dp/B00003WGP5

OneMatt Posted at 2017-8-17 18:33
I do agree with this.

When you consider that this is designed to work without any connected controller, there is no excuse for this thing, with its multitude of sensors, to fly away..

Or lose RC connection when it is les than 15ft away from you direct vlos in the boonies where there are no wifi networks and barely cellphone signals. Unless pine trees and oak trees are made of metal or magnets no reason for interference or for the compass to malfunction.

I wrote once in this forum. There is no reason for flying away scenario. Flight controller is designed good enough to detect the compass interference and give instruction to the pilot to move the aircraft or does the compass calibration. Moving the aircraft away from the interference, the flight controller can do by itsself as well as rotating around itsself trying to solve the tempolary error. If every attempt should fail, the aircraft should try landing safely using built-in camera and terrain recognition rather than flying away. The programmer must be crazy to do that, direct the aircraft to fly away. But I think the flight controller goes crazy.