[Cold Fusion]

Once again, need facilitates creativity. I've got lots of parts that are 2d with lots of holes and odd pockets. This would be a pain to do with point cloud system. Perhaps we could bring the operator into the picture a bit more. If the entire system could be moved by hand, it would definitely improve the quality of the data received because it would only be the necessary points determined by the user.

So here is what I thought of. A free sliding 2 axis frame similar to a tabletop cnc router. The Z axis could be a simple sliding axis with a leadscrew for raising and lowering a stylus. The key to the design is low friction and minimal skew. There are two options for encoders. Linear encoders could be installed along the rails, or rotary encoders could be attached with timing belts.

Now, what I'm confused about is inputting the data into Cad. With this type of system, isn't it possible to get arcs and lines instead of a point cloud system?

Have I completely lost my mind, or does this idea have potential?

[HuFlungDung]

Andrew,
If you do the joining of the points manually, you can probably do a much better job of reconstructing the object. Otherwise, you are likely to end up with a bunch of splines or a mesh to mess with, and most likely you would end up wanting to straighten up and simplify the model anyways.

I like the idea of the 3 axis linear scale idea, in some kind of gantry setup, guided by hand. However, in real life, I wonder if a person would find it too tedious to do?

I wonder what the users' working impressions are, who have used a Microscribe or Faroarm digitiser?

[Cold Fusion]

True. It might be a bit time consuming to operate, but for the hobby guys, that's less of an issue. In any case, the idea is to have something of similar capability but 1/10 the price of the Microscribe.

What is not clear to me is the software end. Is there a program that can take a direct signal from the encoders and reverse interpolate that into a managable format. I know mach2 can control a cloud point type probe, but that's going against the whole idea of this thing.

[DieGuy]

Think spherical coordinates. an encoder on each pivot point and a known length between pivots and you are in business.

[turmite]

Andrew are you using mach software or Desk?? If so why not just program in the envelope and let the machine digitize the part for you. As Hu said far less boring and you can be doing something else while the work is being done????

Mike

[Cold Fusion]

Mach2. Because I'm trying to avoid huge mesh files and wasted time scanning unneeded areas. The parts I need to trace are quite simple, except when you're triangulating every single hole by hand with calipers and parallels... With some sort of hand controlled gantry tracer, I could probably do the whole part in 5 minutes. Here is an example of what I would be doing.

[turmite]

Andrew all you need for that is 2 axis of recording with a accurate and quick way to move your probe up and down. Mach3 or Newfangle has wizards for digitizing but I am not sure it would do that.

Mike

[HenryCNC]

There are a lot of manual CMMs out there... if you could lock your Z axis and then manually push the X-Y axis around (with a touch probe, of course) you have basically the same thing. The software should just sit there are save XYZ points to a file.

Unfortunately, I can't answer the question of collecting coordinates.

You could bring that XYZ text file into any CAD software (I use Rhino) and make a 2D profile with very little trouble.

[vmax549]

Andrew I have posed the same question to the Mach group. I also do a lot of obsolete part replicating. If someone could write a wizard it could be a simple task. Jog over the hole insert probe, mach then does a probe cycle to determine center position and size of hole, move to next hole and repeat. You could also save points on other vertical surfaces. Then Mach would save out the points to be used in a cad/cam program. You can do the same thing manualy if you need to. You just have to do the work and center calculations but it is a lot better than having to draw it out manually. If I could program we would have such a critter. Mach is capable I am not. "MACH 3 DMM" (:~)=TERRY

[JBV]

Why not use a ordinary scanner? Just put the part on the scanner and a ruler, put some cloth over and scan at highest optical resolution...

Then you load the picture into a cad program and scale it correctly by using the ruler as reference. Then its easy to trace the lines and hole locations...

600 dpi = 0.001666" (0,042333 mm) resolution...

[Cold Fusion]

I've tried the scanner approach, but it never works out cleanly. All the scans have a slight shadow on one side, which ruins the image when I convert to grey scale and do a raster/vector conversion.

[greybeard]

If the scanner could lose the shadows..... ?
How about filling all the holes with a contrasting colour wax, then scan ?
John

[vmax549]

Andrew try the manual DMM approach with Mach it works fine. It is easy to pickup all the major points and write down a few notes. Then all you have to do is connect the dots. You do not need a expensive probe, A $20 led touch probe from Enco will work fine. Mach will give you all the info, you just have to write it down Terry

[Ed_R]

Where can I find more info on the DMM approach and the $20 LED touch probe?

[vmax549]

Ed the manual approach has been around as long as DROs. You simply insert a probe into a circle, move +x until you touch, -x until you touch, divide the distance by 2 and return to that point. Do the same in Y and you have the center position of that circle. plus the last moves gives you the diameter if you apply the offset of the probe tip. To do a straight line you only need the starting point and ending point then connect the dots. On an arc you need the starting point ending point and the outer most edge or the arc and cad can complete the arc line ( or calculate by hand ) On all your points you need to note whether it is an outside point or inside point to know how to apply the offset of the tip. On the LED probe it is just a low cost electronic edge finder. When you touch the part a light comes on. (This only works on metal parts) Hope that helps. This is only effective on simple 2d or 3d parts. Terry

[Ed_R]

I think for what I'm doing that approach will only work for 20-30 percent of the surface. The rest is pretty complex curves and arc.
I have a headache.

[vmax549]

Ed how complex is your part? Have you tried taking a digital picture in grey scale and try to convert it. Sometimes that works quiet well. I use it to make cookie cutter dies.( don't laugh too hard). I take a picture and convert it to grey scale adjust the contrast and convert it to a vector drawing. It gives me a very good head start on Gcoding the part program. (:~)= Terry

[Ed_R]

That's a great idea, Terry. It's a top carve on a guitar, it has concave and convex surfaces in multiple dimensions, but I might have to try that.

[ViperTX]

Well guys it all has to do with resolution.......do you manually scan at 0.100 intervals when you're looking for 0.010 fineness......"Do you all get the pic....I'm sure you do."

[HansM]

I've been experimenting with the process of using digital photos or scans as a starting point for creating CNC sources. I have found that processing the image prior to the trace process is essential. I bring the images into Photoshop and use tools such as Threshold and Posterize to simplify them. Then I bring the touched up images into CorelTrace.

The settings in CorelTrace are important, too. I keep the "Complexity" and "Max Colors" low (around 9), and the "Node Reduction" at about 25. I do not think I have the process down to the point that I can accurately reproduce machine parts, but if pretty close is good enough I can go from a snapshot to cutting very quickly! Here are some images of a knife handle I am making from Rosewood right now.

Lance