Hi,

Today I upgraded my DJI SPark to the latest firmware V.01.00.0900 and did some practice flight indoors.
I have DJI Go 4 App v.4.1.15 since I am weary of any new versions after I experienced issuse with V.4.1.18.

I fly indoors using Tripod mode. GPS goes on and off. Home point not enabled which is just fine.

During one of my flights DJI Go 4 App crashed. Spark kept hovering while I took my hands out of controls and restarted application.
When application restarted I got connection with Spark again and resumed flight.
After perhaps 10 seconds application gave low battery warning and after another several seconds Spark started return to home procedure.
At that point Spark was at about my eye leve and started to raise. I pushed left stick all the way down (Spark was perhaps 2 metres from me) to force it to land.
It reacted at first by stopping its ascent and lowering ever so slightly for a fraction of a second. Than it resumed its ascent even though I was pushin left stick full force down.
It hit the ceiling and continued to grind aginst it. After perhaps two seconds I abondoned the RC and my attempt to bring it down since it was clearly not responding to my commands even though connection with RC and my mobile and app was fine. I jumped, caught the Spark and flipped it sideways to turn it off.

When I reviewed my flight log, it shows clearly initiation of RTH and my frantic attempts to bring it down by pushing left stick down, so I did not err here.

What could have gone wrong? Why would Spark refuse to listen to RC in RTH mode? Needless to say I did many flights at home before and it never initiated RTH procedure so I never had issues. Perhaps something was reset when my application crashed mid flight but it still does not explain why I lost control of Spark when in RTH mode.

Experiences like this severely erode my trust in Spark and any new firmware upgrades. But perhaps I do not understand something here.
Please help.

Mirek

What was/is your RTH altitude setting and was RTH at current altitude enabled? Maybe it was trying to go to the set RTH altitude.

You have to cancel the RTH in DJI Go4 before the AC will become responsive again.

I don't believe stick input overrides a RTH unless you purposely cancel the RTH
Since you weere flying indoors, you probably didn't have enough gps reception for the home point to be set.
The AC was climbing to 30m (default for RTH height) and was probably heading for the last home point set, where ever that was.

never initiate RTH at home.
it is impossible spark initiated a rth becos if the drone is withing 20 meters, there is no program written into spark to do rth...

becos of internal router and concrete, compass errors happen and the spark moves here and there..typically what we call as going nuts

S.J Posted at 2017-12-23 03:35
never initiate RTH at home.
it is impossible spark initiated a rth becos if the drone is withing 20 meters, there is no program written into spark to do rth...

It had the old Homepoint from the last GPS flight before loaded!
Wherever that was
-'GPS goes on and off. Home point not enabled which is just fine.'-

Yes, in the app I did have RTH at current altitude enabled. Since I was flying indoors and the ceiling is at 2.3 metres, it would be possible that if Spark was over 3 metres from home point, when low battery RTH was initiated, it tried to climb to over 2.5 metres to return and hit the ceiling. However, I still do not understand several things:
- why my commands to lower it down were ignored while it was climbing?
- I certainly did not initiate RTH as S.J. suggested. Also I may expect compass going nuts inside, GPS going on and off, Spark switching from Visual to ATTI when too dark and being less stable etc. but I do not expect that Spark stops listening to input from my RC when I do have connection.
- Is it true what S-e-ven suggests that Spark keeps old Home Point settings? This would be a major disaster (in my opinion) if Spark retained old Home point between flights and battery changes. This would basically mean that you could absolutely never trust it without setting home point for each flight. Please confirm that this is the case and explain why DJI would make such a bad design decision (or maybe again, I do not understand something here?).
- why Spark suddenly decided to RTH when it never did that before when I was flying inside? I did many flights inside before, always without setting home point due to unstable GPS reception and when battery was getting low it always started to lower itself in the spot where it was flying and never started to climb. I understood that without Home Point set, this would be expected behaviour, and until yesterday, it was.

I am trying to understand as much as I can about what is and what is not expected in Spark behaviour and thank you for your replies and help.   

S.J Posted at 2017-12-23 03:35
never initiate RTH at home.
it is impossible spark initiated a rth becos if the drone is withing 20 meters, there is no program written into spark to do rth...

Without stable GPS the Spark could not know where it actually was. And there is RTH within 20 meters range. Only within 3 meters range it will not RTH but land on the spot. The move to the ceiling quite likely was an attempt to RTH and as reported by the OP this is also reflected in the flightrecord.

If you wish us to have a closer look please upload your flightrecord here: http://www.phantomhelp.com/LogViewer/Upload/ and share the link with us.

Thank you Wachtberger. Yesterday's incident left me with many questions. Let's try to answer them one by one.
I uploaded my flight record and you can find it here: http://www.phantomhelp.com/LogViewer/OJNMLDV8JVHCJ2GZ5S4R/

My first and main question is: Why Spark ignored my commands to go down when it initiated RTH procedure?

Please note that the flight record starts in mid flight. My DJI Go 4 application (Android v. 4.1.15) crashed mid flight and I needed to reopen it while Spark was already in flight and hovering in front of me.
20 seconds after restarting application, it gave me low battery warning and 10 seconds later Spark started its RTH and ascent. This is when I slammed down on my left stick to try to get it down. The connection between Spark and RC was healthy so Spark certainly "saw" my command to go down. After few seconds of doing this and Spark struggling to bore a hole in my ceiling I gave up, put the RC down and retrieved Spark manually from my ceiling (I jumped up, caught it and flipped it to the side shutting down engines).

One more note - my flight record IMU altitude when Spark started RTH is certainly wrong. When RTH started Spark was at about my eye level - about 5 feet. My ceiling is at about 7 feet. This is where it got and was struggling to go higher but ceiling prevented it.

Mirek6 Posted at 2017-12-23 07:33
Thank you Wachtberger. Yesterday's incident left me with many questions. Let's try to answer them one by one.
I uploaded my flight record and you can find it here: http://www.phantomhelp.com/LogViewer/OJNMLDV8JVHCJ2GZ5S4R/

Thank you for sharing your flightrecord Mirek. It's a pity that the first part is missing, thus we don't know if a homepoint was properly recorded at take off time. But to me this does not seem to be the really relevant bit in this case. To answer your question raised a bit further above, yes, it is vital to have the homepoint recorded properly at take off when flying outside. Otherwise the Spark would have no reference where to go in case RTH is triggered and might well choose the last home point it can remember from a previous flight.
But now to the more important issues in this particular case of indoor flight. Until now it has also been my understanding that the Spark would react to my stick movements, even when in RTH mode and I have successfully done that already but outside and in different flight situations. The user manual e.g. mentions in one part that you can make your Spark ascend in order to fly to a more suitable landing spot which at first glance would confirm that stick movements should translate into flight action. But, and here I am a bit puzzled, there is another note in the user manual that is difficult to understand for me (see extract below). It could be read in a way that in case of RTH is triggered, Spark would first insist to climb to at least 20 meters altidude before the pilot could intervene with stick movements. But I really don't know if I read this correctly and somebody else here might know better.
In your flightlog I have noticed two rather short connection losses between the RC and the Spark, which is a bit surprising given the short distance between them. Maybe at times the antennas were in the wrong direction? Nevertheless this raises the question whether your stick movements to bring the Spark down were properly received. I have not yet aquired the ability to read the stick movements from flightrecords, but again there are more experienced users in this forum who can.
In summary the best way to handle the situation in my opinion would have been to cancel RTH as soon as possible after it had been triggered.

Hi Wachtberger.
Many thanks for your thoroughfulness and quick responses. I was searching the forum to answer the RC control question during RTH and found this post: https://forum.dji.com/thread-38269-1-1.html
This post was created by "DJI Team" two years ago.
Here is a copy of a section which caught my attention:

Once RTH is pressed, this is what the drone thinks:
“Master is commanding me to return home.” (When receiving the RTH command)
“Alright, I’m ascending to the preset RTH altitude! Nothing can stop me from ascending!’ (To the predetermined height)
‘Ok, now I’m adjusting my nose towards master.’ (Adjusts direction)
‘I’m on my way back now!’ (Returns to the Home Point)
‘I’m going to land now!’ (Descends after reaching the Home Point)
Please remember that the drone can’t be controlled while rising to the RTH altitude!

This explains why my attempts to bring the Spark back down did not work while it was trying to ascend during RTH.

The new question is "Why would DJI design team make such a decision?". This is very, very counter intuitive and I would like to hear explanation from DJI team. This may also be dangerous when you fly your drone inside, under a tree or whatever obstacle. User actions to control drone should always override drone's automatic actions unless there is a very specific reason not to allow user control - for example when obstacle avoidance is turned on and a user tries to ram a drone into a wall.
Yes, I understand that there is a cancel RTH button but, when you see drone raising up and you do not hear a warning from the app (speaker is down or your full attention is on drone), your first reaction is to move control sticks to control it, not search for cancel button. This is what I did. TO amke things worse, I had perhaps half a second to react between I saw the drone rising and it hit a ceiling two feet above. Half a second is enough to slam on left stick down but may not enough to fumble for cancel button.

I have more follow up questions which I will post below but to make things simple and clear, here is a question I would like DJI developers to answer:
"What is the reason why Spark user cannot control Spark when it is raising when RTH sequence is triggered?".

BTW - to be clear. I did not hit RTH button. RTH was triggered by low battery. Still, the bolded sequence in the post above stands for all RTH scenarios and the question is "why?".

So I have interpreted the user guide correctly, at least something ;-) And to your latest questions there is a simple answer that I had given at the end of my little analysis: All remains in your hands when you react appropriately, and in your situation this would have been to cancel RTH and fly manually. Luckily no damage was done, thus take it as lesson learnt and keep it in mind for next time. Unless there are no serious malfunctions, the ultimate flight decisions (and responsibility) remain with the pilot.

Starting with a series of follow up questions to make sure I fully understand RTH mechanism and algorithms.
In his post above S-e-ven writes "It had the old Homepoint from the last GPS flight before loaded!"

The question is : When is the home point reset?
In other words:
- Home point is successfully set for flight X
- I finish flight X and replace battery.
- I start next flight.
- GPS is weak or non existent so new Home Point is not set.
- During my flight battery runs low and Spark starts RTH sequence.
Will it attempt to land in home point established during previous flight (flight X)? Or should it simply think that home point was not acquired so it will land where it currently is.

I can hardly believe that DJI team would make a decision not to reset home point between flights. This could lead to very dangerous situations.

So the question is: Does Spark keep old home point from previous flight when it cannot acquire a new home point for next flight? And if so, what was the logical reason for such a design decision?  

I have several more questions relevant to my yesterday's situation but let's try to get answers for the above before I ask another one.
In the meantime I am trying to repair my ceiling which has proven to be much weaker than Spark. Ceiling suffered - Spark survived.

I also think the altitude stick should work during rth.
So the pilot in the intuitive way can prevent crashing, like in this case and when under trees.

You could always set the RTH option to "RTH at the current altitude"

I've run the Spark battery down to less than 10% indoors and it's never triggered a low battery RTH maybe it's because the "home point" was  NOT set.

Also I notice  when I fly the Spark indoors that as I get close to the ceiling it keeps on ascending, I suspect due to interference from the rebar in the concrete slab used for the first floor of my house.

So I'd put your problem down to  compass interference or wi-fi interference,
It's not much of a problem if you have the genuine DJI prop guards which give you some protection vertically.

OK - so first question is semi-answered and there is no point in digging deeper without DJI support involvement.
DJI team post from 2 years ago and DJI Spark manual seem to suggest that  RC controls do not work when the drone starts its RTH sequence and it is ascending - at least first 20 metres.
However, the post was from 2 years ago and manual may be old too since Dirty Bird posts suggests that it is not so and that, in his experience, DJI drones do react to RC in RTH.
Certainly it was not my experience yesterday so manual seems to be correct.
While I can assume that manual is right and old DJI team post is right, and learn that I should first hit cancel when the drone starts ascending before attempting to control it via sticks, I still think that it is a legitimate question for DJI to answer if this behaviour is still as described, and if so, what is the reason behind it, since it definitely can cause problems.
I will e-mail DJI support and notify community here of DJI answer.  

lannes - it was not wi-fi or compass interference. Spark started RTH sequence - it warned me about and there is a log for it. It was either DJI designed behaviour to ignore RC commands as it was ascending (as per manual and DJI post) or, if DJI fixed it later, it was a firmware bug. In any case I learned my lesson (to use Cancel button in such situation) and I will follow up with DJI support to get definite answer to their reason behind this design.

Mirek6 Posted at 2017-12-23 10:50
Starting with a series of follow up questions to make sure I fully understand RTH mechanism and algorithms.
In his post above S-e-ven writes "It had the old Homepoint from the last GPS flight before loaded!"

Please do not focus on this homepoint issue, it is not at all relevant for your case. What is relevant is the behaviour of Spark in well defined situations that all of us need to get familiar with and be aware of. And there is a very reasonable logic behind this automatic behaviour if the pilot does not decide otherwise and intervenes because it might be more appropriate for a given flight situation. You first of all need to understand that your flight situation is not really a typical one for the Spark, but rather for small toy drones. Spark is made for outside operation and there the failsafe programming has to focus on avoiding damage to others first with Spark only being the second in priority. If in no matter which outside flight situation automatic RTH is triggered the very best action can only be to bring the Spark outside a range where it could hurt others. For this goal there naturally can be only one direction, which is upwards. And here the pilot's responsibility kicks in. He or she has to determine whether upwards is really the best way out or if in the specific situation other actions are more appropriate. I fully agree that you have tried to do your best in this extremely short timeframe but please also accept, that this was not a common flight situation either.
I wish to add that through your case, I have personally learned a lot myself on certain details that I had not been aware of before.

lannes Posted at 2017-12-23 12:32
You could always set the RTH option to "RTH at the current altitude"

I've run the Spark battery down to less than 10% indoors and it's never triggered a low battery RTH maybe it's because the "home point" was  NOT set.

The drone keeps ascending because it gets sucked as much as it gets bounced when extremely near to the ground. It’s the air flow and a basic principle of the rotating wing cheers.

Mirek6 Posted at 2017-12-23 13:06
lannes - it was not wi-fi or compass interference. Spark started RTH sequence - it warned me about and there is a log for it. It was either DJI designed behaviour to ignore RC commands as it was ascending (as per manual and DJI post) or, if DJI fixed it later, it was a firmware bug. In any case I learned my lesson (to use Cancel button in such situation) and I will follow up with DJI support to get definite answer to their reason behind this design.

Ah right, you must have ignored the low battery warnings until it initiated a RTH and even after that you did not cancel the RTH with 10 seconds to do so. So it ascended to the 20m RTH height which is perfectly normal and as per the design.

Yes I agree the main issue is the lack of control while the Saprk is ascending to the RTH height, but that is overridden if you hit the cancel button.
So I'm thinking it is functioning as designed

Yes - all points valid and appreciated. Yes Wachtberger and lannes - you are right, if it is functioning as designed - no problem - I learned my lesson. DJI designers may have had a good reason for this design (although I still do not understand what it might be).

Dirty Bird threw a bit of a curved ball in our discussion with the claim from his experience that his DJI drones do react to stick controls even when going up in RTH mode. However, he also added that he does not own Spark but other DJI drones. It is quite possible that Spark, being a beginner drone, has limitations or overrides which other, higher end, DJI drones do not have. One of these limitations may be to dissallow joystick controls while Spark is raising first 20 metres to RTH height. But this is just my speculation.

For now I accept that Spark behaviour is as described and the Spark did what it was supposed to do. But I will still follow up with DJI in an attempt to understand the logic of blocking pilot from controlling drone in first few seconds (which may be critical) of RTH sequence even though they added Cancel button. I can clearly imagine situation where I do not want to cancel RTH sequence but want to react with my joysticks to get the drone out of difficult situation and then let it continue what it was supposed to do and return home. For example, I am 100 metres from my drone and see clearly that it starts RTH under a canopy of a large tree. Instead of cancelling RTH I may want to steer the drone gently from under the tree and then let it continue RTH.  

In summary - I do understand what happen, but want DJI team to help me understand what is their reasoning behind such design.

I am also analyzing all my yesterday's logs to see why RTH was not triggered in my previous indoor flights and I could always control my drone safely inside and what was different about this one particular situation when it decided to do RTH and hit the ceiling.  

Mirek6 Posted at 2017-12-23 16:35
Yes - all points valid and appreciated. Yes Wachtberger and lannes - you are right, if it is functioning as designed - no problem - I learned my lesson. DJI designers may have had a good reason for this design (although I still do not understand what it might be).

Dirty Bird threw a bit of a curved ball in our discussion with the claim from his experience that his DJI drones do react to stick controls even when going up in RTH mode. However, he also added that he does not own Spark but other DJI drones. It is quite possible that Spark, being a beginner drone, has limitations or overrides which other, higher end, DJI drones do not have. One of these limitations may be to dissallow joystick controls while Spark is raising first 20 metres to RTH height. But this is just my speculation.

Very good conclusions to start with and you are absolutely right to seek answers to all your questions. I have done exactly the same when I started and continue to do so. As stated above, from your case I have learned myself too.
Just keep one element of your case in mind. It all happened indoors which is not the normal flight environment. I am pretty sure that if all that had happened outside, you rather would have been amazed (or at least not worried) about Sparks behaviour.

Thank you Wachtberger.

With continuous learning in mind, here are my other conclusions based on analysis of yesterday's flight logs which may shed some light on why my previous indoor flights were successful and the last one not and what to expect indoors and what to watch out for.

Key Lesson: all my indoor flights had setting "Return Home at current altitude". This is a mistake. I should have disabled it.
The reason is that if this setting is on the drone will attempt to raise up to at least 2.5 metres if RTH is triggered and drone is more than 3 metres away from the home point.
If this setting is disabled, than drone will land where it is if it is up to 20 metres away from the home point when RTH starts. It will not go up. Only if it is further away than 20 metres, it will go up.
In regular house the ceiling is likely lower than 2.5 metres which means that if drone go ups "at current altitude" or not, it will hit the ceiling. But what really may make a difference here is 3 versus 20 metres away from home point. It is quite unlikely that flying in the house you will get more than 20 metres away from home point, but it is possible to get 3 metres away or, even if you are within 3 metres, as I was, the GPS and compass errors inside the house may skew that to go over 3 meters. I am showing it below.


Indoor flight one log: http://www.phantomhelp.com/LogViewer/P3BTLWKEKE5JCZV7XOKY/
- Uneventful - GPS on and off.
- Home point updated 3.5 seconds into the flight. I did not notice that - the message on DJI GO 4 app did not pop up. Nor it is popping up when I play it back from the log. Perhaps DJI Go 4 app bug?
- When the first message about critical battery power appears ( 6m and 41 sec into the flight) the drone distance from home point shows 10.5 ft but when message to return home appears 1 second later, it is within 9.9ft which is exactly 3 metres.
- Drone starts landing where it is since the distance is 3 metres or lower

Indoor flight two log: http://www.phantomhelp.com/LogViewer/NDXIK664IOR8JWQV4LEN/#
- Taking off without home point - GPS signal low
- 5 mins and 17 seconds within the flight the Home Point is acquired with message "Home Point recorded. Return-to-Home Altitude:30M" This message is different than "Home Point Updated" message from first flight - not sure why.
- At 8 min 55 sec the message to return home is received but no action initiated since drone is less then 3 metres away from Home Point. I believe I kept it up for a bit while it was trying to land to see what will happen with battery and how long it can hold

Third flight - application crashes and is restored by me mid flight:
Part 1 of log:
http://www.phantomhelp.com/LogViewer/6ZJG3SR11BC51IBSN70K/
- At 1 m 1 sec home point is set
- The log ends mid flight - not sure why. I think I briefly landed but the log does not show the landing. Perhaps DJI Go 4 bug?

Part 2 of the same flight:
http://www.phantomhelp.com/LogViewer/AMS77F0T5XMX54LKRU9B/
- Seems that home point continues to be set (it is not reset) when I landed but did not switch off the power and took off again.
- I am saying that because the home point is not set in this log, yet from the beginning of the log the distance to home point shows 3.5 feet. Which leads me to believe that home point setting persisted through landing (but, as I said, the power to drone was not shut off)
- At 7 m 50 sec my DJI GO 4 app crashes and log ends mid flight
- Drone hovers while I restart application

Part 3 of the same flight log: http://www.phantomhelp.com/LogViewer/OJNMLDV8JVHCJ2GZ5S4R/
- Drone hovers while application restarts.
- Just from the beginning log shows that drone is now over 18 feet from home point.
- This is inconsistent with what is shown just before app crash but, given I am indoors and GPS and compass may be flakey I accept that this is an error caused by my flying environment which ultimately led to RTH and drone hitting ceiling.
- Drone continues to fly when at 8 m 41 second return home is triggered due to low battery. At this point distance from home point reads 17.8 ft which is totally wrong (home point was perhaps 2 metres away) but, as said above, may have been caused by indoor GPS and compass disturbances.
- Since my drone has "Return Home at current altitude" set and thinks that it is further away from home point than 3 metres, it tries to raise to over 2.5 metres to return home and hits the ceiling.
- I try to get it down using left joystick but it does not respond as it was programmed to do while going up in RTH mode.

The above makes sense to me know although I did notice several inconsistencies, potential DJI GO 4 app bugs and observed DJI Go 4 app crash in mid flight which, perhaps somehow, contributed to skewed home point reading. Not sure.

So: if flying indoors :
- disable "return at current altitude"
- when you hear low battery warning RTH, use Cancel button immediately to cancel any potential RTH sequence (just in case)
- don't think that drone will listen to your controls when it starts going up in RTH sequence - it will not.

Thanks to all.

Mirek

Mirek6 Posted at 2017-12-23 19:36
Thank you Wachtberger.

With continuous learning in mind, here are my other conclusions based on analysis of yesterday's flight logs which may shed some light on why my previous indoor flights were successful and the last one not and what to expect indoors and what to watch out for.

I believe what’s missing in this whole saga is responsibility, it is your responsibility to know what is going to happen in all situations you fly in , case in point you should know that in critical battery mode your aircraft will rise to 2.5 metre you were operating inside with ceiling level lower than this so it’s your responsibility if something goes wrong and it did accept it and move on. You are clearly told in your manual not to fly in unsuitable environments.

If you were in a situation where critical RTH could be canceled and you accidentally done this causing aircraft to run out of battery you would be complaining that critical RTH should not allow pilot to preform this.
You and only you should know before you fly what the capabilities of your aircraft are.

It is not reasonable for the spark to assume neither where the home point is or how far it is from it when it is indoor without proper gps reception.
Trying to return to go home in that situation is bogus. If that’s what it tried to do. I can’t believe it’s that stupid, though. Must have been some other malfunction.

heliman Posted at 2017-12-26 09:57
It is not reasonable for the spark to assume neither where the home point is or how far it is from it when it is indoor without proper gps reception.
Trying to return to go home in that situation is bogus. If that’s what it tried to do. I can’t believe it’s that stupid, though. Must have been some other malfunction.

Or else it picked up gps signal, this is not a malfunction . OP did say gps was coming and going.

Mirek6 Posted at 2017-12-23 07:33
Thank you Wachtberger. Yesterday's incident left me with many questions. Let's try to answer them one by one.
I uploaded my flight record and you can find it here: http://www.phantomhelp.com/LogViewer/OJNMLDV8JVHCJ2GZ5S4R/

In your log  at 514 sec you got low battery warning you continued to fly ignoring low battery warning until 521 sec, before getting final RTH , you must take responsibility for this, I know it’s tough but I’m afraid this is down to yourself.

Mirek6 Posted at 2017-12-23 19:36
Thank you Wachtberger.
https://forum.dji.com/home.php?m ... op=rule&fid=115
With continuous learning in mind, here are my other conclusions based on analysis of yesterday's flight logs which may shed some light on why my previous indoor flights were successful and the last one not and what to expect indoors and what to watch out for.

When flying indoors make sure that.

1/ you are not going to receive signal interference and loose RC signal or your drone will auto RTH and you will hit the ceiling again, and we know flying indoors there is potential for great interference

2/ you can set RTH at 20% battery, if you get that far you can just press pause button or X and cancel you can then land.

3/ if your prepared to go to critical battery , you might be better raising your ceiling.

You can always cancel RTH with pause or X and take over manually, but critical battery ( only enough to get home) which in your case because of no homepoint being recorded your Aircraft was using its calculations to get back to a previous homepoint. So letting it get that far was the real problem.

I had similar problem on my first day with the spark indoors. It tried to ascend and hit the ceiling, and ground away at the props when they hit the ceiling.

It made me realise the prop guards sit lower than the props, so only protect from side collisions,
there's no protection for impact above.
It might be a good idea to 3D print some prop guards with protection higher up ?