3D Scanning

Werner Almesberger werner at openmoko.org
Thu Jul 8 09:41:45 EDT 2010

Daniel Clark wrote:
> Also it's possible Wolfgang and I may be able to get into an MIT lab
> with a 3d scanner during his visit to Cambridge :-D

By the way, I have a Roland Modela MDX-15, which has "2.5 D" scanning
capabilities. It works by gently passing a needle over the object's
surface and detecting when the needle touches something. The needle
exerts almost no mechanical force on the object.

Since the MDX-15 is a 3-axis mill, it can only scan elevation above
the X/Y plane, but it can't look into lateral holes, peek under
overhangs, or scan the bottom of an object. One can of course
manually rotate the object and align the scanned surfaces in a CAD

Nominal X/Y resolution is about 50 um, but I would expect to get
about the same real-life accuracy with 100 um steps. The needle's tip
has a diameter of about 100 um. Z axis resolution is 25 um and is not
affected by the X/Y step size.

Wolfgang has a few test scans on his laptop which I hope he can post.
The scanning software supports various output formats, including DXF
and just a CSV text file with the point cloud.

My main obstacle at the moment is that the scanning software, called
Dr. PICZA, runs only under Windows, while I have a pure Linux shop.
Worse yet, the protocol used for scanning is not documented.

So my choices are to either figure out why one version of Dr. PICZA
has major troubles communicating with the mill/scanner when running
under Wine while another has problems too, but much later; reverse
engineer the protocol and write my own scanning software; or bite the
bullet and see if I can find some infidel who'll lend me a Windows PC
for a few days.

My Bens should get delivered around next Tuesday, so I still have
time to find a way to get this scanner to cooperate. I'd expect each
scan at 100 um X/Y resolution of a large surface (rear, front, rear
case plastic, etc.) to run for about 2-3 days. Perhaps I'll start by
making "draft" scans at, say, 1000 um, which would be considerably

- Werner

More information about the discussion mailing list