Last fall I had an epic crash with my semi-scratch built 750mm quad after a propeller came apart in flight. This was only my 3rd flight with the quad. I watched helplessly as the quad tumbled out of the sky. I can still hear the thud as it impacted the ground. After the crash, I collected the pieces and my pride. I went home to contemplate the gravity of the situation and I decided to look into a parachute system. This quad was my first endeavor into the Arducopter platform. I was using an APM 2.6 on the quad that crashed. I did some research on the Ardupilot web site and found that the platform was beta testing a parachute command with the Pixhawk flight controller. This peaked my interest. My intention was to eventually upgrade to a Pixhawk controller after mastering the cheaper APM 2.6. Based on the information provide on the ardupilot page, I scoured the internet and found that commercially available parachute system were expensive. The cheapest unit was over $150. If I were flying an expensive AP platform I could justify spending $150 or more to protect my investment. It is hard to justify the expense when the parachute system was half the cost of my quad. I researched building my own system. None of the systems I could find were very elegant in design. They ranged from a Tupperware box with a rubber band to having a parachute balled up and rubber banded to the bottom of the quad. After my research, I decided, in Flite Test fashion, to build my own parachute system that would rival and cost less than the commercially available systems. Here is what I came up with.
36in Parachute System
42in Parachute System
One of the biggest challenges in designing the system was the parachute. I immediately found that commercially available parachutes are also expensive. In fact, the companies linked on the Arducopter site were charging over $100 for parachutes. Are you kidding me? That was not going to cut it. I did some more research and I found a couple of YouTube videos describing how to build your own. Hell, the Flite Test guys used a garbage bag for a parachute in one of their videos. I definitely need to build something reliable that wouldn’t come apart if I was going to spend all this time and effort in building my own system. After watching several videos I decided to give the internet one last search. I came across a company called Sperachutes that produces parachutes for model rocketry. http://spherachutes.com/
To my surprise their parachutes were relatively inexpensive when compared to the other sites I had found. They provided a chart on what size parachute you’d need for a given mass and elevation (MSL) you would be flying at. I decided to try the 91 cm (36in) parachute (Size is equal to ½ circumference). There chart showed it worked for up to 1.27kg (2.8lbs) at my elevation (1524m/5000ft). The best part was the price. With shipping (within the US) it cost about $36. At this cost, I decided to purchase the parachute rather than make it. I knew that I could probably make a “usable” parachute but decided to save time and purchase a professionally made one. I speculated that it would cost around $10 or less plus time and effort to make my own parachute.
An important step with the parachute is how to fold it. If it is not folded correctly or wrapped tightly enough it could hang up in the tube and not deploy correctly. Folding takes practice. Do it several times until you get it right and tight. This is how I folded my parachute:
Fold both sides the same. Each side should have the cords together.
Fold in half so all cords are now together.
Fold in half again making sure all cords are together. You may need to make another small fold before you fold lengthwise depending on how big the tube allows for.
You can either fold accordian style or roll it up from the top of the parachute down towards the cords. Make sure the cord is wrapped tightly around the parachute.
I decided to go with a spring and tube launch system. The end result was a system with a mass of 142g (5.0oz) and performance similar to commercially available parachute systems.
After I received the parachute, I measured it when folded. I determined that I needed a tube that was approximately 5cm (2in) in diameter. I found that the treats I feed my cat came in a plastic tube that was about the sized I needed.
To determine the length of the tube, I measured the folded parachute and added that to the compressed spring and spring top measurements. The spring needed to be longer than the tube in order to push the parachute out of the tube efficiently. The tube for the 91cm (36in) parachute is approximately 8.9cm (3.5in) long and the spring is approximately 15.2cm (6in) long.
3D Printed Parts
I went with 3D printed parts for the top of the launch tube system. I have a XYZ Davinci 1.0 printer. Linked below are the files on the Thingiverse web site. I used Tinkercad to design the parts. This site is free and really easy to use. With the 3D design file, it would be easy to disassemble my parts and enlarge to diameter to fit a different size tube if need be.
Depending on the 3D printer used, some of the parts may need some sanding or trimming in order to get smooth operation and fit. I used pushrod wire to make the hing and safety wire. I heated the wire up and melted it through the plastic in order to install the wires.
3D Parts: http://www.thingiverse.com/thing:796612
Since my spring needs were very specific, I made my own spring. The following link will describe how to make a homemade spring.
I was not able to find spring steal and ended up just using some stiff 14ga wire that I found at Home Depot. A spool cost about $10 for 30m (100ft). As you can see in the videos, this wire works well enough to launch the parachute an acceptable distance. The other thing I learned was to use a cylinder that is smaller than the diameter than the launch tube you will be using. There is some rebound after the spring releases when making it. I ended up using 3.1cm (1 1/4in) PVC pipe for the spring builder.
I topped the spring with a 3D printed plunger to push the parachute from the tube. I used some orange p-cord to tie the plunger to the bottom of the tube to prevent it from shooting off when the parachute is deployed. There needs to be enough slack to allow the spring to go past the top of the tube.
I used orange p-cord for the attachment points to the quad in my experiment. It is important that the vehicle hangs level over the CG. I used a small screw lock oval to attach the parachute to the cord. This screw lock oval is way overkill for this but I like the ease of being able to detach the parachute from the copter when it comes time to do any work on the copter. I intend to replace this cord with something smaller like light cable or Kevlar cord in the future to clean the look up.
The completed launch tube is attached upright on the tri-copter using on screw and locking nut. The tube itself doesn’t take any load other than just securing the system to the frame.
I used a small 3.7g Turnigy servo (TGY-1370A) for the build. This servo can be found for about $3. The biggest concern I have in using the cheap servo is the potential for the servo being damaged by the force of the spring pushing on the lid. After a couple of months of testing, the servo has held up. I did have to modify a large servo arm to work with this servo. The ones that come with the servo are not rigid enough. I cut the supplied servo arm down and glued this piece into a larger servo arm.
I used a safety wire to prevent the parachute from launching by accident while not in the air and to remove the pressure off the servo while in storage. The wire goes through the two nipples on either side of the lid.
For my tests, I used a Taranis radio. I programmed two switches to control the parachute. Both switches need to be activated for the parachute to launch. Since radios are not universal, you will need to do some research on how to set this up on your particular radio.
I decided to use my FT Bat Bone Tri-Copter for the test vehicle. I only chose this vehicle because it was at its end of life. I had unintentionally done some drop tests with it, sans a parachute, several months back. The frame was cracked in several places and the bottom portion of the tilt mechanism was missing. The servo had to be hot glued in place in order to function properly. The tilt mechanism was the original design from Flite Test and does not hold up well to abuse. This proved to be the weakest link in the drop tests. I destroyed two servos during the drop tests (Well, actually three. But the third one broke when the tri-copter fell from about 6m (20ft) when a plug came undone from one of the motors due to violent vibration from the failing tilt mechanism. I didn’t have enough time to deploy the parachute from that altitude). I currently have a Dragonfly with a Tough Tilt on it. I believe that the Tough Tilt would do a lot better than the original FT tilt mechanism. In addition to the sad shape of the frame, I also had issues with the motors. The Turnigy Park 300 motors were burned out. They were unable to lift the tri-copter high enough to do the tests. When they would heat up, they could no longer product enough lift to keep the copter in the sky. I ended up replacing them with DT750s and 11x4.7 props that I had on hand. Unfortunately, before the swap I crashed twice cracking the frame again. The test vehicle became the biggest headache for the whole drop tests. By the end, hot glue and duct tape were the only thing holding it together. Between the spongy frame and wild vibration from the tilt mechanism, the tri-copter was a challenge to fly and film at the same time.
I used plenty of padding to protect my flight controller (RTF Flip 1.5 MultiWii) and receiver. The end result was a test vehicle that weighed 1.16kg (2.55lbs).
I conducted several test on the parachute system. My first test was to determine how well the parachute would launch while in flight. I statically tested the launch tube while on the ground and it seemed to work well, but I needed to see it work in flight. I set up a down comforter in my living room and flew the tri-copter above it about 2m (6.5ft) and activated the parachute. As you can see in the video I did a relatively poor job of dropping the tri-copter on the comforter. As a side note, it did not sustain any damage from those drops. Based on the slow motion video the parachute system performed well on these test drops.
I also did static drops at both 2m (6.5ft) and 4m (13.1ft) to measure the drop time from a given height. I used this information to estimate the drop distance in a real world drop before the parachute fully deployed. Based on the video, it took approximately 0.5 sec to drop 2m and 0.8 to drop 4m. Using the 4m drops, this would be equivalent to approximately 5m (16.4m) in the first second of drop time. Since I believed the parachute should open within 2 seconds, I figured this calculation would provide a large enough margin for error for two reasons. First, these drops did not take into account the drag created by the spinning propellers. Second, they did not take into account the drag of the parachute as it is deploying. My end results are an estimation based on these numbers and are used only as a buffer to determine the lowest altitude for the parachute to fully deploy.
I conducted four drop tests. All of the tests were done at an altitude of greater than approximately 40m (131 ft) based on a variometer used to calculate altitude. Due to issues with the on board cameras and the difficulty of filming and flying, the first video above was pieced together from several tests to demonstrate the results.
Based on an onboard camera, on one of the drop tests, the system was able to fully deploy the canopy of the parachute in less than 1.5 seconds. This was the time from the servo activation to the parachute coming into frame and being completely open. Other videos revealed that the parachute canopy began to open within one sec after activation. Using 5m of drop, as calculated above, for the first second of drop time and considering the drag created by the parachute as it opens during the next 0.5 seconds, it is reasonable to assume that the drop distance will be no more than 7.5m (24ft). This distance does not take into consideration the reaction time (time to recognize a problem and then activating the parachute). If I added a second for reaction time, that would result in an additional drop of over 5m (as the vehicle continutes to drop the velocity will continue to increase past the calculated 5m/sec). The vehicle now has close to two seconds to accelerate before the parachute starts creating drag. If I stick with the 5m/sec, 12m is “probably” a reasonable buffer to deploy the parachute with the understanding that an increase in altitude would provide more time for the parachute to decelerate the test vehicle to a stable decent rate before it hits the ground. My calculations are of course a “best guess” as they are loosely scientific in nature. I would welcome anyone to take my data and calculate a theoretical drop distance and compare that to what I calculated.
Three of the fours tests resulted in the tri-copter coming down in the correct orientation landing top up. A fourth test resulted in tangled attachment cords which cause the tri-copter to land tail first (servo destroyed). Ultimately, the vehicle sustained minor damage in each drop. The parachute used (91cm/ 36in) according to the manufacturer was at is limit for mass at the elevation tested. A larger parachute would help to lessen the impact force upon landing. The tradeoff is mass and size of the launch system. With the understanding this is a last ditch effort to prevent major damage to the multirotor, I would stick with this system for the test mass used. The tri-copter was easily repaired after all of the drop tests (even with the poor condition prior to the testing).
The cost of the system will be varied based on the availability of parts and access to a 3D printer. My costs were as follows:
3D Pinter material: $3
Spring: $0.50 (wire spool cost $10 for 30.5m (100ft) and the spring uses about 5 feet of wire)
Keep in mind, if you don’t have items on hand as I did, like the PVC pipe, quick link and p-cord, the initial cost could go up by~ $20.
Even with a total cost of ~$62, this system is far less than any commercially available system and it performed well. Based on this design other size parachutes can be used with slight modifications to the system. I currently have a system set up with a 106cm (42in) parachute.
With all the time and effort I put into this, how necessary is a parachute on a hobby level multirotor? If using quality parts is it really necessary? If I had used higher quality propellers, would I have even gone through all this? I can come up with two use cases for a DIY parachute. First, if you are flying hobby level AP platform with an expensive Gopro style camera/gimble and you regularly fly over 12m it may be worth it to protect your investment(see note below). The second reason occurred to me last week when I was flying line of sight and I lost orientation. The wind picked up my FT Dragonfly and started to carry it away. I did not have a parachute attached at the time. The wind was pushing it towards houses. I was seconds away from cutting the throttle so it wouldn’t come down over houses and people. Thankfully I regained orientation and didn’t have to ditch the Dragonfly. Had I had a parachute attached, I could have activated it instead of ditching the aircraft. Similarly, it could be used as a last ditch effort with loss of video signal during FPV flights or during a flyaway, assuming the receiver was still connected to the transmitter.
So, do you really need a parachute on your multirotor? Maybe not, but parachutes are cool!
*Note: My testing did not take into consideration the impact force on landing. With a top up orientation, a camera and gimble would be the first part of the multirotor to hit the ground after the landing gear/skids etc. If you are trying to protect an investment like a camera and gimble, a larger parachute would be recommended based on the force your landing gear/skids can absorb to prevent any significant impact on your camera setup.