Cuda416
Well-Known Member
Exactly, the point cloud is generated from the scan, the STL generated from the point cloud, which is what I said.
=C
Young Mopar fan bringing the A-925 to life.
- Thread starter A925guy
- Start date
-
Exactly, the point cloud is generated from the scan, the STL generated from the point cloud, which is what I said.
=C
Phreakish
Well-Known Member
Exactly, the point cloud is generated from the scan, the STL generated from the point cloud, which is what I said.
=C
Not to be pedantic, but not in all cases. Modern systems use additional video data context to produce surfaces rather than simpler point-in-space data that older scanners did(often laser, or sometimes physical touch systems). This sophistication cuts down on the clean-up and geometry creation after the fact.
My original point was that the work necessary these days is vastly less than it used to be. The first cmm-scanned surface I dealt with took a crew of 3 more than a month to turn into something usable. The last set of parts I had scanned took an afternoon and we had parametric solids by the next day with a crew of me
Cuda416
Well-Known Member
Not to be pedantic, but not in all cases. Modern systems use additional video data context to produce surfaces rather than simpler point-in-space data that older scanners did(often laser, or sometimes physical touch systems). This sophistication cuts down on the clean-up and geometry creation after the fact.
My original point was that the work necessary these days is vastly less than it used to be. The first cmm-scanned surface I dealt with took a crew of 3 more than a month to turn into something usable. The last set of parts I had scanned took an afternoon and we had parametric solids by the next day with a crew of me
Ok, I'll play along... the scanner doesn't know what an edge is, that is a single break in a plane. it "sees" a crap ton of points that can be used to "guess" with a high probability of being correct, that a bunch of points, situated closely together, within a certain tolerance, "could" be an edge . further, point clouds, that are turned into STL files are just a collection of triangles that make up a mesh...
Scanners don't "scan" STL's anymore than a radar knows what and aircraft "looks" like. The scanner could have software internal to it that does a conversion to an STL (a shell) but even if you scan a "knife" you are likely to get an "edge" made of a billion points.
And yes, I 100% agree that things have gotten easier, but the beginning "point" (pun intended) is still a point cloud, regardless of what the machine spits out.
Fair?
-=C
A925guy
Member
I didnt thank about the macro approach. Right now, I extruded the overall solid. Inventor has a feature where you can paint brush select multiple stl triangles and the program can generate a spline surface. Then i use that to remove material. Just discovered it last night.
Phreakish
Well-Known Member
Can't 100% agree based on the tech we've been employing recently. Iterative surface mapping basically bypasses the point collection and streamlines the scan process. The technology can in fact find edges, create curves, and generate contours. Not precisely in every instance (some geometry just isn't 'regular' enough), but it works well enough to save hundreds of hours of clean-up time when dealing with a large number of parts.
Strangely enough, one project in particular was actually a knife - the edge came in with two curves, a straight line and a spline which fit the curved tip. It took a few tries to set the right level of detail to avoid getting jagged results but we eventually got a very good surface model with only a few hundred more surfaces than a proper parametric model would have had (mostly due to the sharpened edge).
The trade-off is that accuracy isn't as exact, and operator proficiency plays a bigger role than when starting with point data.
Cuda416
Well-Known Member
Good points, but the term "Iterative surface mapping" itself, infers that multiple points are measured/scanned etc.
I'm done being a pain, you clearly have loads more actual real wold experience. I'll go out saying this debate is largely semantics. No matter what process is applied, the scan still begins as points, but to your point (there's that pun again) you are able to bypass the point cloud. I however, would rather have the point cloud as raw data and apply an "Iterative surface mapping" process as a second step, if even manual.
I look forward to seeing what is done with this.
Phreakish
Well-Known Member
Oof. I'm a fan of 'use what works', but I've found over the years it's often better to not use the original data to constrain model features/geometry.
Often making a 'fresh' model and massaging it to match the scan data will result in something more usable for your manufacturers. If in doubt, share yo
You're not being a pain - I find this field pretty fascinating and was surprised by how far it's come since I did my first scans almost 20 years ago.. so I share my more recent experience since you obviously have some interest/use. It definitely IS largely semantics, but the newer process doesn't necessarily use discreet 'points' these days. To me, a point implies that specific xyz locations are tracked on a part, which is how the older methods DID work (and for laser CMM, and plenty of other scanning tech in-use now). But that came at a cost of having to have a fixed part/scanner and lots of movement tracking which made large or very complex surfaces difficult to manage.
But the newer stuff is done with moving parts/moving scanners and so tracking xyz points-in-space is not as feasible. However, if 'points' is used as a descriptor for surface/depth geometry, then yes - all scanners are basically 'point' collecting
The point cloud definitely has it's uses, and whether I'd want one would depend on the project. For very precise stuff that happens to have some complex geometry I too would want both because I know laser-mapped points are going to be accurate to typically about .002, whereas the iterative surfaces are often .020 at best. But the surface mapped stuff will get me close in shape/function with far less work/effort. Usually the precise stuff I can measure directly with conventional tools and simply model that into the curved solid. But for me, having 'both' means having to pay two contractors to do it and increases the schedule too. Even if one place could do both for me, I wouldn't trust it because there would be an incentive for them to make one match the other - with no guarantee either is precise (fell for that before, lots of money wasted and I still couldn't get my stampings made).
Phreakish
Well-Known Member
Oof. I'm a fan of 'use what works', but I've found over the years it's often better to not use the original data to constrain model features/geometry.
Often making a 'fresh' model and massaging it to match the scan data will result in something more usable for your manufacturers. If in doubt, share your model/geometry with the shops you plan to work with to see if they can use your data and adjust accordingly. Most shops will be able to use almost anything, but if they have to rebuild a whole new model - they're going to build that into their costs. Just something to be aware of.
Where'd everybody go?
- Joined
- Apr 16, 2009
- Messages
- 4,186
- Reaction score
- 5,988
Still looking for the missing files to build a head that was designed before the file existed.
Would be cool to see this but I fear it is a dead project.
Hellrats
Just another dumbass
Somehow I don't think it's gonna happen
Well, let’s look at it this way…..
A 25 YO comes on board and being probably the only honest thing he says is “I was on the B body site and got shoved off.”
Is claiming to have a never seen working HEMI engine that has the amazing dual overhead cam set up and is working on parts to get it running.
Smells like bullshit to me.
He hasn’t posted in two years and only posted n the thread he started.
A 25 YO comes on board and being probably the only honest thing he says is “I was on the B body site and got shoved off.”
Is claiming to have a never seen working HEMI engine that has the amazing dual overhead cam set up and is working on parts to get it running.
Smells like bullshit to me.
He hasn’t posted in two years and only posted n the thread he started.
But I think in this day and age of computer-controlled machinery and 3D printing and all this other new technology that it would be possible to build these heads. I'm going to keep my hopes up.
-
