Making the transition from N (Normal) / S (Sport) Mode to M (Manual) Mode is not easy. I needed a lot of train with simulators like DJI Virtual Flight and Liftoff until I understood the behaviour of the drone. Hovering at low speed is always difficult, when you go faster, with higher camera angles, it's easier to learn the reaction of the drone to each stick axis.
Then changing the speed is a mix of the camera angle and a vertical reference position in the screen where you point the drone to make it go slower or faster.

My first experience in limited M Mode:


But this is only the first step, because attitude restriction is still enabled!
No loops, no rolls...

I'm still wondering why can't some of the popular FPV tricks be done with the Motion Controller (MC)?
An Acrobatic A Mode should exist for the MC leaving free unlimited restrictions to the drone attitude.
A loop or roll would be done just as it is done in a airplane with a single stick.
I believe some FPV tricks may be limited, but most of them may be acomplished by the Motion Controller stick since it can measure 3 axis of motion, yaw, roll and pitch, and a fourth axis, the speed trigger. So why learn the M Mode? Guess it will be replaced sooner or later.

Part of the reason why the motion controller will never work in  Manual the way it's currently designed is because of the mixing.
It can't go sideways and that's very important in manual
It has no way to know whether it's doing a flip or roll.
Or it's just going up therefore it wouldn't be able to do a role properly.
It cannot fly backwards also very important when flying a manual.

So without some modifications to the engineering and the mixing algorithms maybe some switches it is very unlikely a one handed motion controller can control the different axes that a drone can move in witch are much more than a plane's.

Cool that you made the jump. It will soon get better and is so much more fun and much cooler flying than N/S and MC!

I would switch out of the limited M mode as quickly as possible. It's not much more dangerous and flies so much better!

Nice flying.  

Nice flying.    Good to see you make the transition to M mode.  

SgtRay Posted at 8-17 17:17
Part of the reason why the motion controller will never work in  Manual the way it's currently designed is because of the mixing.
It can't go sideways and that's very important in manual
It has no way to know whether it's doing a flip or roll.

I can't understand why there is this design limitation. It's all software that can be updated with firmware upgrades. In fact the current mixing of channels is very well done, with very good calculations so that the drone can move in the direction where the camera is pointing. Excellent, but overlapping two input channels, the yaw and tilt of the MC.
Traditional method has 2 sticks, meaning we have 4 axis or 4 degrees of freedom, yaw, power, roll and pitch. With the Motion Controller we have one stick, from gravity acceleration sensor we can obtain the 2 axis for roll (left right tilt) and pitch (up down tilt). From magnetic sensor and angular rotation of the vertical axis we obtain the rotation for yaw motion. Next we have the accelerator for motors power. Making all the necessary control axis. Some calculations would be needed to mix left right tilt with the rotation resulting in the yaw and roll mix that we use with the 2 traditional sticks. Motion Control Acrobatic Mode would work like an helicopter stick for low speed (adding the MC rotation as the helicopter pedals) and progressively more like an airplane stick in high speed (adding MC rotation as the rudder control).

I understand all that.
you've seen my videos.
You cannot do those manovers with one stick. Not with the current sensors.. They're just not sensitive enough. And when I need to fly sideways I need to do it at all speeds at all angles. How would the software know when I need it to go sideways and when I needed to mix.
The 1st couple of chases I did I did with the stick. It just wasn't respond to enough. It just couldn't keep up.
That's not saying it never will. But there wouldn't need to be changes.

SgtRay Posted at 8-18 08:52
I understand all that.
you've seen my videos.
You cannot do those manovers with one stick. Not with the current sensors.. They're just not sensitive enough. And when I need to fly sideways I need to do it at all speeds at all angles. How would the software know when I need it to go sideways and when I needed to mix.

You may be right. Current sampling rate and processing power may be a problem in MC (no available specs). However accelerometer and gyroscopes sensors have very high sampling rates, for example MPU-9250 can work at 8KHz in normal mode with filtering capabilities. Not bellow the hall sensors of traditional sticks. For example, if remotes are using magnetic encoders like the AS5045 for sticks, sampling rates can go up to 10KHz in fast mode, similar frequencies. The problem may be how fast can we move our hand to keep the desired fly effect and if we have the ability to move it the right way. Gyroscopes are so fast and precise that they control the quad by themself. User pilot inputs are much slower!

EiraSYS Posted at 8-18 14:10
You may be right. Current sampling rate and processing power may be a problem in MC (no available specs). However accelerometer and gyroscopes sensors have very high sampling rates, for example MPU-9250 can work at 8KHz in normal mode with filtering capabilities. Not bellow the hall sensors of traditional sticks. For example, if remotes are using magnetic encoders like the AS5045 for sticks, sampling rates can go up to 10KHz in fast mode, similar frequencies. The problem may be how fast can we move our hand to keep the desired fly effect and if we have the ability to move it the right way. Gyroscopes are so fast and precise that they control the quad by themself. User pilot inputs are much slower!

Yet still it can't fly sideways. It would have to know that I wanna go sideways and not mix the sticks. There's no way it can know that.

Doing a Chase video is much like doing acro in manual. If you look at the flight data that I have on some of the videos you'll see that I am not pointing straight for most of the maneuvers. There is no amount of software that will know what I'm intending unless it's connected to my brain. But my fingers are connected to my brain and they know.

Unless there is a redesign that allows me a switch to dampen or remove the mixing it will never happen. Once you get further into manual flight you might see this intuitively.

Look at 1 of my last videos the one where I fly with a Cessna and you'll see immediately that there's no way the software will know when I change from sideways 90ﾟ to backwards 180ﾟ and that motion controller which I own by the way will never be able to do that because it will never know.

EiraSYS Posted at 8-18 14:10
You may be right. Current sampling rate and processing power may be a problem in MC (no available specs). However accelerometer and gyroscopes sensors have very high sampling rates, for example MPU-9250 can work at 8KHz in normal mode with filtering capabilities. Not bellow the hall sensors of traditional sticks. For example, if remotes are using magnetic encoders like the AS5045 for sticks, sampling rates can go up to 10KHz in fast mode, similar frequencies. The problem may be how fast can we move our hand to keep the desired fly effect and if we have the ability to move it the right way. Gyroscopes are so fast and precise that they control the quad by themself. User pilot inputs are much slower!

As for the sensors they are not precise enough. Which is why they're not used for anything precise. And the ones that are precise enough would be far more expensive than a couple of $100 that they ask for in that motion controller. The ones in the motion controller are barely as sensitive as the ones they use in virtual reality gear. To fly a manual precisely you need a lot more fine tuning. The problem is if you so much sneeze something that precise would crash

EiraSYS Posted at 8-18 14:10
You may be right. Current sampling rate and processing power may be a problem in MC (no available specs). However accelerometer and gyroscopes sensors have very high sampling rates, for example MPU-9250 can work at 8KHz in normal mode with filtering capabilities. Not bellow the hall sensors of traditional sticks. For example, if remotes are using magnetic encoders like the AS5045 for sticks, sampling rates can go up to 10KHz in fast mode, similar frequencies. The problem may be how fast can we move our hand to keep the desired fly effect and if we have the ability to move it the right way. Gyroscopes are so fast and precise that they control the quad by themself. User pilot inputs are much slower!

Just to make sure I understand. Backward flight would be achieved by pointing the MC up by some amount and pulling the trigger. Forward flight would be achieved by pointing MC down and pulling the trigger. The white circle wouldn't be used to know where the FPV will be going when the trigger is pulled.

I can see how this would work but the question is why. Is there some advantage in having one "stick"?

BudWalker Posted at 8-18 15:46
Just to make sure I understand. Backward flight would be achieved by pointing the MC up by some amount and pulling the trigger. Forward flight would be achieved by pointing MC down and pulling the trigger. The white circle wouldn't be used to know where the FPV will be going when the trigger is pulled.

I can see how this would work but the question is why. Is there some advantage in having one "stick"?

Yes, that’s the idea. To move sideways you tilt right or left and pull the trigger enough for the desired thrust. You still have all the necessary axis for control. The more tilt in MC the faster will be the maneuver.
For example a power loop can be accomplished by a small long tilt up of the MC and pulling the trigger, when upside down, releasing the trigger, then MC back to vertical leveling. No circles in the screen to point any direction. A small loop is obtained by a large and fast tilt up.
A roll right would be a MC tilt right releasing trigger when upside down. A yaw spin is obtained by rotating the MC around the vertical axis. The more you rotate the faster will be the spin. You need to rotate it back to stop spinning.And another design spec, if you released the trigger you would stop the motors for a dive.
Most of the people will only want those easy tricks, just like me, no need of complex rubik’s cubes.

EiraSYS Posted at 8-18 16:17
Yes, that’s the idea. To move sideways you tilt right or left and pull the trigger enough for the desired thrust. You still have all the necessary axis for control. The more tilt in MC the faster will be the maneuver.
For example a power loop can be accomplished by a small long tilt up of the MC and pulling the trigger, when upside down, releasing the trigger, then MC back to vertical leveling. No circles in the screen to point any direction. A small loop is obtained by a large and fast tilt up.
A roll right would be a MC tilt right releasing trigger when upside down. A yaw spin is obtained by rotating the MC around the vertical axis. The more you rotate the faster will be the spin. You need to rotate it back to stop spinning.And another design spec, if you released the trigger you would stop the motors for a dive.

I'm not sure how to text it and say it in typing. But I know I could show you in person. None of what you said would be good for manual.

There are more reasons to be a manual than tricks. And manual is really the only time some moves can be done. You absolutely cannot have mixing going on while you're doing a maneuver on a different axis.

I'm trying to think of a way I could possibly do a video on it. But I'm not really sure I have the proper skill. But I may try. Without modifications to that motion controller it will not be done. The fact that it hasn't been done is partial proof.
Right now this is the only example I can think of period but if I can make a video I'll make one later.

https://youtu.be/lO8B9Csd6Eg

These maneuvers are not complicated not one of them. There's a young lady at this event who I was teaching half of the maneuvers for the 1st time. She flew in manual for the very 1st time. She did half of them.

EiraSYS Posted at 8-18 16:17
Yes, that’s the idea. To move sideways you tilt right or left and pull the trigger enough for the desired thrust. You still have all the necessary axis for control. The more tilt in MC the faster will be the maneuver.
For example a power loop can be accomplished by a small long tilt up of the MC and pulling the trigger, when upside down, releasing the trigger, then MC back to vertical leveling. No circles in the screen to point any direction. A small loop is obtained by a large and fast tilt up.
A roll right would be a MC tilt right releasing trigger when upside down. A yaw spin is obtained by rotating the MC around the vertical axis. The more you rotate the faster will be the spin. You need to rotate it back to stop spinning.And another design spec, if you released the trigger you would stop the motors for a dive.

As for you saying it'll go sideways if you tilt left to right. No it won't. It will turn left to right. And you cannot shut the mixing off. And then when will it know to turn it back on. Thereby you cannot do it without modifications to the controller most likely a switch.

EiraSYS Posted at 8-18 16:17
Yes, that’s the idea. To move sideways you tilt right or left and pull the trigger enough for the desired thrust. You still have all the necessary axis for control. The more tilt in MC the faster will be the maneuver.
For example a power loop can be accomplished by a small long tilt up of the MC and pulling the trigger, when upside down, releasing the trigger, then MC back to vertical leveling. No circles in the screen to point any direction. A small loop is obtained by a large and fast tilt up.
A roll right would be a MC tilt right releasing trigger when upside down. A yaw spin is obtained by rotating the MC around the vertical axis. The more you rotate the faster will be the spin. You need to rotate it back to stop spinning.And another design spec, if you released the trigger you would stop the motors for a dive.

Again, I can see what you're describing and that it would be the equivalent of using the standard two stick controller. However, the advantages of the MC wouldn't exist in this new mode - no white circle to establish direction and no algorithm to interpret the MC orientation. That's OK - nothing wrong with being different. But, what would be the reason for doing this. Making a simple turn requiring yaw and roll would be harder in this mode than in the standard two stick controller.

I will try to explain with another language.



Imagine that Maverick was piloting an acrobatic drone quad as well has all his students, not F14s.
And they need to make the chase along the hills as well as the drop inside the montain to explode the base and get out.
If they had to mix yaw and roll with two sticks all the time then they would get something similar to SgtRay Cessna chase, a big difficulty to follow the valley pointing and correcting the sticks depending of speed and attack angle of the quads.
A single stick will be like fly by wire, you can process information with algorithms and adjust the desired quad reaction to the stick direction in function of the speed and attack angle. The result will be a previsible and much smoother movement feedback of the controls.

EiraSYS Posted at 8-19 00:08
I will try to explain with another language.
[/img]

Exactly. This comparison to fixed wing aircraft illustrates the point I'm making. That pilot has one stick that's identical to the right stick on a two stick controller. For the left stick the pilot uses his feet for yaw and his other hand to adjust the throttle. He is not using a single stick that combines roll, pitch, yaw and throttle.  

Try thinking about the wrist-hand-forearm motions required to achieve a simple turn with the new MC mode.

BudWalker Posted at 8-19 01:07
Exactly. This comparison to fixed wing aircraft illustrates the point I'm making. That pilot has one stick that's identical to the right stick on a two stick controller. For the left stick the pilot uses his feet for yaw and his other hand to adjust the throttle. He is not using a single stick that combines roll, pitch, yaw and throttle.  

Try thinking about the wrist-hand-forearm motions required to achieve a simple turn with the new MC mode.

Ok, I may be willing to transform the behavior of a drone and turning it into an airplane flying style when in reality it is not. The fact is that learning M Mode is not easy. For me it’s not obvious the need of pushing two sticks in opposite directions to roll around the direction I’m flying.

EiraSYS Posted at 8-19 01:55
Ok, I may be willing to transform the behavior of a drone and turning it into an airplane flying style when in reality it is not. The fact is that learning M Mode is not easy. For me it’s not obvious the need of pushing two sticks in opposite directions to roll around the direction I’m flying.

I also struggled with understanding axial rolls. This YouTube helped a lot


Since  the gimbal has fixed Yaw the only way that an axial roll can be  achieved is by Yawing the FPV. I suppose it's possible to have the  Flight Controller make any roll be an axial roll by determining and  applying the necessary opposite yaw. But, is that really what you want? During a  simple turn both roll and yaw in the same direction are required. The FC  would be applying yaw opposite that intended by the pilot. I.e.,  turning would be difficult, maybe impossible.


Your  comments earlier about the difficulty of transitioning to full M mode  reminded me of my own difficulties at that point. My advice? Remove the M  mode attitude restrictions as soon as you can. Don't waste time trying  to hover. You can do that later. Have some fun doing some high speed  turns, etc. Get comfortable in a simulator doing barrel rolls and loops.  Then do those with the FPV with plenty of altitude. And, don't try to  land it in Manual mode - just switch to N mode to land it.

BudWalker Posted at 8-19 06:56
I also struggled with understanding axial rolls. This YouTube helped a lot

Thank you BudWalker, this is an excellent video. It explains exactly my difficulty with the controls and the relation with what is seen by the drone camera. The direction of flight is not always the same of the pointed by the camera, so we need to fly thinking about the drone orientation and not by what is seen by the camera. The camera gives only relative pictures to what is happening to the drone. Guess that is why racer videos are so difficult to understand, they catch only some instantaneous moments of the images seen by the camera to act automatic sequences over the sticks. All in between video frames don’t even care for them.

Second flight in M Mode, full manual, no attitude restrictions as suggested by Depp. It makes all the difference!
There was less wind than in previous flight, so it was easier to control. Guess the simulator has no wind at all, how did they forget that?

Back to the Motion Controller proposal for the Manual Acrobatic Mode! (M Mode or A Mode)
The information is directed for DJI engineers for the development of a new firmware software for the DJI FPV drone and the Motion Controller.
This also valid for the new Avata drone that may be controlled with the Motion Controller.
The desire of having an easy to use controller that can make impressive and spectacular movies with simple hand gestures may attract many users.
This way there is no need to learn complex combinations of sticks in manual mode to make a simple loop or roll.Hovering control will also benefit from this solution, approaching it to an helicopter, with precise motion and allowing backwards displacement.
Most people prefer standard acrobatic moves that look great in a movie, with stabilized and smooth sequences.
This is not directed to racing drones that need technical complex moves to obtain good racing times but loosing in image and movie quality that are difficult to perceive by a normal spectator.
A camera angle must be calculated as a function of drone speed and direction.
Motion Controller axis remains equal to the calculated angle of the camera.
If speed=0, pressing the accelerator will make the quad go up and behave as an helicopter.
As speed goes up the pitch angle of the quad will decrease while the camera angle will increase up until the quad will behave as an airplane.
Gimbal tilt control should be used as a displacement from the reference camera axis, to show upper or lower footage relative to the calculated angle of the camera.
In this case the Motion Controller reacts to the axis of the calculated angle of the camera and not to the axis of the camera with the tilt displacement.

EiraSYS Posted at 8-24 04:31
Back to the Motion Controller proposal for the Manual Acrobatic Mode! (M Mode or A Mode)
The information is directed for DJI engineers for the development of a new firmware software for the DJI FPV drone and the Motion Controller.
This also valid for the new Avata drone that may be controlled with the Motion Controller.

Interesting. Just to make sure I understand. Yawing (blue) is translated to standard stick yaw and roll to achieve a yaw spin; i.e. a normal turn. Rolling the MC (red) is translated to standard stick yaw and opposite roll to achieve an axial roll. There is an ambiguity that arises when the MC is both rolled and yawed. What is the standard stick yaw response in this case? Not a significant problem.
Don't quite understand "MC axes are always aligned with camera axes". To achieve a barrel roll or loop the MC roll and pitch would have to be rate mode for those axes.

Is the specification that camera angle depends on drone speed an attempt to keep the camera angle fixed relative to the earth frame?

BudWalker Posted at 8-24 06:27
Interesting. Just to make sure I understand. Yawing (blue) is translated to standard stick yaw and roll to achieve a yaw spin; i.e. a normal turn. Rolling the MC (red) is translated to standard stick yaw and opposite roll to achieve an axial roll. There is an ambiguity that arises when the MC is both rolled and yawed. What is the standard stick yaw response in this case? Not a significant problem.
Don't quite understand "MC axes are always aligned with camera axes". To achieve a barrel roll or loop the MC roll and pitch would have to be rate mode for those axes.

No ambiguity at all, you just need a orthogonal transformation matrix of a 2D space to achieve this conversion.
Yes, roll, pitch and yaw must be all rate mode axes. The angles of the Motion Controller define the angular speed of the camera axis of the drone, minus the desired tilt angle of the camera. The reference frame should not be the earth, but the direction and speed of the drone.

EiraSYS Posted at 8-24 07:00
No ambiguity at all, you just need a orthogonal transformation matrix of a 2D space to achieve this conversion.
Yes, roll, pitch and yaw must be all rate mode axes. The angles of the Motion Controller define the angular speed of the camera axis of the drone, minus the desired tilt angle of the camera. The reference frame should not be the earth, but the direction and speed of the drone.

The ambiguity arises when the MC is both rolled and yawed. E.g if it's rolled CCW and yawed CCw what happens to yaw? Is it CCW as it would be if the MC was only yawed CCW? Or, is it CW as it would be if the MC were only rolled CCW?
MC axes can't be "aligned with camera axes" if "angles of the Motion Controller define the angular speed of the camera".


What happens when the MC is pitched down? Does this affect just the camera pitch, just the drone pitch, or some combination?

BudWalker Posted at 8-24 08:00
The ambiguity arises when the MC is both rolled and yawed. E.g if it's rolled CCW and yawed CCw what happens to yaw? Is it CCW as it would be if the MC was only yawed CCW? Or, is it CW as it would be if the MC were only rolled CCW?
MC axes can't be "aligned with camera axes" if "angles of the Motion Controller define the angular speed of the camera".

You will get something in between according to the transformation equation that I updated in previous image. Matrix can be optimized and made non linear.
MC axes are aligned with the camera axes so that angular speeds are applied to the drone over the camera axes (according to what is saw by the camera), not the drone reference. It was a bad english translation, should be "The angles of the Motion Controller define the angular speed of the drone over the calculated camera axes, minus the desired tilt angle only used for footage frame alignment."
Pitch down will increase the angle speed of how drone will pitch. The more you pitch down the faster it will rotate and allow to make a loop. The matrix can be further refined so that the typical left and right turns don't need pitch down compensation. The camera pitch is not affected by the Motion Controller pitch. It must follow the current drone speed. If hovering you have drone pitch=0, camera=horizon line-gimbal tilt. If fast forward you have the camera at 0º - gimbal tilt as shown in figure. The faster you go the higher the camera angle, just as it is working in current firmware. The camera will follow the drone reference, not the earth reference, so the camera will always try to look where the drone is going when moving forward. Going backwards and lateral motion will allow better footages that are not possible with current MC configuration.

The formulas for Yaw_Stick and Roll_Stick gets the signs correct. I.e rolling the MC adds yaw with the opposite sign whereas  yawing the MC adds roll with the same sign. But, there is more to it than just getting the signs correct.

I suspect that what you want is the Flight Controller to compute the FPV's direction of travel and1) if the MC is just rolled (no yaw input) then the FC will apply the yaw required to achieve an axial roll about the direction of travel
2) if the MC is just yawed (no roll input) then the FC will apply the requied roll to achieve a yaw spin with a constant pitch set when the yaw spin started.

BTW, my experience has been that if a transformation matrix is mostly just the identity matrix then using a matrix only makes it harder for most people to understand what you're trying to say.

BudWalker Posted at 8-24 16:38
The formulas for Yaw_Stick and Roll_Stick gets the signs correct. I.e rolling the MC adds yaw with the opposite sign whereas  yawing the MC adds roll with the same sign. But, there is more to it than just getting the signs correct.

I suspect that what you want is the Flight Controller to compute the FPV's direction of travel and1) if the MC is just rolled (no yaw input) then the FC will apply the yaw required to achieve an axial roll about the direction of travel

You got the all picture and summarized it very well! This matrix is only a simplified way to look at it, we have a control vector with four variables that can be used as we want, more or less complex, to control the drone actuators, in this case the speed of the four motors. The Flight Controller is already making most of the necessary calculations, with impressive results! Pointing where to go and pressing the accelerator is already a very complex task for the FC. This upgrade can only be made by people that really understand what is happening inside the FC code. Most people don’t need to understand… just need/want to move the MC!

EiraSYS Posted at 8-25 03:11
You got the all picture and summarized it very well! This matrix is only a simplified way to look at it, we have a control vector with four variables that can be used as we want, more or less complex, to control the drone actuators, in this case the speed of the four motors. The Flight Controller is already making most of the necessary calculations, with impressive results! Pointing where to go and pressing the accelerator is already a very complex task for the FC. This upgrade can only be made by people that really understand what is happening inside the FC code. Most people don’t need to understand… just need/want to move the MC!

I noticed that the DJI Avata has the "HorizonSteady" feature that "uses DJI's latest stabilization algorithm to lock onto a leveled horizon in every frame, regardless of the camera's orientation". I'm thinking this may be effectively what you're trying to achieve with the new A mode.

I love these "first m-mode flight videos". I have mine recorded somewhere as well (with shaky thumbs) which was exhilarating...it's like taking off the training wheels for the first time

"Once you go Acro...you never go back..."

BudWalker Posted at 8-31 05:40
I noticed that the DJI Avata has the "HorizonSteady" feature that "uses DJI's latest stabilization algorithm to lock onto a leveled horizon in every frame, regardless of the camera's orientation". I'm thinking this may be effectively what you're trying to achieve with the new A mode.

That’s good, so the processing power is there. If it can stabilize video in every frame than it can easily calculate the drone direction and speed and reaction of the motors to make a roll or loop as a response to dirac’s/impulses of the motion controller, on every frame!