On the Road to a Full-CF Aircraft

The main driving force behind my current UAV experiments is lowering aircraft weight and noise. In principle though, this should not come at the expense of other factors, such as flying time, rigidity, responsiveness and crash-worthiness. So with these in mind, and after the eventual success of my last build, I’ve been planning for some time to build a quadcopter frame that can combine most – if not all of these good characteristics.

The choice of material? – You guessed it, Carbon Fiber.

It’s almost the end of 2016 and by now CF frames are becoming more and more popular in the DIY world, but also in commercial and pro solutions. All these 250-class racers, the Tarot frame series, even DJIs own Inspire drones – they all make use of CF. Sometimes just few parts here and there, other times major parts of the frame. In most cases though, what you have is a combination of CF plates, CNC cut to shapes of varying ornamental quality, and CF tubes for attaching the motors and optionally landing legs.

Ok so you ask, how do you bring these CF parts together? CF is notorious for its unfriendliness to being processed. Well, the answer is one that usually disappoints me. What most manufacturers do is to join the various CF parts using CNC cut aluminum or steel connections; clamps, plates, rods.. you name it. But I believe this is sub-optimal in two ways: First, connections between CF panels and hardware are point connections that use a screw to hold the pieces together. The connector is usually a rod. This type of connection is quite poor structurally, since loads parallel to the CF plate induce large moment loads at the joint. Secondly, the weight of the connecting hardware usually cancels the weight benefits of using CF in the first place. In the end, one could achieve similar performance with a well-designed plastic frame1.

For my next build, I’d like to go all CF. For this reason I have been looking in making my own parts by laying up CF fabric with epoxy resin. I started with a design for an H-Quad, which I chose for the additional space it offers. Additionally, the arm connections can be made quite simple and strong, since they are coaxial and non-overlapping.

My initial design was a U-shaped central body that supported lateral beams for the motor mounts. I made the design for the mold in Rhino, and 3d printed it. The 3d model included supports for safely 3d printing the delicate 0.8mm thick sections.

Mold model being 3D printed in PLA

The next step was to correctly place and glue the printed model on a solid board.

3d printed positive for the creation of the (negative) mold, on which the CF will be laid up. I shrink-wrapped it while waiting for the mold cast to prevent dust from setting in.

Casting of the mold followed. There are specialized products that allow one to create molds suitable for CF lay-up. I found the following video to demonstrate the process in the most comprehensive manner:

The process is as follows:Before starting with the casting, the model needs to be prepared with several layers of mold-release wax, and a layer of PVA, which is a release agent, and let to rest until the PVA dries (20 minutes should be about enough). Then the actual mold casting begins. First, a layer of gelcoat is applied to the model, which helps capture the finer details and provides a smooth foundation for post-processing, such as sanding and buffing. The gelcoat is a two-component mixture, consisting of a resin and hardener. After waiting for a while, so that the gelcoat reaches a gelatinous state, a clay-like mixture is spread on top. This is again a two-component mix that reinforces the weak gelcoat layer and stiffens it to a rock-hard object.

Read also:  Tinymovr: An Integrated Brushless Motor Controller
Cured mold, showing the green-colored gelcoat. (c) Easycomposites


3d printed positive with the gelcoat applied
3d printed positive with the reinforcement applied

The mold is set to cure for about 24h2, and then separated from the positive original. Owing to the use of wax and PVA, separation was a snap in my case.

Separated mold (negative)

Post-processing included cutting of excess reinforcement, smoothing out of the gelcoat and a bit of sanding of the bottom to ensure a firm and flat contact with the supporting board.

Processed mold (negative)

After the mold is prepared, the actual lay-up process begins. Again, there is a great video on that, as a continuation of the previous one, below:

First, the mold is covered in several layers of mold-release wax and  a layer of PVA, as in the previous steps and left to dry. Then, the resin+hardener are mixed to proper ratio, and a first layer of resin is added to the mold. The resin needs to rest for a while (around 20min), till it gets the consistency of a gel.  In the meantime the CF cloth is cut to size. After the waiting time has passed, the CF cloth is laid up layer by layer, adding a thin layer of resin in between, every time. It’s a good idea to make sure your CF layers are in good contact with each other, otherwise you will get a structurally weak outcome. Pay attention especially to any sharp corners, and push against the mold lightly to bring them close together.

CF Lay-up process

The frame was quite simple and easy to manufacture, and the lay-up went well. However, it turned out to be a complete failure due to structural reasons, because I did not factor in torsion. Open sections such as the one used are susceptible to orders of magnitude more torsion than closed ones. I tried to remedy the situation by reinforcing the top part of the body, but it never really became as stiff as I would like it.

The cured and post-processed frame


The frame together with a 3d printed reinforcement piece, which added significantly to it’s weight.
Top: The new all-carbon frame. Bottom: The old frame.

As a conclusion, this was my first experience with manual CF lay-up, and it was quite enjoyable, and a great experience overall. It would be very productive as well, hadn’t it been for the excessive flexing and twisting of the final product, which made it practically unusable as a stiff quad frame. In the next post, I will outline the second iteration of the CF frame, which, contrary to this one, was a success. Stay tuned!

For more exciting experiments and tutorials, subscribe to our email:

  1. An example is the DJI Phantom drones
  2. actual curing time may depend on the type of resin you get. I figured out that mine did cure completely in around 12h
5 replies on “ On the Road to a Full-CF Aircraft ”

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.