CNC Making Discussion

[scordatura]

On 3/10/2019 at 7:34 PM, HoGo said:

I've done some vector modelling of F-5 mandolin in RHino but I don't think it would be too diffrent in Fusion. I had to learn on my own mistakes but the result was good. What I found is that the result is more in the technique used to create the model than in the tool (Rhino or Fusion).

Very cool mando modeling. If you are into mandolins, you might be interested that my college roommate was Edgar Meyer. It was very cool to have  Bela Fleck and others hang out in our apartment on occasion. Very interesting conversations and music making to say the least.

[HoGo]

7 hours ago, scordatura said:

Very cool mando modeling. If you are into mandolins, you might be interested that my college roommate was Edgar Meyer. It was very cool to have  Bela Fleck and others hang out in our apartment on occasion. Very interesting conversations and music making to say the least.

:-)  Skip, Hop and Wobble (by Meyer, Barenberg and Douglas) has been one of my most favorite recordings even without mandolin (except one or two tracks), as well as New Grass Revival with Bela - legendary music for sure.

I dug out some old violin drawings I made before family and paid jobs sucked me out of time for violin hobby work and gave it a quick go in Rhino. I will post some pics for ideas.

[HoGo]

First was the basic drawing I took... I had outline of the plate and few crossections (I guess I had three across and one along the top to start with, rest was derived later). Outline was created in Photoshop from scan or photo, I don't remember, you can resize the scan to real size or correct minor distortions in PS very easily and trace using paths (basicly same vectors as Illustrator curves) and exported to Illustrator.

From these crosssections I created the number 8 shaped topographic lines. I did it in Illustrator as I like the way the curves are edited and created much more than in Rhino (RHino also has similar handlebar editing, but I haven't mastered that yet and the clean simple interface of Illustrator makes it whole lot easier). I took each crossection and simply drew line offset by 4--6-8-10-12-14 mm from base (2mm alevation steps) and projected the intersections of the lines with crossection profile perpendicularly onto the base line. Repeat for all crossections  and you have points that the 8 shape curves should pass through (do this on one side of plate if you want symmetric shape, if you have asymmetrical crossections you can either correct the crossection before yu start or just copy in a mirrored crossection and take pair of intersections and your 8 shaped lines should pass somewhere between the pairs of points for averaged shape)

For drawing smooth 8 shaped lines I start at one end (top or bottom, still working on just one half of top) and make sure the direction of curve from that point is prefectly horozontal (in AI holding shift key will lock the direction in 45 degrees increments), then proceed to the next point just eyeballing the direction and shape, in the waist (that is narrowest part of the 8-shaped line) tha direction should be perfectly vertical and at the end again horizontal (so there won't be a kink when you mirror the two halves). smoothing of the curves is especially easy in AI (I haven' found similar editing way in Rhino or even Correl draw) as you can adjust the shape of curves between points without changing position of points or direction through them (angle of the "handles"). By the lengths of the adjacent "handles" you can see how nicely the lines are following the neighboring line. Use as few nodes as possible for smoothens of the shape. I used 5 nods total per (half) one 8-shaped line, I had to add one extra nod near tailpiece on the two outermost lines to follow the body shape closely.

From this map of lines it is relatively easy to create any other crossection (like I did at the corners) to check visually the shape and adjust if needed.

This is the 2/3 of the whole  modelling work. The better source curves you have the easier it is to create the model in CAD.

 

 

[HoGo]

When I import the curves into Rhino I have to check all starts and ends of (half) 8-shape lines if they really start exactly at the centerline - accuracy of AI is low and they typically miss the line by 0.05mm or so which is above typical threshold set in CAD (0.001 or less). I go one by one and move them to centerline (in CAD they will just snap on the line) so when I do the mirrored other half I can join them together into one (thats where the threshold is used).

I lift the topographic lines to corresponding altitude and rotate the crossections 90degrees to create skeleton of the shape.

You can see that I lifted one copy of the outline up - I decided to use the Hargrave style edgework (since this was loosely based upon early Guarneri) and lifted it 5.3mm in this case (just eyeballed where the  crossection curve  would end at outline). so the channeling will be "carved" all the way out towards edge and then rounded off that will hopefully create just about correct edge heights.  Otherwise I would heve to lift the "crest" of the edge to correct height - higher for c- bouts and smoothly connect it to lower height at lower and upper bouts - which would add complexity I wanted to avoid for this one, or I could just make the edgework the same height or for even more simplicity create a flat platform all around....

[HoGo]

Now goes the real modelling... You have to create completely new set of crossections as the original do not create real "mesh" or "network" of curves (intersecting at nodes) due to imprecision in creating the 8-shaped lines (in AI and even in CAD the transformations and curve editing could lead into less than optimal precision, the lines would miss by 0.1mm or less if you were good enough in step 1) and we wanted the crossections go right to the new edge. In the pic you can see the old crossections in red, new in black. (I used the colors to check if I'm close enough)

I could use this imprecise network for creating simple surface using "drape", but I don't like that tool as you have to adjust parameters till you get good enough fit and on "recurved" surfaces results are not optimal (but with the flat platform around, it would be quite OK for CNC roughing).

For each step I create new duplicate f the curves in new layer and leave the previous untouched if I need to step back and adjust something.

First I created new long crossection (interpolate curve through points), starting at the edge (using project-snapping options of CAD to get points at the exact positions where the edge crosses above the centerline) following through the ends of the 8-shaped lines and also center of the arch (highest point). This creates nice smooth long crossection from edge to edge that is almost identical to the original. I could use the same for the perpendicular crossections now (but using the NEW long arch - so we get intersecting network - I colored the old to red and hid them in temporary layer so they were not in the way) but I wanted to do just one half (and save time needed for clicking on all the points across top to half) so I created dummy lines (blue) parallel to base lines (that I created where I wanted crossections) starting at the long arch right above the base lines (you have to work in top viewport to snap the points in correct place (crossections of lines) and look at the side or perspective viewport to see if you are using the proper points). The dummy lines serve to start the crossection parallel to base plane (using intepolate through points & start tangent to curve) and follow through the points where the base line of crossection intersects the 8-shape lines in top view (also looking at the other viewports so you know you are snapping to the correct curve).  Proceed with the rest of curves. I also have one 8-shaped line roughly at bottom of channel I could use similar dummy line at that point to make sure the bottom of the channel will be right there precisely, but this simplified method works good enough if the original curves were well drawn except the areas near points that distort things a bit - I left that alone but I would use "cornerless" outline to solve this and create the corner surfaces separately later to get better results.

You can see how close the resulting network is to original. at the edges the new provide some extra meat which is OK for me...

Final network consists of perfectly intersecting curves which can be used to generate surfaces.

 

[HoGo]

For this one I chose to mirror the network first and use "curve network" command that generates surface from series of intersectng curves in two directions. One direction is "around the top" (the 8-shaped curves) the other are the crossesctions (halved, I had to cut in two the long arch as well). Theis way I created the two ring shaped surfaces around (yu can discern them by density of their network curves). In this command you can connect surfaces seamlessly if you use edge of previous surface instead of the orginal curves and check preserve tangency (or even curvature continuity) with the neighboring surfaces. I did two such surfaces (for the inner one I couldn't use the upper and lower bout crossections so I did them separately) and the center ws capped using patch command with the network of curves inside. I got soem distortion at the lower points because I did the whole channeling in one sweep but it would be better to do arch just down to bottom of channel (that would require using  dummy surface around the channel bottom to preserve direction of the arching at the bottom and ability to connest the outside smoothly as well) and the outside in 4 separate pieces, perhaps the points as separate triangular pieces too.

The last thing was rounding of the edges which is simple drawing one curve that defines the rounding (that crosses the generated surfaces) and generate surface along rail (body outline). Rest is simple trimming of the surfaces.

Fro even better results I would suggest creating crossections not perpendicular to centerline but as "rays" from center of upper and lower bouts (or suitable points closer to center of top) so they would be rather square to the 8- line shape or perimeter.

[Fiddlemaker5224]

Thanks HoGo, that answered many questions I have been trying to solve for a few years. Using the Probe helped me out a lot. 47hrs of scanning and I have a Strad style arching STL file. Now it is a matter of setting the carving software to cut the pattern correctly. I made some cam clamps for the CNC and they work well. I also used some two sided tape to hold the plate to the spoil board. Problem was encountered when the tape released and the edges curled up slightly. The test carve ended up with 0.063 thickness at the edge. Still more work to do. Photo 7 is my attempt at carving the inside. That pattern needs more work. The basic carving for the back worked well, the top not so well.

[scordatura]

Interesting idea Fiddlemaker to use cams to hold the wood. For another approach, here is what Mike Molnar (hope Mike does not mind me posting his pic) does. The hold down tabs can be removed after the CNC does it's thing.

[Fiddlemaker5224]

I think I will have to look into that idea. Thanks to you and Mike.

[carl1961]

23 hours ago, HoGo said:

For this one I chose to mirror the network first and use "curve network" command that generates surface from series of intersectng curves in two directions. One direction is "around the top" (the 8-shaped curves) the other are the crossesctions (halved, I had to cut in two the long arch as well). Theis way I created the two ring shaped surfaces around (yu can discern them by density of their network curves). In this command you can connect surfaces seamlessly if you use edge of previous surface instead of the orginal curves and check preserve tangency (or even curvature continuity) with the neighboring surfaces. I did two such surfaces (for the inner one I couldn't use the upper and lower bout crossections so I did them separately) and the center ws capped using patch command with the network of curves inside. I got soem distortion at the lower points because I did the whole channeling in one sweep but it would be better to do arch just down to bottom of channel (that would require using  dummy surface around the channel bottom to preserve direction of the arching at the bottom and ability to connest the outside smoothly as well) and the outside in 4 separate pieces, perhaps the points as separate triangular pieces too.

The last thing was rounding of the edges which is simple drawing one curve that defines the rounding (that crosses the generated surfaces) and generate surface along rail (body outline). Rest is simple trimming of the surfaces.

Fro even better results I would suggest creating crossections not perpendicular to centerline but as "rays" from center of upper and lower bouts (or suitable points closer to center of top) so they would be rather square to the 8- line shape or perimeter.

Awsome work HoGo, thanks so much for this work, this helps me a lot. Please Keep us updated. 

[scordatura]

Yes thanks to HoGo for contributing to the cause!

[Michael_Molnar]

On 3/14/2019 at 12:22 PM, scordatura said:

Interesting idea Fiddlemaker to use cams to hold the wood. For another approach, here is what Mike Molnar (hope Mike does not mind me posting his pic) does. The hold down tabs can be removed after the CNC does it's thing.

Hey. Imitation is the sincerist form of flattery. Thanks.

[Don Noon]

It will be a while before I get the plate modelling done, but I made a little fixture to use CNC for one of my more unpleasant jobs... trimming the ankles and bottom curve of the bridge.  I wan't sure if this was going to work, but I was impressed with the quality of the finish, using a 4-flute 1/16" end mill for the tight spaces, and a 6-flute 1/4" end mill for the bottom of the feet.  

I use pins to locate the bridge on the fixture, clamp it down, and then remove the pins before milling (you  can see the pin holes in the photo).  I have 4 different bridge sizes so far, each requiring a different pin spacing.  That only needs a different adapter piece (light-colored plywood).  I'll make more to deal with viola bridges, and maybe some day I'll make an entire bridge of my own design.

[carl1961]

I'm impressed too Don,  that great! 

[Michael_Molnar]

Great idea, Don. Someday I too might give that a shot. Anything that eliminates repeated drudge work ranks high with me.

[scordatura]

Good idea Don. I have been using a pin router to do the same. There is not a lot of room in that area in some places.

[Fiddlemaker5224]

I also like that idea, but I have some more learning to do before I tackle Don's approach.

[nathan slobodkin]

I passed over  this topic for several days and now that  I have read it have good idea of how Mr. Ludd felt. The use of this type of machine seems like an awful lot of work to create something which completely bypasses the artistic possibilities of the medium. There is a reason we value Italian Baroque era instruments which were made by or under strict control of one man over French 19th century instruments essentially made by a committee over  machine made German and Japanese instruments of the 20th century. The essence of art is the individuality and personality of the work which unconsciously reflects the society and experience of the artist as well as his health, the weather, what he ate for breakfast and how friendly his wife felt the night before. While small mistakes might occur in hand work the artist accepts that possibility in exchange for having the freedom to push an arching slightly fuller or swing the curve  a bit faster on a corner depending on his mood. This is how one  produces a body of work where each piece has a unique personality which reflects that of the artist or craftsman. For those with physical or fiscal issues which require that a machine replace the traditional apprentice in rough wood removal there is nothing wrong with machinery but the individual traits of the instrument do not come from the final scraping or last knife cut. They are also dependent on the direction and vigor of tool use at a much earlier stage of production.

 

 

 

 

 

 

 

 

 

[Michael_Molnar]

Here we go again. I was hoping that we could talk in this thread about CNCing without suffering someone's romantic indignations. 

Anyhow, HoGo, I also lay out my plates much like your method. Doing this in CAM makes me think hard about graceful continuity, and gives me room to fix mistakes before ruining any wood. It's nice to flip the plate about and evaluate the arching. Nevertheless, I have ruined a lot of wood until I mastered the CNC tool as it were. 

From my experience in my father's tool & die shop, I learned how important it is to first decide how to hold the work. Once you decide on this, the workflow can then happen. I chose a system based on 1/4-20 sockethead bolts that are color-coded for their length. I put paint on the socket Heads and keep them separated in plastic bins. Then I made a carrier base from MDF board with 1/4-20 threaded inserts. I think I posted photos of this somewhere. I pick up the hole positions to align my endmill. My point is that your fastening and alignment system always needs attention first. 

[uncle duke]

1 hour ago, nathan slobodkin said:

IFor those with physical or fiscal issues which require that a machine replace the traditional apprentice in rough wood removal there is nothing wrong with machinery but the individual traits of the instrument do not come from the final scraping or last knife cut. They are also dependent on the direction and vigor of tool use at a much earlier stage of production.

I feel and understand the aura of violin making in your words but what is one to do if V.A. administered Colchicine, Tramadol and Allopurinol just doesn't seem to make things better for the hands?

[scordatura]

Here is the evil machine I will be using to rough out parts. I will have pleanty of opportunity to individualize the final product (make mistakes) with the detail work. 

[Michael_Molnar]

That is a beautiful CNC. How do you handle dust collection?

[scordatura]

If you look in the left corner of the pic, there is a brush that attaches to the cutting assembly

[nathan slobodkin]

Michael, no romantic indignation here but with no offense intended to Don Noon if the picture of the bridge ankles in his post are of a finished bridge it kind of proves my point. And if not surely the few knife cuts to get to that point could be done faster than setting up the CNC.

[carl1961]

2 hours ago, nathan slobodkin said:

Michael, no romantic indignation here but with no offense intended to Don Noon if the picture of the bridge ankles in his post are of a finished bridge it kind of proves my point. And if not surely the few knife cuts to get to that point could be done faster than setting up the CNC.

nathan slobodkin, your days coming, you either give up violin stuff or improvise like Don is having too. I rather see Don keeping on the go