I just wanted to know what the Spark is supposed to do in this very situation, because I could not find it in the manual:

What happens when a compass error occurs making the Spark switch to ATTI mode and then the signal/connection to RC also gets lost?

In the manual I only found the case for what happens when home point is recorderded and compass and GPS signal strength is good -> Failsafe RTH. But RTH is not possible in ATTI mode. So what will the Spark do? My guess is that it will initiate landing after 3 seconds. But can this be found somewhere in the manual?

RTH is not possible in ATTI mode because is no GPS and Spark don't know where he is... I think after losing RC signal Spark in ATTi would fly same direction when flew in the moment when signal was lost. Probably till battery is critical landing or crashing. Please correct me if its different.

I don't know.  This is the worst case scenario, and I can't think of a way that this can be reliably tested to see what the behavior would be.  I reviewed the manual (just before answering your question) and I don't see where it suggests that it would attempt to land as a part of RC disconnect.  This is entirely guessing and speculation, but here's what I think would happen:  I think that the craft will remain flying (though not necessarily stationary depending on environment conditions) until the battery reaches critical low, at which point the craft will attempt to land automatically.

How would it initiate landing if it’s moving which it would be in Atti mode, landing will only occur straight down this would not be possible unless you had gps or you had control of the AC in Atti mode.

I guess the question needs to be answered by DJI since even the experts here are only guessing. ;) And I thought that maybe I missed out something in the manual.

@hallmark: by "landing" I mean descending until the vision sensors detect ground. It does not matter whether the aircraft is moving or drifting away in ATTI mode. It should still be able to descend.

hallmark007 Posted at 2017-9-29 13:21
How would it initiate landing if it’s moving which it would be in Atti mode, landing will only occur straight down this would not be possible unless you had gps or you had control of the AC in Atti mode.

We might have a way to test it... (though there are reasons why this might not be a valid test).

Suppose we try to launch the spark indoors, in an environment where we are certain it can't get GPS (like a basement).  If we are using palm gestures only, we wait and see what happens when it runs out of battery.  This may not be a good test because of gesture mode (and the vision system).

We could try to take off with the remote, keep it out of gesture mode, and turn off the remote.  Though it might still be using the vision system.

This is my efforts at a thought exercise to meet the challenge of figuring out how to try and test this.  Still, since my ideas involve indoor flight, it's not a very "real world" test.

Charles Adams Posted at 2017-9-29 13:34
We might have a way to test it... (though there are reasons why this might not be a valid test).

Suppose we try to launch the spark indoors, in an environment where we are certain it can't get GPS (like a basement).  If we are using palm gestures only, we wait and see what happens when it runs out of battery.  This may not be a good test because of gesture mode (and the vision system).

It would not be the same if using vision system as this would be holding position so when it reaches low battery it will almost certainly just land it will not be in Atti mode but rather opti mode,  you could try but make sure RTH is set to land.

hallmark007 Posted at 2017-9-29 13:48
It would not be the same if using vision system as this would be holding position so when it reaches low battery it will almost certainly just land it will not be in Atti mode but rather opti mode,  you could try but make sure RTH is set to land.

Nah...  if it isn't truly testing ATTI mode then no.  I'd just be sitting and watching my craft for 10-15 minutes, and the results wouldn't answer the real question.

ciesnik Posted at 2017-9-29 13:32
I guess the question needs to be answered by DJI since even the experts here are only guessing. ;) And I thought that maybe I missed out something in the manual.

@hallmark: by "landing" I mean descending until the vision sensors detect ground. It does not matter whether the aircraft is moving or drifting away in ATTI mode. It should still be able to descend.

If there is a system programmed into spark to land after 3 seconds or go to force landing why can this not be programmed to do same when AC goes in Atti mode and why would dji leave such an amazing failsafe out of the Manual.
From what your saying it would then prudent to just switch of radio control if your aircraft goes into Atti mode and it will just land so now they have left out two failsafe’s.

hallmark007 Posted at 2017-9-29 13:48
It would not be the same if using vision system as this would be holding position so when it reaches low battery it will almost certainly just land it will not be in Atti mode but rather opti mode,  you could try but make sure RTH is set to land.

OH!  (I'm not letting this go)

We could try using a craft that does in fact support manual atti mode (such as my P3S), and on an extremely calm day launch it, switch to ATTI, and turn off the RC to see what happens.  Here the assumption is that one DJI craft probably behaves like another DJI craft (but that's not necessarily a great assumption).  And now that I think of it, I would be most disappointed if it just floated there until battery ran out, and then it drops and breaks something.

Charles Adams Posted at 2017-9-29 13:51
Nah...  if it isn't truly testing ATTI mode then no.  I'd just be sitting and watching my craft for 10-15 minutes, and the results wouldn't answer the real question.

If I understood all of the above correctly, we are discussing what would happen in ATTI mode combined with lost connection to the RC. It is clear that RTH is no longer an option in such a scenario. Drifting depending on wind speed is inevitable too. In my understanding a somewhat "healthy" reaction of the Spark should be landing on the spot as soon as possible. And if not at that "early" stage, it should do so at the latest when battery gets to critical level.

Well... If Spark lost GPS but still can receive command from the pilot, let the pilot controls the Spark is, in general, good choice. Unless the pilot does not capable in controlling the aircraft or the Spark lost communication with rc then it should land safely by itself, meaning is use its camera and object recognition to avoid crashing into people or vehicles.
But if all system is functioning and it just lost communication with rc, landing or rth are all good choice.

In each scenario, some should be mandatory and some should be optional. Loosing compass and rc communication together, only choice is landing. Loosing only compass, the options of landing or ATTI soft switch should be provided. Loosing only rc, rth or landing or hoovering in place are options. Loosing only GPS, entering ATTI or landing. etc.

Wachtberger Posted at 2017-9-29 14:09
If I understood all of the above correctly, we are discussing what would happen in ATTI mode combined with lost connection to the RC. It is clear that RTH is no longer an option in such a scenario. Drifting depending on wind speed is inevitable too. In my understanding a somewhat "healthy" reaction of the Spark should be landing on the spot as soon as possible. And if not at that "early" stage, it should do so at the latest when battery gets to critical level.

What will happen is Atti mode drifting on the wind loss of radio no gps continues drifting if it regains gps it will RTH if not will drift until battery dies or crash. I can’t be sure if it will land on critical battery but I don’t think thr aircraft is equipped to preform this without gps .

hallmark007 Posted at 2017-9-29 14:28
What will happen is Atti mode drifting on the wind loss of radio no gps continues drifting if it regains gps it will RTH if not will drift until battery dies or crash. I can’t be sure if it will land on critical battery but I don’t think thr aircraft is equipped to preform this without gps .

Just to clarify: I did not attempt to describe what "would" happen in my post above. I tried to describe what "should" be a possible healthy behaviour to be implemented in the future.

Wachtberger Posted at 2017-9-29 14:36
Just to clarify: I did not attempt to describe what "would" happen in my post above. I tried to describe what "should" be a possible healthy behaviour to be implemented in the future.

Watchberger your very forward thinking, and yes it is redundancies like these that might make drone flying a safer hobby for all of us.

hallmark007 Posted at 2017-9-29 14:40
Watchberger your very forward thinking, and yes it is redundancies like these that might make drone flying a safer hobby for all of us.

Thank you hallmark007! Seeing all these posts of "fly aways" and "sudden falls from the sky" which at a closer look mostly appear to have rather very "individual backgrounds" and not necessarily inbuilt technical faults of the Spark, I started thinking about failsafe functions that should not be too difficult to implement and make life for DJI support much easier.

All we do is guessing according to my opinion. One thing i really do not understand is: at this time dji has got enough examples of fly aways, compass errors etc etc. Why do we (sparkusers) have to keep testing and testing if we have the sparkbuilders/inventers/software producers and if not being fired by dji , the dji testers. So far i only read posts about errors and that we have to test, and not a word from dji

The only way to test it is by using the "DJI Assistant 2 development module" and set the sports mode value to 3, so that sports mode will automatically default to ATTI

fans61699a75 Posted at 2017-9-29 14:59
All we do is guessing according to my opinion. One thing i really do not understand is: at this time dji has got enough examples of fly aways, compass errors etc etc. Why do we (sparkusers) have to keep testing and testing if we have the sparkbuilders/inventers/software producers and if not being fired by dji , the dji testers. So far i only read posts about errors and that we have to test, and not a word from dji

Please accept my apologies that I have to disagree to a certain extent at least. First of all I am not a DJI staff member who would try to defend the quality of their products. But the number of reports of so called "fly aways" and "falls from the sky" you are referring to are at the beginning most of the time without a clear description of the precise circumstances of the flights. Secondly they almost always come from pilots for who the Spark is the very first AC they are flying. And here we come closer to the root of a large part of the problems at least and one share that DJI is not completely innocent of: The Spark is marketed as a perfect beginners drone that you can fly without RC and even without a mobile device. "Just unpack, switch on and take fantastic "selfies" by just directing it with gestures". Such a marketing is definitely condemmed to create huge disappointment on all sides because it does seriously underestimate the complexity of the Spark and the requirements to operate it safely as well as with its full capacities. Another share goes to all those who "just unpack and fly" without prior reading the manual, taking some time to train and gather additional information in this forum or the very many national ones available. But here the somewhat "vicious circle" closes again. The Spark is probably the first high tech drone affordable for a larger audience. If one buys a Mavic or above, the level of investment most probably automatically leads to a much more careful and attentive approach of the buyers with more preparation before launching it into the field.
I certainly do not want to criticise the many Spark buyers! But I strongly believe that we should all accept that also the Spark is a highly sophisticated device that needs some attention and time invested by the pilot to properly understand all its functions and behaviour.

Well i am iceland now for a few days with my phantom 4 pro and spark  and within these days i already met 2 experienced spark droners who lost their ac here. Also i had twice compass atti and gps lost. And reading all these reviews about spark crashes, i start to wonder if these drones are reliable and flyworthly. I am also reading phantom forum and comparing the phantom crashes and the spark crashes , i can tell that spark is the winner of all crashes. And dji is nowhere to find with comments or explanations what to do , how to prevent and what they are doing about it. The only thing they come up with are new fw and app updates, but the spark flyaways remain and not solved

As discussed above, the drone will enter ATTI mode when GPS signal is weak and VPS does not work. In ATTI mode, the drone could not hover or brake, but the pilot can control the drone fully. I know it is hard for some beginners to fly in ATTI mode, so I suggest to fly the drone in line of sight and keep away from interference.

DJI Susan Posted at 2017-9-29 19:34
As discussed above, the drone will enter ATTI mode when GPS signal is weak and VPS does not work. In ATTI mode, the drone could not hover or brake, but the pilot can control the drone fully. I know it is hard for some beginners to fly in ATTI mode, so I suggest to fly the drone in line of sight and keep away from interference.

This does not answer the initial question: what happens when the connection to the RC is lost while in ATTI mode?

ciesnik Posted at 2017-9-29 20:58
This does not answer the initial question: what happens when the connection to the RC is lost while in ATTI mode?

There is probably no good solution in this difficult scenario. The only one would be Spark to descent and to be smart enough to select by itself appropriate spot for landing – not real.

Another approach is to work on prevention such a scenario to happen. Spark’s GPS receiver definitely needs improvement. I have read many cases of weak or lost GPS signal without visible reason.
Here are my observations in comparison to Phantom 3:
Being at one place – flat terrain with some trees and small buildings around – P3 records the Home point about 30 seconds after turning on, while Spark is not able to record Home point at all. Sometimes it happens after a few minutes of hovering in the place at some height.
In the air P3 keeps constant 15-20 satellites lock as usual, while Spark - inconsistently 10-12 satellites and sometimes loses strong GPS signal and GPS related functions do not work. It has happened even in wide open areas:

I have observed similar performance of Spark in different locations.

Guys, I am not asking “what should happen” or “what do you think happens”. I am asking what is actually the built-in algorithm for this situation running in the Spark, because the manual does not address this case.

ciesnik Posted at 2017-9-29 20:58
This does not answer the initial question: what happens when the connection to the RC is lost while in ATTI mode?

The AC is uncontrollable then. You should direct the Antennas precisely towards the AC and hopefully will get the RC connection back. If this fails and also GPS link does not come back you have a serious problem and can only hope that there is no strong wind and that the battery will go empty soon with the AC not being above water.
@Others: Please correct me should my assessment be wrong.

Wachtberger Posted at 2017-9-30 02:33
The AC is uncontrollable then. You should direct the Antennas precisely towards the AC and hopefully will get the RC connection back. If this fails and also GPS link does not come back you have a serious problem and can only hope that there is no strong wind and that the battery will go empty soon with the AC not being above water.
@Others: Please correct me should my assessment be wrong.

Your right logic will tell you what happens, if there was hidden technology in spark , we would all know about it but there is not, it is all explained in the manual you just have to look at it logically.

hallmark007 Posted at 2017-9-30 02:41
Your right logic will tell you what happens, if there was hidden technology in spark , we would all know about it but there is not, it is all explained in the manual you just have to look at it logically.

What is explained in the manual is the situation what happens when the connection to the RC is lost while home point is recorded and compass AND GPS is ok -> then the AC initiates RTH after 3 seconds without connection to the RC (or after 20 seconds if you are connected with your phone).

I tried this indoors: switching RC off while aircraft is hovering - it initiates landing after 3 seconds.

Question to DJI: Is this the same behavior when outdoors and in ATTI mode when connection is lost to RC?

ciesnik Posted at 2017-9-30 03:22
What is explained in the manual is the situation what happens when the connection to the RC is lost while home point is recorded and compass AND GPS is ok -> then the AC initiates RTH after 3 seconds without connection to the RC (or after 20 seconds if you are connected with your phone).

I tried this indoors: switching RC off while aircraft is hovering - it initiates landing after 3 seconds.

Indoors in opti mode is not the same as outdoors in Atti mode dji Susan already explained this to you , you will have no vision sensors so nothing to hold AC  in place for landing.
You already know what happens in Atti mode Aircraft will drift on the wind you still have control, when you lose controls it will continue to drift with only barometer to hold its horizontal, I’m sure if there was a failsafe landing we would know about it, but there is not so it’s just a matter of doing the math.

hallmark007 Posted at 2017-9-30 03:34
Indoors in opti mode is not the same as outdoors in Atti mode dji Susan already explained this to you , you will have no vision sensors so nothing to hold AC  in place for landing.
You already know what happens in Atti mode Aircraft will drift on the wind you still have control, when you lose controls it will continue to drift with only barometer to hold its horizontal, I’m sure if there was a failsafe landing we would know about it, but there is not so it’s just a matter of doing the math.

And can you explain why the AC then simply does not descend until vision sensors can kick in and then initiate landing in this case?

ciesnik Posted at 2017-9-30 04:22
And can you explain why the AC then simply does not descend until vision sensors can kick in and then initiate landing in this case?

Because vision sensors are off in Atti mode.

fans61699a75 Posted at 2017-9-29 14:59
All we do is guessing according to my opinion. One thing i really do not understand is: at this time dji has got enough examples of fly aways, compass errors etc etc. Why do we (sparkusers) have to keep testing and testing if we have the sparkbuilders/inventers/software producers and if not being fired by dji , the dji testers. So far i only read posts about errors and that we have to test, and not a word from dji

I would love it if there was a DJI representative who participates in testing the products and who participated in the forums.  It would be enjoyable to throw these scenarios at him/her and he/she could report on his/her experiences.

Alas we don't have this, which is a bummer.  I think DJI could benefit if they had some "technical" representation participating in the forums.  I know the forum community would benefit.  Presently DJI has "support" representation only, which is essential.

I personally like to test.  For whatever reason, I've found that I don't fly my drone just for the sake of flying.  I'm much more motivated when I fly with a purpose, and testing out scenarios and behaviors gives me that purpose.

This particular use case is difficult to test, because we can't presently control atti mode.  As stated earlier, we (I) could test on another craft where atti mode is controllable (my P3S).

My hope would be that it tries to perform an emergency landing when the batteries are critically low.  My suspicion is that it will just run until it drops if it doesn't crash first.

fans61699a75 Posted at 2017-9-29 15:59
Well i am iceland now for a few days with my phantom 4 pro and spark  and within these days i already met 2 experienced spark droners who lost their ac here. Also i had twice compass atti and gps lost. And reading all these reviews about spark crashes, i start to wonder if these drones are reliable and flyworthly. I am also reading phantom forum and comparing the phantom crashes and the spark crashes , i can tell that spark is the winner of all crashes. And dji is nowhere to find with comments or explanations what to do , how to prevent and what they are doing about it. The only thing they come up with are new fw and app updates, but the spark flyaways remain and not solved

You forget that many if not most of spark flyers are beginners, and all of this is quite comparable to Mavic last January. Along with new firmware/ software comes experience.

Charles Adams Posted at 2017-9-30 05:05
I would love it if there was a DJI representative who participates in testing the products and who participated in the forums.  It would be enjoyable to throw these scenarios at him/her and he/she could report on his/her experiences.

Alas we don't have this, which is a bummer.  I think DJI could benefit if they had some "technical" representation participating in the forums.  I know the forum community would benefit.  Presently DJI has "support" representation only, which is essential.

To test it with a Spark maybe you could intentionally decalibrate the compass, fly in ATTI mode and then turn off the RC mid-flight?

ciesnik Posted at 2017-9-30 07:27
To test it with a Spark maybe you could intentionally decalibrate the compass, fly in ATTI mode and then turn off the RC mid-flight?

As in calibrate it in a very poor location?  Maybe.  Or purposefully launch it from a poor location (starting from a location with lots of metal).

I should point out an observation that just occurred to me:

The fact that we are trying to figure out a way to make the spark fly less-than-optimal, AND the fact that we are struggling to come up with a reliable way to do so is a testament to spark engineering.  Yes it happens when you don't want it to (it always does), but look how difficult it is for us to figure out a way to make it happen on purpose.  The Spark WANTS to fly right.

Charles Adams Posted at 2017-9-30 07:42
As in calibrate it in a very poor location?  Maybe.  Or purposefully launch it from a poor location (starting from a location with lots of metal).

I should point out an observation that just occurred to me:

Just tape a magnet to the body of your spark, this will throw off your compass , risky particularly when we already know there is no failsafe landing in Atti mode.

hallmark007 Posted at 2017-9-30 07:58
Just tape a magnet to the body of your spark, this will throw off your compass , risky particularly when we already there is no failsafe landing in Atti mode.

Yeah....   no..... , but yes it makes a lot of sense.  Again, if this is what we have to go through to try and make the spark behave poorly....

Charles Adams Posted at 2017-9-30 07:42
As in calibrate it in a very poor location?  Maybe.  Or purposefully launch it from a poor location (starting from a location with lots of metal).

I should point out an observation that just occurred to me:

the only reason why it is so hard to simulate this is because the software does not have a switch to enter ATTI mode on purpose. I would not confuse this with sophisticated engineering...

I'm trying to post results of me flying my P3S, but I'm getting a response that my post fails because of bad information.  This is a test post to see if the same thing occurs to me regardless of the content.

I'll try again with different words.  My test of these circumstances with my P3S failed.  The reason is that the P3S set in manual ATTI mode still knows its location.  It only changes flight behavior (it lets the elements have influence over the flight), but it is still able to identify and map its coordinates.  Therefore the P3S in manual ATTI mode cannot be used to perform this test.