Re: digitizing

From: Steve_Farentinos (
Date: Thu Jun 06 1996 - 00:31:17 EEST

Kurt Skifstad made some very good points about laser (non-contact) vs. CMM (contact)
scanning. Here's two cents more:

> In fact, the "higher density" data you typically
> get from a non-contact scanning device may allow you to much more
> effectively "digitize in stages" as you describe above. The coordinate
> systems for the multiple scans can in many cases be more accurately aligned
> (registered) in software, with or without datums.

A good point. When dealing with "datumless freeform" parts, we sometimes have to rely
on a (nearly) pointed CMM stylus to pick-up an exact location on the part. These are
hard to calibrate and therefore inherently inaccurate.

> Certainly, either
> technology is adequate for the range of manufacturing application we deal
> with daily in the automotive, aero, and consumer products industries.

We inspect some parts on the CMM with tolerances in the .001 inch (or less) range. Is
there a laser that can give me this kind of accuracy in a real-life situation?

> I'd rather have the option of ignoring extra
> data, than not having the option at all.

Not me. Especially when "extra data" means thousands of points. Of course, If you're
digitizing a plane with three points and the planar feature isn't quite flat, which
three points do you digitize? This where the skill of the CMM operator is critical.

> The dense data approach actually took one third of the time (8 days vs. five
> weeks), but was more successful in finding manufacturing defects, due to the
> more complete description of the shape provided by the dense data.

Sounds interesting. Was it a comparison of part inspection capabilities? If so, this
can only be true if the scan data were being compared directly to a surface model as
opposed to engineering drawings. Then, when all is said and done, do you really know
what dimensions to change to fix the out-of-spec. areas? Dispositioning a part based
only on 3-d point clouds won't fly on the shop floor, at least not yet.

> >6) Since the CMM can measure size and location of primitives, it is better
> >suited to
> >working with feature-based solid modelers.
> This seems to me to be more of a process issue, rather than a sensor
> technology issue (see above comments)

For me, this is the most important point of all. If my goal is to create a 3-d solid
model of an existing part, I want to build-in as many constraints and as much parametric
history as possible. In the end, I want to be able to easily modify features. The CMM
(and its software) were designed to produce the best possible answers for size and
position of prismatic features. For example, I can ask the CMM for the best-fit
cylinder through 20 points. For these features, measurements are taken and the answers
are keyed-into the CAD system.

> Here's a question for the CMM guru's out there: all the accuracy numbers
> you see from CMM's are based on benchmarks against (NIST ?) standard
> geometry like blocks and ball bars, etc, right? Since, like Steve says, you
> have to offset the data to get true surface location, can you have the same
> level of confidence in feature measurements, like hole locations or edges?
> Don't the offset calculations add an uncertainty to the data? and, doesn't
> the process then become highly dependent on the skill of the operator (to
> align the probe as optimally as possible)?

I'm not sure I understand the question. If 5 different people can measure the same
feature 5 times with 5 different probe orientations and get the same answers (to within
an expected accuracy), isn't that enough?


Steve Farentinos
PML, Inc.
201 W. Beach Ave.
Inglewood, CA  90302
310 671-4345
310 671-0858 Fax
310 671-1862 BBS

This archive was generated by hypermail 2.1.2 : Tue Jun 05 2001 - 22:37:23 EEST