If you take off on the top of a mountain and raise the spark for example two meters, but then you fly out the cliff where there may be many more meters to the ground, would the spark change the distance shown to the ground? How does he know the real distance from the ground?

X-Files for me....

It uses barometric pressure

Measured distance will show a minus, if you fly over cliff, it's measured by barometer, from take off your Mavic will always use AGL from homepoint .

It is always counting meters above/below the start height,

Yes, relative to takeoff position. It is not absolute height from sea level

Cool, that means it will still hold its altitude even though you fly it over a cliff right?

If you take off from the top and head off the AC to the cliff, it does not make much sense to think that it will indicate in negative or that he knows that the meters below have increased.

I understand that when the Spark takes off, he calculates the ambient pressure and take that value as 0 (ground), but if we fly it 3 meters above 'our ground' and get it out of the top, where you may have 50 meters below, the ambient pressure it is still the same, so it is not logical to think that he will notice or show any change in the altitude indicator (he may think he's still at 3 meters high, instead of 53 m.

I am not sure if this is a real problem, but it frightens a little, since in cases of flying between mountains and cliffs you would never know the real distance to the the ground.

Do you see it logical or am I messing?

There is no other way to figure out altitude using atmospheric pressure. If you take off from an elevated position and head down hill or beyond a cliff it will be negative. There is no toerh way but that's is useful data anyway.

The only way to know actual altitude over ground woudl be to use an electronic signal which woudl increase weight, complexity and battery life

I am not an expert in aviation or in physics... so please someone correct me if I am wrong:

Is it correct that atmospheric pressure is the same at point A and B?

Accepting that it is correct, the only difference is that at point A we have 5 meters below us, and at point B we have 30 meters, but spark thinks he is at 5 meters.

So, at point B spark will tell us he is at 5 meters high, and if we descend -10 meters then he will say he is at -5.

That could explain a negative height on the app, but in my case the AC was hovering at the same height and the indicator gone crazy changing values all the time, undecided.

If you mean a thing the top of the vertical lines, then yes. Barometric pressure will be the same at point A and point B

In airplanes the altimeter is manually set either to current sea level pressure to get altitude above sea level or it can be set to current local airport pressure. If you take off from a runway at an elevation of 3000' and head towards sea level, you altimeter will read negative if you don't reset the altimeter to sea level pressure.

As to why your altitude readings were erratic, it had to be a fault. Either the sensor or a  software glitch

I hope it is something related to software :s

Great explanation, thanks!

Nemroig Posted at 2017-8-24 06:38
I am not an expert in aviation or in physics... so please someone correct me if I am wrong:

Is it correct that atmospheric pressure is the same at point A and B?

And now think it this way:
You start at sea level, in position B, and let the Spark climb up the hill, to 30m.
There are trees, so 35-40m.
You loose the Spark behind/over the trees and have a RTH - height set to 30m
Spark will go to that set height, no matter what!

Nemroig Posted at 2017-8-24 06:38
I am not an expert in aviation or in physics... so please someone correct me if I am wrong:

Is it correct that atmospheric pressure is the same at point A and B?

This might explain a little for you although I first used it for Mavic it is the same for Spark.

The technical term in aviation is ATO - altitude measured above take off.

The Mavic system measures all altitudes relative to the takeoff point.

The default setting for maximum altitude in the controller/aircraft may be something like 120 meters ATO; this means it won't accept a command to rise higher than 120 meters (400 ft) above the takeoff point. You can change this limit however you like, up to a maximum of 500 meters ATO.

Some jurisdictions' laws may specify a ceiling of 400 ft AGL (above ground level) as the US currently does. This may include a reduced ceiling across canyons or a raised ceiling around buildings and mountains. The Mavic has no idea where the ground is, unless it happens to be hovering within about 2 meters of some object. It does not understand geographical dropoffs or hills at all. Your compliance with the 400 ft AGL rule is entirely on your own mental calculations.  If standing on a bridge or cliff, you can fly below the take off altitude, for example to negative numbers like -50 ft ATO.

The Mavic actually has an internal barometric altitude device, and the GPS/GLONASS receivers, both of which measures altitude relative to the average sea level. Thus your hometown may be at 2000 ft AMSL (above mean sea level) and the Mavic would use that measurement at the time of takeoff. However, NOWHERE on the controller screen nor the DJI Go app screen will you see altitude described this way. This is only used for internal calculation.

no shows where you took off from

As far as I can tell it is relative to your takeoff point. On the face of it it would seem that the Spark has no idea of what the actual altitude is. However, it does record the takeoff elevation. But after that it has no way to determine actual AGL unless you're close enough to the ground to use the vision system.

So if you take off from a mountain and fly over a valley you may have taken off from say, 500m and flown to 10m AGL (above ground level), so 510m total, but then you fly out over a valley that has a low point of say, 300m and now you're technically 210m AGL but the Spark still thinks it's only 10m up. :-/

In order for it to be able to accurately calculate AGL it would need the current altitude (which it can calculate from the GPS signals) and it would need to determine where ground level is. For that it would need either a very detailed terrain map or a downward facing sensor that would work downward several hundred meters, at least. It has neither. Only the vision system which iirc is limited to 10m AGL.

I would just be happy with a current altitude display which it already knows from GPS.

Nemroig Posted at 2017-8-24 06:38
I am not an expert in aviation or in physics... so please someone correct me if I am wrong:

Is it correct that atmospheric pressure is the same at point A and B?

@Nemroig - a barometer is a poor way to calculate altitude. Yes, in your limited scenario the barometric pressure at A and B should be the same but with a minimum of 4 GPS satellites you can get much better accuracy and since the Spark uses at least 10 satellites it should know to within inches its actual altitude.

As I mentioned in my previous post, actually calculating AGL is more complicated as the Spark doesn't have the right sensors or data to understand where actual ground level is (once its above 10m AGL).

TranceMist Posted at 2017-8-24 10:14
@Nemroig - a barometer is a poor way to calculate altitude. Yes, in your limited scenario the barometric pressure at A and B should be the same but with a minimum of 4 GPS satellites you can get much better accuracy and since the Spark uses at least 10 satellites it should know to within inches its actual altitude.

As I mentioned in my previous post, actually calculating AGL is more complicated as the Spark doesn't have the right sensors or data to understand where actual ground level is (once its above 10m AGL).

Its a pity that such a complex device doesn't have some way of knowing the actual distance to the ground at all times (gps, cartography, atmospheric pressure ...).

The app only shows the relative altitude of the takeoff point, will not change the altitude to the current ground. So you need to fly carefully.

What Thor is saying is that you MUST be aware of your surrounding's when you fly and understand the way your drone works in failsafe and RTH.
You should Always reset your RTH To a height that is appropriate to your surroundings
I.E large level area with no obstructions 30m is fine, wooded area with 40m trees on a rolling landscape 100m mite be more appropriate.
There should be a better explanation in the Manual

I have no doubt it works, but wouldn't GPS do just as good a job as a barometer?  If not better?  What about a storm front coming in, affecting the local air pressure?  High winds disturbing the immediate area of air pressure?  None of that stuff would be affected by GPS.

I'm actually curious to know if, say, a large 10'x10'x50' acrylic tube were made in a warehouse, with plenty of GPS signal, and the drone were set at 30 feet above the ground.  The tube is then sealed, and then decompressed to create a vacuum, would the Spark begin to FALL because the air pressure changes to make it think it's too high (thin air), or would the GPS take over, and tell it that it's actually in the right spot?

Sounds like a sweet test on YouTube.  

Exept for high end rtk gps which requires differential corrections, gps provides a much noisier vertical measurements than barometer.
Barometer are the standart for all flying machines as they will give same info for different aircrafts flying at the same time and neighboorhood. This allows ATC to manage vertical separation.
Now regular AC fly close to ground only near airports which are usually flats and the reference pressure (QFE) is broadcasted by radio or ATC.
For drones an intermediate solution would be to embed a dtm (digital terrain model) and the use gps to compute dtm delta to takeoff point and add it go barometric delta.
This will not give info about treetops and free dtm have an horizontal resolution and accuracy that would greatly reduce the interest. And global files are huged (several GB at 90m horizontal resolution). They could be downloaded on the fly/prior flight at app level and used only for display/warning

Pilots have been flying real airplanes using atmospheric pressure for about a century... it works. Why the heck would you need anything more fancy to fly a remote controlled toy?

PascalG Posted at 2017-8-25 16:39
Pilots have been flying real airplanes using atmospheric pressure for about a century... it works. Why the heck would you need anything more fancy to fly a remote controlled toy?

That's true, but as I explained, manned AC most often fly close to the ground only near "flat" airports... and are mostly concerned with ground height security aspects around those areas where they have a mean tu turn to "QFE" mode from the ref pressure given by ATC.
For them, the main advantage of barometric altitude is to guarantee consistent altitude separation between close aircrafts.

Not tht is does no for regular AC, but for UAV leisure pilots, a ground detection / height above ground measurement system makes sense. And indeed, DJI has implemented a very short range one.

First we are usually flying much closer to the ground. Second in most countries, the rules are expresses in terms of max height above ground.

There are sensors that could provide such info at 100m for quite cheap. If you type laser rangemeter in amazon, you'll find a ton of cheap devices (<40€) that can measure accuratlery a distance up to 100 m at say 1 measure per second. This is all technically possible. For now, it's probably a little too big for the Spark, add weight and extra consumption. For sure will fit in some future DJI product 'if not already in pro geas)... Just we are still now at the begining of UAV era.

Low cost solution is to do it "offline" from a database :
you have your barometric delta height from take off.
you poll DTM database from horizontal GPS information at take off and current point to have the terrain delta height. Add both and you'll be able to retrieve your current height above ground. Accuracy of this method depends :
- on horizontal accuracy of GPS vs DTM horizontal resolution
- on accuracy of the DTM file

Does anybody knows if there ss there a kind of API at DJI GO4 level to poll current AC coordinates (e.g. a TCP/IP server where the datastream would be available to some extra apps ?)