Half Blind Dovetails

[mrdovey]

Jay Mack sent me some pix of some CNC-produced dovetails and got me thinking (re-thinking, actually).

I ended up writing a pair of programs that allow specification of angle, height, maximum tail width, and displacement from the bottom of the board - and which are capable of producing common, half-blind, and full-blind dovetail joints using only a straight bit.

I thought I'd share a photo of the first test result:


2502

[stickman]

Morris,

I knew if I could get you started thinking on it. Something would come about. Did you do these both flat, or have one upright in your jig. I haven't had time to take the info you gave me and work on it much. But I plan to later on.

Looks good...

Jay

[mrdovey]

Jay...

Tails flat and pins upright (in the jig). I used a 1/8" center cutting end mill.

Thanks for the inspiration!

...Morris

[paco]

Hey guys!

I assume there's radius... and they are matching...?

[mrdovey]

Paco...

Of course there's a radius - and yes, wherever there's a radius in one part, there's a matching radius in the other part. :-)

The trick is to make the radius small relative to the size of the tail. The 1/16" radius in these parts isn't very visible; and I guarantee that if I'd cut the parts with a 1/32" bit, you'd never be able to distinguish the 1/64" radius from a sharp edge.

What's fun is that only six ShopBot commands are needed to cut the outline (external) of a tail or (internal) of a pin gap! The space between pins takes a few more M2 commands to do the area clearing, but since this area won't be visible in the assembled joint, a large stepover can be used.

...Morris

[mikejohn]

Morris
You say "Tails flat and pins upright "
I can see how the joint can be made pins flat and tails upright, but I am at a loss how to do it the other way around. I am assuming the pin is on top in your picture.
............Mike

[gerald_d]

Mike, the top part looks the nearest to a dove's tail, so I thought that was the tail and the other part the "pins"?¿?

Some ideas for round cutting tools.....link (http://www.thailandtradenet.com/photos/catalog/the-wood/dovetail.joint.jpg)

[mrdovey]

Mike...

My understanding is the same as Gerald's - but I could have it backward.

To remove ambiguity, here's a photo of what I've been calling the "pins" side still in the 'Bot jig:


2503

...Morris

[mikejohn]

Morris and Gerald
You are right.
If you imagine through dovetails, then cut off the tail at its base, the cross section will be rectangular. Cut off a pin it will be trapezoid at its base. Note when you cut them off, they both have the same shape!

Do you cut full depth with a 1/8" (3mm) bit?
............Mike

[drodda]

Morris,
Looks great. Do you allow some clearance so that they will go together after a day or so laying in the shop or do you have to assemble them immeadiatly?

Dave

[mrdovey]

Dave...

Well, I just did a test and can report that they fit just like they did yesterday.


Hmm. It never dawned on me until just now to add a parameter to specify the width of an optional glue gap. Oh well, back to the drawing board...

...Morris

[mrdovey]

Mike...

For now, I've set up the programs to make multiple passes, stepping down the diameter of the bit for each pass. I could probably get away with stepping down as much as three diameters; but I only had one sharp 1/8" bit on hand.

The part programs adjust step down value and compute points on the toolpath based on the current bit diameter (gotta love access to those system variables!)

Just out of curiosity, what were you thinking?

...Morris

[mikejohn]

Morris
I am looking hard at furniture joint making over the holidays. (Frowning if I trundle off to the workshop from nearest and dearest, so its computer theoretical stuff).
My thoughts so far are:
Precise positioning of materials square to x,y and z axis, and precise to 0,0,0.
Getting a joint that fits well, but not to tight.
Wondering if traditional joints are still the best joints now we have the capabilities of the 'bot.
Wondering if its possible to build a library of .dxf joints to use in different situations.
Trying to balance accuracy with speed.
Which is where the depth of cut came in.


..............Mike

[mrdovey]

Mike...

Ahh... Exploring the zen of woodworking - one of my favorite recreations! Then let me share a few thought fragments...

My positioning efforts have been much facilitated by adding a reference edge to my table. It can be removed when necessary; but for 99% of what I do on the 'Bot, it provides dead-on alignment with the x-axis. Since I machined the reference edge in place, I know exactly where it is and use it as a reference point for y=0.

Dave Rodda inspired me to cut a set of dummy bits from drill rod. I have 3" lengths in 1/8", 1/4", 3/8", and 1/2" diameters. I can chuck one of these into my most-commonly-used collets and use the 'Bot itself as a sort of stop block for precise (to the limits of 'Bot precision) workpiece positioning. You'd probably benefit from making a similar set in metric sizes.

I like joints that almost don't need glue. I've discovered a couple of CNC-only joints that tighten with any change in moisture content without causing fracture - there's one such on my web site. This is the joint that Dave was yanking my chain about (he's visited the shop and seen it first-hand). [That joint came about as a result of Dale Kerr's nudging to produce a tilt-top material handling cart here on the forum.] From my perspective, the only way a good joint can be detected is by the interruption of the grain pattern at its boundries.


The traditional joints are good; but modern glues provide a strength that in the past could only come from the joint structure itself with a much lesser strength contribution from the glue. I like the idea of making high quality dovetails - not because they're (any longer) stronger than other joints - but because by producing them well I feel as if I'm honoring the incredibly skilled craftsmen of the past who didn't have precision robots and glues that bonded more strongly than the wood fibers themselves.

I took a shot at building a library such as you suggest a couple of years back and gave it up because I could imagine no way to relate the drawings to the hugely different behaviors of different woods in different environments. The effort to build a meaningful library struck me as overwhelming. Worst of all - the more I learned, the less I knew.


After all of that, it's got to sound pretty lame when I say that I went with a shallow depth of cut because I didn't want to risk breaking my last 1/8" bit on a preliminary test...

...Morris

[charles_o]

Morris
A couple of questions

How long does it take to cut a joint like the one above.
Have you tried to do this with a dovetail cutter to speed production

Charles

[mrdovey]

Charles...

This first test took almost six minutes on my PRT. I think that could be cut in half with a more agressive stepdown value; but that'd be pure conjecture at this point. I will guess that it could be done some faster on a PRT with a faster computer and (perhaps) very much faster on an Alpha.

I haven't tried to do this with a dovetail cutter on the 'Bot because I already have a conventional dovetail jig. I was after some of the things that aren't generally options with normal dovetail cutters:

[1] The higher strength and greater gluing area of the full rectangular cross section joint elements.

[2] The ability to use dovetails with small angles - one of the hallmarks of hand-crafted dovetails.

[3] I wanted the ability to use dovetails of different geometries in a single joint - which includes not only mixing wide and narrow dovetails; but also asymetrical dovetails.

I've seen photos of attempts to use specialized dovetail bits on CNC machines and I really didn't care for the quality of the resulting joint. The photos Jay sent, for example, show a CNC-produced joint that (1) left a tail-sized semi-circular void hiding behind the tail and (2) the tail part required a shoulder large enough to cover an unsightly top radius.

My reading is that a specialized dovetail cutter could be used to speed production and lower costs - but that the quality of the result would make the product not worth buying.

If I were making speed a priority, I'd go for a second (tandem) z-axis and a second stepper for my y-axis - and if I had the money to spend, I'd opt for a tool changer so I could use a large bit to hog out the bulk of the material and only use the small bit to dress the inside angles.


...Morris

[stickman]

Morris,

What you have got there, looks really good, I like how you used just one bit to get your results. It looks as though you might have something there you could easily convert to a through type dovetail joint. Even with the slight radius, that could be part of the joints feature.

Look forward to seeing more.

[gene_marshall]

Morris Doveytail joints, I love it.

I have a few drawers to make this week and would like to try this.

gene

[mrdovey]

Thanks, both Jay and Gene.

Jay, both through (common) dovetails and full-blind dovetails can already be achieved by changing the existing cutting parameter variables.

I don't particularly care to have noticably radiused edges; but they're controllable to a reasonable degree by choice of bit diameter. With a single bit, there's an inverse relationship between size of radius and cutting time. I rather like that the radii in the finished joint will clearly signal the maker's prioritization of quality.


Gene, there've been photos of my first two prototype jigs in plain sight on my web site for well over a year. The jig isn't complex, but needs to be built and trued with exacting care. My software won't be available until it's been thoroughly tested; but there's nothing to stop you from whipping up a jig and using your favorite CAD package and PartWizard to turn out dovetailed drawers for your project.

You'll probably have your drawers finished and installed before I'm able to satisfy myself that I have a safe and accurate general solution - and I'm not sure what the impact of incorporating a specifiable glue gap [Thanks a lot, Dave!
] is going to be.

...Morris

[mikejohn]

Morris
For a specific glue gap, can you not tell the program your bit diameter is smaller than it really is, by half the glue gap?
.............Mike

[mrdovey]

Mike...

Because both joint elements (pins and tails) contain both "inside" and "outside" arcs, we'd produce a shape mismatch between tails and pins. If each element contained only "inside" or only "outside" arcs, then your shortcut could be used to good advantage.

There are two other (completely independent) considerations that contribute to making joinery on the ShopBot "interesting":

The first is step resolution - there is no way to, for example, accurately split 0.001" (or even 0.005") on the 'Bot.

The second comes into play because each tail, and each pin, is located with a single reference point and all other points in that tail or pin are computed from that reference point and each point is computed relative to the preceeding point. This, in combination with the fact that the ShopBot control program uses a low-precision floating point format, requires that, to avoid unacceptable cumulative errors, I perform those computations outside the ShopBot environment using a higher precision floating point format.

I discovered this when I observed a difference between geometrically identical toolpaths produced by PartWizard and those computed (using sine, cosine, and tangent functions) within an SBP program.

All of this together means that the "glue gap" value needs to be examined and either applied to just one of the joint elements - or (perhaps unevenly) split and applied to both tails and pins.

And no one is likely to specify a gap as large as 0.01" unless they plan to grout/spackle the joint.


In any case, it's a "solve once and forget" kind of problem - it just introduces a bit of complexity to the development that I hadn't anticipated.

...Morris

[mrdovey]

Just to provide a bit of closure, I've completed construction of the 3-1/2 axis joinery machine. Along the way, I posted pictures of each stage of construction on my JBot web page (http://www.iedu.com/DeSoto/JBot.html).

All three axes have a step size of 1/4800" (0.00023") in a 12x12x4 workspace to address the "glue gap" problem.

I've been busy learning g-code.

Morris

[drodda]

Morris,

Good to see that this is coming together. Do you have any pictures of your first test cuts with the machine?

Looks great.

Dave

[mk_elliott]

Why not use the Dovetail bits, and avoid the round corner problem????

[mrdovey]

Mark...

Sorry for the slow response.

I wanted more options - and I wanted a stronger joint than dovetail bits allow and control of the "tightness" of the joint.

Not to hijack the thread - but I'll mention in passing that I built a (possibly interesting to some) fixture that allows me to cut angled tenons with my bot. There are a few photos on this web page (http://www.iedu.com/DeSoto/Bevel.html).

...Morris

...Morris