Why am I gaining altitude in a turn?

[BudWalker]

When turning my FPV will often gain altitude. I was assuming that it is because I inadvertently was moving the throttle while giving it yaw input. But, that doesn't appear to be the case. I'm hoping someone can tell me what I'm doing wrong.

Below are plots that illustrate this with the turn taking place between the two dashed lines.


The top plot shows the stick inputs. Note the pitch and throttle are held constant while the roll and yaw are given positive input.
The second plot shows the altitude gain of about 30 meters.
The third plot shows the roll dipping for about 0.75 secs and then slowly rising in the turn. In the last plot the pitch rises at first but then starts to decrease.
@Kilrah

 

[okw]

I don't have a DJI FPV and probably can't help you figure this out, but what mode were you flying in (N, S, or M)? That may help others give you some ideas?

My understanding, as someone on the outside looking in, is that Normal and Sport modes act similar to the Mavic line where it holds altitude if the throttle stick isn't moved. Or at least is supposed to.

 

[BudWalker]

I don't have a DJI FPV and probably can't help you figure this out, but what mode were you flying in (N, S, or M)? That may help others give you some ideas?

My understanding, as someone on the outside looking in, is that Normal and Sport modes act similar to the Mavic line where it holds altitude if the throttle stick isn't moved. Or at least is supposed to.

Sorry, should have mentioned the FPV was in M mode.

 

[Kilrah]

A bit hard to know without seeing the matching video and speed you're flying at... I couldn't tell what I'm doing when flying a level turn without actually making one and concentrating on it

One point that probably affects it is the difference in axes due to camera tilt. When your camera angle isn't 0 then what you see isn't aligned with the aircraft anymore, so that when you aim to roll the camera view the command you have to give the aircraft essentially becomes a 3-axis mix. If you're not commanding all 3 you're not doing what you think.

BTW it seems the RC pitch value doesn't read correctly from the logs, always 0 on your log above, and on mine I see an occasional 1 when I give full input but nothing in between, seems it might drop the decimals

 

[BudWalker]

A bit hard to know without seeing the matching video and speed you're flying at... I couldn't tell what I'm doing when flying a level turn without actually making one and concentrating on it

One point that probably affects it is the difference in axes due to camera tilt. When your camera angle isn't 0 then what you see isn't aligned with the aircraft anymore, so that when you aim to roll the camera view the command you have to give the aircraft essentially becomes a 3-axis mix. If you're not commanding all 3 you're not doing what you think.

BTW it seems the RC pitch value doesn't read correctly from the logs, always 0 on your log above, and on mine I see an occasional 1 when I give full input but nothing in between, seems it might drop the decimals

I think my problem is because I haven't fully transitioned from flying previous DJI drones. On those drones the gimbal is compensated in both roll and pitch so it always looks like a turn is made where the drone doesn't roll. In FPV M mode I was doing a turn starting with yaw and then applying just enough roll to bring the FPV to level. More roll needs to be applied so that the camera view shows that the FPV is leaning into the turn.

Consider the extreme case where enough yaw is applied to rotate the FPV 90° then enough roll is applied so that the camera view shows a level horizon. Then all the thrust is pointed down and not to the side causing a gain in altitude.

Reading this I'm realizing that I'm thinking a level horizon is just roll. I need to include the height of the horizon (i.e. pitch) as well. I'm guessing you would make the point that a proper camera tilt will make this easier .

Anyway, thanks for the bug report. That will be fixed in the next version. In the interim you can select "DatDefined Signals Only"


and look at mr_controller:ctrl_pitch

 

[rktman]

Sorry, should have mentioned the FPV was in M mode.

Well there you go.

You will never, ever, ever be able to hold altitude like the Flight Controller can.

EVER.

 

[thetechnobear]

you are probably overcompensating.....

when you turn , you are applying roll, so you are not turning on a single axis - this means the drone is pitching forward, so diving (and gaining speed)
so to compensate, you pull back on the stick, and also adjust your throttle - you will do naturally (as otherwise you'd hit the deck )
but of course, its easy to overcompensate and so gain height.
(we dont have an artifical horizon , so it can be a little hard to see this whist flying, and the VSI is a bit small to 'keep an eye on')

with practice of course you can get much better at it - especially for shallower turns - with sharp turns Ive still a habit of diving a bit.
though, I thinks thats partly because I love to do a rise n' dive turn, with the turn leading me into a dive down a mountainside ... looks really dramatic

 

[BudWalker]

you are probably overcompensating.....

when you turn , you are applying roll, so you are not turning on a single axis - this means the drone is pitching forward, so diving (and gaining speed)
so to compensate, you pull back on the stick, and also adjust your throttle - you will do naturally (as otherwise you'd hit the deck )
but of course, its easy to overcompensate and so gain height.
(we dont have an artifical horizon , so it can be a little hard to see this whist flying, and the VSI is a bit small to 'keep an eye on')

with practice of course you can get much better at it - especially for shallower turns - with sharp turns Ive still a habit of diving a bit.
though, I thinks thats partly because I love to do a rise n' dive turn, with the turn leading me into a dive down a mountainside ... looks really dramatic

But as I said in my OP both pitch and throttle were held constant. Here it is again

Throttle (blue) and pitch (green) don't change but altitude (red) does increase.

As I said in #5 the problem is that I need to input more roll.

 

[BudWalker]

Well there you go.

You will never, ever, ever be able to hold altitude like the Flight Controller can.

EVER.

I agree. I'll never be able to hold altitude in a turn as well as the FC in N or S mode. But in the OP you can see that the gain was 30 meters. I should be able to do better than that.

I bet that the more experienced M mode flyers will be able to hold altitude so that it's indistinguishable from the FC in N or S mode.

 

[thetechnobear]

But as I said in my OP both pitch and throttle were held constant. Here it is again
View attachment 5281
Throttle (blue) and pitch (green) don't change but altitude (red) does increase.

As I said in #5 the problem is that I need to input more roll.

ok, so some assumptions...
- you doing a co-ordinated turn
- you have a postive pitch on your camera.

the more you roll, the more height you lose in a turn ... (yaw is irrelevant)
you cannot gain height in a turn, you'll always lose unless you have throttle , or reduction in pitch.
its all just about how much downward thrust you get due to the pitch/roll of the drone. it gets max thrust at level flight, and less at any other angle.
... if your data says otherwise, Id start suspecting the data.



(if you're doing a kind of yaw-spin to turn, then things are a little different - but thats not really the way you turn a quad.)

 

[Kilrah]

I was asking about the speed because at speed you start getting aerodynamic effects, you can't just assume the props' static thrust is the only thing at play. When you do a flat turn with speed you give roll and yaw towards the turn and will need to reduce throttle not to climb. Turning means the bottom of the quad is facing the current direction of travel, which basically "cushions" the props. A component of that extra force is directed upwards and makes you climb.

 

[thetechnobear]

Not something I’ve noticed.
This is no wing, so I’d have said the lift would be negligible.

 

[BudWalker]

..........

the more you roll, the more height you lose in a turn ... (yaw is irrelevant)
you cannot gain height in a turn, you'll always lose unless you have throttle , or reduction in pitch.
its all just about how much downward thrust you get due to the pitch/roll of the drone. it gets max thrust at level flight, and less at any other angle.
... if your data says otherwise, Id start suspecting the data.
.......

Here is the .DAT for anyone wanting to have their own look at the data.
FLY036-039-20210509104624.DAT

I agree (except that you probably meant to say that an increase in pitch will cause an altitude gain). Thinking about what you said I noticed that the going into the turn that roll was -5.8°. This was because the FPV was headed north and the wind was out of the west.at 6.8 m/sec. The right turn was done by providing right ctrl-yaw and right ctrl-roll which caused the roll to increase towards 0.0. I.e., less absolute roll which provided additional thrust downward.

The top plot shows the gradual roll increase towards 0.0. The middle plot shows the resulting altitude gain. The bottom plot shows the FPV heading relative to the wind direction. About 5 secs after the turn was started the wind was coming from behind the FPV.

 

[BudWalker]

ok, so some assumptions...
- you doing a co-ordinated turn
- you have a postive pitch on your camera.

I'm not sure what coordinated turn means for a quad drone. I was providing both right ctrl-yaw and right ctrl-roll.

For a fixed wing AC co-ordinated flight means that nose is pointed into the air stream. (I can still hear my flight instructor hammering it into my skull that air flowing over the fuselage doesn't provide lift.) But that doesn't apply to a quad.

Without the benefit of a yaw string on the canopy the fixed wing pilot relies on the slip ball (inclinometer) to know if there is lateral acceleration. The .DAT log file provides accelerometer data. So I suppose that if IMU:accelY is 0.0 that could indicate coordinated flight. But, if the quad is holding some roll to compensate for wind then the Y axis will be partially aligned with the earth's gravitational field. This will result in a non zero IMU:accelY not caused by any lateral movement.

 

[rktman]

Well, gotta say I'm very confused. I'm new to acro/manual flying (very new), so I lack experience to comment on this.

What I DO have is a physics degree, am a retired engineer, and private pilot. So I think I understand the basic physics and forces in play here.

For any aircraft, there are 4 forces involved in movement, two pairs opposing each other: Gravity and lift, forward thrust and drag. The plane moves level at constant speed when each "pair" of forces is balanced.

When pitch or roll are non-level, a portion of lift or thrust is "diverted" to a horizontal component relative to gravity, which results in horizontal movement relative to the gravity vector. This reduces the amount of that force component balancing it's partner. For example, without throttle changes to a quad when pitching forward, it should start and continue to lose altitude, due to less thrust going to fighting gravity, since now part of the thrust is opposing drag. As such, to maintain altitude, theory says you should have to INCREASE throttle to greater you make the pitch or roll angles.

Yet people here, experience pilots, are saying the opposite -- reduce throttle when pitching forward and translating forward? I'm confused.

 

[BudWalker]

......

Yet people here, experience pilots, are saying the opposite -- reduce throttle when pitching forward and translating forward? I'm confused.

Where do you see this?

 

[rktman]

I'm not sure what coordinated turn means for a quad drone. I was providing both right ctrl-yaw and right ctrl-roll.

Same as it means for a fixed-wing plane, or a helicopter. It's the difference between sliding around a turn with the vertical axis of the aircraft parallel to the center axis of the turn, vs. the vertical axis intersecting the center of the turn axis.

The control inputs are the same, and produce the same result. Just like with an airplane, you can turn uncoordinated successfully, but there are a bunch of reasons coordinated turns are better, chief among them being they are much tighter (i.e. you can make smaller radius turns, and get to your intended direction faster).

It's not necessary, but it's rather sloppy not to learn to and turn this way.

 

[rktman]

Where do you see this?

Well, several places, the most recent here:

I agree (except that you probably meant to say that an increase in pitch will cause an altitude gain).

My understanding, based on the physics I described, is that an increase in pitch will cause a LOSS of altitude, without increasing throttle.

 

[BudWalker]

Well, several places, the most recent here:
My understanding, based on the physics I described, is that an increase in pitch will cause a LOSS of altitude, without increasing throttle.

Sorry for the confusion. I was assuming that the quad was moving forward requiring a negative pitch. Increasing pitch towards 0.0 will result in more vertical thrust resulting in an altitude gain.

Another way to look at is to consider the tilt angle with 0° being aligned with the Z axis. If tilt angle > 0° there is a loss of downward thrust resulting in an altitude loss. This perspective also includes the situation where the quad is moving laterally, i.e. along the Y axis.

 

[BudWalker]

Same as it means for a fixed-wing plane, or a helicopter. It's the difference between sliding around a turn with the vertical axis of the aircraft parallel to the center axis of the turn, vs. the vertical axis intersecting the center of the turn axis.

The control inputs are the same, and produce the same result. Just like with an airplane, you can turn uncoordinated successfully, but there are a bunch of reasons coordinated turns are better, chief among them being they are much tighter (i.e. you can make smaller radius turns, and get to your intended direction faster).

It's not necessary, but it's rather sloppy not to learn to and turn this way.

I believe the rest of my post beyond what you quoted shows I know what a coordinated turn is. But, more to the point is how can a FPV pilot know if the turn is coordinated. There is no indication about the alignment with the air stream. And, the accelerometer data is not displayed in the Fly app.