This project was started in , electronics and control loops. Because I always need a cool project to learn new things, it was clear that something that can fly had to be built.
The project started as a "tricopter-only" project, but as I wanted to build smaller vehicles with more payload capacity, I decided to make some quadrotor, hexacopter and Y6 hexacopter firmwares too. My main interest is to build very small MAVs that fly as good as larger ones (or even better) and that can be controlled by wireless video link. I also experimented with autonomous flight in GPS-denied areas (video), and with GPS assisted autonomous hover (video).
-- William

Contact: Shrediquette @ g m x . d e --- All content published under CC Attribution-Noncommercial-Share Alike 3.0 Germany

Some brushed action for the winter

It has been a while since I posted something here. It is not because my project is dead, it is more because I spent my hobby time with flying and not with making new things. The derbe evo is almost perfect and I am very happy with it, so I don't see the need to design a new frame currently. But I am currently rewriting my flightcontroller code in order to achieve a faster control loop frequency. 1000 Hz should be possible on my 8bit ATXmega32A4. I will achieve this by using integers instead of floating point variables, and by removing every unnecessary line of code.


But recently, I bought an Eachine QX90 (always wanted to adapt my own flight controller for a small brushed copter, but there was not enough time...). And it flies great. I however changed it a bit, here is my setup:
  • QX90 original FC (stock firmware, see settings in the pictures below)
  • Standard motors
  • Gemfan 65mm propellers
  • Custom frame (weight: 5.5g, see pictures below)
  • Eachine TX01 cam with dipole antenna
  • Turnigy graphene 1S 600 mAh lipo (see picture)
  • Spektrum satellite receiver (the original FRSky receiver only had 50m range, now I have >500m)
Have a look at these videos for the flight performance:



Some pictures of the frame and the cleanflight settings:








Derbe evo available on Shapeways.com

Hey, I decided to make the derbe evo v2 available on Shapeways.com. So I don't have to send the files manually by e-mail. You can download or order using this URL:
https://www.shapeways.com/shops/shrediquette

The fleet...

Derbe evo - Version 2

I've been flying the derbe evo for a while now, and it is very agile and fast. I had to reduce the PD parameters by 30%. This is due to the fact that the mass is closer to the centre of the copter, and that reduces the moment of inertia. I was flying a competition with the derbe evo already (Hannover Cebit FPV Cup, here's a video). Unfortunately, the derbe evo crashed very hard during the qualifying (the gates were made out of massive steel, and the video reception was very poor). I think, I was frontally hitting a metal tube, and the 3D printed frame broke (two Cobra motors also have a bent shaft - I had to replace them as I do not manage to remove the shafts).
I implemented some modifications in the new version. The camera angle (wide angle lens) is now 35° (instead of 25°) to allow for higher pitch props. The 3D printed frame is thicker. And the power conncetor is now integrated in the frame. Additionally, I added a small button for the TBS unify pro to change channels comfortably.

There are many competitions this year that I am planning to attend, therefore I will most likely make another derbe evo frame as backup.

  • Flying Fischkopp vs. Kloppokopter vs. FPV Nutz (16. April 2016)
  • FPV Drone Master (30. April 2016)
  • FPV Race Friedewalde (07. May 2016)
  • Copter Clash Hannover (28. May 2016)
  • Bexbach German Master (27. August 2016)
RGB LEDs are mandatory on some events...

Increased camera angle. The ultra-large capacitors
might keep my ESCs and motors from dying all the time
(I lost 4 motors due to ESC malfunction already). I also added a
pull-down resistor (10k) to the signal wire of each ESC.

Pushbutton for channel selection & integrated power plug.

'Last christmas, I gave you my....'


First flight - derbe evo

The copter had its first flight yesterday. With the 5.5" Graupner Race-Props, the speed is absolutely amazing - I still need to learn how to control such a fast device properly.

Here is a quick + dirty video:
https://www.youtube.com/watch?v=V6I8i4Hk2J0 

And Shrediquette is still open source of course, so here are the 3D files:
william.thielicke.org/multirotors/derbe_evo_v97.zip

The arms are from the Claerials HUMs quadrotor: http://claerials.com/shop/hums-ersatzarm/






Shrediquette derbe evo progress

Some images of the assembly process...:




There is not much space inside the frame, but everything fits. The assembly also went quicker than I thought. I just need to replace all the motor shafts of the Graupner Ultra 2806 2300 kV motors (due to a hard crash), and then I am ready to go.

The take-off weight is just 447 grams.

derbe evo FPV racing quad



I think I now have all the parts that I need to start assembly. The frame weight is 77 grams (without electronics).

Here is the latest teaser:
https://youtu.be/no3FbcdH4LE

My next FPV racing machine...

Here are some renderings of my latest frame concept. It has a very small frontal area and should reach high speeds. The diagonal motor distance is 210 mm and I will use 5.5 inch propellers (Graupner Race-Prop or C-prop).
















Explanation video: FPV camera tilt compensation

This is a follow up to the previous post on tilt compensation.

I made a video where I explain what FPV camera tilt compensation is needed for, how it works and how it looks like in flight:

https://youtu.be/gXeaINFpvow

In FPV camera tilt compensation, control inputs (RC transmitter --> FC) are transformed to the local frame of reference of the camera.

The control outputs (FC --> motors) may additionally be transformed to the local frame of reference of the propellers (if e.g. the props are tilted or the FC is mounted at an angle - use parameters like 'board align' for control output transforms).

Some thoughts on FPV camera tilt compensation

Almost every 'serious' FPV racer is tilting the FPV camera up in order to have a better image during fast flight. Angles between 10 and 35 degrees seem to be common. E.g. I am using 25 degrees. But tilting the camera with respect to the multirotors horizontal plane has a side effect on the controls (maybe you never thought about it, because you are so used to flying with this side effect).

Just to make the effect easier to visualize: Imagine you drank a beer too much and now you suddenly think you are Charpu, Mr Steele or Mattystuntz. You mount your FPV camera with 90 degrees tilt (pointing up vertically) because your best friend told you that all the pros do it like this. In this case, your roll and yaw control will be interchanged: When you move your RC sticks to roll left, your copter will roll left, but looking through the FPV camera, it appears as if you are yawing to the left! And if you move your RC sticks to yaw left, your copter will of course yaw left, but the image from the FPV cam looks like you would roll to the right!

The strength of this effect is somewhat proportional to the camera tilt angle. But already at 25 degrees it can clearly be noticed. It doesn't really make sense to have control in the copters frame of reference when flying FPV. It does make more sense to shift the frame of reference to the FPV camera and to eliminate the effect of camera tilt: Moving the roll stick of the transmitter should roll the camera image only, it shouldn't add some undefined amount of yaw.

Luckily, it is easy to compensate these effects (you might even do this in your transmitter by adding some mixers). I however hardcoded this in my latest flightcontroller firmware. The code is like this:
Roll_output = cos(camera_tilt) * Roll_input - sin(camera_tilt) * Yaw_input
Yaw_output = sin(camera_tilt) * Roll_input + cos(camera_tilt) * Yaw_input
'camera_tilt' is the angle (in radians) that your FPV camera is tilted up with respect to the propeller disks. It is a constant. I am using 25 degrees of camera tilt (= 0.436 radians).
'Roll_input' is the roll rate that you steer with your remote control.
'Yaw_input' is the yaw rate that you steer with your remote control.
'Roll_output' is the setpoint for the multicopters roll rate.
'Yaw_output' is the setpoint for the multicopters yaw rate.
  
I am not familiar with other flight controllers, but I guess this feature might have been implemented already in some of them. Otherwise it really should be implemented...! 

Here is a video explaining the above:
http://shrediquette.blogspot.de/2016/01/explanation-video-fpv-camera-tilt.html

Aerodynamics in racing multirotors! Part 2.

Introduction

In part 1 of this article on multirotor aerodynamics, some ideas on how to reduce the aerodynamic drag of racing multirotors was presented. I was also designing a tilted-body racing quadrotor called "Shrediquette DERBE". There were not yet any flow measurements of multirotors flying at high speeds. Therefore, I had to make quite a number of assumptions on the aerodynamics of a racing copter. This time, I am presenting some flow measurements, along with some potential optimizations for the next version of the "Shrediquette DERBE"

Methods

Recently, I had the opportunity to do some flow visualizations in a large wind tunnel at the Bremen university of Applied Sciences / Dept. of biomimetics (which is the place where I worked one year ago). I brought the Shrediquette DERBE and mounted it inside the wind tunnel. Prior to the measurements, I did some tests to determine a realistic flight speed and the appropriate pitch angle.
The measurements were performed at a pitch angle of 45 degrees with the motors running at full throttle (control loops are all turned off completely). Wind speed was set to 30 m/s. The Shrediquette DERBE was equipped with Graupner C-Prop 5.5x3", 4S 75C, Ultra 2806 2300 kV.

DERBE mounted inside the wind tunnel
The flow was visualized with a method called Particle Image Velocimetry (PIV). This allows to measure space- and time-resolved flow velocities in fluids: Some very small tracer particles are added to the air (e.g. oil droplets). These particles are illuminated by a laser in a very thin sheet. A camera is mounted perpendicular to this sheet, recording the motion of all the tracer particles. The discplacement of the particles is used to calculate the velocity of the fluid. During my PhD research, I was developing a tool (called 'PIVlab') to perform these kind of measurements within Matlab (see this article for more information).

Particle image velocimetry: Setup consisting of high-power laser, high-speed camera and particles in the fluid.


PIV measurement with a laser sheet
Here is a short clip that shows the copter 'flying' in the wind tunnel: Youtube Video

I measured the flow at four different locations (labelled A/B/C/D):

Top view of a multirotor. The green lines show the locations where flow velocities were determined.
This image sequence shows the laser sheet in position B.

Results

The flow over the main frame is mostly horizontal - hence a tilted-body concept really makes sense. This concept aims to align the main frame parallel with the flow to reduce the frontal area and hence the aerodynamic drag (which is proportional to frontal area). The following image shows the flow velocities around the main frame (position A). The arrows indicate the direction, and the colors indicate the relative velocity magnitude:
Warm colors: Flow > 30 m/s
Cool colors: Flow < 30 m/s
Dark red: shadows / no measurements possible

PIV measurement around the main frame (position A): The flow is mostly horizontal in this plane.
 The flow under the propellers is actually highly assymmetric. It does not make sense to assume a constant flow velocity below the propellers. In measurement position B, the blades of the front propeller move forward, and the blades of the rear propeller move backward through the measurement plane (see the animated image sequence above). Therefore, the front propeller blades experience much higher flow velocities than the rear propeller blades. Hence, only the front propeller blades generate thrust in this plane. Furthermore, the assumption that the flow is perpendicular to the propeller disk below the propellers is only true for the regions of the propeller disk where the blades move forward.
The velocities in measurement position B are shown in the following image. Note the high flow acceleration (warm colors) behind the front propeller. Also note that the rear propeller does not accelerate the flow at this measurement position - it is almost passive.
Measurement position B: Only the blades that move forward through the plane (= the front propeller) generate thrust. 

Together with the results of the other measurement positions (not shown here), we can safely assume that only parts of the propeller (disk) generate thrust in high speed forward flight. This is very similar to the aerodynamics of large helicopters, where also the advancing and retreating blades experience very different flow velocities and generate different amounts of lift and drag). The importance of this effect (sometimes also called P-factor), is linked to the advance ratio of the propeller. Earlier, I actually thought that this effect might be negligible on multirotors, but this is clearly not true. A multirotor in fast forward flight only creates noteworthy lift in the green areas shown in the following image. In the centre of the red areas, the propellers might even create additional drag:

Multirotor in fast forward flight: The green areas create most lift, parts of the red areas might even create drag.
Note that this is only true for the propeller rotational directions shown in this image. If all propellers
rotate in the other directions, then red and green areas need to be inversed.

Conclusions

What does this mean for our racing multirotors? Tilted bodies make sense, as the flow is really horizontal at the main body. Vertical arms (as in the Shrediquette DERBE, see image below) are however problematic for two reasons:
  1. The flow will only be parallel to the arms at the front propellers. Because only these propellers do really accelerate the flow in fast flight. Below the rear propellers (the flow is not accelerated here), the flow will actually hit the arms from the side - which causes a large drag penalty.
  2. Multirotors use their motors to rotate around the pitch / roll / yaw axes. The force (or better: the moment) to rotate around the roll and pitch axis is induced by generating differential thrust between opposing propellers. Very large forces can be generated around these axes. As an extreme example: If both front motors run at full throttle, and the rear motors are off, then the pitch moment (moment = radius * force) could be around M = 0.12 meters * 12 Newtons = 1.4 Nm. But the force around the yaw axis is much lower because it is induced only by the torque of the motors - not by the thrust. If we assume the propellers to have a lift-to-drag ratio of 10:1 and a diameter of 5 inch, then the moment around the yaw axis is about 20 times lower than the moment around roll or pitch axes. Due to their orientation, vertical arms can however generate large forces around the yaw axis which can not be compensated by the small torque of the motors. In order to have a better control around yaw, it makes more sense to not rotate the arms.


Do vertical arms make sense...? Not really...
In the last weeks, I designed a standard racing quadrotor (called 'Shrediquette 0815') that I use to test the properties of 3D printed frames. I learnt how to design a very rigid and lightweight plastic frame. I will use this knowledge to improve the design of the Shrediquette DERBE II.

Shrediquette 0815, a standard racing quadrotor that I designed for training and to learn more about
lightweight and rigid plastic construction.