Official Luthiers Forum!

Modeling off of a point cloud is something that is done quite a bit in this industry. Many of the prototype parts are handmade and it is always a challenge to capture the carver's intent when building a CAD model. Here's a typical process for converting a hand carved part to a CNC part:

1) Hand carve & sand the part
2) Digitization - Some people use Faro arm type digitizers and some use their CNC equipment with a touch probe. Some of the laser scanning technology is coming along nicely but it is generally more suited to measuring larger objects with less precision. .010" can make a ton of a difference on something like a neck carve....and for this reason most builders rely on CNC mounted touch probes for digitizing. The end result of digitizing is a point cloud. While it is possible to write G-Code directly from a point cloud, it is generally best to fully surface the part in CAD. Once it is in CAD you have complete flexibility to add or remove pickups, f-holes, switches, etc..

Here's a pic of a point cloud with the contour of the actual part:


3) Surfacing - This is the stage where we try to form 3D surfaces to match the point cloud as close as possible while I "cleaning up" the little imperfections that are usually present in any hand carved part. The point cloud of a part like the shank of a neck can generally be matched within about .005" or so. A carved top is challenging to get much below .010 or .015. The surfacing process is pretty iterative....you model one thing...check it against the point cloud....tweak it...recheck it...etc..

Rhino has a great tool for checking parts against point clouds. Click on Analyze>>Surface>>Point Cloud Deviation will bring up the interface for checking a part.

Here's a screenshot of a top carve surface being checked against a point cloud using this tool:


4) Compensate the surface for the probe diameter - Most probes have a spherical surface that actually contacts the part when digitizing. It can be challenging to compensate for this. Some digitization software takes care of this...but for the rest of us here is a great way to compensate for the probe diameter using something like Rhino:

http://www.emicroscribe.com/uploads/file/Ball%20compensation.pdf

5) Toolpaths - Using CAM, it is usually a pretty trivial task to write a toolpath. Once the toolpath is done, it's usually prudent to check it out by carving a part, sanding the cutter marks and checking it over to make sure it is nice and smooth and that there aren't any funny transitions or anything.

I get a kick out of this stuff, I hope you find it interesting as well!

Best,
Trev

_________________
http://www.PeakeGuitars.com

Hi Parser...

is this a scan of an actual LP top, or something which you simulated?

_________________
-CyborgCNC
http://www.cncguitar.com
https://www.facebook.com/CncGuitar

It's an actual scan - and surface - for an archtop top plate

Trev

_________________
http://www.PeakeGuitars.com

Hello Parser,

Very cool indeed and thanks for the tips.

I am still deciding on what Cad software at this time and I'm leaning toward the Rhino. What Cam program are you using?

It's hard to look at so many choices, when your trying to stay within a budget.

Thanks again,
Mike

_________________
I'm outside looking in, just farther from the window than most.

Hi Mike,

I use Rhino and RhinoCAM and couldn't be happier. I've used higher end stuff (solidworks and MasterCAM) extensively and I don't think it is possible to beat the value of Rhino and RhinoCAM. Really powerful and cost effective stuff. RhinoCAM is very similar to MasterCAM. Rhino is most similar to AutoCAD...with a bunch of extra surfacing & curve control stuff that AutoCAD missed the boat on.

None of this stuff is cheap, but they are both great tools....and you're only as good as your tools!

If I did have the money, I would get SolidWorks....but I would have to do a lot more CAD work/ mechanical design to really justify the $5k expense.

Best,
Trev

_________________
http://www.PeakeGuitars.com

I digitized the 16" archtop back plate on my Techno using the Mach Digitizing Wizard.
My probe was acting up and giving a lot of false triggers at the top of the Z axis stroke.
This resulted in all the holes you see in the above point cloud.
I've since replaced the probe with this one and it works great.
http://www.wildhorse-innovations.com/index.php?_a=viewProd&productId=80
Trev did an amazing job of developing the surface model in spite of the "holy cloud".
He's programming the job for me now as we speak. I'm impressed with his service!
Nelson

Thanks Trev,
I'm thinking though I might go with Madcam vs RhinoCam as I can get the Rhinocad and Madcam both with an educator license. Whereas the RhinoCam does not appear to give the educator discounts. Teachers don't make much but sometimes there are benefits to having the Teaching certifications

MK

_________________
I'm outside looking in, just farther from the window than most.

how long did this take to scan?

_________________
-CyborgCNC
http://www.cncguitar.com
https://www.facebook.com/CncGuitar

I believe it was about 5.5 hours on this particular scan.
This depends on several factors including the stepover in X and Y between each "hit".
The Mach3 digitizing wizard is basically the "rectangular bed of nails" which is basically a rectangle in X and Y large enough to cover the entire specimen.
I used .200" in X and .100" in Y stepover which resulted in about 16,000 hits.
The Z axis returns to it's Z0 position, after each hit, which is set at a point high enough to clear everything that the probe might contact.
A lot of time is spent probing points in the rectangle that are unnecessary because they are "outside the target".
The return to Z0 each hit also wastes a lot of time.
The Digitizing/Probing Plugin http://www.artsoftcontrols.com/plugins.php is supposed to eliminate a lot of this "air time" by backing off only a small amount after a hit before moving to the next hit point. I tried it but can't get it to run on my machine.
What is needed, and I hope that Henny can help us here, is a probing routing that first 2D probes the perimeter of the specimen and then does a 3D probe bed of nails within this perimeter. This combined with only a small retract of the probe after each hit would be super!
There is a beta probing routine "out there" that does this 2D/3D probing but it is buggy.

http://www.razordance.co.uk/probe25D.htm
This is the 2D/3D probing I mentioned in post above.
Nelson

Here's a pic of the surface without all the point cloud info on it...

_________________
http://www.PeakeGuitars.com

Very nice Parser. If your services are for hire, we may have to talk in the new year.

Trev--What is the process in Rhino for trimming a point cloud?
For example, trimming a rectangular bed of nails with a 2D profile?
Thanks
Nelson

Sheldon, he's rather grouchy to deal with.

Grumble Grumble...OK, I guess I'll reveal a few secrets...

There is actually no secret...I just manually clean up the point cloud by deleting that points that don't appear to be part of the real point cloud. This means deleting points that represent the fixture surface as well as the points that didn't register correctly. I just keep spinning the model in 3d and looking for points that are too far from the rest...window them..and delete them. It takes a bit of time but it's a good first step.

One trick that Rhino gives us is that they actually have a point cloud object. This allows you to add any number of points into a point cloud object. When measuring how far the cloud deviates from a surface, I actually select the the point cloud object I've created and then selecting the surface.

Trev

_________________
http://www.PeakeGuitars.com

Trev I have seen your models and you aren't through with this one...are you? There is a worm hole in it!

BTW, just to butter you up a little.....your neck tut is still the best thing going on the net for Rhino use and guitars!

Mike

Thanks for your kind words regarding the neck tutorial & I hope that has helped a few people out!

Are you referring to those little triangular shadows that pop up sometimes? It always occurs on 3 sided surfaces (I always model 4 sided surfaces whenever i can..but sometimes you have to fill in 3 sided surfaces in some areas)....usually it seems to be a rendering issue only. I check them by running section lines across those portions to make sure that the surface is consistent. Sometimes I can make them go away by resetting the surface tangencies, etc....but sometimes they are persistent.?!? As long as it appears it is only a rendering issue I don't worry too much about it.

Trev

_________________
http://www.PeakeGuitars.com

How did you create the surface from the point cloud? Did you just surface the Rhino Point Cloud Object?

The reason I ask these questions is all the iso curves and multiple segments. I have in my possession (not mine) a model of an arch top plate that is very similar in shape to yours, but there are almost no iso lines on it compared to the one you have here, and I am curious if it is a difference of modeling technique or Rhino tool chosen. (edited) or if this is a rough draft, thus my earlier question about not being finished?

Please don't take this as being critical either, since I am trying to learn about Rhino......and dude is there ever so much to learn about Rhino!

If you like, I will see if I can get permission to post an image and maybe even get the owner of said model to come by?

Mike

It's most likely a difference in modeling technique. I can do one with less "sections" but I'll be sacrificing some accuracy to do so. I try to make the surfaces match as closely as I can to the point cloud while still "smoothing" things out where needed.

The easiest way to model this type of thing...but not the most accurate...is to simply select the point cloud object and the profile and to generate a patch surface to fit it all. It will fit fairly close, but since the patch command doesn't realize it is drawing a carved top guitar it will have little bumps and dips where it shouldn't. Very quick and dirty. I do use this sometimes to visualize the point cloud and to help develop reference geometry (often using the Section command).

The surfacing process for something like this isn't too easy. The basic process consists of modeling a 3D wireframe from which to construct your surfaces and then to surface them out (using network surface, lofts, etc..). I use a lot of the same commands that I used in that neck tutorial. You can make some pretty good guesses about the framework that I used just by looking at the arrangement of the surfaces in that pic. You can even see where some surfaces have been split in order to use portions of them to aid in creation of adjacent surfaces.

Trev

PS - this is not a rough draft..this is the finished part. I just modeled the outside surface of the top plate on this part..on others I've modeled inside, outside, f-holes, pickup cavities, and whatever else you could want on a top...!

_________________
http://www.PeakeGuitars.com

Thanks Trev,

I was just curious if it was indeed a difference of modeling.

Thanks for the continued support of the undernourished Rhino users here!

Mike

There are definitely a lot of ways to skin a cat when it comes to these things. I'm big on using minimal numbers of surfaces, but it can be more work initially and it doesn't effect the end result (cut parts) if the 'real' geometry is identical. Fewer surfaces makes the model easier to work with or rework later, but if the odds of changing anything are very low then it might not be worth the time. More surfaces make accuracy easier, and fewer make tangency and smoothness easier.

I spend a LOT of time working on curves for most models I make, and the trickiest part is keeping up accuracy without ruining surface smoothness (when working from points to curves to surfaces). Surface creation is pretty much all NetworkSrf, and I often create a surface to pull isocurves from and refine into a better surface.

_________________
Bob Garrish
Former Canonized Purveyor of Fine CNC Luthier Services

I do the same with regard to the curves. That's what really controls the surfaces....so if you get those setup right, then surfacing goes pretty quickly. It's usually a pretty iterative process

_________________
http://www.PeakeGuitars.com