SBD Dauntless (from scratch)

[hendie]

a long time ago I used to work in 3D Studio Max to do animations and walk throughs for customers so I have a basic understanding of what you are doing. My skills were never this good but I do understand how much time, effort and skill goes into a project like this. Great work !

 

[(ex)Sgtrafman]

Just incredible - I don't pretend to understand any of it but I'm in awe.

Iain.

 

[Witold Jaworski]

hendie, (ex)Sgtrafman: thank you for following! I will do my best

[Witold Jaworski]

In this post I will work on the weathering effects of the color texture, while in the next one I will add scratches and some other remaining details.

The weathering effects that you can observe on the aircraft from WWII era are quite “dramatic”. The paints used in mid-20th century were not as chemically “stable” as the contemporary coats, thus they could change their hues in few months of intense service. The archival color photos below show an extreme case of this effect:



These photos were taken by Frank Sherschel on 14th November 1942, for the “Life” magazine. The SBD-3s depicted on the pictures belonged to VMSB-241 squadron, stationed at Midway in that time. Marines received these aircraft in July 1942, but all of them were already used before - most probably on the U.S. Navy carriers. I think that in November 1942 these SBDs had accumulated about 10-11 months of the war service. I will use them as an extreme case of the weathering. (It is always good idea to recreate such an ultimate case in the texture, because you can always make your model “newer” by decreasing intensities of the weathering layers. On the other hand, you cannot use more than the 100% of their intensities, thus you cannot make your model “older” than you initially painted).

Having historical color photos of several airplanes used by the same squadron, you can easily determine the general pattern of the stains, smudges and scratches. This pattern repeats with some random variations on every aircraft. It seems that such a radial engine like R-1820 emitted a lot of oil – in their exhaust fumes and in the air flowing from the NACA cowling. (Because of the high “oil consumption” of the R-1820 the oil tank in the SBD was quite large). The non-specular paint of the two-color Navy camouflage absorbed this oil mixed with the soot from exhaust fumes, creating characteristic dark traces along the fuselage and the center wing. The crew most often walked on the center wing, thus you can see on its upper surface the lighter traces along dome rivets seams and darker, “trodden” spaces in between. There are also some scratches in the paint. Some of them exposed the yellow primer, while the others reached the bare metal of the aircraft skin. (Thus I assume that the first layer of the Douglas primer had a yellow/orange color).

You can also notice some white splashes on the outer wings (traces of the coral sand from the atoll?), as well as the repainted areas around the tactical numbers. (On a black-and-white photo, it would be extremely difficult to distinguish these repainted areas from the oil/fumes traces).

What is interesting – in spite of the non-specular camouflage of these aircraft, you still can see a specular highlight at the wing root (as in figure "b", above).

The key elements of the weathering effect still depend on the technical details of the aircraft skin: rivets and panel seams, bolt heads, inspection doors. However, their pattern is so random, that I cannot recreate them in the SVG image, as I did in the case of the reflectivity map. Thus we need a reference picture of the skin details for painting the color (diffuse) texture. It has to be mapped in the same UV layout (UVMap) as the color texture. To do this, I composed from the key layers of the SVG drawing an auxiliary image, mapped in the UVTech layout. Then I quickly transformed it into UVMap layout using the Bake feature:


 

Quote

The old Blender Renderer engine is a better tool for such a direct transformation of a single texture than the Cycles Renderer. In the Blender file that accompanies this article, I created a dedicated screen layout for this purpose. It is named Texture Baking. You can find there a simple script that I used to switch the UV layout of all model meshes between UVTech and UVMap. To bake a texture from UVTech to UVMap layout, you have to switch the current rendering engine to Blender Renderer and disable nodes in the B.Skin.Camouflage material. Then assign the image (mapped in the UVTech coordinates) to its single Blender Render texture assigned to B.Skin.Camouflage material, named Image for Baking. When you click the Render:Bake button, it will generate the resulting texture (in the UVMap coordinates) in the Test image, which you can see in the UV/Image Editor. You can use Image:Save As command to save it to an external raster file.

 

I created such a reference image for each SBD version I have modeled (SBD-1, SBD-3, and SBD-5), and placed all of them in GIMP:



The darker areas in this weathering appear between the rivets seams. I decided that it will be easier to recreate them using one layer of darker camouflage color, overlaid by partially opaque layer that contains the weathering pattern (in white). First I painted in this way the right outer wing:



I used this “sample” for testing if such a pattern looks good in the final image:



I think that it looks acceptable. Thus I started to paint in this way the weathering of the whole upper surface.
I create the basic pattern of the lighter traces along rivet seams in four steps:



First I painted the “overall noise” with an “acrylic brush” tool (figure "a", above). Then I changed the tool shape to “pencil” and draw thin lines along rivet seams (figure "b", above). In the next step I used Eraser to make this pattern more random (figure "c", above). Finally, I filled it again with light touches of an irregular brush, to lighten the overall effect (figure "d", above). (This last step is optional, depends how this fragment looks like on the reference picture).

Figure below shows the resulting weathering pattern on the fuselage:



As you can see, it differs from the pattern on the center wing. The lighter traces along rivet seams are thinner, the color is more uniform. Using a separate layer, I added a yellow tint to the darker areas.

Unfortunately, in Frank Sherschel’s collection there is only one, small photo which shows the bottom surfaces of an SBD from this squadron:



You can identify there some smudges on the outer wing and rear fuselage. However, you cannot determine the dirt pattern for the center wing and the engine cowling. (You can only say that they are dirty).

Well, in this case I had to use another, black and white photo as the reference. This photo shows clearly the center wing section of a typical aircraft:



As you can see, I recreated in my image the exact copy of the smudges from this SBD-4 center wing. They are placed on another layer (Flow) in the Stains layer group.

Finally, I enriched this basic dirt pattern with all the additional details visible on the reference photos: white “burnouts” on the fuselage sides, discrete traces of soot. There are also irregular, darker lines along some of the panel seams:



Each of these elements has its own color and layer. Note that the darker lines along panel seams extend across the upper and lower surfaces. I painted them using the same color, but on the bottom surfaces they are more transparent, to obtain the appropriate contrast. (Painting them, I used a 50% opaque mask for the bottom surfaces).

Figure below shows this “weathered” diffuse texture on the model:



Frankly speaking, I am still not satisfied with these results. This weathering requires some minor adjustments. For example: on the reference photos it has a slightly different hue. The fuselage below the tailplane also requires some fixes.

While painting all these weathering effects, I came to conclusion that I cannot re-use them without any modification in the three-color camouflage, used in the SBD-5s and -6s. Thus I will not split the color texture into three interchangeable parts, as I announced in my previous post. I will have to prepare few alternate color textures instead:

• A colorful pre-war painting scheme (orange wings!), without weathering for the brand-new SBD-1s from the Marines squadrons (+ eventually the later single Light Gray color scheme, as visible in color photos preserved in the Smithsonian Air And Space Museum);
• Two-color Navy scheme for the SBD-2,-3 and -4s (described in this post);
• Three-color semi-gloss Navy scheme for SBD-5s and -6s (+ eventually the white variation of this for the SBD-5 and -6 scheme, using on the Atlantic areas);

Of course, I will reuse fragments of the weathering pattern described here in the three-color scheme for the SBD-5s and -6s. However, before I do it, I have to finish this color texture. Thus in the next post I will fix the minor flaws described above, and recreate the scratches visible on the reference photos. Then I will apply the “decals” – national insignia, tactical numbers, etc.

In this source *.blend file you can evaluate yourself the current version of the model, here is the GIMP source file of its textures. Because of the large size of the original GIMP file (*.xcf), this post is accompanied by its smaller version (2048x2048px), packed into *.zip file. I think that such a version is sufficient for checking all the details of this image (the structure of its layers, their opacities and mixing functions). The resulting textures (4096x4096) are packed into accompanying Blender file.

[Gorby]

I'm not sure there are enough 'wows!' in existence for me to say how impressed I am by this! It's more than stunning.

[Vadim]

I think it should be viewed by any modeller who is interested to simulate an aircraft weathering. This is a perfect guidance to understand the weathering and its application on a model.

Thanks Witold - very impressive!

[Witold Jaworski]

Gorby, Vadim - thank you!

Below another post, this time about specific detail: recreating the aircraft skin abrasions. To provide a complete picture to eventual other 3D modelers, I described here the "kitchen details" of the Blender node setups. (I know that it may look like a "rocket science" for those who never used it :). However, this is one of the last posts about "painting". I will finish this phase soon, then will provide posts on modeling more classic elements when I start to work on the details: undercarriage, engine, etc.)

_________________________________________________________

It seems that Douglas used a high-quality paint for their SBDs, because I cannot find any trace of chips/flakes, even on such a worn-out aircraft as this from VSMB-241 (see figure below). However, you can see some scratches on the center wing, trodden by the crew:
 

In the photo above, the minor scratches are yellow, because Douglas used a yellow layer of Zinc Chromate primer below the camouflage paint. (The interiors were painted with another layer of the Zinc Chromate, mixed with Lamp Black to obtain a darker, greenish hue).

However, the larger area along the leading edge was often trodden to the bare metal, which you can see in the photo. This scratch has a typical, irregular band of the primer around its borders. In this post I will recreate these abrasions.
The aircraft producer anticipated this kind of damage (at least to some extent), placing a thick, rough, anti-slip strip along the fuselage. It seems to be made of a black, rough material, glued (?) to the aircraft skin (as in the picture above), and spans from the trailing edge to the main spar. However, the most “visited” by the servicemen area in the front of this spar had no such a cover. On some SBDs (including the one on the photo above) you can find a glossy, black continuation of the anti-slip strip. I suppose that it was added in the Navy workshops. It seems to be simply painted using a typical black paint, thus was more prone to the abrasion (as you can see in the photo).

To recreate the effects from this archival photo, I had to paint this glossy black area first. It requires changes in the two textures: color map (because it was black) and reflectivity map (because it was glossy). I created in the source GIMP file separate layers to reproduce these shapes. This way I can simply turn them off while painting another aircraft that did not have this feature (like most of the SBD-5s and -6s). Figure "a", below, shows the color texture image, while figure "b" shows the additional reflectivity map (they share the same GIMP file):
 

Note that I introduced here another part of the reflectivity texture: the map of the areas that are more specular than other. This is a natural addition to the ref-aux.png map that I used in this post to make the remaining anti-slip strip more rough. There are also some other areas that look more “wet” (specular) than the others. For example – the fuselage sides (because of the small amounts of the oil, spreaded from the engine). I painted them in GIMP in a lighter gray, on a separate layer named Wet area (see Figure "b", above). I saved this image to a file named ref-aux-spec.jpg. However, I could not simple merge it with the ref-aux.png map, because it uses the different UV layout (UVTech). Thus I had to join these images in Blender, using special node that mixes two grayscale pictures (as in Figure "a", below):
 

For the brevity I renamed the ref-aux.png file to ref-aux-rough.png. Figure "b", above, shows these updated textures on the rendered picture. Frankly speaking, the weak increase in the specularity of the fuselage sides is not visible here, but you can see the much more accented difference between the reflectivity of the forward and the rear anti-slip strip segments.

I am going to recreate the scratches that you can see in the archival photo using so-called shader mask. Thus, in addition to the weathered camouflage shader that I used so far, I need a “bare metal” shader, which looks like a rough duralumin skin. Figure below shows such a thing:
 

Note that this shader uses the ref_details.png image (see Figure 73-4 in this post) as the base color texture (for the Diffuse as well as the Specular colors). The dark lines from this image create appropriate shadows along the seam lines. I just made this input image somewhat darker (using a Multiple node), to transform the semi-white background color of the original ref_details.png image into metallic gray. What’s surprising, it’s better to use a fixed Reflectivity for such rough (low-reflectivity) metallic shaders. When I tried to use the same ref_details.png image as the reflectivity map for this case, it made the panel seams lighter, reverting the effects of the color texture.

To share the dirt between the “bare metal” and the camouflage shaders, I separated it from the color texture. I named the two resulting images as color-camo.jpg (the weathered camouflage) and color-dirt.png (the soot traces and some stains on transparent background). I used the Stack Image nodes (see Figure "a", above) to combine them with corresponding backgrounds, like the layers in GIMP. Figure below shows these two components of the color texture image:
 

In the color-dirt.png image I initially grouped only the soot traces. It is possible that I will also transfer other layers there. (We will see if I have to do this during the work on the pre-war “natural finish” painting scheme of the SBD-1).

Then I prepared a B/W “shader mask”, painting white, feathered scratches on a black background:
 

As you can see in Figure "a", above, I decreased the opacity of these shader mask layers while painting, using as the reference the camouflage from the layers below. I also used a copy of the reference image (layer Rivets and seams) to precisely recreate technical details visible inside this scratch. (SBD had flush rivets on this area. Usually the paint remains on their heads even when there is bare metal around). Figure "b", above, shows the final material mask. I saved it to a file named mask-scratches.jpg.

Below you can see how I used this mask in the material scheme:
 

Note that I placed a Color Ramp strip between the mask and the Mix Shader node. The pattern, set in this node, controls the ultimate size of the scratches, and the width of the transition area. (The transition area is “feathered” border of these bare metal scratches, where the share of both shaders is greater than 0). For example, if I want to get the “chips” effect (bare metal areas with sharp borders) I should switch this color ramp mode from Ease to Constant. (It creates a strip that contains just sharp black and white spans, without any “gray transition” between them).

The result of this shader mask looks like the test render below:
 

The bare metal scratches look good on the dark background (like the anti-slip strip). For certain viewing angles they disappear in the camouflage, because of their gray color (as in Figure "a", above). Still, they can shine (as in Figure "B", above) when you view them from other sides, especially when you do it from greater distances. (This shining effect looks quite convincing – this is the advantage of this method over scratches painted directly on the color texture).

Now, let’s add the primer to these abrasions! This means that we have to add yet another shader to our material scheme. It is similar to the camouflage shader, with one exception: it has to have a uniform yellow color, instead of the camouflage texture. To reuse all the other settings, I decided to “extract” the final Gloss Paint shader from the X.Textured Skin group. Figure below shows the modified material scheme:
 

I “cascaded” the camouflage, primer and bare metal shaders, using two Mix Shader nodes. In general, each Mix Shader could use a different mask image. In this case both use the same input mask, but each one modifies it in a different way, using its Color Ramp node.

You can see the result below:
 

I think that it looks good enough. These and other details bring the final effect closer to the original photo from the beginning of this post (compare the picture below with the first picture in this post):
 

Of course, there are still differences. (For example - I have impression, that I should use stronger sunlight in my scene, to obtain such a high contrast between the lighter and darker areas as in the original photo). Well, I will work on these issues later.

Using the shader setup presenting in this post, painting all other scratches is a breeze: you just have to add another white splash on the shader mask image.

In the next post I will prepare the “decals” texture, containing national insignia, tactical numbers, and some service labels. (I will add more of these labels during the detailing phase, when I review each area of the aircraft skin).

In this source *.blend file you can evaluate yourself the current version of the model, and here is the GIMP source file of its textures. Because of the large size of the original GIMP file (*.xcf), this post is accompanied by its smaller version (2048x2048px), packed into *.zip file. I think that such a version is sufficient for checking all the details of this image (the structure of its layers, their opacities and mixing functions). The resulting textures (4096x4096) are packed into accompanying Blender file.

 

[Witold Jaworski]

The last texture for my model contains various elements that in the plastic kits are delivered as the decals: national insignia, radio-call numbers and various service labels. I prepared it as another vector drawing in Inkscape:

I exported this picture to a raster file named color-decals.png. It has transparent background, because I will combined this image with the other components of the color texture, prepared in previous posts.

 

The U.S. national insignia passed various transformations during the WWII. Between December 1941 and May 1942 the roundels on the Dauntless wings were enlarged, so they spanned over the ailerons (see figure "a", below):

However, as you can see in the photo of the USS Enterprise deck, there were exceptions: some aircraft preserved the older, smaller roundels. After 6 May 1942 all the roundels reverted to their “standard” size (72 in). Note that in this case they did not “touch” the aileron, but still their outer edge was very close to the leading edge (as in figure "b", above).

All of this means, that I cannot use for these wing roundels the same UV map as for the camouflage (UVMap). Although in this default UV layout the ailerons are in-line with the main wing surface (so they also fitted for the “decals” image), the problem occurs on the leading edge. In the UVMap layout its seam runs on the wing bottom surface, along the edge of the first panel. (I masked this seam on the camouflage texture in this way). Such a layout would split the bottom roundel into two parts – as marked in Figure "a", below):

To keep these roundels “in one piece”, I had to create another copy of the default UV layout (UVMap). I named it UVDecals. Then I modified it, adding an additional seam along the leading edge, and shifting appropriate faces from the upper to the bottom surface (as in figure "b", above).

In fact, I created this new UV layout only for the two objects: the outer wing meshes. This is possible in Cycles thanks to a special “fallback” node. In my previous model I worked out a node group, which can deliver the default (UVMap) coordinates for the all meshes that do not contain the requested (UVDecals) map. Such a group greatly simplifies using alternate UV layouts. You can find it in the material scheme as the UVFallback node:

Conceptually the color (diffuse) texture is composed from three images. The Decals image is placed over the camouflage (Camo), while the Dirt image is placed over them. Technically, the X.Textured Skin node internally places the decals image over the camouflage. Thus in the scheme you can see the Dirt image placed over the Camo and Decals images (it uses two Stack Image nodes for this purpose). If you want to learn more about these group nodes, see vol. III of the “Virtual Airplane” guide).

In the GIMP source, I shifted all the Stains layers from the color-camo.jpg into the color-dirt.png (see Figure 76-5 in this post). It allowed me to use the white stains layer for recreating weathering on the roundels located on the upper wing surfaces.

The stars on the wing bottom surface were also painted inside the letterbox slat. Initially there was something wrong with my UV mapping of this element (figure "a", below):

The pictures of the star on the slat inner surfaces were distorted (shifted). To fix this issue, I copied current UV layout (UVMap) into the UVDecals layout, and then shifted some of its inner UV vertices (figure "b", above).

Below you can see the first test of the “decals” texture:

At this moment it only contains the national insignia.

Now it is the time to “personalize” this aircraft. Let’s recreate the “black 4” from VSMB-241. As the first thing I added the radio-call numbers:

The single-digit number (“4”) was painted using the standard stencil. There was no problem in recreating this detail using the USAAF stencil font. (In fact – its vertical “stroke” was shortened. To recreate such a shape, I transformed the text into path and made appropriate modification). Then I exported from Inkscape the resulting color-decals.png picture and placed it as the reference in the source GIMP image, above the Camouflage layers. Finally I painted the darker background behind the radio-call number on a separate layer, as a new part of the camouflage image.

Using the USAAF font I am able to quickly recreate various service labels. In fact, most of them disappeared from this war-weary “black 4”. On the archival photo I can see only one label, on the life raft door (above left horizontal arm of the star – see figure "b", above). It is interesting to note how this detail appears on the restored aircraft:

As you can see, the labels on the restored aircraft are too large, and located in the wrong places. I recreated these elements in Inkscape, on a separate layer.

Quote

Restored aircraft can differ from the original in many details. In particular, their painting (the hue and the gloss of the camouflage, service labels fonts and sizes) leave much to be desired.

 

The last elements that I have to draw my “decals” texture are: the serial number (on the fin) and the model description (on the rudder). Unfortunately, the serial number is too small to be readable on the reference image, and the rudder is clipped out of its photo frame. All we can do is to use the photo of another aircraft from the same flight (as in figure "a" below):

Among the historical photos, I had only close shots of these numbers on some SBD-5s (figure "b", above). They showed me the proper font and size of these labels. For this aircraft I could only use a random serial number (figure "c", above). I chose one of the lesser ones from the two polls of the SBD-3 serial numbers (3185-3384 and 6492-6701). I suppose that the aircraft from this first pool were delivered before December 1941. On the historical photo (figure "a", above) of another aircraft from VSMB-241 you can still see the traces of the red and white stripes on the rudder, painted in December 1941. Then, in May 1942, the rudders were covered with the standard camouflage. You still can see these stripes behind the Blue Gray paint, because it was impossible to scratch the previous paint from their fabric skin. I reproduced this effect on my “decals” texture, drawing seven highly transparent stripes on the rudder.

Below you can see the final render of the “black 4”:

 

I also prepared for this model an alternate, sea environment:

I think that this picture of the Pacific Ocean creates a more familiar surroundings for such a naval aircraft. You can find the definition of this environment in the World tab of my Blender file. I named it Sea. Consequently, I renamed the previous environment to Land. You can easily switch between these two “worlds”.

In the next post I will work on the three-color Navy camouflage, used after January 1943 (you can find it mainly on the SBD-4s and SBD-5s). I will re-use in that new color scheme most elements from this two-color painting (the dirt texture, some of its weathering). Thus it will be a much quicker work.

In this source *.blend file you can evaluate yourself the current version of the model, and here are the Inkscape and GIMP source files of its textures. Because of the large size of the original GIMP file (*.xcf), this post is accompanied by its smaller version (2048x2048px), packed into *.zip file. I think that such a version is sufficient for checking all the details of this image (the structure of its layers, their opacities and mixing functions). The resulting textures (4096x4096) are packed into accompanying Blender file.

Edited July 22, 2017 by Witold Jaworski

[hendie]

stunning as always

[Witold Jaworski]

hendie: thank you for following!

Today I am describing my work on the tri-color Navy camouflage:

__________________________________________________________________

 

In my previous post I finished the case of so-called “two-color” U.S. Navy camouflage, which was used between September 1941 and January 1943. You can observe on the archival photos that its non-specular Sea Gray / Light Gray combination was especially prone to weathering, and accumulated every grain of the soot and drop of the oil stains. Simultaneously the weathered Sea Gray paint became more and more white. The new, “tri-color” camouflage, introduced in January 1943, fixed these flaws, and provided better protection on the vast, dark waters of the Pacific. You can see an example of this pattern on an SBD-5 from VB-16:
 

However, this historical photo has a technical flaw: its colors are “shifted toward blue”. You can unmistakably see this “shift” in the color of the bottom surface (it was Intermediate White). I was not able to correct this deviation, finding acceptable. Below you can see another photo of a SBD-5 from VSMB-231, which colors are more balanced:
 

There were two variations of the tri-color painting scheme. While the most probably the “white 35” from the first picture represents the painting applied in the factory, the photo below shows a case of another variation:
 

The main difference is the dark Sea Blue section below the cockpit. It is creating a “bridge” of the Sea Blue color between the upper areas of the wing and fuselage. Most probably such a camouflage was applied by the Navy workshops, when the older aircraft were repainted from the “two-color” scheme. Note that all of the SBD-5s on this photo have larger national insignia than the “white 35” from the first picture. Their stars have precisely the same size and location as those in the two-color scheme. (It seems that the workshops just painted the two rectangles on each side of an existing roundel). You can also encounter aircraft that had the “bridged” camouflage and the smaller (i.e. standard) insignia, but it seems that all aircraft without the “bridge” below the cockpit had the standard roundels. This fact seems to confirm the “workshop” hypothesis of the “bridged” camouflage origins. Many modelers think that this variant of the tri-color scheme was created in the main Navy overhaul facilities at Norfolk.

In this post I will recreate the “white 35” shown on the first picture. This particular aircraft belonged to VB-16 squadron from USS Lexington (CV-16), and was flown by Lt. (Jg) George T. Glacken, with RM Leo Boulanger at the rear gun. There is another close-up photo of this aircraft, most probably taken during the same flight (early April 1944, over New Guinea):
 

This SBD-5 seems to be n much better condition that the weary SBD-3 from my previous post. From the photos of the other VB-16 aircraft it seems that the crew of this squadron had enough time to take care of their machines. All of them had uniform squadron emblems, the flying staff names were painted below the cockpits, and every mission was marked with a small “bomb” on the fuselage.

Unlike on the SBD-3, on this SBD-5 the anti-slip strip ends at the main spar (there is no forward part, painted in the glossy black). There are no visible deep (“bare-metal”) scratches on the center wing upper surface. Just some irregular areas and a few seams of the dome rivets are brighter. Most probably the paint was scratched from the heads fo these rivets. (There is no such a thing in the front of the main spar, because its seams were made of the “flat”, countersunk rivets).

I started my work on this camouflage by creating a new copy of the previous source GIMP file (Color.xcf). Then I modified its contents by repainting some key layers. Finally I exported the resulting pictures, overwriting the existing images (texture components in the skin material of my model).

The first repainted elements were the basic layers of the camouflage (color-camo.jpg image) – as in figure "a", below. This is one of the three color texture components. I simultaneously modified the ref-specular.jpg component of the reflectivity map, providing the “gloss” to the dark Sea Blue surfaces (figure "b", below):
 

I left the weathering layers of the camouflage image intact (hey are the same as in my SBD-3). You can see the first test render of this new camouflage (combined with modified color-dirt.png image) below:
 

This first render revealed that while the non-gloss surfaces look quite convincing, I had an issue with the more glossy upper surfaces. The dirt pattern disappeared on the highlighted areas. They look unrealistic smooth and clean (like on a polished airliner!).

The remedy for this issue is yet another texture, which will “modulate” the color of the specular reflections, making some areas darker than the others. It is quite simple – a neutral gray background and just some darker splashes. I named it color-specular.jpg. Figure below shows this image and its place in the material schema:
 

 

Quote

I also could put these splashes on a white background. However, I did not know if I would need some lighter elements. That’s why I used a neutral gray here.

Figure below shows the test render of this updated material:
 

I reproduced the scratches on the center wing in the same way as in the two-color SBD-3: using the scratches mask (mask-scratches.jpg image):
 

In this case the only bare-metal spots are the heads of the dome rivets. I recreated them using an inverted copy of the reference image. I also added some partially scratched areas in the front of the anti-slip stripes (they “reach” just the primer color).

Finally I prepared the “decals” picture in Inkscape, then exported it to the color-decals.png file. Analyzing various photos of other aircraft from the same squadron, I determined that the serial number on the fin was black, and small radio-call numbers (“35”) were also painted on the wing upper surface. I repainted in GIMP the VB-16 emblem (it seems to be in the contemporary cartoon style). Then I exported to a *.png file, and placed it in the SVG source image as a linked picture:
 

Below you can see another test render, featuring the complete texture set:
 

In overall, the tri-color painting looks acceptable. However, this model badly needs the details: the cockpit interior, radial engine, crew… Thus, in this post I am finishing the third phase of this project (“working with textures and materials”). Now I am starting the last, fourth phase: detailing. For most of the small parts that I will create in this last phase, I will use simpler materials that do not require any UV-unwrapping and texture images. For example – on the picture above the propeller hub requires different material (in this “white 35” it seems to be painted in a glossy Sea Blue). At this moment I kept the hinges and canopy rails in the natural metal color. I will have to “repaint” them, using simpler versions of the camouflage colors. Finally, it seems that I have to improve the glass material of the cockpit canopies (comparing with the archival photos, they are too “clear”). Anyway, I will describe my solutions to all these issues in the future posts.

In this source *.blend file you can evaluate yourself the current version of the model, and here are the Inkscape and GIMP source files of its textures. Because of the large size of the original GIMP file (*.xcf), this post is accompanied by its smaller version (2048x2048px), packed into *.zip file. I think that such a version is sufficient for checking all the details of this image (the structure of its layers, their opacities and mixing functions). The resulting textures (4096x4096) are packed into accompanying Blender file.

[hendie]

excellent work there. I'm looking forward to seeing the details being added in

[Bandsaw Steve]

This is awesome - and no-one has to breathe any sawdust!😀

[Witold Jaworski]

hendie, Bandsaw Steve - thank you for following this thread!

 

_____________________________________

 

I published my previous post a month ago, but the current stage of this project – detailing – requires less frequent reports. (Otherwise the posts would become rather monotonous: week after week they would describe making similar things, using the same methods). I started this last phase of the Dauntless project by recreating its main landing gear. First, I had to finish it, then I am able to write about this process. Thus I will describe it in this and next two posts. (I will publish them in a short sequence, week after week).

The retractable main landing gear of the SBD was probably a direct descendant of an experimental solution used in the Northrop 3A fighter prototype. In general, it looks quite simple:
 

The upper part of the landing gear was an “L” – shaped tube, mounted between two wing spars. The lower part, visible below the wing, was a simple shock strut mounted to the wheel axle (see figure "a", below). The axis of the landing gear retraction was parallel to the thrust line and perpendicular to the walls of the spars (see figure above). The shock strut is deflected (by 6⁰) from the vertical axis, so that in the open position the wheel is directly below the axis of landing gear retraction (see figure "a", below):
 

Figure above also shows various treads of the SBD tires. The tires of the earlier versions (SBD-1, -2 and -3) had no tread pattern (figure "b", above). The simple “straight grooves” treads appeared on the SBD-4 wheels (figure "a", above), while in the SBD-5/6s we can find a more elaborate, “brick” (figure "c", above) or “honeycomb” tread patterns.

Another interesting thing is the lack of the torque arms, that connect the cylinder and piston of the shock strut in most of the other aircraft (figure "a", below):
 

The SBD manual explains that the designers used splined cylinders and pistons in their shock strut (figure "b", above). It looks like a quite elegant solution for the blocking random torsions of the shock strut piston (less outer parts that are prone to the eventual dust and jams). I did not find any complaints for this landing gear in the veteran memoirs and technical reports (usually they praise the “rugged structure” of the SBDs). However, all other aircraft designs use the torqe arms in their landing gear. Maybe they were just cheaper (i.e. easier to produce)?

Basically the work on the details means that you have recreate “in the mesh” all the parts you can see on the photos. Below you can see how I recreated the upper part of the landing gear leg:
 

Recreation of such a die-cast part, with all of its additional walls, roundings, is a small challenge. To make it with as simple mesh as possible, I used several modifiers. First, I used the Mirror modifier to automatically generate the symmetric half of this object. (This symmetry was only possible because I decided to split this “L” – shaped part into two objects: this complex die-cast and a simple tube behind it. (This tube is not present in the picture above). Then I recreated all the rounded edges on this object using dynamically-generated fillets. Another modifier (Bevel) creates fillets along all the edges that I assigned a nonzero bevel weight. (The fillet radius is controlled by the value of this weight).

When this upper part of the landing gear leg was finished, I created the cylinder of the shock strut. It was just a tube with an octagonal flange *– nothing difficult. Then I had to create the lower part of the leg:
 

It was another die-cast, which shaping required some time. (As you can see in the figure above, I formed this mesh from two crossing tubes).

While creating such a detailed assembly, I prefer to model each of its parts as a separate object. It gives me the opportunity to take advantage of its local coordinate system, when I need it. For example – the shocking strut is a tube rotated by 6⁰. When I formed it, I often extruded its faces along this local axis. Another advantage of such a model structure is the possibility of quick, “natural” adjustments of various parts. (For example – piston movement along the cylinder. In my model it occurs along its local Y axis).

Building the landing gear, I tried to check its retracted position as early as possible. Figure below shows first of these trials:
 

As usual, a small fragment of the retracted landing gear leg protruded from the upper wing surface. I had to re-examine the photos, find which part has the wrong shape, and fix it.

I also used my photo references to recreate other landing gear elements, like the wheel brake disks:
 

During this work I also found some differences between my model and the reference photo (see the notes in blue frames in the figure above). It seems that my landing gear is somewhat shifted forward.

Such a finding led to many rearrangements in the geometry of this assembly:
 

In the background of figure "a", above) you can see my original drawing from 2015. After some deliberations, I decided to leave the center of the wheels at its current location, because it was dimensioned on the original general arrangement drawing. However, after measuring tire proportions on various photo, I decided that the SBD used slightly wider tires (30x7.5”) than the size (30x7”) specified in one of the comments placed on the original drawing from the SBD manual. (Sometimes draughtsman could make such a mistake). To fit the photo, I adjusted location of the shock strut, moving it slightly toward the fuselage. I also shifted downward the axis of retraction (rotation). Figure "b", above, shows how the updated model fits the reference photo. The tires on the photo still seems somewhat smaller than those from my model. However, I decided that this restored CAF aircraft could use a slightly smaller tires (29x7.5”). (This particular SBD-5 also uses at least another non-original part: a different version of its Hamilton Standard propeller).

Another element that requires some adjustments are landing gear covers. Although I created them during the modeling phase, now I have to compare their shape with the reference photo. Preparing for this test, I placed in my model a simple “stick” which I use as the hinge (I set it as the parent of the landing gear cover):
 

The hard part was to determine the proper axis direction and the angle of this rotation. Surprisingly, the hinge was not lying directly on the aircraft skin (it would be the simplest solution). While it was relatively easy to find for a given hinge orientation a rotation which angle placed the left cover as on the photo, the same rotation applied to the right cover did not match the reference. It required some hours to find a combination that produced an acceptable (although not ideal!) match.

As you can see, I performed a lot of various checking, adjusting and matching while forming the key elements of the landing gear. All of this because it is still quite easy to correct the geometry of this assembly while it is relatively simple. It would be a nightmare, when I did such a thing on the final, detailed assembly, which you can see in figure below:
 

However, before I finished this landing gear in such a state depicted in the figure above, I had to create several dozen bolts of various size, as well as other details. I also discovered some small secrets of its retraction mechanism. You will find a short description of all of these findings in my next post (to be published next week).

I decided to not enclose the source Blender file to this post, because it would contain just these few basic landing gear components. I will add it to the next post, which describes this assembly in the finished state.

 

[Witold Jaworski]

The SBD shock absorbers had to disperse a lot of the kinetic energy of landing aircraft, minimizing the chance that the airplane accidentally “bounce” back into the air. (This is a key requirement for the carrier-based planes). For such a characteristics you need a relatively long working span between the free (i.e. unloaded) and the completely compressed (i.e. under max. load) strut piston positions. Indeed, you can observe that the Dauntless landing gear legs are much longer in the flight than in their static position on the ground:
 

The working span of the SBD shock strut piston was about 10” long, while the difference between the static and the free (extended) piston positions was about 7.5”.

In the most of the aircraft the landing gear retracts with the shock struts fully extended. When I tried to do such a thing with the Dauntless landing gear, as in the figure above, I discovered that it definitely does not fit its recess in the wing! (see the figure below):
 

It seems that there was something that compressed SBD landing gear during retraction by about 7”. In the case of the Republic P-47 (which landing gear leg shortened during retraction by 9”), such a thing was widely discussed as an exceptional achievement of its designers. However, nobody even mentioned that the same issue was already resolved several years earlier by the Northrop/Douglas SBD team.

The SBD landing gear leg was made shorter during retraction by compressing its shock strut piston. This “compressing” mechanism starts with the cable that pulls the piston upward. The cable is attached to a quadrant, which rotation is controlled by the rail, fixed to the wing spar:
 

An additional spiral spring, attached to this quadrant (as in figure above), tightens the cable when the landing gear is extended. In this landing gear position the spring “freely” rotates the quadrant, just following the shock strut piston movements:
 

(You can also see how it works in this short video sequence). Note that this spring ensures that in the fully extended (free) position of the landing gear the tip of the quadrant arm nearly “touches” the rail.

This is the starting position for the piston compression. During the retraction the quadrant rotation axis is elevated upward, while its arm is dragged along the rail (there is a small roller on its tip). In the effect, the quadrant rotates, pulling the cable that compresses the shock strut piston:
 

You can also see how it works in this short video sequence. (Note that in this video the path of the quadrant arm tip is not perfect. This is the result of a relatively simple real-time rigging. Anyway, it gives the general idea how this mechanism works).

I also recreated the inner details of the landing gear recess:
 

During the modeling phase I already recreated two spars and two main ribs in the wheel bay. Now I refined their shape, following the available photo reference. I added also the remaining ribs as well as the stringers and the covers (at the rear part of the bay).

As you can see in the picture above, I did not model the lightening holes: as in the case of the wing flaps, I recreated them using textures:
 

I unwrapped meshes of all these new wing structure elements into the available space on the UV layouts that I used in the outer skin material (B.Skin.Camouflage, refined in my previous posts). The Navy painted the inner space of this wing recess in the same color as the wing bottom surfaces, so there was not any problem in assigning the B.Skin.Camouflage material to these objects. As you can see in the picture above, I also recreated the internal reinforcements of the landing gear cover. The inner side of the aircraft skin is covered with a generic material, named B.Skin.Details.Bottom. (It is assigned to the second material slot of the wing skin mesh, and set in the Material Index Offset field of its Solidify modifier). This simpler material is intended for the smaller details, and uses exclusively the generic, procedural “noise” textures for the dirt/ref effects. Thus it does not require any time-consuming UV unwrapping. In the figure above I used it also in the cover fittings. I only have to remember to alter the basic color of this simpler material when I switch to another camouflage scheme. (In the future, I will also create a similar material for the details on the upper surfaces – it will only differ in the basic color).

Because this B.Skin.Details.Bottom material has no regular bump map, I had to recreate more landing gear details “in the mesh”. In particular, I had to add the bolts (and nuts) visible on the reference photos. I created each of them as a separate object:
 

All these bolts are clones of a few basic original bolt meshes (one with the octagonal head, others with the flat one). Among these originals there are also two or three variations of the bolt length. By default these bolt meshes are covered with a generic “steel” material. When I needed to “paint” them into different color, I alter their material assignment. (In such a case, I had to switch them into the object – based material mode).

Quote

Sometimes some of these bolt objects are also useful as the reference objects in the rigging, but I will discuss it in the next post.

I also like to have a look at the retracted landing gear inside the wing structure. When you can see these two assemblies together, suddenly the reason for some strange features of the particular rib or spar shape becomes clear:
 

In figure above you can also see a couple of bolts forming an octagonal group on the rear spar. In general, I recreated most of the bolts and rivets on this spar using the bump texture. However, this was a special case: in the original airplane these bolts had quite large heads. What’s more, they were placed on a “stack” of two subsequent panels. I simply run out of the available grayscale of the bump texture in this particular place, thus I decided to recreate these “topmost” details in the mesh. (Just an exception from my general modeling tactics of using the textures as often as possible).

Figure below shows the complete landing gear in the open position, the shock strut fully extended:
 

The decision to use a simpler materials that do not require UV-mapping has certain disadvantages. The most important of them is that I am not able to paint the small stains around the landing gear bolts and other kinds of the local dirt. But this is just the consequence of the level of details that I assumed for this model.

Quote

I have to reveal that this project has a certain deadline: I promised a local modelers’ magazine to deliver detailed SBD scale plans and a couple of color profiles in January 2018. Thus currently I am focusing on the external details, because I have to recreate them to the end of this year. It includes the landing gear and a simpler version of the engine (not intended for the extreme close-ups). When it will be done, later in 2018 I will recreate the cockpit interior, as well as the detailed engine compartment and wing flaps mechanism. This more enhanced version of this model will be intended for a SBD monograph (a book). I will publish it much later.

Among the materials that I have to deliver in winter 2018 I will not enclose any close-up pictures of the landing gear or other details, thus I can use the simpler materials here. However, in such a harsh light as in Figure 80‑10 the elements covered by the B.Skin.Details.Bottom material seem too “clean”. I will definitely work on this issue, increasing the contrast of its “noise” textures to give these landing leg and brake disk a more “dirty” look.

In the next post I will describe how I rigged this landing gear. (I will describe building a kind of “virtual mechanism” that allows me to extend/retract the undercarriage using a single slider. I already used it, making the video sequences presented in this post).

In this source *.blend file you can evaluate yourself the current (i.e. non-rigged) version of this model.

[hendie]

nice explanation of how the landing gear issue was solved.   simple but effective

[Witold Jaworski]

In previous post I discussed how the SBD landing gear retracts into its wing recess:
 

In principle, it is simple: the landing gear leg rotates by 90⁰. However, the parts responsible for shock strut shortening during this movement increase mechanical complexity of this assembly. The figure above does not even show the deformations of the brake cable, which follows the shock strut piston movements.

For some scenes I will need the landing gear extended, while for the others – retracted. In practice, moving/rotating each part individually to “pose” my model would be a quite time-consuming task. That’s why I created a kind of “virtual mechanism”, which allows me to retract/extend the landing gear with a single mouse movement. In the previous post I already presented its results in this short video sequence. In this post I will shortly describe how I did it.

In general, I coupled some key elements (objects) of the landing gear using so-called constraints. For example, I connected the rotation of the landing gear leg with the movement of a special “handle” object. To do it, I used a Transform constraint, attached to the parent object (the axle of the retraction) of the landing gear leg:
 

I created an auxiliary (non-rendered) “handle” object (X.600.Wheel.Handle). The Transform constraint of the landing gear leg axle (0.604.Strut.Axis.L object) converts the handle linear movement into landing gear rotation. Thus when I shift the handle object upward, landing gear leg rotates, retracting into its place in the wing. To restrict the range of this rotation, I assigned to the handle object additional Limit Location constraint. It restrict its possible movement to a 40-unit long span along local Z axis.

The more detailed explanation of my methods for the “mechanization” of the landing gear would take too much space in this post. However, some years ago I published an article on this subject (in “Blender Art Magazine”):
 

To read this article, click the picture above or this link. I hope that this publication will explain you the general idea and typical implementation of such a “virtual mechanisms”.

The format of my posts (10-11 pictures) allows just for a quick review of the SBD landing gear constraints (one picture per a subassembly). Thus the next element coupled with the handle object (using another Transform constraint) is the landing leg cover:
 

In the previous post I added many bolt objects to the landing gear, and mentioned that some of them will have an additional use. So this is just such a case: I set the forward bolt of the cover hinge as the parent of the whole cover assembly. (Because it lies on its rotation axis). A Transform constraint, assigned to this bolt, forces it to rotate in response to the handle vertical movements. Note that the range of rotation of this cover (101⁰) is greater than the range of the landing gear leg rotation (90⁰).

Another Transform constraint converts the rotation of the quadrant object into movements of the shock strut piston:
 

This relationship seems quite straightforward: when the quadrant rotates upward, the piston shifts up, when it rotates downward, the piston shifts down – as if they really were connected by the cable. Rotation of the quadrant is forced by its Locked Track constraint, which arm “tracks” the auxiliary (non-rendered) target object (the red, small circle in the figure above). Effectively, location of this quadrant target controls the shock strut position.

The full motion path of the quadrant target object contains an arc and a straight segment:
 

The arc corresponds to possible piston positions for the extended landing gear (see this short video sequence). The linear segment corresponds to the forced compression of the shock strut during retraction, when the quadrant arm tip slides along the internal rail. (You can see this motion here, although in this video the path of the quadrant arm tip is not ideal).

I did not want to use the animation motion path here, because it creates a “deterministic” movement (“frame by frame”). Instead, I wanted a general solution, controlled by a handle object instead of the animation frames. Thus this is the most complex subassembly in this landing gear rig. I will describe it in two pictures: one for the extended landing gear (implementation of the movement along the arc), the other for the landing gear retraction (movement along the linear segment).

When the landing gear is extended, the rotation of the quadrant target is forced by its “grandparent” object. This is an Empty instance, named after the source of the original movement it mimics: X.Quadrant.Spring.Control:
 

A Transform constraint converts the vertical movement of the additional handle object (X.600.Strut.Handle) into rotation between -24.7⁰ and +118⁰. There is another Empty object (0.607.Target.Left.Parent), attached (by the parent relation) to the X.Quadrant.Spring.Control. Simultaneously, 0.607.Target.Left.Parent is the parent of the quadrant target object. (The reason for such an indirect relationship will become clear in the next figure). When the landing gear is extended, this chain of “parent” relations forces the quadrant target object to move along the arc path when the handle object moves up and down. To not exceed the minimum and maximum angles of this movement (and in the effect – the lowest and highest shock strut piston position), the location of the handle object is restricted by a Limit Location constraint.

Note that this smaller handle is the child of the main handle, used for the landing gear retraction (X.600.Wheel.Handle). Note also that the Transform constraint that forces this rotation, evaluates the handle position in the World Space (figure above, bottom right). This means that regardless of the position of the smaller handle, the shock strut will be fully extended after moving the main handle along the first few units along its way up. (So that I do not have to care about the initial piston position when I am starting landing gear retraction: it will set up itself).

During landing gear retraction the locations of the quadrant target, its parent, and “grandparent” become dispersed. Figure below shows how it looks like in the middle of the main handle (X.600.Wheel.Handle) movement:
 

The parent (0.607.Target.Left.Parent) of the quadrant target object just “delivers” it to the rail line and stops there. (This is the end point of the rotation of its parent: X.Quadrant.Spring.Control object, as you can see in the previous figure). Then the quadrant target object is “dragged” by the landing gear retraction along the rail. I tried to obtain this effect by assigning it two constraints:

1. Limit distance, which forces a fixed distance between the target object and the quadrant axis;
2. Limit location, which restrict the possible location of the target object just to a certain span along its parent X axis (which I set parallel along the rail);

In theory, for such a constraint combination there is always just a single possible location of the target object (marked in the figure above). Unfortunately, it seems that Blender treats these constraint with certain “flexibility”. In the effect, the quadrant arm tip “sinks” into the rail. This effect is especially visible at the beginning of the landing gear retraction. Well, I tried hard but I could not find a better solution for this movement. Finally I concluded that I can left it in this state: this is a small part, which movements are partially obscured by the wing recess. When I need to make a close up, I can prepare a “deterministic” animation motion path for the quadrant target object.

The last rigged subassembly of this landing gear was the elastic brake cable. Because I used in this implementation another (better) solution than described in my article and book, I will discuss it shortly below
 

First, as in my previous model, I formed the curve (figure "a", above), that deforms the simple tubular mesh into the brake cable. (I used a Curve Deform modifier here). Then I created a simple armature consisting two bones: upper and lower (figure "b", above). The armature object is attached (by a parent relation) to the strut cylinder. Thus the origin of the upper bone is also fixed in this way. I assigned to the lower bone an Inverse Kinematics constraint, and set its Target to an auxiliary Empty object at its end. This target object is indirectly (via the brake disk) attached to the piston. In the effect, this armature follows every piston movement in a natural way – like a pair of the torque scissors. Finally I attached subsequent curve vertices (control points) to the nearest armature bone (figure "c", above). I also attached (via parent relation) the ankle object (at the bottom end of the brake) to the bottom bone. (So that it will follow its rotation). In the effect, the brake cable deforms when the shock strut piston is shifting, following its movement in a natural way. See this short video how it works.

In my previous model (the P-40B) I used for the same purpose a different, less effective combination of the constraints. I think that in the future I will always use the solution as in figure above, not only for the brake cables, but also for the torque scissors (the SBD landing gear did not have such an element).

Finally, I applied my Handle Panel add-on to create a convenient landing gear controls among the Blender panels:
 

See this short video how it works. Now I can extend/retract the landing gear with a single mouse movement. I even do not have to know, where their handle objects are – the add-on will discover them automatically. (I just have to arrange them in the model space in a certain way – see the add-on description for details. This is a general utility, you can use it in your own Blender models).

The next part to recreate is the tail wheel assembly. I will report this step within two or three weeks.

In this source *.blend file you can evaluate yourself the current version of this model.

[Witold Jaworski]

The Dauntless had fixed tail wheel of a typical design among the carrier-based aircraft. The tail wheel assembly consisted a fork connected to two solid-made beams, which movement was countered by a shock strut. The beams and the shock strut were attached to the last bulkhead of the fuselage:



The bottom part of this assembly was covered by a guard and a fairing. Both of these elements were attached to the lower beam. The archival photos reveal that the bulky fairing was often removed:



There were two tail wheel versions: the smaller, solid-rubber wheel for the carrier-based aircraft (as in figures above), and the larger, pneumatic wheel for the ground-based aircraft. As you can see on the example of a SBD-1 (below), they could be replaced in a workshop:



Figure above shows two SBD-1s, which were exclusively used by USMC squadrons in the continental United States. All of these 57 aircraft were built in 1940. As you can see in the photo above, they were originally delivered with small, solid tires. However, in the photo below, taken a year later, one of these SBD-1s has the large, pneumatic tail wheel. It seems that Douglas delivered these large wheels to the Navy / USMC as an optional kit, which could be mounted when needed.

The smooth tip of the tail was mounted to the last bulkhead as a light, easily detachable tail cone. It covered the tail wheel shock strut, as well as the rudder and elevator control mechanism. There was a large opening in the bottom of the tail cone. For certain sunlight directions (for example, in a dive) you could “look inside” the fuselage:



As you can see in the figure on the left, the tail wheel fairing effectively closed most of this view (figure "a", above), so that you could see just the catapult holdback fitting, two ribs and their stringers. However, when you remove this fairing, you can see many details inside (figure "b", above). Of course, they are not visible on most of the photos. These details are hidden in the pictures taken on the ground. In most of the in-flight photos this opening seems to be too small to reveal the interior details. However, thinking about the future scenes of diving bombers, I decided to recreate at least the key internal elements of this tail cone.

While the tail wheel in the SBD does not retract, it follows the shock strut compression. Thus recreation of this assembly requires some initial adjustments of these moving parts. (For example: I had to make sure that the wheel leg will fit into the tail cone). I am shortly describing this process below:



I started with the tire, its fork and the guard, which are present on my reference photo (figure "a", above). I also added the stringers along the border of the opening in the tail cone. Then I placed the simplified shapes of the tail wheel beams. At the beginning I did not know their proper width, thus their initial mesh was a simple, four-vertex trapeze, ready for eventual adjustments. In figure "b", above, you can see also a bolt at the end of each beam. They were important parts of this “virtual mechanism”. No armature was needed in this case. I decided that the lower beam will be animated by the handle movement. Thus I forced the upper beam to follow its rotation using a Copy Rotation constraint. Each of these two bolts is attached (by parent relation) to the corresponding beam. The tail wheel fork is attached (by parent relation) to the lower bolt. Simultaneously this bolt rotates, tracking the center of the upper bolt (using a Locked Track constraint). Thus, when I rotate the lower beam, it rotates the whole assembly (as in figure "c", above). Then I adjusted the width of the guard and the beams so that in the fully deflected position it fits the opening in the tail cone (figure "d", above).

Once the overall size of the beams was determined, I recreated their shapes, as well as the catapult hold back fitting, shock strut, and the tail wheel fairing:



The catapult fitting was attached to the ribs of the tail cone. Its side arms were fitted between the tail wheel fairing and the stringers running around the border in the tail cone opening. As you can see in figure "a", above, I started with the conceptual lines of this part. Then I used them to form the final “Y” – shaped fitting (figure "b", above). Using similar methods I recreated the beam details. (I could do safely, because they were already fitted to the tail cone, and I do not expect any further changes in their shape). Finally I also added the internal ribs of the cone (figure "c", above).

In the next step I recreated the rudder/elevator torque tubes and their fittings:



Figure "a", above, shows the rudder torque tube. It was mounted on a tripod. Note, that the rudder rotation axis lies in the front of this tube. (It rotated around the bolt that joined the torque tube with the tripod). For this level of detail I did not recreate the control cables. (Perhaps I will do it in the future). I also formed the elevator torque tube (figure "b", above). Its rotation axis also lies in the front of the tube. According the maintenance manual, there were also two pulleys for the elevator trim tab cables (figure "c", above). Unfortunately, I do not have any precise photo of this detail. I recreated their brackets using two rough sketches from the maintenance manual, but about 50% of their size and shape is just my guess. If any of you have a photo of the details from figure "c") (for example – from a restored SBD), let me know. I would appreciate any of such pictures.

Figure below shows all of the tail cone internal details that I have recreated so far:



For the assumed level of details I simplified some small elements, like the trim tab pulleys. I also did not recreated other minor, less visible elements like control cables, electric cables, some additional fittings and bolts. I will recreate them later, when I decide to make cutaway pictures of this model.

To recreate the alternate, “solid” version of the tail wheel, I switched to another reference photo:



The SBD on this reference photo was restored for Pacific Aviation Museum Pearl Harbor. Although the restoration teams do their best to recreate their “birds”, sometimes you can find a non-original part on their aircraft. In this case it was the tail wheel fairing: it is smaller and simpler (figure "a", above) than the original part (figure "b", above). However, the rest of the tail wheel assembly seems to be original, thus I used this photo as the reference for the “carrier-deck” version of the tail wheel and its fork.

Finally I also added the tail hook (it was removed only from the A-24s, delivered to the Army). As you can see, it was accompanied by some minor details:



Figure below shows the final result: the complete tail wheel and hook assemblies:



As you can see, I initially painted the internal surfaces using the standard Interior Green color, but then I was starting to have doubts. In the restored aircraft these surfaces are usually painted in the gray camouflage color (the same as used for the aircraft underside). Unfortunately, I did not have any historical photo of this area to determine how it was painted in the original SBDs. Most probably I will update the color of this detail according the restored aircraft.

The next part I am going to recreate is the R-1820 engine (a simplified version, for the external pictures). However, for reasons beyond my control I have to take a break from this project. I will write next post in the spring (2018). So, for time being my SBD will look like this:



In this source *.blend file you can evaluate the model.

 

[NickD]

Excellent as always. I am just a dabbler in comparison.

 

ATB Nick

[Ex-FAAWAFU]

I fully understand about ⅓ of this - but it is 100% fascinating

[Witold Jaworski]

NickD, Ex-FAAWAFU: thank you!
After the long winter, finally we have fresh spring, and I am back at my old desk!
I started working on the most complex part: the Wright R-1820:
__________________________________________________ __________________________

The engine is the heart of every powered aircraft. In the case of the SBD it was the Wright R-1820 “Cyclone 9” (the “G“ model). In fact, this engine was one of the “workhorses” of the 1930s: designed in 1931, it was used in many aircraft, especially in the legendary DC-3. “Cyclone” was a reliable, fuel-saving unit for the Navy basic scout type. (Remember that the “Dauntless” was not only the bomber: it was also a scout airplane^[1]). In general, the R-1820 is a classic nine-cylinder, single-row radial engine:



The R-1820 G had been produced for over two decades, not only by the Curtiss-Wright, but also (under license) by Lycoming, Pratt & Whitney Canada, and Studebaker Corporation. Thus various less important details of this engine “evolved” during this period. In this post I would like to highlight some of these differences. I will focus on the forward part of this engine, because at this moment I am going to create a simpler model of the “Cyclone”, intended for the general, “outdoor” scenes. Inside the closed NACA cowling, you can see only its forward part. (Thanks to the air deflectors, placed between the cylinders - see picture above). In such an arrangement, the visible elements are: the front section of the crankcase, cylinders, ignition harness, and the variable-pitch propeller governor. While the front section of the R-1820 crankcase remained practically unchanged in all versions, and the governor depends on the propeller model, I could focus on the cylinders and their ignition harness.

Identification of the version differences is the basic step, because otherwise you can build a model of non-existing object that incorporates features from different engine variants.

Quote

BTW: do you know, that the R-1820 design had remarkably long life? The United States factories produced the last batch of these engines in 1964. The metric version of the earlier “F” model had been produced in Soviet Union under Wright’s license since 1934. A few years later Soviet engineers developed its enhanced version: Ash-62 (resembling the “G” model of the “Cyclone”). Ash-62 was widely used in 20th century aircraft of the former eastern block (especially – in the popular Antonov An-2), and had been produced under Soviet license in many countries. Actually the last factory that still produces these engines is PZL WSK-Kalisz in Poland. They provide new units for the last flying DC-3s, An-2s, and M-18s, as well as the overhauling services. Who knows, if this “eastern branch” of the R-1820 will last long enough to celebrate the 100th anniversary of the famous Wright design?

While looking for the reference materials, I have also found an interesting article about the development of air-cooled aviation engines (more precisely, their most important parts: cylinders). I think that it provides a valuable “technical context” for the visual differences that I am describing below.

Searching for the reference photos, I have identified two basic variations of the “Cyclone” cylinder shape:

Figure "a" above shows the classic version, produced to the end of the WW2, while the cylinder from Figure "b" comes from the post-war production. I will refer this earlier one as the “classic” version. This is the engine used in all SBDs. You can quickly identify this version by the characteristic “L”-shaped fins on its cylinder head (Figure "a"). The “classic” head has also curved contours, while the head of the post-war version has different style, and its contour is based on the straight lines. Both heads are aluminum die-casts. The critical element in this design was the overall area of their fins. Greater cooling area of the cylinder head allows you to obtain more power from the same piston volume. Thus the fins of the “classic” head are small wonders of the 1930s metallurgy: they are evenly spaced at 0.2” (5mm) along the head, and the widths of their tips do not exceed 0.05” (1.2mm). The fin at its base is about 0.1” (2.5mm) wide. Die-casting of such an object is extremely difficult. It requires not only the “written down” engineering knowledge, but also individual artisanship of the key workers. Note that the spaces between the fins of the post-war head are two times wider than in the “classic” version. However, between each pair of these “full-size” fins there is a smaller, much shorter “inner” fin. It is much easier to die-cast such a head. I suppose that the post-war heads are cast from an aluminum alloy that has better heat transfer characteristics. It would allow their cylinders to maintain similar power output using somewhat smaller cooling area.

The cylinders of the last R-1820 versions had yet another, conical shape:

In this photo you can also see here the propeller governor (in the first photo in this post it is hidden behind the propeller blade), and another version of the ignition harness.

The “classic” and the “post-war” cylinder heads have different orientations of their intake openings, which results in different shape of the intake pipes:

The classic version has a simpler, L-shaped intake pipe, which fits to the oblique opening of the intake valve (Figure "a", above). In the post-war version planes of both valve openings (exhaust and intake) are parallel (Figure "b", above), thus the intake pipe has a more complex shape (resembling “S”).

Quote

In fact, Figure "b" above shows the smaller, 7-cylinder version of the Wright Cyclone (R-1300). However, it used the same cylinder and intake ducts as the late R-1820. (I just could not find a shot similar to the Figure "a" of the late R-1820 version, so I used the picture of its “smaller brother” instead).

There are also minor differences in the rocker covers:

The classic version has simpler, four-bolt rocker cover (Figure "a", above), while the post-war covers uses two bolts more. The head of the post-war engine has some additional features (Figure "b", above), which do not exist in the classic head.

Finally, the ignition harness:

Classic ignition harness has a “collar” shape, smaller radius, and individual spark plug cables organized in pairs (Figure "a"). Post-war harness has a ring shape, somewhat greater radius, and evenly spaced spark plug cables (Figure "b"). Although each of these photos is taken from different side, it seems that both versions use the same propeller governor.

Having all these issues identified, I could select appropriate reference drawings. They came from “Cyclone 9GC Overhaul Manual”, published in 1943. I expect that even the simplified model of such an engine can have many hundred thousand faces, thus I decided to build it in a separate Blender file. I use the same “scale” as in the SBD model: 1 unit = 1 inch. When it is ready, I will import it into the SBD model.

In this new Blender file I decided to give chance to the alternate method of setting up the blueprints: using Empty objects with the attached image:

First I placed on the perpendicular planes the four views of the original installation drawings (Figure "a", above). Note that they contain a lot of the explicit dimension values – such information is an invaluable help in recreating this engine.

I quickly realized that the Empty objects with the reference images allow you to use simultaneously several alternate sets of the blueprints. Just place each of them on a separate layer. It will be a great tool in the Blender 2.8, which has to have unlimited number of layers. While working in the actual Blender 2.7, I placed these planes on layers 7…10, practically reserving them for the reference pictures. The second blueprint set contains the images from the original “Limits and Lubrication Chart” (Figure "b", above). These two views (side and rear view) are much more detailed than the installation drawings (presented in Figure "a", above). Of course, these images do not match each other in a perfect way: there always are some differences. However, I did not fix them, as in the case of the SBD planes, because all the key dimensions of this engine are specified in the installation drawings. I will just use these explicit values.

Following the standard of my posts, I am enclosing the current state of the source *.blend file. While there is no model, yet, you look inside to check the arrangement of the reference pictures. Next week I will report the first stages of building this model: forming the central crankcase and the basic cylinder shape. (Cylinders of this engine are identical with each other. Once you prepare one of them, you can quickly “populate” the crankcase with its eight clones. However, as you will see in the next posts, the die-cast, air-cooled cylinder head is one of the most complex objects to model…).

---

^{[1]The SBD Dauntless was a new implementation of the US Navy carrier doctrine, worked out in the preceding decade: in the clash of the carriers always wins those, who first finds carriers of their opponent. In fact, the best option was to find, report, and immediately make the first attack – that’s why all SBDs carried a 500-pound bomb on their scout missions}

Edited April 18, 2018 by Witold Jaworski

[Witold Jaworski]

In this post I will recreate the main and the front sections of the R-1820 crankcase, and the cylinder basic shape. Let’s start this model by forming the main crankcase:

This section is always obscured by the cylinders, so you cannot see it clearly on any photo. That’s why I used here the original drawing from the manual. Generally, this barrel-like shape contains nine cylinder bases. It is formed by two steel castings, bolted to each other. (These bolts are hidden inside the crankcase, between the cylinder openings).

It is always a good idea to start with a simplified model. It allows us to check all constrains of the geometry that are not obvious at the first glance from the reference drawings. In this case started by forming a symmetric half of the crankcase:
 

This is a simple barrel, smoothed with a Subdivision Surface modifier. Then I placed the flat piston bases along the circumference of this crankcase. I quickly realized that the side contour of this barrel depends entirely on the size and shape of these piston bases. After a few quick adjustments of the control edge loops, the barrel surface “touched” the outer edges of the piston bases along their whole length (as in figure above).

Note that these piston bases are so tightly packed around the crankcase, that they nearly join each other along a short, straight edge:

This means, that the crankcase barrel contains a cylindrical strip in the middle, which matches this straight edges on the piston bases. In fact, the sharp corners of these edges forced similar sharp edge on the barrel side contour.

When the general shape of the crankcase barrel looked right, it was time to create the final mesh. I decided that I will not use the dynamic effects of the subdivision surfaces for such a complex objects as the engine parts. (Because I want to keep the polygon count of this engine model below 1 million). Thus I “fixed” this subdivision effect, converting it into the normal faces (by “applying” the modifier). Then I took take the advantage of the “repeatability” of this shape. I deleted all the faces of the original “barrel”, leaving just the 20⁰ “slice” (as in figure "a", below):
 

The opposite 20⁰ of this “slice” is generated by the Mirror modifier. Then I made further modification to this mesh, removing all the faces from above the piston base (figure "b", above). I also copied and inserted into this mesh a quarter of the piston base contour. Then I started to join this contour and the mesh around it with new faces. You can see the result in figure "a", below):
 

 

As you can see, I also recreated the rear part of this crankcase section, just adding another symmetry axis to its Mirror modifier. The whole body of this crankcase can be built from 9 clones of such an object (as you can see in figure "b", above).

The shading of the crankcase faces is set as Smooth, except the faces around the piston base (which are marked as Flat).

I would like to mention a little “trick”, which can be useful in many other cases. To obtain a seamless join between the crankcase “slices” (as in figure "a", below), I added an additional, thin “strip” of the faces around the slice edge. These faces are parallel to the faces of similar strip in the adjacent slice (as in figures "b" and "c", below):

 

Once the middle section of the crankcase is ready, I started working on its front section. Generally speaking, this part looks like a combination of a cone and a cylinder, with many “protrusions” of additional details:
 

Actually, I recreated the basic shape of this section. (I will recreate the remaining details later). I did it using the same workflow as in the case of the previous section. First, I made a simple, “conceptual” model of this part. It was smoothed using a Subdivision Surface modifier. When the shape seemed to be OK, I converted the result of this modifier into normal mesh faces. Then I removed all the unnecessary edge loops and created the basic 20⁰ “slice” of this section:
 

To obtain the smooth shading between slices, I also created the additional thin strips of parallel faces along their adjacent edges. The basic slice of this section was easier to form than the one from the middle section, because it did not contain any opening.

Just to make the front of the engine more complete, I created the front disk (in a classic way, no “slicing” here) and the propeller shaft.

Finally I started working on the cylinder. Because all cylinders of this engine are uniform, I will complete a single (the topmost) one. The complete cylinder will be an assembly of many objects. Then, when it is finished, I will clone it around the crankcase.

As the first object in this assembly I created the simple, basic cylinder (i.e. the cylinder and its head without the fins and rocker covers):
 

It will be the parent object of all further elements of the cylinder assembly. As in the case of the crankcase, it does not use any Subdivision Surface modifier, just the “fixed” mesh faces with the Shade Smooth option (and Mirror and Bevel modifiers).

You can check details of this model in the source *.blend file. In the next post I will model the rocker covers and the covers of the intake/exhaust valves. (In the R-1820 they are just fragments of a single-piece cylinder head).

Edited April 29, 2018 by Witold Jaworski

[NickD]

Witold, a clear elegant description as always. I look forward to seeing how you approach the fins on the cylinders. I found this very time-consuming on another project I did without finding a particularly elegant solution.

 

Regards

 

Nick

[Witold Jaworski]

NickD - thank you!
I will create the fins in the next post. In this one I am wrestling with the partially hidden shape of the cylinder head:

_______________________________________________________________________________________________

One of the most prominent features of the R-1820 engine cylinders are their rockers. More precisely – their covers, cast as the part of the cylinder head:

 

 

The R-1820 was a classic four-stroke engine. Its cylinders had two valves: single intake valve, connected to the supercharger via a wide pipe, and single exhaust valve. Movements of these valves were controlled by cams, via pushrods and rocker arms mounted in the cylinder heads. The covers housing these valves and rocker mechanisms were placed on the right and left side of the cylinder head.

 

To simplify my model, I decided to separate the cylinder fins from its “solid” body (i.e. to create them as separate objects). However, because in the reality the cylinder head was cast as the single piece, it is very difficult to precisely determine its shape hidden between these fins:

 

 

While you can see the upper parts of the rocker covers on the reference photos, you can only guess their contours below the “fin surface”.

 

There is a blueprint that provides some additional clues:

However, I have some doubts about details of the contour that you can see on the rear view above. (I marked it with thick dashed lines in the picture). Look at the lowest part of this top contour: it should correspond to the upper (outer) surface of the combustion chamber. According other drawings, the shape of this chamber resembled a regular dome. If so, why the fragment of its contour visible in this drawing seems to be (a little) oblique? In the cutaway depicted in the first photo in this post I cannot see such an oblique shape. And why the side contours of this heads (the vertical dashed lines below the valve openings) are not symmetric? Thinking about it, I concluded that this drawing was not focused on the precise representation of the cylinder geometry: its main goal was to show the lubrication areas. Thus all these details, which we can see here, were drawn thanks to a “good will” of its draughtsman. They were hand-made, ink-traced drawings, and we can be just thankful to this technician for such a detailed piece of work. Still, I assumed that these lines can differ a little from the real contours – just because of the plain human error.

 

I formed the basic shape of the rocker cover using two clones of the same mesh: I placed one instance on the auxiliary drawing, while the second instance is located in its proper position on the cylinder (Figure “a”, below):

 

 

Modeling this cover as a separate object allowed me to switch between its local (along the valve axis) and global coordinate systems. I could also modify this mesh switching between its clones. I used the instance, located on the cylinder, to fit its base into the combustion chamber dome. The other instance of this cover, placed over the auxiliary drawing, allowed me to follow the shape of this element. (In fact, I could also put another instance of this mesh over the top view of the rocker cover. However, I did not do it - just because I used this view only during the initial phases of the modeling, and it was relatively easy to rotate the modeled object and move it over the side view).

 

This is the initial, “conceptual” model of the cylinder head, so I split it into the key “solids” and formed the semi-spherical cover of the exhaust valve as another object (the red one in figure above). Such an arrangement allows for easy manipulating of these parts. During this phase I have to determine their most probable sizes and locations. For example – following the precise location of the exhaust opening, I discovered that for the size as in the “Lubrication Chart”, it has to be placed in a slightly different position (as in Figure “b”, above). Otherwise, the right-bottom corner of the rim around exhaust opening would “sink” into the combustion chamber dome. (Of course, I also checked multiple times the most probable radius of this dome!).

 

When the whole thing seemed to match the photos, I made the rocker cover asymmetric (by “applying” its Mirror modifier and modifying the resulting faces). Then I modeled the oblique pushrod base (Figure “a”, below):

 

 

To avoid some potential errors in the future, I started with placing the pushrod (another object) in the proper position, then formed the base around it. Figure “b”, above) shows the resulting mesh. Note the sharp edges in its upper part. In the next step I rounded them, using a multi-segment Bevel modifier (Figure “a”, below):

 

 

To have more control over these fillets, I used the weight-based version of the Bevel. Figure “a”, above, shows the mesh edges that have a non-zero bevel weight marked in yellow. However, even in such a case, I could not avoid an artificial sharp edge between two fillets that were too close to each other (Figure “b”, above). Well, in this situation I had to “apply” this modifier, and manually introduce small fixes to the resulting faces (Figure “c”, above). I also dynamically created a “rim” around the upper edge of this cover. It is generated by the Solidify modifier, assigned to the thin face strip around this edge. Figure “d”, above) shows the final result of these modifications.

 

While working on these parts, I simultaneously “scanned” the Internet, searching for more reference photos. Sometimes they just expose details, which were obscured in the reference materials that I already have. In this case – it was a protrusion on the rocker cover around the first and the last bolt (Figure “a”, below):

 

I just had missed this tiny detail while forming the upper part of the rocker cover! Now I had a headache, how to fix it in a quick way. Ultimately I prepared two reference “cylinders” (I marked them in red, as you can see in Figure “b”, above. Fortunately, there were many faces around the area that I had to modify. I placed these faces on the corresponding reference cylinders using the Blender Sculpt tool. (It allows me to push/pull multiple faces at once in a gradual manner).

 

You can see the final result of this modification in Figure “a”, below:

 

 

Frankly speaking, I can see now that this protrusion had somewhat smaller radius. Ultimately I decided that it is “good enough” for the assumed level of details.

 

In the next step I cloned the rocker and valve covers onto the opposite side of the cylinder head: over the intake valve (Figure “b”, above). In this first approximation of these parts, I rotated the intake valve cover (marked in red in the picture above), trying to find the proper location and angle of the intake opening. To fit it better, I also placed in this model the intake pipe. I knew, that in the future I will adjust its shape multiple times. That’s why I crated it initially as a simple cylinder, smoothed by the Subdivision Surface and bent along a parent curve using Curve Deform modifier. By controlling the location, rotation and shape of the parent curve I had full control over this pipe.

 

The intake rocker cover had also a unique feature: two bolts on its front and rear walls (Figure “a”, below). They were intended for mounting around the engine an eventual NACA cowling. (Wright added these bolts on the Army request). For this “conceptual” stage of the modeling, I decided to add the bases of these bolts as a separate part. (Because I expect that I will move/modify shape of this element many times, before I reach the result that matches the reference photos). I will eventually join it with the cover (and add appropriate fillets around its edges) when it fits well.

 

 

In the next step I transformed the clone of the intake cover into a completely separate object (marked in blue in Figure “b”, above). I also added the bolt bases around the exhaust and intake openings. (As you can see in the figure above, there are four of them on the exhaust cover, and three on the intake cover). Initially I created these bases as separate objects.

 

Once I verified their location, I joined these bolt bases with the cover mesh (Figure “a”, below):

 

 

I joined these objects by applying a Boolean (Union) modifier. However, after such an operation the resulting edges required some manual “cleaning” (removing doubled vertices and edges).

 

I also formed an initial approximation of the rocker upper cover (Figure “b”, above). I just placed it over the left rocker. The front contour of this part had to fit the circular contour of this engine (dimensioned on the original installation drawing). I also rounded its upper edge using a multi-segment Bevel modifier, but I can see that this part will require further modifications.

 

While working with these rocker covers, I discovered that I made a mistake in reading the original blueprints! I thought that one of the exhaust rocker cover elements was a cross-section, while it was oblique view of one of its fins (Figure “a”, below):

 

 

On the reference photos I can also see that the bottom pushrod base plane was bent, with sharp side edges (Figure “b”, above). Thus I had to modify accordingly the bottom part of this cover (Figure “c”, above).

 

Well, such “discoveries” slow down the overall progress of the work, but they are inevitable, if you want to build a close copy of the real object. They happen all the time, as I am collecting growing number of the reference photos. In fact, I have measured that I spend at least half of the overall time on analyzing the photos. (Sometimes I also sketch on a paper the most complex shapes, before I start to model them in Blender. These sketches help me to better “understand” the objects that I want to recreate). The complex details of the cylinder head are often obscured by the fins, which makes this element an extremely difficult case. I am sure that I will identify and fix many of similar mistakes in the nearest future. For example: I will shift and rotate the cover of the intake valve multiple times, and then have to adjust the intake pipe after each of these updates. That’s why I prefer keeping this cylinder as an assembly of multiple, relatively simple objects. It would be much more difficult to modify this head, if it was a single, complex mesh. (In such a case you would have to care about all of its intersection edges!).

 

You can check details of this model in the source *.blend file. In the next post I will model the cylinder head fins.

[Azgaron]

Very interesting to see the 3D model being made!

 

Håkan

 

[NickD]

Withold,

 

That's sooo ambitious. It really helps my motivation to tackle similar complex arbitrary shapes. Thanks for sharing.

 

Nick