Tricopter build log


#1

Hi guys, I hope this is not a totally inane thing to post about but I thought some people might find my build-log interesting.

David Windestål’s (rcexplorer) build video on flitetest is largely responsible for me having a go at this, I’ve been researching it on and off over the past few months. I decided on a tricopter rather than a quad based on lower part count and thus lower initial investment. As a noob I don’t want to dump a lot of money into this before I learn to crawl.

Parts list:
Tx: Turnigy 9X
Rx: Bundled 8 channel.
Motors: DT750’s
ESCs: HK 20A UBEC’s
Flight controller: KK2.1 HC
Servo: HiTEC HS-82MG
props: 9047SF Slow Fly
Battery: Turnigy 4S 3000mAh (I also picked up a Accucell 6 to charge the batteries)

I decided to use the ‘optimised’ variant of the rcexplorer coffin-body frame (http://www.undcon.com/colorex/) making sure to adjust the screw hole positions to allow for the booms I have. I cut the plates from 3mm plastic sheeting - a £1.80 clipboard from Tescos - allowing me to get 4 of them per clipboard; I’m sure any other brand clipboard would work just as well. :wink:

Frame is built from 15 * 12mm pine stripwood from B&Q. I couldn’t find anywhere locally that stocked 12mm square stripwood and I don’t have a table saw that would have allowed me to square the booms up.

Note that with the ‘optimised’ design the front booms rotate around the outside screws rather than the inside screws.

I’m working on the yaw mechanism, I had hoped to replicate the rcexplorer design but HobbyKing was out of the landing gear parts in the UK store so I went looking for alternatives and found the Simplecopter design. I could have purchased the parts from the international HobbyKing store but postage costs would have been 10x the cost of the parts…
I’ve built the rotating platform and am testing that today. If all else fails, some friends have a 3D printer, I may experiment with some of the designs I’ve seen on Shapeways if I get a chance to get myself up to speed with the nuances of using the device.

I built a very expensive test rig (see below) after I had soldered bullet connectors to the motors and ESCs to ensure everything was working. Thanks to everyone who contributed to the KK2 thread on the Multirotor UK forum, it was very handy and informative.

I decided to flash the KK2 board after reading a number of reports that the base firmware is barely fit for purpose. I used the version 1.18S1 Pro firmware created by Steveis. I initially attempted to flash using an arduino but the kkMulticopter tool couldn’t detect it under linux. I tried again on windows but the avrdude flash tool didn’t play well with the arduino so I bit the bullet and picked up a USBasp programmer for a couple of quid on ebay. Flashing was thereafter painless with one exception - the aperture for the 6-pin header on the kk2 HC is not large enough for what appears to be a standard-sized plug on my USBasp. I used some servo extensions to ‘bridge the gap’.

I appreciate that the ESCs can be programmed using the beep codes but after a tedious time programming the first one a programming card was promptly purchased. You plug it in, the ESC beeps once and that’s it. For less than £5 it’s a no-brainer. :smiley:

After messing with the simplecopter design for a bit I was unhappy with how much slop was in the mechanism. I saw a very simple yaw design on the Multirotor UK forum. It was such a good idea I shamelessly pinched it (thanks to user DarrellW for the inspiration). I didn’t have the L hooks he used so I used a bit of clothes hangar wire. It’s pretty stiff wire and should handle a crash fairly well. If it doesn’t there’s plenty more. Here’s what the assembly looks like:

Here’s a quick video of the prototype (warning, the motor sound is quite shrill):

You may notice a little bit of play in the servo - the 15mm wide boom means that the zipties aren’t pulling directly down. I think a few notches on either side of the boom will resolve this. Regardless, I think this is a promising design as the servo directly drives the yaw reducing slop and it’s super cheap to build whilst being pretty structurally sound.

N.B. Whilst I haven’t experimented, it makes sense that the optimal axis of rotation is through the combined centre of mass of the motor and prop. A high angular acceleration in the yaw mechanism may present a risk to the servo in the long-run. I’ll put his theoretically better design on the drawing board for now, though. :slight_smile:


#2

Brief update:

Everything ‘zip-tied’ together and a few short test flights (and the odd crash or three) suggest that balance around the COG is not too far off. I’ll have to take it to a larger area to do some more testing.

My yaw mechanism works reasonably well but it’s not terribly strong. I had to reinforce it a bit after a few heavy bumps. I might investigate some of the 3D printable designs available, e.g. http://www.thingiverse.com/thing:227642

Finally, I have balanced the props to reduce vibration but I’ve yet to balance out the motors.


#3

Looks good, I considered self building a quad but in the end bought a DJI Phantom, a custom build might be the next step!

If you’re taking it for a flight and want any in flight shots let me know and i can bring my phantom.


#4

Looking good :slight_smile: Great build!
Keep it up!


#5

David visited!

Awesome. :smile:


#6

Based on the prototype yaw mechanism I have refined the design a little. The new one is more robust.

I was charging the batteries from an old 12V wall-wart, decided that an upgrade was in order.

I need to properly package the PSU, some connectors and a nice, sharp metal bit should make short work of that. :slight_smile:


#7

Wow. I just had a great day flying.

A few comments and observations:

There is a decent amount of oscillation in the yaw before the motor gets up to take-off speed, I’ll have to investigate the cause as I don’t want to damage the servo. The props are reasonably well balanced but I need to find an Android application that will allow me to record accelerometer data from the motors and balancing them out.

Despite feeling heavy, this thing catches the wind! I was in a large park, far away from trees. I brought the
tricopter to about 20m above the ground and the wind took it horizontally into trees about 30m away in seconds! Luckily I managed to get the tricopter out of the branches with a bit of bamboo (seriously thinking a pole should be standard in my toolbox! :smiley: ). David’s design is pretty damn robust, other than zipties failing as intended the tricopter is still intact after falling out of the tree. I think a more luminous frame colour would be a good idea as it took a short while to spot it in the foliage.

I dialled in expo (30%) and dual rates of 50 on the elevators and ailerons. With the self-levelling on it feels like I really need to push the tricopter hard to get it to change direction. I guess this is by design but control authority (if that’s the right term) needs to be increased to get the tricopter to move responsively.

I am only slightly short-sighted but it was very difficult to ascertain orientation. I think some lights would greatly assist but I suspect that against the bright background of the sky even those are going to be
hard to see. I’ve seen people mounting a softball to the ‘front’ of quads and tricopters but I’m hoping to mount an inexpensive camera on the frame eventually so I’ll have to think of a temporary solution.

Lastly, I had a hell of a lot of fun. :slight_smile:


#8

Orientation will become easier if you paint the beems brigh colours and/or put a tennis ball (it’s bright) at the front.
if you balancing propellers - you need to balanse with the bell of these motors.
In order not to damage the servo I would advise to make a coupler out of high pressure rubber air tube
Is the 3D printer back?
I have some ball berrings, could donate em for improvement of the mount at the back.


#9

Cheers Costa, I’ve seen equivalent good advice on many of the multicopter forums, the flitetest guys cover these topics as well.

I have some ball-bearings here from many projects that might make their way into the tricopter eventually but at present I’m very happy with the yaw mechanism (bearing in mind that I’m not pushing the tricopter very hard); it has no perceivable slop as a result of being directly driven from the servo. I think it’s a reasonable trade-off as the servos are cheap. Other builders have reported that the cheaper metal-gear 9g servos stand up pretty well so I think mine should be more than capable for the easy flying it’s currently getting. Many thanks for the offer of some ball-bearings, though.

How is your build coming along? Are you waiting for the 3D printer to create the frame or motor mounts?


#10

Yeap, building a quad planning to print motor mounts with integrated wiring (to simply plug in motors when required.
Got 4pcs 5010 motors - 50mm diameter, 10 high and 15x5.5 propellers.
Should be fine for 2kg quad. Waiting for controller and battery to come.


#11

Sounds good! Can’t wait to see what you build. :slight_smile:


#12

Bad weather means more hardware hacking!

I decided to do a couple of modifications to my transmitter. The transmitter is a Turnigy 9X nine-channel unit with decent build quality and it’s terrifyingly cheap. It’s a clone so there are many rebadged variants. Here’s a picture of a stock unit:

So, why modify it? Well, the standard firmware is pretty feature-filled but the interface is an absolute dog to navigate. Thankfully, the internal microcontroller is an ATmega64. A couple of FOSS projects (TH9X, er9X, Open9X) building upon common source exist so I decided I’d try out the widely recommended er9X.

First problem is the transmitter module; in theory one can unplug the unit and replace it with, for example, an FrSky module that incorporates telemetry. I’m aiming to do this at some point in the future so I decided to modify the stock module. The problem is that whoever designed the transmitter ran a coax from the module, up through the body of the transmitter to the aerial on top. Here’s an image of the stock module:

I opened the module, desoldered the coax and then drilled a hole in the face plate to run the coax through. I then fed the coax through the plate, resoldered it and glued the aerial to the outside of the module. This is the result:

With that done it was time to tackle programming the transmitter. Loads of good how-tos available on the 'tubes so I’ll skip that and just describe my method of adding a port. Servo extensions are dirt cheap and have the same pin-spacing as a standard programming header. They have three conductors. The SPI interface needs six so I chopped a 10cm one in half, superglued the headers together and soldered the conductors onto generously sized pads on the PCB. A small cut on the base of the transmitter chassis allows external access to plug the programmer in. The servo-headers were hot-glued in place:

After putting the whole thing back together this is the result (looking at the bottom of the transmitter:

I tried to flash the firmware using AVRDUDE on debian but I’ve yet to have any success with it. Alas, I ended up doing it via a USBasp programmer via WinXP. I now have a very easy to use interface on my transmitter:


#13

Just for fun I modified the transmitter’s TX module further. There’s a trainer port on the back of the transmitter in the form of a standard 3.5mm stereo headphone socket. Normally this is used for helping new pilots; a veteran can connect their controller to yours and in the event of a loss of control they can take over at the flick of a switch and save the plane/multirotor.

Unfortunately, due to the way the port is designed, one normally needs to remove the transmitter module[1] as it grounds the signal otherwise. Thankfully a 1K resistor is all that’s needed to avoid this issue.

With the fix in place one can then connect the transmitter to the mic or line-in on a PC soundcard and use the transmitter as a controller in crrcsim, a FOSS RC simulator. The physics engine isn’t perfect but it at least allows one to practice with a range of planes, multirotors and a helicopter without fear.

  1. which, if you remember from the above post is attached to the transmitter via a somewhat fragile coax cable

#14

Decided to try some LED strips as a method of determining orientation:

I haven’t determined how much it impacts the flight-time yet but it’s at least clear which direction the tricopter is pointed. If I decide to fly it in low-light I’ll add port and starboard red and green lighting to the end of the front booms.

I decided to shorten the booms a few cm, I can’t see any negative effects and the frame is a little lighter.

Lastly, I flashed the speed controllers with SimonK’s modified firmware. The firmware has a much faster refresh time making input far more responsive and auto-levelling a lot more effective.