3D modeling hardware and accessories for archiving

When we talking about dumping and backing up games and prototypes, it seems we might not consider the physical hardware and accessories that all contribute to the history of gaming. After all, it's much simpler to copy a disc than copy the physical attributes of a prototype PlayStation controller.

Herein, I would like to write a brief overview of my experience in 3D scanning with an outlook to it's application in archiving physical gaming hardware and accessories.

[GALLERY=media, 2527]Doctor V64 enclosure upper half in wireframe and solid 3D model by Shane Battye posted Mar 2, 2018 at 10:43 PM[/GALLERY]
Figure 1: Doctor V64 enclosure upper half in wireframe and solid 3D model​

Firstly, there are many options for automated 3D modeling/scanning physical hardware or gaming paraphernalia but some important factors to consider include:

• Most inert hardware and accessories are made of solid colored, synthetic material with little to no texture.
• This means photogrammetry methods (such as structure from motion photogrammetry, e.g. Agisoft PhotoScan) will perform poorly as they rely on detecting thousands of unique keypoints across a photoset. The surface of a an Xbox One S is essentially uniformly white – one point on the surface may look identical to another – photogrammetry would fail to model this.
• Projected light scanners are typically better performing for scanning uniformly coloured/textured objects like plastic or metal parts; these systems are able to determine the shape of a uniformly coloured object by analysing the contour of projected light/shadows across the surface.
• However, it should be noted that with visible spectrum light projectors, only light coloured objects can be captured (dark or black objects absorb too much light so there will not be enough contrast between light and shadow; one workaround is to apply a non-permanent matte white spray coat to these dark objects).

[GALLERY=media, 2531]Some Methods Of Constructing 3D Models by Shane Battye posted Mar 2, 2018 at 10:43 PM[/GALLERY]
Figure 2: Some options for constructing 3D models of an object.​

For illustrative purposes I’ll move forward from here with a stock David-3 SLS 3D scanner with the front shell of a Nintendo 64 controller as our subject.

Firstly I would not recommend this particular scanner to anyone before trying alternatives – it is incredibly tedious to set up for beginners and a lot of time can be consumed by calibration and re-calibration. The system design is essentially several years old now so there may have been significant advances in SLS systems since I acquired this one; if you are aware of any, please mention them in the comments below.

It’s ideal to use a stable turntable of some description when rotating an object for scanning. David offer a rather expensive digital turntable but I just use an Ikea SNUDDA ‘Lazy Susan’ turntable with a black velvet cover. I have marked a start point on the edge of the turntable and an approximate ring of 16 radial markers around the base which allows for fairly evenly distributed views between captures.

Ambient lighting must be low. First you ensure the projector is in focus across the subject and the camera can see all parts as it rotates on the table. Next calibrate the scanner with the supplied 90-degree calibration chart. After this point, anything above the ball/socket joint on the tripod must not be altered (focus, camera position in relation to projector etc).

[GALLERY=media, 2528]Structured light projection across an N64 controller by Shane Battye posted Mar 2, 2018 at 10:43 PM[/GALLERY]
Figure 3: Structured light projection across an N64 controller. Ensure the projector is in focus across the subject and the camera can see all parts as it rotates on the turntable. Light coloured, matte finish, objects such as this provide a good surface for structure light projection.​

The David 4 software is reasonably good at aligning each capture for models smaller than 15cm (6”); objects larger than this I find suffer from misalignment due to the stock projector losing focus across a longer distance. Users within the David forums have suggested projector enhancements to overcome this deficit but that is beyond the scope of this overview.

[GALLERY=media, 2529]3D reconstruction external surface of N64 controller by Shane Battye posted Mar 2, 2018 at 10:43 PM[/GALLERY]
Figure 4: 3D reconstruction external surface of N64 controller. The smoothness of the surface can determined by the number of polygons used to render the mesh.​

After one surface has been scanned, the part can be flipped and the process repeated. You may need to manually suggest the orientation for this second half to help the David software understand that it is the under surface of the previous scan, but after the first capture is aligned to the previous set, the subsequent views will all auto align to the previous.

[GALLERY=media, 2530]3D reconstruction internal surface of N64 controller by Shane Battye posted Mar 2, 2018 at 10:43 PM[/GALLERY]
Figure 5: 3D reconstruction internal surface of N64 controller. 3D scanners may struggle with fine details like the vertical screw shafts, particularly when they have a somewhat reflective finish. In this case, however, the N64 controller internal surface has modeled quite well.​

When all captures are complete a watertight mesh can be fused and the result exported to formats such as .obj or .stl. The latter commonly used by 3D printers. For 3D printing, the next question becomes how to orientate the model for the best quality finish while minimising support structures; and there's really no right answer. I left this model right side up just because on my 3D printer, removing support structures can make a mess of the surface they touch and I wanted to preserve the exposed surface as much as possible - this did mean wasting a lot of plastic on support material however.

[GALLERY=media, 2532]3D Printed N64 Controller by Shane Battye posted Mar 2, 2018 at 10:43 PM[/GALLERY]
Figure 6: 3D Printed N64 Controller. The scanned upper half of the controller has been 3D printed on a household printer using ABS plastic; the scale is almost an exact replica of the original and fits together with the original lower half and buttons.​

The process of just capturing a 3D model with SLS using the method above, from start to finish, is not quick. Expect to spend a good half hour to an hour just getting the captures and then longer refining your 3D model; even then you may not be entirely pleased with the result. There's a lot of trial and error involved in high quality 3D captures with such a system.

I hope this short overview is of interest to some of you; I’m sure many will have delved far beyond the basics suggested here, possibly used alternate methods of 3D modeling such as infrared (hey Xbox Kinect sensor bar right!) so please share your experience with 3D scanning and modeling retro hardware.

Don’t forget, preservation and archiving isn’t such for software or circuits!

 

Very interesting, I never replicated something with 3D scanners (neither I have one), but once tested the original Kinect for MoCap with also scans a bit of the area as a 3D, I also have the Xbox One version althrough never got to to test as the MoCap software is now paid.

 

Yeah they're not quite main stream yet but HP just acquired David and I believe had been incorporating structured light scanners into some of their desktop PCs for the past couple of years so the technology is becoming cheaper and more accessible. The David scanner was about $3k when new a few years back. I've seen relatively cheap handheld scanners (under $1k) but have no experience with those. The Kinect was a really interesting concept, infrared overcomes a lot of problems suffered from visible light spectrum, but not quite a high enough resolution in its original form for archiving.

 

This is really impressive work, Shane! I actually believe in the potential of 3D scanning on physical hardware for archival purposes as well. I've been using photogrammetry to archive some hardware myself, but the results you have gotten with your N64 controller are extremely impressive.

 

Thanks! It did work pretty well the N64 controller. The Doctor V64 enclosure was less accurate though because of its size; the projector just couldn't focus across the whole thing so when captures were aligned there ended up being misalignments on corners that were out by up to 5mm; it can smooth this out but not great for archiving. I might try it again with a 2x projector setup and higher quality projector some day.
I've considered 3D modeling game boxes/packaging too but it may be best quality to simply scan the six sides on a flatbed and apply as a texture to a scale cube in something like Blender and for reproduction just print as a flat foldable design on lightweight cardboard.

 

I wonder if photogrammetric software would respond well to your scanner's technique of projecting a pattern onto the surface. Photogrammetry works best on surfaces with a lot of texture detail (like stone or wood), so usually the flatly colored, reflective plastic surfaces of game hardware doesn't scan perfectly. Perhaps a combination of these two processes would work better.