SW – ANH (5 Foot) - Studio Scale Millennium Falcon Build

Don't want to see this ad? Sign up for anRPF Premium Membershiptoday. Support the community. Stop the ads.

vfxsup64

Well-Known Member
RPF PREMIUM MEMBER
Thought I would share some initial and early progress on my ‘5 foot’ Millennium Falcon dimensional survey and eventual studio scale build.

I come from a Visual Effects/Design Engineering background and I’ll be leveraging techniques and tools I use every day to help define the forms and overall dimensions of the model – as an initial first step.

I decided to start with the ‘core shape’ – the forms that have no kit-bashed or scratch-built parts.

onValjean_000.jpg


The next step was to analyze the image and strategically define converging perspective lines to estimate the ‘camera lens - field of view’.

onValjean_perspectiveLines.jpg


By defining the lens ‘field of view’ and later establish some base-line dimensions of key forms, one can infer other details and start to build additional core shapes.

At this point, I had to make a few assumptions and begin the process.

For example, I felt safe in assuming that the docking hatch and gun turret forms were symmetrical in the X and Y-axis, respectively. In other words, the docking hatches are mirrored along the X-axis and the gun turret forms are mirrored along the Y-axis across a central origin buried in the middle of the model’s armature.

I also referenced my old Star Wars sketchbook and noted the 2” diameter shaft call out on the gun turret Falcon sketches and worked out that the permanent sleeve shaft diameters are probably 2.25” (1/8” wall thickness). For the model mover shaft, they most likely used a 2” diameter steel shaft and an ID collet chuck to affix the model to said 2” shaft.

MF_modelMoverShaft_003.jpg


One of my first design engineering jobs in visual effects was at Boss Film Studios working for Richard Edlund and the model mover shafts there were often of the same configuration.

ModelMoverMockUp_001.JPG


By breaking down the major mechanical shapes and assign those variables with assumed values, I derived other dimensional forms and measurements for my initial mock-up.

Further, based on very detailed and researched information posted on this forum (i.e. 3” mandible hole diameters, 2” mandible thickness in the Y axis etc.) I arrived at an initial form guesstimate:

onValjean_cgModelOverlay_001.jpg


I even found a vintage Travel Lodge Motel Room glass on a certain Internet auction site and decided to use it as additional dimensional cross-reference – my humble contribution to the studio-scale pantheon of reference objects:

travelLodgeGlass_003.jpg


After a fair amount of time researching, making assumptions, modeling, cross-referencing and estimating…

PH_MF_0001.jpgJJ_MF_0001.jpg


Parts of it seem to be lining up – other areas…not so much.

I then decided to grab a short image sequence from Star Wars (ep IV) and track the MF in 3D as it escapes the Death Star. I wanted to see how my initial assumption would compare against the studio-scale model in an actual shot from the film.

escapeTheDeathStar 001 on Vimeo


Encouraging - but too soon to call it solved. I decided to try a slightly different tack.

More to follow...




Regards,

Andre
 

Don't want to see this ad? Sign up for anRPF Premium Membershiptoday. Support the community. Stop the ads.

eagle1

Sr Member
RPF PREMIUM MEMBER
Hi Andre,
Cool, another Falcon nutter to join the fray!.

While I think the work you're doing is awesome, it takes me back to when I was trying to work out dimensions.
I feel that actually researching each section & building it fully as much as possible, such as the gun turrets, walkways, docking rings, loading ramps & rear underside main gear box & bringing together as a whole will more or less give a fairly accurate hull dimension.
The greeblies are the key, if they fit on each section you have made, then all should marry together, albeit with a little tweak here & there.

An example of it all going wrong & having assumptions was my early work on the cockpit tube. I trusted photo ref too much & got the size wrong. In the end I pulled out 3 sizes of tube & wrapped around certain donor parts to see which fitted. My third try came up trumps, thanks to the greeblies.

This is the way I am approaching the build, but hey, really keen to see what you come up with & if you want to chat about it, cool!.
Bests,
Stu
 

Scott Graham

Sr Member
RPF PREMIUM MEMBER
I love the way you're analyzing this Andre! These factors always seem to be the most important things that we often don't think about - what's under the surface. This will one to watch.
 

vfxsup64

Well-Known Member
RPF PREMIUM MEMBER
Stu and Scott,

Thank you for the kind words and advice. I appreciate it!


About 3 months ago, after laboring for a while with the above-mentioned workflow, I pulled the e-brake a bit on this generalized approach to the overall core.

I couldn’t get close enough – at least with the image reference I currently have access to. Too much chance for error and with this size a model, error compounds on error quickly.

You are right Stu, focusing too much on the photo reference can lead one in wrong directions and assumptions – fast. It amazed me time and time again how easy it was to convince myself of just about anything regarding an overall dimension or measurement looking at just a 2D reference image...

I needed to get closer – more accurate and more detailed. I decided to work from the micro to the macro, so to speak.

Moving forward, I decided to build on what I know – or rather, what I can know with the actual kit part measurements as a guide.

Recently, I’ve been collecting key donor model kits and have been hand modeling and digitally scanning them to aid in the sectional and overall layout:

MF_donorPartSamples_001.jpgmodoWireFrames_001.jpg


Each of these parts is a measuring stick and when certain parts are laid out and related together, they can help define a major dimension or form. For example, I chose many of the parts in the above image strategically as they are key to the long form dimension of the outer mandible walls and help define the placement of the mandible holes (1/24 Panther G, M23 McLaren, Entex Wankel, M42 Duster, Kettenkrad, Kenworth W925, California Hauler etc.)

portSideTopView_001.jpg
(Initial Test Layout)

Using 3D tracking and CG photogrammetry tools I’ve been working in sections and using my digitized donor kit parts solve for camera variables (position, rotation, scale, field of view and lens distortion) and as the parts come together individually, I will group them together and then relate them back to the major form – cross checking and double checking as I go. I would do this digitally and also work with the actual parts as a critical triple check – especially in hard to see areas (plating thicknesses, for example).

SW_exhibitComp1_001.jpgSW_exhibitComp2_001.jpgSW_exhibitComp3_001.jpg



I’ll document further progress as I go and explain in more detail the 3D process.

One of the advantages of some of this workflow is that it can be applied to any model one is trying to make. I lieu of actual blueprints (or having in-person access to the actual miniature), there is quite a bit of information encoded in a perspective still image – parallax is critical to the success of the photogrammetry process.

But as Stu mentioned, there be peril in relying too much on photo reference alone. The actual donor kit parts and how they relate in the real world - are the most important consideration.

More to follow…



Regards,

Andre
 

Don't want to see this ad? Sign up for anRPF Premium Membershiptoday. Support the community. Stop the ads.

eagle1

Sr Member
RPF PREMIUM MEMBER
Andre, loving the dedication you are putting into this, the computer stuff is beyond me, so if it works for you, great man!.
Another approach to it is always a good thing, I'm sure some good info will be found mate!.

Stu
 

vfxsup64

Well-Known Member
RPF PREMIUM MEMBER
Stu,

Thank you very much.

My plan moving forward is to share derived measurements with the forum and see how it jives with other people's data points. I'll be bouncing back and forth between the 3D computer and the real-world donor part pieces so as to help keep myself honest!

What would be ideal is an accurate "blueprint" for the core form that we all agree is spot on. One of my assumptions that I haven't mentioned yet, is that most of the core measurements are rounded off to the nearest reasonable fraction. In other words, dimensions like 2.50 inches as opposed to 2.457" or 2.523" for a major form relationship. Your measurement for the cockpit diameter (5.75") and other known measurements (2.00 inch think mandibles, 3.00 inch diameter mandible holes) seem to indicate that assumption might be a reasonable one.

When I was at Boss Film, our engineering department interfaced with the model shop from time to time and that was how a lot of the inner armatures and mechanisms were made for the miniatures. Rounded off "cardinal" measurements make life easy for you down the road assuming the resultant forms match the overall approved design. Exceptions were made but in general, measurement relationships were drafted that helped simplify the manufacture of things on a lathe, mill and drill press.



Regards,

Andre



Andre, loving the dedication you are putting into this, the computer stuff is beyond me, so if it works for you, great man!.
Another approach to it is always a good thing, I'm sure some good info will be found mate!.

Stu
 

Don't want to see this ad? Sign up for anRPF Premium Membershiptoday. Support the community. Stop the ads.

bwayne64

Well-Known Member
This is good stuff ! I'm curious as to how you'll get the dims on the curvature of the main hull. I'm working on a project that has a domed shape and trying to get the curvature of it. Have precious little photos of it. It's right on the tip of my brain how to do it, :) If I knew the height of the dome I could get it. I just have the diameter. Great work so far Man, can't wait to see more,

Joe
 

vfxsup64

Well-Known Member
RPF PREMIUM MEMBER
Joe,

Yes, the hull shape is critical to get right.

Looking at the overall MF design, it is a major form that helps unify a lot of the other forms - both large and small. It allows the eye to move from one shape to another efficiently - not unlike a well drawn figure sketch that has great gesture and rhythm (to use a drawing metaphor). This ship design is by far my favorite so getting the hull shape and form intersections correct is extremely important.


Now, as for how - that is two fold.

First I've made a critical assumption that the spherical convex hull forms are indeed that - spherical (radius of constant value).
From a layout and manufacturing point of view in the 70s, this would make things easier to shape than a compound form that these days is trivial with 3D modeling programs (B-Splines, NURBs Curves etc.) and Rapid Prototyping.

Second, I remembered a workflow from math class that if you knew the location of three points on a circle, one could derive the radius and center of that circle relative to the perimeter.

After a quick search on the 'net, I found a web page that does just that: mlm - Circle radius and centre calculator given 3 points

Input the X and Y position of the three points and the tool outputs the circle center location and radius. For my initial shape guesses, I chose a point that hits the bottom edge of the top gun turret and two mirrored points that make up the intersection of the top convex hull and the side wall perimeter (the waist line center core wall that spans between the two convex clam shell hull forms). Once I derived the radius and location center, I lofted the circle into a sphere (in Y) and cut it down to make the two convex shell forms.


One test was to compare the intersection lines between my convex hull and the docking hatch walls (circled in green), for example:

JJ_MF_0001a.jpg

This particular intersection "curve" looks to line-up reasonably well to the picture so that is promising.


I'm not 100% sure on everything yet as the overall shape is just a temp mock-up. So, in the next few days, I'll be updating this thread with my continued "micro" surveys based on modeled donor kit parts (i.e. Tamiya 312B engine) and start working my way around the perimeter of the front mandible walls - inside and out...

312B_eM_001.jpg





Regards,

Andre




This is good stuff ! I'm curious as to how you'll get the dims on the curvature of the main hull. I'm working on a project that has a domed shape and trying to get the curvature of it. Have precious little photos of it. It's right on the tip of my brain how to do it, :) If I knew the height of the dome I could get it. I just have the diameter. Great work so far Man, can't wait to see more,

Joe
 
Last edited:

joberg

Master Member
Love the ''calculated way'' you're using on that project; keep up the good work and looking forward to more updates.
 

Don't want to see this ad? Sign up for anRPF Premium Membershiptoday. Support the community. Stop the ads.

stonky

Sr Member
RPF PREMIUM MEMBER
Good stuff!

I'm curious, maybe I missed it - did you undistort the photographs/ANH footage?
 

bwayne64

Well-Known Member
Thanks Man, Awesome stuff. I should have paid attention in math class, :) This helps alot. Cheers,

Joe
 

vectorzero

Well-Known Member
RPF PREMIUM MEMBER
Andre

This is fascinating read. It's great to see industry guys helping out the fans.

Are the software tools that you are using proprietary? (I'm guessing so due to your day job.) The open source photogrammetry tools that I have discovered so far (excellent article here) rely on you taking overlapping photos from multiple different angles of an object to which you already have access. These tools seem to rely on identifying the same points from different angles, and they then build up a representative 3d mesh. When using them they specifically state not to hold the camera still and revolve the subject.

When trying to determine dimensions using analysis of still frames you are often dealing with a situation where the camera is locked off, and the model moves. Your original analysis seems to have been done from still frames and/or stills - suggesting a different approach or maybe different and more powerful software.

Right now I'm trying to work out the proportions of a model which doesn't have much in the way of greeblies to create a base reference. However, there is an existing kit which seems reasonably accurate, so I'm using that as the reference. I'm trying to infer from the stills what details I need to change, and its proving much harder than I thought - and the shapes are very simple rectilinear forms (viz. the upper part of ED-209's legs.)

Any tips? Or is this software only in the price range a VFX house can afford?
 

vfxsup64

Well-Known Member
RPF PREMIUM MEMBER
Franz - thank you!


Joberg,

Indeed - but, as I've learned over the years in my work and that has been articulated by many here on this forum - it doesn't always have to be right, it only needs to look right!


Stonky,

Thank you - and yes, I did undistort the stills but not the ANH SW short film image sequence.

I will undistort everything as I get a more refined form and want to compare and double-check.

The undistort step in the software is heavily dependent on accurate 2D feature points and accurate 3D survey points - not to mention the algorithm is an estimate albeit a reasonably accurate one - but heavily "image-based". I have no idea what variables the vintage images have baked into them. For example, many of the ILM shop images look to have come from Ektachrome or Kodachrome slides shot with a single lens reflex 35mm camera. Over time, slide film emulsion warps and distorts. Not to mention film scanning/cropping/rescaling and other variables.

The current day MF images from the SW exhibits are a bit more predictable as there isn't the film/scanning variable to deal with. But there is the issue of widely divergent optical system accuracy with the digital cameras. Not to mention rolling shutter (if the camera is moving even a little during exposure - the image will warp) - I'm even cognizant of the issue of refraction - - the MF is currently encased in the big plexi box and the light rays will diffract a little at shallow glancing angles.

All reasons not to be too dependent on the images alone. The donor kit parts and a real world layout are essential and the key to the whole thing.

I'm using the computer tools to try things out quickly and non-destructively...to experiment and add another way to look at things.



Regards,

Andre



Love the ''calculated way'' you're using on that project; keep up the good work and looking forward to more updates.
Good stuff!

I'm curious, maybe I missed it - did you undistort the photographs/ANH footage?
 
There were several scale models for Millienium Falcon, including a full size.

I wonder if anybody had tried to collect many pictures of each scale model and tried to find major diferences between each one.
 

Don't want to see this ad? Sign up for anRPF Premium Membershiptoday. Support the community. Stop the ads.

Don't want to see this ad? Sign up for anRPF Premium Membershiptoday. Support the community. Stop the ads.

Top