And I thought I had OCD to details…you just are an aircraft engineer in disguise…LOL
I totally agree with @clockmender.
Hi Witold, she starts great and she will be perfect, for sure 
I will follow this one.
clockmender, tommy1441, ookka:
Thanks guys! I will do my best!
I think that within a month I will finish all the drawings I need to start the real modeling. I will report here the progress of this work (on a weekly basis).
Nice explanation of the 3d geometry. 
In this post I will shortly describe how did I create this top view. Drawing such vertical views (from the top or bottom) of the SBD Dauntless is more difficult than the side view, because there are no “vertical” photos which you can use to verify and enhance the available plans. The methods presented below can be useful when you want to draw or verify blueprints of an aircraft.
I started my top view using everything I could, for example some photos from the restoration done by the Pacific Aviation Museum:
The photo on the picture above has a strong barrel distortion. We cannot effectively “revert” it as we did for the side view. Why? Because the photo of the side view all contours of the aircraft lie on a single plane (the symmetry plane). This one contains are at least three important planes: the edges of the cockpit, the center of the fuselage (along its maximum width) and the wing contour. Each of them is located at a different distance from the camera, and each requires different distortion (fixing one of them you would spoil the others).
Nevertheless, taking all of this into account, this high-resolution photo is still useful to determine the rivets pattern of the center wing section, as well as the width of the cockpit frame. The edge of the Dauntless cockpit is formed by an important longeron: it determines the fuselage shape in this area. To precisely estimate the width of the cockpit canopies I draw auxiliary contours of their cross sections (you can see them on the picture above as the blue lines). Positions of the bulkheads are copied from the side view. On this top view I roughly approximated positons of the longerons below the cockpit edge. This is just a “workshop drawing”, not a regular scale plan: I will form the fuselage following its contour on the side view and a few key cross sections which I will draw later. Because of the barrel distortion of the reference photo I was not able to check the contour of the fuselage in the top view. This is the only element I had to redraw without any verification from the Douglas general arrangement drawing.
In next step I used dimensions from the Douglas diagram to draw the trapezes of the outer wing panels and horizontal tailplane:
Picture above shows all the lines which you can deduce from the general dimensions provided by the manufacturer. We can further enrich it using the information from the stations diagram:
The station diagram provides precise position of all wing ribs. Most of them are just a row of rivets, but along some of them you can find the panel seams.
All right, but this wing drawing is still missing its “vertical” elements: rivet and panel seams along the spars and stringers. How to determine their locations?
I had to review all the collected photos. Ultimately I chose one of the pictures from the web page of Chino Planes of Fame Air Museum:
I rotated this photo, aligning the wings of this airplane to the vertical guides. As you can see, it is made with a telescopic camera, so that it is very close to a perfectly orthographic projection. (The guides of the tailplane are not ideally parallel to corresponding guides on the wings, but this difference is minimal). The left wing is depicted at a relatively high angle, so you can see clearly the rivet seams along the spars and stringers. I decided that I can use this picture to map these lines onto my drawing.
I flipped this image from right to left, and stretched it, fitting its wing into the basic trapeze:
It allowed me to recreate the wingtip curve. In such a highly-deformed image the rib lines are bent. They match their “true” positions only on the wing edges. However, we can easily map from this image the spar and stringer lines. All of them continue from the center wing section. Combined with the ribs these lines form a kind of the “reference grid”, which cells allowed me to draw all the remaining details: the circular holes in the flaps, fixed slats openings, etc.
I used similar method to map the tip of the horizontal tailplane as well as its two spars. In the effect I obtained a detailed top view of the SBD Dauntless.
In the next post I will publish the bottom view.
I like the way you think. The principles are perfectly clear and explained very well. From looking at some of the other models in this board, I get the impression a lot of people aren’t aware of how this stuff works.
The photo on the picture above has a strong barrel distortion. We cannot effectively “revert” it as we did for the side view.
Actually, you could. It would just be a bit more involved. If you can establish the viewpoint, which should be possible, then by correlating the overhead photograph with the dimensions from the profile and elevation drawings, it should be possible to get a decent correction for the distortion in the overhead shot.
It may even be possible to get someone to code an add-on which would do basic photogrammetry for you, which could be very handy. The geometry is not that complex, it’s just the iterative side of it which would be tedious. That’s an ideal job for scripting.
[Nitpick] It’s due to perspective. Even a perfect lens would give you the same image. A real camera can’t switch to ortho view the way Blender can.
Steve S
@Steve S: I know it. I use terms “barrel distortion” and “perspective distortion” interchangeably, as the synonyms.
@Gumboots: Thank you! I just report my progress in this way because it may be useful as a reference for the others. On the other hand, I am a scale modeler, and this fact sets my standards. I built my first decent model when I was ten. I am sure that you do not need such a high precision for most of the pictures created by artists on this forum . The most demanding audience are the other modelers or similar buffs. That’s why the most demanding images are the pictures for the model kit boxes or their niche aircraft/armor/ship magazines…
I also thought about your comment concerning the reversion of perspective distortion (as for such an image as in the first picture from the post #27). I realized that it would be possible to build in the Blender composer a set of the nodes that would perform the required math. However, to keep the error level of such a reversion small enough, you have to deliver a kind of the “height map” of the depicted object. And the only reliable source for such a height map would be a 3D model… Well, it seems to be a typical “chicken or the egg” problem :).
Hmm. Well you have a pretty close approximation of the “model” from the profile drawings, etc that you have managed to create. If you can get factory drawings for some things, even better. This could give you sufficient “height map” data to get a reasonable correction. It wouldn’t be accurate to the millimetre, but it probably doesn’t have to be. It would be a fairly boring iterative process to get it to all come together, but you’re stuck with that anyway. If some of that can be handled by a script people might find it useful.
During previous weeks I was working on the bottom view and other details of the SBD Dauntless. For example — I added a modified side view that reveals the engine and the cowling hidden under the NACA ring:
Because of the formatting issues of this post I had to split the original square drawing into two parts:
(Click here to get these drawings as a single, high-resolution image). As in the case of the top view I draw the outer wing panel without its dihedral.
Detailing of the bottom view resulted in minor updates of the side view:
(See its high-resolution version).
I have already started working on the front view. One of the elements I need for the model are the key cross sections, thus I identified their shapes, and incorporated them into this drawing:
I did not draw the first sections of the NACA cowling here, because they will be visible on the front view. As you can see there are large gaps between sections 2 and 3 and between 8 and 9. Why I did not add these intermediate contours? Because nothing special “happens” between these bulkheads: the resulting shape will be automatically interpolated during modeling.
I sketched the engine and the inner cowling, because I am going to model these parts. Analyzing this area I discovered many differences between the earlier versions (SBD-2, -3, -4) and the later versions (SBD-5, -6) than were not mentioned in any previous publications about the SBD:
- Different cross section B (in the SBD-1…SBD-4 it had wider, elliptic shape);
- Different widths of the oil radiator scoop;
- Yet another carburetor air scoop: you can find in the books that in the SBD-5 it was removed from the NACA cowling and replaced by two intakes located between upper cylinders of the radial engine. However, they did not mention that they were just additional intakes for the filtered air (for the takeoff/landing from provisional ground airstrips). The main air scoop was still at the top of the fuselage, but since SBD-5 it was hidden behind the NACA cowling!
In the next post I will elaborate about these unpublished differences between the SBD versions, showing them on drawings. I will also prepare a simplified front view (for my model I do not need to redraw all the minor details there).
The drawings of this aircraft will be complete soon. I think that I will start building the first part of the model within two weeks.
SO MUCH REFERENCE
Seriously though, your final result should be spot on accurate to the real thing. Kudos!
You are welcome!
However, when you are a “born scale modeler”, you will always worry whether you have complete information about every nook and cranny of the modeled object :). [SUB]I have about 1500 photos of this aircraft, but I still miss some details - for example the shape of the upper part of the wall behind the engine, in the SBD-1…SBD-4. On all the wreck/restoration photos it is always obscured by the large cylinders of the R-1820 radial, or the exact shape of the carburetor air duct behind the NACA cowling in SBD-2…SBD-4…[/SUB]
To recapitulate my work on the Dauntless plans, I decided to draw all the external differences between its subsequent Navy versions. Because of the numerous changes that occurred in the SBD-5, I decided to split this description into two posts. This is the part one describing changes from the SBD-1 to SBD-4. The part two (about the SBD-5 and SBD-6) will be ready in the next week.
NOTE: All airplanes on the drawings below are equipped with the small tailwheel with solid rubber tire for the carrier operations. However, for ground airfields Douglas provided alternate, pneumatic, two times larger wheel. These tail wheels could be easily replaced in workshops.
Starting from the beginning: here is the SBD-1, the first of the Douglas Dauntless series:
(See the high-resolution SBD-1 left & top view).
US Navy originally ordered 144 SBD-1s in March 1939. The first of these aircraft took off from Douglas airfield in May 1939. However, the Navy was not satisfied with their relatively short combat radius. Probably the outbreak of the war in Europe (September 1939) forced the Navy to accept first 57 SBD-1s “as they were”, assigning them to the Marines squadrons. For the 87 remaining airplanes from the original contract, the Navy requested longer range. To improve Dauntless combat radius, Douglas installed additional fuel tanks in the external wing panels. They also equipped these airplanes with the Sperry autopilots. This new variant was named SBD-2. It was delivered in 1940 to carrier squadrons of the US Navy. Externally, the SBD-2 had lower carburetor air scoop than the SBD-1:
(See the high-resolution SBD-2 left & top view).
The next Dauntless version — the SBD-3 — was originally ordered in 1940 by French Aeronavale. SBD-3 was updated for the identified requirements of contemporary battlefield. It had armor plates protecting pilot and gunner seats, armor glass plate inside the windshield (I did not draw this and other cockpit internal details). Douglas installed also the self-sealing fuel tanks. After June 1940 all 174 ordered aircraft were taken over by the US Navy, which then ordered additional 411 airplanes. The Navy workshops doubled in these machines their rear guns. This modification was adopted by Douglas in the later series of this aircraft. Externally — the boxes containing flotation gear (“balloons”) were removed from the engine compartment:
(See the high-resolution SBD-3 left & top view).
The side slots of the SBD-3 cowling were slightly larger than those in the SBD-1 and SBD-2:
The next version — SBD-4 — received new, 24V electric installation, which allowed for installment of the radar and broader range of other electronic equipment. However, in the 1942 the Navy was short of these devices, and the factory-fresh aircraft did not have any of them. (The Navy workshops installed radars on some SBD-4s later). Externally you can recognize this version by the new Hamilton Standard Hydromatic propeller:
(See the high-resolution SBD-4 left & top view).
The previous SBD versions (-1, -2, -3) used the Hamilton Standard Automatic propeller. As you can see in drawing below, the blades of these propellers had different shapes:
(See the high-resolution SBD-4 front view, SBD-3 front view).
Below you can see another drawing of the SBD-4, consisting the bottom view as well as the side view without the NACA cowling:
(See the high-resolution SBD-4 bottom view).
Comparing it to similar drawing of the SBD-5 published in the previous post, note the different profile of the internal cowling (the cowling behind the engine cylinders). For this version I had no photo of its upper part! The shape of this element is deduced from the shape of similar part in the SBD-5 and from the size and location of the Stromberg-Bendix injection carburetor, located just behind this cowling.
Next week I will describe the external differences between SBD-4 and SBD-5. It will be the last post about the “general” reference drawings. Then I will report my progress on the first element of this model: the wing.
As always, an impeccable detailed research on those subtle differences between variants and amazing draftsmanship to being able to sift through all the needed information out from some very difficult reference photos.
I salute you my friend :eyebrowlift:
Thank you!
I will do my best in the further posts :).
In this last post about scale plans I will write about the modifications introduced in the SBD-5 Dauntless version.
For the reference, I placed below the drawing of the previous version: the SBD-4:
(See the high-resolution SBD-4 left & top view).
In February 1943 Douglas started to produce another Dauntless version: the SBD-5. It used more powerful Wright R-1820-60 engine (performing 1200 HP on takeoff: 20% more than the R-1820-52 used in the SBD-4). The engine was moved a few inches forward, and the whole area in the front of the firewall was redesigned
The old telescopic sight was replaced by modern reflector sight. The SBD-5 had heated windscreen (because it sometimes misted over in dives). (See the high-resolution SBD-5 left & top view).
The engine in the SBD-5 was moved forward by 4 inches, together with its NACA cowling. The overall shape of the NACA ring was the same as in the previous versions, except the removed carburetor air scoop. (The cross sections A are the same in both versions):
The shape of the firewall (section C in the figure above) remains unaltered. However, there is a difference in the width of the gap behind the NACA ring. In the SBD-1 … 4 this gap was relatively narrow, and the cross section of the fuselage below (section b in the figure above) forms a regular ellipse. Thus in the previous versions the upper part of the NACA ring had six flaps that controlled the flow of the cooling air through the engine. In the SBD-5 the fuselage was a little bit “thinner” here, and the bottom part of its cross section (section B in the figure above) had slightly different shape. The larger gap between the NACA cowling and the fuselage increased the constant amount of the incoming air that cooled the engine. It allowed Dauntless designers to reduce the number of cowling flaps from 6 to just 2.
The figure below reveals more differences between the SBD-4 and SBD5 engine cowling:
(See the high-resolution SBD-5 bottom view).
Some of these changes are well known, like the removal of carburetor air scoop from the top of the NACA cowling or the different shape of the side ventilation slots. However, while studying the photos, I have found two minor differences that were not yet mentioned in any source:
- The oil radiator air scoop was in the SBD-5 was wider than in previous versions (as well as its panel);
- The bottom seam of the NACA cowling was in the SBD-5 shifted left, while in the previous versions it was running along the symmetry plane;
Finally, I would also like to share with you my findings about the carburetor air intake in the SBD-5. As I mentioned earlier, it disappeared from the cowling, as you can see it on the front views:
(See the larger SBD-5 front view).
But where did they place this air scoop in the SBD-5? Studying the photos and descriptions in the books you can find two air intakes located between engine cylinders (as in figure a, below). However, in the original SBD Dauntless maintenance manual I discovered that the central air intake remained — just hidden under the NACA cowling:
The side air scoops were filtered, while the central air scoop was not. I used the Pilot’s Manual to find that there was a switch to flip the carburetor air intake between the filtered and non-filtered air. The filters were auxiliary devices, intended for takeoff and landing on dusty ground airstrips. (You can see similar solutions in contemporary designs from 1943: P-40L and P-51C). In the Pilot’s Manual you can read that you should switch into the non-filtered (i.e. central) air scoop to get the full power from the engine.
I must say that I was used to more streamlined carburetor air ducts. Such a location of the main air scoop is quite strange. It seems that the designers of the SBD-5 concluded that there is enough air behind the single-row radial engine to feed its supercharger. (In an airplane flying 100mph or more the amount of the air passing around the engine is several times larger than during takeoff. Thus such a solution could work if we assume that for the takeoff pilots used the less obscured side air scoops).
I did not prepare drawings of the last Dauntless version — the SBD-6. It had even more powerful engine (R-1820-66, rated 1350 HP on takeoff). Douglass built 450 of these airplanes between April and July 1944. Their radars were fitted in the factory. However, there is no external difference between the SBD-5 and the SBD-6!
In the next post I will report my progress in building the first part of this airplane — the wing.
Wow,amazing work Witold, Your an inspiration to all. a huge thank you. (on my knees bowing with arms outstretched, I’m not worthy, I’m not worthy:D:yes:
kenD, you are definitely worthy :yes:!
These drawings are just - well, “workshop drawings”. They still contain errors, which will appear during modeling.
I have some experience in making professional drawings for modelers’ magazines and books, thus I can draw them relatively quickly… (“relatively” - this is the proper word! There are better modelers around, they can create beautifully detailed models in the time I spent just for drawing these references). The modern software like Blender or even 2D graphic programs combined with the vast Internet photo resources give scale modelers great, powerful possibilities I am just starting to explore.
I started by setting up the initial scene:
Although Blender allows for arranging the reference drawings on the three perpendicular planes like in the 3D Max, I prefer the alternate way: the Background images feature. Using them, I can assign appropriate image to the corresponding view, and simultaneously use all the six views (bottom, top, left, right, front, rear). They appear just when I set appropriate projection.
This is also the moment to determine the “scale” of this model. Because in the SBD drawings that I have all the dimensions are in inches, I decided to assume that 1 unit in this Blender scene = 1 inch on the real airplane. However, I have no experience with the Blender Units setting, so I left them set to None. If you want to check details of this setup, here is the original *.blend file.
I started modeling the wing by forming the contour of its root rib. (For this purpose I draw the shape NACA2415 airfoil on the reference drawing). I smooth most of the model meshes with Subdivision Surface modifier (it uses the classic Catmull-Clark scheme). The shape of a single edge loop smoothed by this scheme is a piecewise Bezier curve (or, if you wish, a NURBS curve – this is just an alternate math representation). The edge vertices are its control points, so I can easily shape this contour. You can see the result in the figure below. (In this image you can see that the vertices lie on the rib contour, because the mesh drawing mode there was switched to draw the resulting surface):
The theoretical shape of the NACA-2415 airfoil has a thin, sharp trailing edge. However, in the real airplane it was rounded because of the technological reasons. I tried to determine its radius from the photos. As you can see in the enlarged fragment of this picture, it forms a small wedge with rounded corner. It is shaped using five vertices. (Their number corresponds the number of the leading edge vertices — I will explain the reason further in this text). The Dauntless inherited many solutions from its Northrop Delta lineage. For example — its wing spars are not perpendicular to the wing airfoil chord. Instead, they are perpendicular to the fuselage centerline. (In the SBD, like in the earlier Northrop designs, the center wing panel and the fuselage form a single unit. I suppose that it was easier to put together the wing spars and fuselage bulkheads when they shared the same technological bases).
To provide as many “technological bases” for my model as possible, the X axis of the wing object is parallel to the wing chord. I can set it “in the Northrop way” by setting the object incidence angle to 2.5⁰. In this position I can work with the wing mesh, moving vertices along the global coordinate axes (i.e. the axes of the fuselage), and then switch to the local wing object axes when needed.
In the next step I formed the basic wing trapeze. I did it by extruding the wing root edge, and shrinking the airfoil located at the wing tip:
Now you can see why I draw this wing section on the plans without dihedral. This drawing would be useless if it depicted the wing “properly”! From the reference images and descriptions it seems that the wing tip had the NACA-2409 airfoil. In the first approximation I scaled down the rib of the tip, fitting it to the reference drawing. (To fit this mesh to the front view I temporarily rotated the wing by its dihedral angle — 10⁰ 8’ — as in the figure below). However, although scaling down the original NACA-2415 coordinates produces the NACA-2409, it does not work precisely for the airfoil shape recreated with the Bezier curves. To fix these small differences I prepared an auxiliary “guide” rib of the NACA-2409 airfoil and placed it in the tip. (see the figure above). Then I modified the wing tip airfoil, fitting the wing surface to the contour of this guide rib (you can see on the picture that it minimally protrudes from the wing – as a very thin line).
Then I rotated the root airfoil, adjusting it to the wing dihedral:
In the SBD Dauntless all the wing ribs were perpendicular to the wing chord plane, except the root rib of the outer panel. To easily insert properly oriented ribs in the middle of this wing, I inserted another rib after the skewed wing root rib. It is perpendicular to the chord plane. I marked this rib edge as “sharp” (by increasing its Crease weight to 100% — you can recognize it on the picture by different edge color). In this way I ensured that the skewed root rib has no influence on the new edges I will add in the middle of this mesh.
In the Catmull-Clark subdivision surfaces, you can use the Crease weights to obtain a local sharp edge or to separate a mesh fragment from the influence of the outer mesh vertices. I learned this method from a Pixar paper, presented on SIGGRAPH 2000 by Tony DeRose. (Before I started my first model, I studied the subdivision surfaces math, to know better properties of the basic “material” used in the digital modeling).
I had an occasion to learn that it works as expected in the next step: forming of the rounded wing tip. First I inserted into the tip area a few new ribs (using the Loop Cut command). Then I started bending their trailing and leading edges, to finally join them into an arch:
As you can see in this picture, I also removed some of the internal mesh faces. I did it because I had to alter the topology of this area. (It is easier for me to determine the new faces when the old ones are removed).
Note that it was a good idea to have the same number of vertices on the trailing and leading edge. Now I can easily join them at the wing tip.
The figure below shows the resulting surface:
Note that the wing tip edge lies on the wing chord plane. As we can see from the reference drawing, in the real airplane the wing tips were slightly bent upward. We can easily obtain such an effect by moving upward (and slightly rotating) last vertices of the tip:
In the figure below you can see the control (i.e. not subdivided) mesh of this wing:
Note that I tried to align as many “longitudinal” mesh edges as possible to the stringers and spars visible on the reference drawing. This will be extremely useful when I draw skin details on the wing surface unwrapped in the UV space (for texturing).
In this source *.blend file you can check any detail of the mesh presented in this post. The next post will describe further steps of the wing modeling: separation of the aileron and forming of its bay in the wing.
This thread provides just an overall picture of the process. If you want to learn more about digital aircraft modeling and subdivision surfaces, see this guide: “Virtual Airplane” (vol. II).
Excellent work Witold and thanks for the Blend file!
The Cutmull-Clarke and crease weight stuff is straight-out-of-the box with the Blender??