We have a Microscribe articulated arm 'sorta-CMM'. We got the 'high precision' and 'long reach' model, claimed accurate to +/-.009", and large enough to touch all of a basketball or about half of a car tire. The whole setup cost about $10,000, including a nice refurbished laptop, Rhino software, and two days of training for Rhino.
We get a lot of use from the laptop, and Rhino. The Microscribe pretty much gathers dust between dog- and- pony shows.
The Microscribe sends a series of characters describing the point location in
3D space on a button press. The data can be directed into AutoCAD with a TSR. Rhino receives it directly.
With a little practice, you can build a decent _looking_ image of most anything you can touch, but it's not accurate enough to use for inspection or for capturing the actual geometry of an object.
Yeah, Rhino can fit a surface to a point cloud, build circles or planes in space from three touches, stuff like that, but the generated model always comes out a little rough, with bumpy surfaces and edges that don't meet, stuff like that. Not Rhino's fault; the arm's data has a lot of scatter. Basically, you can build a sort-of 3D of an object, but then you need to go in and manually build a coherent and cohesive model over it, or spend a lot of time fiddling with individual points.
I think the original plan was to wear it out with use to justify the need for a bigger, more accurate Faro Arm, but we can't get anyone to use it twice.
I.e., 'not too accurate' is 'not too useful'.
-Mike-