First mold of cabin

Last weekend, we vacuum resin infused the first mold of the cabin. It’s the first and most crucial one of three pieces, because it will be used to make the die for the clear canopy. We had to take two shots, as the first vacuum infusion failed. More on the reasons below.

Everything began with the dry lay-up of the sandwich. We made some simple tempates using Compoflex® SB 150. This peel ply is ideal for making templates because its drapable, but it does not deform as easily as woven fabrics, and it’s a consumable.

Diese Diashow benötigt JavaScript.

After having made the sandwich’s dry lay-up, we used the Compoflex® templates to make the peel ply layer. There is a Compoflex® variant specifically desgined for vacuum resin infusion (RF 150), but we have only the one for vacuum bagging (without flow mesh).

We used this type of Compoflex® as a substitute for usual peel ply, as the forces needed to peel it are much lower (prevents premature sepration of the mold). Though it has no flow mesh, we didn’t expect to get into trouble, beacuse the Soric® XF core should act as a flow medium. We were proved to be wrong, as we painfully recognized later…

As always, we used the MTI® vacuum line by DD|Compound. It makes a resin trap unnecessary, and we’ve used it in all our infusions.  Highly recommendable.

Having vacuum bagged everything, we started to mix the resin and degas it. We used HP-E120RI from HP-Textiles. It’s an epoxy resin with 200 minutes pot life. About a month ago we made an infusion test together with the Soric® XF core. It worked pretty well, but we saw that it takes quite some time to degas. Degassing of 1.7 kg (3.7 lb) of HP-E120RI took us about 30 to 35 minutes, and we needed roughly 6.8 kg (15 lb) for the mold! To make the story short, it took ages to get the resin ready:

After opening the feed line, it became clear that we’ll get into real trouble. The flow speed was incredibly slow. If everything works well, it’s possible to infuse larger amounts of resin before it starts to heat up. We’ve done this, for example, in the mold of the center section.

Epoxy resin reacts with the hardener and creates heat. It’s an exothermic reaction. The larger the amount of resin is, the more it heats up because less heat is dissipated to the environment. Heat speeds up the reaction ending up in more heat being produced (pot life is halfed every 10 °C or 18 °F more). If you have bad luck, it can heat up so fast, that it hardens within minutes, although it takes usually one day to room temperature. This is exactly what happend in our first infusion: We needed very long to degas the resin, infusion speed was too low,  the resin started to heat up quickly, and it „boiled“ at some point. Lucky us that it was cold and snowy outside and the resin did not catch fire.

We had to stop the infusion, although only 1/3 of the mold was ready. Usually this is a killer for a laminate, because infusion is usually used to create nice looking laminates. Here, we are only interested in properly wetting the sandwich, and we do not care about the looks. This is why we decided to let the aborted infusion cure and to infuse on top of it in a second shot.

Before we made a new infusion we had to understand why the first one failed. The main problem was probably the low permeability of the sandwich. If the feed line has direct contact to the Soric® core, low permeability shouldn’t be a problem. Here we had no direct contact to the core: The resin had to flow through a layer of Compoflex® SB 150, two layers of 105 g/m² glass and two layers 380 g/m² carbon. This was proably too much friction for the somewhat viscous HP-E120RI. The solution was, thus, to use a flow mesh the next time. We also decided to take the slightly less reactive and less viscous HP-E300RI with 300 minutes pot life:

The second photo shows clearly how much the infusion fronts outside and inside the core differ. We’ll probably not rely solely on Soric® in our next infusions and will use always a flow meash. We documented both vacuum infusions in a short video:

Here are some pictures of the mold after the laminate cured and the peel ply was removed:

We are making now the next parting surface. This surface will separate the aft part into a left and right mold.


Weak sick week

Not much to report on mold-making today. Since we started with the mold last weekend not much has happend. It has been a weak week, because we have a heavy cold. Last saturday we coated the plug with tooling gelcoat:

It’s strange to put put this intimidating tar-alike stuff on the high gloss plug. The gelcoat works perfectly (Formenharz P from R&G GmbH), but I forgot how nasty it is. Use always a good mask when working with it!

After two hours, the gelcoat geled and we laminated two layers of 105 g/m² glass and one layer of 388 g/m² carbon on top:

The outcome looks quite promising:

This lay-up builds a sort of air tight tub on which the rest will be vacuum infused: Soric® XF 4 mm, two layers of 388 g/m² carbon and two layers of 105 g/m² glass.

Though I’ve been sick this week, I finally found some muse to modernize the vacuum regulator of my small pump. You’ve probably already seen my old regulator:

I know it looks unprofessional, but the features of the regulator were ideal: hysteresis regulator, set-point setup over a potentiometer, display of current pressure and pressure history. To make it short: Great firmware running on an unprofessionally mounted hardware. A comparable commercial product costs around 500 EUR, which is ten times more compared to my custom solution.

Some weeks ago I stepped on the regulator and destroyed the display. Shame on me. This is why it was time to build something new. The pump still looks quite strange, but the regulator got a nice case and new components. I lost the code of the firmware and have to start again from scratch, which is demoralizing. Anyway, it is not always bad to start again, and I use anyway a different microcontroller now (SAMD21 Cortex M0 against two ATmega328P before). The pressure history is still missing in the firmware, but at least the regulator works as it should:


Material finally arrived

Ready and set to make the cabin’s mold! Tuesday was a great day, though the delivery was posponed and split into three trips because the load was too large and contained dangerous goods. More about the dangers later.

We were expecting something big, but this exceeded our expectations:


Inside the smaller and lighter box, some Soric® XF 4 mm core and 105 g/m² glass fabrics were delivered.  It also included some goodies, in particular a very nice spread carbon fabrics with IMS 65 fibers:


It’s very dense and very light (80 g/m² or 2.4 oz/ya²) but still very tough. We bought a couple of meters and a sheet of Rohacell® IG-F 31 to make some tests for the wing shell. Further details on the test will be provided later in another posting.

Back to the large and very heavy box. It contained a large roll of thick carbon fabrics (388 g/m² or 11.4 oz/ya²):

It’s so much, that I had problems to lift it into the house (over 130 m² or 155 ya²!). I leave the maths to you, if you want to know how heavy the roll is. This material is obviously too heavy and thick for the airplane, but perfect for making though and stiff molds. Though it looks coarse, it is a special kind of twill weave and very flat. We got a very good price for it and we had to have it. We’ll probably make all subsequent molds with it and there’s enough for a couple of friends too.

Now to the dangerous goods. I love it when companies are creative while packaging a delivery. I can literally see the employee looking for empty boxes at the grocery store :


I leave it up to you to decide which good on the pallet is most dangerous…

Anyway, we bought some of a very special epoxy-resin system: LG 735 AERO (GRM Systems). This resin is a non-toxic—probably more precise to say less toxic—substitute of the well known Aradite® LY 5052 from Huntsman. It makes extremely tough carbon laminates and the producer states that it’s even tougher than the LY 5052. We’ll test it together with the spread fabric and the Rohacell® core. Let’s see how it performs…


Ready for mold-making

The plug of the cabin is ready for making the mold. After the surface was shot several times with filler and sanded, a layer of clear coat was shot on top to get a hard and robust surface. This layer was sanded with 1500 grit and waxed with a Carnauba wax several times: A thin layer of wax is applied with a piece of cotton and after 10 minutes it is polished with fresh cotton. A couple of hours should be inbetween applications to let the wax harden. One should keep on waxing and polishing until the plug looks glossy:

We made again a video on the whole process:


Last weekend we made the first parting surface, which separates the front from the aft. Another surface, which separates the aft part chordwise into two pieces, will be made after the first mold is finished. The edge of the parting surface can be created by protecting the plug’s surface with tape and using putty to fill the gap. The edge is sanded flush after the putty hardened. Minor gaps can be filled later with some wax dough.


The material for the mold should come next week. The layup will be glass, carbon, Soric, carbon and glass. Part of the mold will be vacuum resin infused. As before, we’ll document the process and provide here information.


Skin white as snow

„Skin white as snow,
lips red as blood,
and hair black as ebony.“
, Jacob Grimm

To be honest, today only the skin of the cabin’s plug became white. The lips and hair of Schneewittchen will follow some day. Promised.

The day started with moving the plug into the other room, where the center section was waiting patiently for the cabin. We had to tidy both the room und the garage up. Having the room empty we had to clean it throughly, so that we don’t get dust on the plug while shooting.

The plug was then ready to get the first layers of filler shooted. Tomorrow, after the filler hardens, we’ll sand the plug and shoot a couple of layers more.




It seems to go on forever

„It seems to go on forever,
but it’s not done
until it’s all one color
and it’s all smooth.“, Mike Arnold

Very true words, those of Mike Arnold, the designer of the AR-5 and 6. Some postings ago I mentioned and included his video on plug making. The Arnold Company has some other very interesing videos online and it’s worth to take a look at their channel.

Finishing the plug truly seems to go on forever: We’ve consumed until now more than 4 kg (9 lb) of putty and the plug is still not „all smooth“. We started by sealing the wooden surface with Clou® Schnellschleifgrund, which is a nitrocellulose based pore filler. After that, we applied two thick layers of putty and sanded everything flush with 80 and 120 grit sandpaper between the coatings:

As soon as wood is seen, sanding should be stopped. This is where abachi venner is better compared to glass fabrics: It’s not too tragic, if some of the venner is sanded.

Having two layers of putty on the plug, the long lasting detail work starts: A dry guide coat from Mirka hepls to make the low spots visible. The nice thing about dry guide coats is that they are ready to use right after application. Guide sprays need to dry before sanding. Stroking with a pencil on the surface is also a good technique to identify large low spots, but you’ll probably miss the small ones:

Merry Christmas!



Hard work and patience pay off

The cabin’s plug is ready for finishing! It still has to be filled with putty and painted, but we have about 80% of the job done. Last time we reported how we started to prefabricate an abachi veneer layer for the plug. Now, the abachi veneer is on the plug and sanded. It was worth the effort:

Though the veneer layer needs time to get ready for cementation, the result after grinding is great. Not much putty will be needed to get a perfect surface:

We made also a video tutorial which provides many interesting and helpfull hints on how to make such a complex plug:




Patience …

… is a gift and usually pays off.

The rough XPS core of the cabin is ready to cover. I hoped to be able to use few pieces of abachi veneer to cover it, but it turned out to be impossible. Thus, I remembered a technique I’ve used several times in rc-models. The basic idea is to use small strips of wood to cover the spherical surface. Take for example the prototypes I made for the S12 (1:6), the Beluga (1:4), and Schneewittchen (1:4):




The principle is always the same: Use thin strips to cover the surface.  If you try to cover a spherical section with a large sheet of abachi veneer, balsa or plywood, you produce huge shear loads in the sheet and most likely you’ll fail to get a good surface. The reason is that spherical surfaces have gaussian curvature. It’s very easy to cover a cylinder or a cone, because these lack gaussian curvature: Cones and cylinders are curved only in one direction, whilst the other is straight. This is not the case for the cabin of Schneewittchen, which has for sure gaussian curvature.

I used the same technique. The difference is that I have to prefabricate the curved venner sheet before it is cemented under pressure to the XPS core. This is how it looks at the moment:

There are still a couple of few square meters to cover. Though it will take a couple of days to finish, it is a nice relaxing job which is best done with some music, and of course with plenty of patience.


Taming the beast

„The beast in me
Is caged by frail and fragile bars
Restless by day
And by night rants and rages at the stars“

Johnny Cash



The „beast“ to tame here is not a kind of Halloween monster, but the structure of Schneewittchen.

If there is anything that makes the design of an all-wing aricraft special—besides their aerodynamics—it’s their structure. Sweep-back is a headake! When you build an rc-model, you just drop some elements into the structure, and voilà it works. When you are going to sit in that airplane and there are regulations in strength and weight to fulfill, things get really complicated. There is not a glance of being able to just drop some wood or carbon into the structure and getting it right. Either it’s not strong engough, or it’s too heavy.

A couple of postings ago  I reportedt that I started to calculate the structure. Most headake has provided me sweepback and the implications of it. One of the largest problems is to account for the change in direction of the central spar. This kink in the main spar sheds a large torque, which tends to rotate it about the span axis and has to be balanced by the pilot’s weight. Sounds easy? Well not really, as this torque is considerably larger compared to the one produced by the wings. At the joint between the center section and the wings, there are merly 0.8 kN·m (590 lbf·ft) of pitching up moment produced by the wings. In comparison, the change in direction of the main spar produces a torque of -2.8 kN·m (2065 lbf·ft). The reason for this high torque is that part of the axial loads of the main spar are shed into torsion. The spar’s caps carry a huge axial load originating from the bending moment, and the kink has about 25°. Hence, about 40% of the axial load is shed into tranversal forces at the joint, while the axial load is only reduced by 10%. That’s what I call a bad deal!

Changes in direction are almost inevitable when there is sweepback. At some point the spar will have to account for the symmetry of the wing. We’ve build several rc-models and tested different structural concepts. Most notably is the structure of the Schapel SA-882 models we designed.

Large turbine acrobatic model:

Small speed model:


If you look carefully, you’ll see that the main spar is curved preventing localized changes in direction. Also, the aft spar is straight connected to the main spar, and builds, thus, a sort of triangular shape when seen from the top. This type of structure has prooven to be very stiff and strong. Take for exaple the flight of John Wright’s turbine rc-model he made from a kit we provided him with:

Or speed flights with the small one built by a friend from a kit we gave him:

So what’s all about the curved spar? Well, this is the best way to prevent strong localized changes in direction. The continuous  change sheds constantly small amounts of shear which are taken up by the spar’s shear web. The aft spar adds a surplus of stiffness to the structure, in particular torsional stiffness.

The main problem with a curved spar is that it is not suited for demountable airplanes. Schneewittchen needs to be demountable and a curved spar is not feasible. The aft spar is fine. Here’s a scheme of Schneewittchen’s structure:Struktur

Essentially, the straight aft spar was kept and the main spar had to be kinked. Believe me, this kink has given me something to brood about…

The forces acting on the center section’s structure look as follows (ignoring the shell):


The change of direction sheds a large torque, which is mostly counteracted by the pilot’s weight (2×2kN) and the batteries (2×0,7 kN). The pitching moment of the wings also slightly contributes. The two beams on which the pilot sits cannot transfer a torque on the main spar, or at least they shouldn’t if we want to prevent that the joint breaks. These beams transfer up/down forces on the spars, but they also join the spars at the bottom. Hence,  the main spar can be assumed to be fixed on the bottom but free to move on top. A rotating main spar is a really bad idea!


The above scheme misses that there is a rib connecting both the aft and main spar. This rib is very stiff to shear (fibres oriented ±45°) and prevents shear deformation. Any kind of rotation of the main spar would produce shear. Even more cleary seen, when we assume that there are stifferners between the spar caps.

schubmittelpunktThus, a the torque will produce a rotation of the structures around the shear center of the section at the joint between wing an center section. The shear center is very near to the main spar, and the spars and the rib will try to rotate around a point inbetween the spar caps. This rotation means, however, that the aft spar will have to bend, as it is supported by the main spar at its ends. Violà, the weight of the pilot and the batteries pushes the aft spar down counteracting this rotation.

Let us recapitulate:  If we have a shear web connecting the spars at the joint between wing and center section, the torque produced by kinking the main spar is converted into a bending moment of the aft spar and the masses in the the center section can balance the torque. Quite strange and not obvious.

So much on the theory. It’s sometimes easier to have a small model in which hypotheses can be tested. This is usual in many fields, such as in analysis of aeroelastic effects. In aerolasticity it is common to build models with realistic distributions of mass and stiffness (bending and shear). I built, thus, a scaled down version of the main structure to see how it deforms when loaded:

The video shows clearly how bending the structure produces a torque a the main spar and tends to rotate it about the lower spar cap. Torsional stiffness is also low.

Folloing the theory derived above, the structure should become much stiffer by adding the ribs at the joint:


Stiffeners were added between the spar caps to prevent these from buckling. We did not make a video of the modified structure, but stiffness increased substantially (also against torsion). Rotation of the main spar is minimal. The whole structure tends to slightly rotate about the bearing, which is good because this means that forces are flowing as expected between spars and the pilot’s mass.

We built another even more realistic mockup:


Adding shear webs to the spars made it extremely stiff and, thus, not suited for seeing how it deforms when loaded. Though not having a closed shell surrounding it, the torsional stiffness is exhorbitant. This are good news for Schneewittchen…



Long rainy weekend

„Anstrengungen machen gesund und stark.“, Martin Luther 1566

It is a grey Sunday with lots of wind and rain. Perfect for building Schneewittchen. Tuesday is the 500th anniversary of the reformation day and, thus, an exceptional holiday in Germany. Happy those like us who can take Monday free and have a long weekend. Having a couple of free days does not mean that we bum or laze around. It is pretty much the contrary: We have finally time to do a couple of things on Schneewittchen that take longer.

Yesterday was a bad day. I thought we could quickly machine a couple of pieces for the next prototype of the elevon mixer. It turned out to be a mess ending in a broken milling tool—a 4mm solid carbide one-flute and by the way one of my favorites—and another shiny new solid carbide tool clogged. You can imagine I was not amused! This is the miserable rest of the one-flute:


Heartbreaking! I did not fasten the material correctly and it started to oscillate at some point. This loss is my fault, but how it came to the clogged tool is not explicable…

Anyway, I finished the pieces and adapted the elevon-mixer:


In contrast to the last version, I used two spherical bearings this time (less parts and a less complicated setup). Tightening the bolt on the left in the picture, which fastens the plate on the stick, results in a tighter fit of the spherical bearings. So this setup has some more friction than the last one, although the bearings run very smooth before installation. This shows that their internal clearance is small, which decreases slackness of the controls and is good when not excessive.

Today was, besides the weather, a great day. We started to cover the cabin’s plug with veneer. It was not the surface, but the bottom of the plug:

Sanding the border is much easier when it is reinforced. This is in particular true for the thin tail, which would else be pushed away when sanding. We used leftovers of abachi veneer and glass fabrics (feels like 80 g/m²). It will have to cure at least overnight before we can trim the border.

In one of the last postings I wrote that sanding the plug is a multiscale process. I made two pictures to show you what kind of scale we are working on now:

There is a change in curvature between the canopy and the aft of cabin. This is the kind of errors that need to be corrected at this step. It’s not about getting a perfect smooth surface. This kink has a large scale and extends over 25 to 50 cm (10″ to 20″). No veener in the world would make that disappear.

Nevertheless, not only large scale imperfections need to be corrected, as the plug has also many small defects. Some can be ignored and others would affect the shape even after covering. Large defects can be improved by glueing small pieces of XPS and sanding the surplus material. The following picture shows, for example, a larger dent which was improved as mentioned:


That’s good enough to be covered with veneer. The glue used was specifically designed for XPS and EPS foams (UHU® POR). There are many more similar defects on the plug, which have to filled and improved before covering. The next days won’t get boring…