Don’t get too attached to the great picture up above, as the quad shooting it was in a death plunge when the frame was snapped. There’s just something tempting about free fall. Nearly every tri/quad/hex/multicopter pilot has the impulse to chop the throttle while flying around. Most quadcopters are fixed pitch, which means that as power drops, so does control authority. When power is cut, they fall like stones. A quick throttle chop usually results in a few feet of lost altitude and a quickened pulse for the pilot. Cut power for much longer than that, and things can get really interesting. [RcTestFlight] decided to study free fall in depth, and modified a test bed quadcopter just for this purpose.
First, a bit of a primer on free-falling quadcopters and their power systems. Quadcopters always have two motors spinning clockwise, and two spinning counterclockwise. This configuration counters torque and allows for yaw control. Most large quads these days use sensorless brushless motors, which can be finicky about startup conditions. Brushless controllers are generally programmed to kick a motor into spinning in the proper direction. If a motor is spinning in reverse at several hundred RPM, things can get interesting. There will often be several seconds of stuttering before the motor starts up, if it starts at all. The controller MOSFETS can even be destroyed in cases like this.
When a quadcopter loses power, the motors slow down and thrust drops off. The quad begins to drop. As the falling quadcopter picks up speed, the propellers begin to spin (windmill) due to the air rushing up from below. If the quadcopter started its fall in a normal attitude, all four of the propellers will rotate reverse of its normal direction. The now spinning props will actually act as something of an air brake, slowing the fall of the quad. This is similar to a falling maple seed, or autorotation in a helicopter. The spinning blades will also act as gyroscopes, which will add some level of stabilization to the falling quadcopter. Don’t get us wrong – the quadcopter can still be unstable as it falls, generally bobbing and weaving through the air. None of this is a guarantee that the quad won’t tip over onto its back – which will reverse the entire process. Through all of this bobbing, weaving, and falling the flight controller has been along for the ride. Most flight controllers we’ve worked with have not been programmed with free fall in mind, so there is no guarantee that they will come back on-line when the throttle is rolled on. Thankfully many controllers are open source, so testing and changes are only a matter of risking your quadcopter.
[RcTestFlight] found that his FPV h-quad quad was relatively stable in free fall, so he began experimenting with falls from high altitude. He found that many tests don’t end well. One crash managed to bend his aluminum frame badly enough that he replaced two of the arms with more forgiving wood. His quad originally had 9 inch props. In an attempt to go for a slower sink rate, [RcTestFlight] switched to 14 inch props. The larger props also needed slower motors with more torque. After these changes, the quad definitely fell slower, however he found the fall was actually less stable than the 9 inch props. Self deploying foam drag flaps slowed things down even further, but the flaps themselves became an issue when a particularly violent fall ripped them all off. After taking his quad up to 4142 feet, nearly losing his quad, and a number of rough crashes, [RcTestFlight] had has his fill of free fall. We’d love to see more testing, especially with collective pitch quadcopters. We’ll keep our own quads safe in powered flight though.
If props are rotating in reverse direction in auto-rotate mode and if transition to stable normal forward operation may not be straightforward (both of which were implied to be the case in typical free fall situations) then flipping the copter on its back and using the prior “autorotate” prop direction as the new flying direction, may save the day. Or not.
…pray tell, how does one get it to flip over on it’s back when it has no power to the rotors?
I think the issue is not ‘no power’, but ‘difficult to start a motor that is spinning in the other direction’? Braking a motor on the other hand is trivial (short it), so brake two out of 4 and it will flip :)
You could spin one side of the quad’s rotors faster, but in the same direction they’re already spinning due to the wind.
“You could spin one side of the quad’s rotors faster, but in the same direction they’re already spinning due to the wind.” – Andrew
You could also use electronic braking on the other 2 motors
another idea is to implement whole setup in computer simulated environment and tel some genetic algorithms teach itself to flip over using minimal power.
….or just start the motor in reverse at its highest speed and slowly ramp down.
Very interesting test, I’ve cut the throttle on quads/tris and thrown them but never went this in depth. I’ve also found out the hard way that the quad wont always respond as quickly as it needs to.
I’m sure that wouldn’t be tough to significantly increase the success rate from my tests. I imagine some PID testing/tuning would help. some of my setups would stabilize but be flying quickly in a seemingly random direction, perhaps indicating a couple of setup issues)
So write the software to NEVER let you cut the motors, always at some percentage of thrust to eliminate the problem unless you activate a “fall to your death, emergency” mode
It might be difficult to do this properly with stock ESCs, but if you can change the firmware, it should be possible to restart the motors reliably. Whenever the motors are turning, the ESC can determine the direction, speed, and phase of the rotor, so it should be able to apply torque in the opposite direction, thus reducing the speed while feeding some power back into the batteries (this energy has to go somewhere). When the motors are slowed down, almost to a standstill, the ESC will lose the ability to detect the zero-crossing events, because the back-EMF is very small. At this point, the normal startup procedure (blind commutation) should work, provided the ESC can provide more torque in this mode than the torque generated by the wind going through the prop.
That assumes a sensored motor whereas the article explicitly states sensorless motor where what you are trying to do becomes nigh impossible.
You’re right there shouldn’t be any back-emf (since no power) – but maybe could you sense something worthwhile from the coils as the motor becomes a generator of sorts (maybe this is what Sparky is thinking?)…
“That assumes a sensored motor” not quite right. If the propellers are turning there’s back-EMF, does not matter if it is the wind turning them or the battery power. So it is possible to do this with sensorless motors.
The practical issue is that the it is proportional the the speed thus might be difficult to measure accurately at low speeds.
If you’re going to develop a “sensor-less” BLDC controller for multicopters, it might help to know how airplane controllers start up and operate. During normal operation, they sense back-EMF from the windings to detect speed, motor position, and when to commutate to the next winding.
During start-up, there is no back-EMF (the motor is not moving, but is free to move), so commutation is applied blindly, ramping up motor speed until back-EMF can be detected. This would work well in a plane, but not a multicopter.
For a multicopter in free-fall, the motors may be wind-milling in the wind in the reverse direction (all motor windings open-circuit). A normal blind start may not work – it assumes a rapid start-up from near zero speed and no reverse torque. The first step would be to apply braking to slow the motors. Simply short all the windings together, and wait for slowest motor speed (still in reverse). Then, drive one winding to lock the motor, followed by a slower blind start to return to normal operation. Depending on the copter Center of Gravity and drag, re-orienting to top-up may occur during braking/locked motors. All this could happen within a second or three…
Or gain control of the copter by assuming it’s flying inverted (top-down). Once the motors are started in inverted flight mode, flip the copter to its inverted position, and stop the fall. When stopped, flip to top-up, and resume normal flight. If your copter has an altimiter, this could be a preprogramed manuver to save the craft, and possibly land at a survivable speed.
Full size (i.e. not radio controlled) single seater aeroplanes can have a parachute in the tail – one that’s for the entire plane. Would something similar for RC models be a bad idea? Triggered by e.g. sensing free fall
Better yet, when flying up, use a 2nd parachute to ditch (and safely recover ;-) ) one of the batteries once it’s drained, just like a rocket :D
Someone help me with the logic. He needed to add an additional battery to break his previous altitude record because… one battery alone went dead due of the time/power required to get up there? Or, does it have something to do with the props needing to spin faster at higher altitudes due to reduced air density?
I expect its battery life as you would want to leave battery power to safely abort or recover the drone after ascending to altitude, Even acending at 8fps (based on flight times and height) to reach that altitue it would take most of your battery to get there.
If a battery weighs ‘A’ grams and has ‘B’ watt/hours, then the power-to-weight ratio of a quadcopter with 1 battery would be B/(A+C), assuming the chassis weighs ‘C’ grams. A quadcopter with 2 batteries would have a power-to-weight ratio of 2B/(2A+C). Since C is constant the ratio for a quadcopter with 2 batteries is higher.
The limiting factor in adding batteries is whether your motors can lift the weight and still have enough extra power to manuver.
i was going to ask if something along the line of Autorotation is relevant in this configuration, but you already touched on the subject. it would be interesting to see if there could be some sort of Dead mans switch cut off where if the mainboard loses connection or power for a certain period of time or maybe setup with input from the accelerometer to a set of standards it deems as “out of control” or “loss of control” that it could through the use of an auxilary power source, blow out a little parachute like they use in model rockets, in larger form of course…
The problem with parachutes is that they take time to deploy. That’s not such a problem if you’re going for an altitude record, but most quadcopters fly low to the ground where their human pilots can see them.
true, but with alot of aircraft crashes its not going to do much if your flying low to the ground, but those model rockets shoot out the parachutes pretty quick, and heck, it could possibly work as low as 15 feet right? lol according to my g.i.joe parachute guy http://cdn.instructables.com/FPR/UJKN/GIYWWB3J/FPRUJKNGIYWWB3J.LARGE.jpg
You have to worry about tangling with multicopters too. The ‘chute has to escape the quad without getting tangled up in the props.
yeah that could be challenging, but thats why i was thinking along the line of those Estes Model rockets, where at the end of the 1st or 2nd stage depending on your rocket, you get a last blast that blasts the chute out from the nose cone and away from the rocket, something similar could possible be achieved through spring or something similar.
Use a straw (or similar light weight tube flexible tube?) from the center point or an end point, cut it so that it can fold back onto it self, back to said center point, this helps all the individual suspension lines clear the blades as a single group for the parachute..
As to a method for deployment, hard to say as it depends on your copter, could use a sub-micro servo to pull a pin in a 35mm still film can lid, that is has a spring under the pack? google “Mini AltiDuo” for a solution too
Hi I am the author of the altiDuo which is used for rocketry.
http://rocket.payload.free.fr/
You will be better of with my altiUno altimeter that has only one output.
If you wnt to drive a servo it is possible with little modifications of the code
Then there’s this little matter of the rocket burning the chute it’s trying to deploy….
I didn’t have time to watch the video so I don’t know how it was mentioned, but how did he see his quad at 4142ft? FPV? I’d be scared of messing with real airplanes and the FAA coming and knocking on my door.
I think the FAA would have a hard time picking that up on radar :P
It looked like he was using one of these:
http://www.eagletreesystems.com/index.php?route=product/product&path=64&product_id=60
I have been curious as to why the struts of most quad copters are not airfoil shaped. I believe it would add lift in forward motion flight, and in the event of a free fall, a simple weight/spring combination could be used to pitch all the struts to a slight angle turning the quad copter into a very large expensive pinwheel.
Because square wood/aluminium is cheap. Carbon Fiber is even cheap compared to airfoils. Also, airfoils need to be wider than their square or round counterparts, as they need a higher ratio of width to thickness to be effective, which means the props are blowing more air onto the booms, which reduces efficiency. This is why many people use tubular booms, they reduce the amount of efficiency loss from the top and the front/back. Vertical airfoils to reduce efficiency loss there might be effective, but their price compared to tubes is just too high to be practical.
It would need more than a slight angle. At least two of the airfoils (if it’s a quad) would need to rotate in complete opposition in order for that to would. With that said, more force is going to be required to rotate it against the forces affecting the airfoil (the differential pressures due to the air passing over the airfoils) than the motors/springs/weights could reasonably accommodate. Add to that the extra weight and moving parts (additional points of failure), the solution makes more problems than it solves.
Then again, if someone happens to pull it off, I’d definitely have no problems eating my own words. :-D
You should slap a HayTag on there when you need to find it in the snow ;)