Saturday, December 13, 2014

Interjection

I mentioned in the first entry on the Auto-Regulator that I would talk about how it brought me to where I am now. The short version is that the clock put me out of business. I'm still a woodworker in heart and mind. But I won't be on my 1040 at the end of the year.

I had this grand plan to do the full write up on the clock, and then use this news as the punchline. But the truth of the matter is that I've been very much aware of all of the writing and woodworking that I haven't been doing for the past few months, and I felt the need to say something. At the end of the day, the clock wasn't the greatest business decision. It was a lot of work. It was both challenging, and rewarding. And I'm proud of the end product. But it won't be going into production. That's as much of the story as I'm going to publish. Shutting down was heart-breaking, and a relief.

Since mid-September, I've been at home, taking care of my son. He's almost 2. He's awesome. I've sold most of the big tools. A few will be in storage for a while. I have my small bench in the basement right now, with my North Bennet Street tool chest underneath it. And I have my Festool stuff down there, taunting me. I still love wood-work with a passion. My perfectionist streak dictates that I will still reserve my love for only the very best. But I'll have to find smaller-scale projects to build and blog about, that still stimulate me, and still satisfy my perfectionist urges. I'm excited to see how tht unfolds.

In the mean-time, the toddler does his work well, and I'm pretty wiped out when he finally goes down. So the write-ups on the clock remain slow in coming. I have a couple of other projects that got done in the waning days of my business, so there's plenty to write about while I marshal my energy to be creative again, and while I get my available space organized.

My heartfelt thanks go out to all who have been reading, responding, and offering support or feedback of every kind.

James



Tuesday, October 14, 2014

Auto-Regulator, Chapter 4: Cutting the arch, part 1





The arch, the upper side horizontals, and the vertical posts come together in a pair of 3-way miter joints at the top of the case. That's the short version. And from an aesthetic point of view, that's really the version that matters. As long as the joint is cleanly made, the eye will freely run along the lines of the case. But from a construction point of view, things are almost never that simple. If there are any gaps, voids, or other breaks in the surface, the eye stops there, and the mind will take note. Much like a shrieking saxophone or clarinet in an orchestra, it won't matter if the melody is miraculous. It's the shriek that you'll notice, and the reverie will be interrupted. So, to make those clean transitions, understanding what's going on is a huge help... and I didn't properly understand what was going on when I got started on this project. So I'm going to break down this deceptively simple looking joint, before we get into how it was done.



On the side of the clock, the vertical post meets the upper horizontal in a 45 degree miter. That's pretty straightforward. I'm going to refer to this as the side miter.

On the front of the clock, the vertical post meets the arch in another miter joint, that's cut at an angle that I've never bothered to measure in terms of degrees. Those miter lines point from the top corners of the case, directly to the center of the clock face. The inner radius of the arch is concentric with the dial, so the miter line runs radially through that edge. I'll refer to this as the front miter.

The curved top surface of the arch meets the upper surface of the upper horizontal members in a 45 degree miter. And I'll refer to this as the top miter. And this is where things start to get funky in the mechanics of the joint.

The plane of the cut for the side miter is at 90 degrees to the plane of the side of the clock. Or, the table saw blade is at 90 degrees to the table, when those miters are cut on those pieces. The cut for the front miter is also cut at 90 degrees to the plane of the surface. That's pretty straightforward. And in my head, that made everything seem very, very simple. That should have been a clue to me that something was awry, I guess. But because the face miter is cut at a different angle as the side miter, the edge where those two cuts intersect gets skewed to one side. So the three-way miter becomes a three way compound miter.



Each cut defines a planar surface. Geometrically speaking, two planes that intersect will define a line along that intersection. Practically speaking, that line defines the edge that's made where the two cuts come together. And for this joint to work, the edge defined by the two cuts made on the vertical post, the edge defined by the two cuts on the horizontal member, and the edge that's defined by the two cuts on each end of the arch... those three edges must come together cleanly along their length, with all of the mating faces coming together fully.









The test joint actually came together cleanly, but if you zoom in on the picture, and see the different surfaces interacting, you'll start to get an idea of just how many things can go wrong in the joint. Oh, and having one of these come together is hard enough. To cut the arch properly, there are two of these joints to consider, one on each end. Which brings us back to the top miter.

To cut that compound miter, the 45 you see on the surface is defined in relation to the top edge of the horizontal, and the back edge of the arch. The angle of the blade during the cut, which is what makes this a compound miter, is defined in reference to the surface of the parts that will lie flat on the saw table.

But the top is curved. There is no reference surface.

Obviously, to be continued...



Tuesday, October 7, 2014

Auto-regulator, Chapter 3: Making the vertical elements



One of the primary design elements of the case was the use of three way miters. In the pedestal, as I've explained, they're not true three-way miters. The movement case is a different story. But there's more to it than that.

Because the facings of the vertical elements are book-matched pieces of solid stock, there's a miter joint around the outer front edge of the vertical pieces. And, because the inside of the vertical pieces will be visibly sculpted, I needed to have matierial to sculpt away. So, the front vertical pieces are both three part laminations: solid core, with book-matched, mitered solid face pieces.

Mitering an end-grain joint is a fairly straightforward affair. Typically, it's used in case constuction, and the biggest miter joints I've usually seen are on things like blanket chests. That may be up to 18" worth of miter, but it's not too hard. In this case, I had a few more variables to consider. First, the stock was re-sawn out of a larger piece. It was mostly flat, and behaved fairly well. Second, the joint has to come together flatly, and mate with the solid core. So, if there was any concavity, or convexity in the joint, I'm going to have issues, because either the middle or the ends won't mate cleanly with the core. Lastly, these pieces are long. 32" of walnut, plus extra to trim back, and 42" of maple, with extra. So they needed to be long, straight, and perfect.

I trust my table saw to do many things. But a few days after being resawn, the pieces were ever so slightly bowed... and I didn't trust that operation to go smoothly on the saw. So, I rough-sawed the miters on the saw, and built a long-miter shooting board to do the rest.

Lastly, I needed everything to be dead straight. Given the thicknesses of everything involved, I wasn't too worried that things would go awry, but to be sure, I made box beams to provide flat reference surfaces for gluing, and glued everything up.


Thursday, October 2, 2014

Auto-Regulator Chapter 2: Below the Waist



I couldn't help myself when it came to naming this entry. The irony is that in this case, everything below the waist isn't where the real action is.

 A note on process: I started this project with a commission to build 2 cases. One in walnut, one in curly maple. So, you'll see parts here for 2 different cases. As I went through these parts, I decided that I would separate the two cases when I went on to build the movement case. The walnut case would be the prototype, (the one where I made my mistakes) and the maple would be the first official 'production' case. Walnut's a little more forgiving to work with than maple, both in the working of it, and in the fact that minor discrepancies are more easily covered up.

The case is divided up into three basic parts: The movement case, the pedestal, and the base. The movement case is what it sounds like. The waist separates the movement case from the pedestal, and the pedestal sits in the base. The base gets leveled before everything else goes up. The pedestal sits on the base, and houses the power supply for the clock. The waist is part of the pedestal, and all of the cables that connect the power supply to the movement pass through the waist, and up behind the rear panel in the movement case.

The base is nothing more than a mitered box. There are grooves inside, and some plywood parts that fit into those grooves, to reinforce the base from inside. Basically, it's a splining technique, but it also allows me to help with mounting the feet. The top of the base is rabbeted to receive the pedestal.

The pedestal is actually rabbeted around the bottom edge, because I wanted the joint line to be horizontal. This is intended to be a production case, and this joint will not be glued. So, in the event of any gaps between the base and the pedestal, I didn't want those gaps to be visible. So, the pedestal lips slightly over the base, and seats solidly in.



The pedestal is, at heart, basically a mitered box, too. The panels that miter together to form that box are mitered frame and panel pieces, that all come together to give the appearance of a three-way miter on the top front corners. It was important to make sure that the vertical pieces are oriented properly, as the grain is continuous with the pieces that will frame the movement case. It's a subtle detail that will probably be lost on most folks, but it's that kind of supporting detail that really makes the difference, and helps to unite the entire piece. (I find that the supporting details, when properly executed, become invisible. But when they're not there, or when they're not done right, they stand out.) The panels that fill the frames have book-matched or 4-way veneer patterns.

Like the base, there are internal plywood frames to reinforce the structure from inside. The bottom frame is open to allow access to the adjustable feet in the base, if needed. That way, if anything settles, the clock can be leveled without having to take it all apart. (That's the theory, anyway.) The top frame is open to allow cables to pass through, and go up into the movement case. The solid wood, mitered top of the pedestal is glued to this frame, and the waist is glued to it, too.

Because the plywood frame is glued into a groove that cuts across the vertical members of each panel, the miters won't actually be supporting the weight of the case, and the movement. The vertical load of the clock will sit on the waist, which sits on top of the plywood frame, which sits in dados that lock that frame directly into the vertical members. So, while it looks like the miter joints are supporting everything, they're not. The load actually hooks into the structure just south of that top mitered panel.

So, as I said, below the waist, the case structure is fairly straightforward. That's not where the real action is.

Thursday, August 21, 2014

Auto-Regulator chapter one: Breakdown.




 I suppose the first entry for this particular series should be about milling up the lumber. There are many, many instances where boards continue around corners, or where the grain continues from the movement case, past the waist of the case, and down the sides of the pedestal. There are so many individual parts to the case, that I didn't want to run the risk of having too many subtle differences in color, or texture, especially around the mitered corners. So, I cut every piece from a couple of huge beams of 12/4 walnut. Well, I tried to, anyway. In the end, I'd cut the front arch 3 separate times before I understood what was going on in the miter joint, so that's actually from a different piece of walnut.


 One of the many grain patterns that I ended up putting to use was to book match the wood around the edges of the front pillars. The pillars started out pretty thick... roughly 3" square in cross section... because they needed to be thick enough to allow for the wide sections at the end, where the miters are cut. At that point, the material is about 2 3/4" from edge to the widest point in the miter. But the blank is also square in cross section to allow for the sculpting that happened on the inside. (Click on the top photo, and take a better look at the contours on the inside of the case)

The diagonal grain near the top was a fluke. That's just what I happened to get when I opened up the beam... the grain went from being in-line with the board to 30-something degrees on angle from the face.

In the end, the pillars are a 3-part lamination: There's a 2 1/4" square core, which is destined to be sculpted down on the inside, faced on the outside with 3/4" thick pieces that are seamlessly mitered around that front edge. Those face pieces continue down into the pedestal, but the grain's pretty straight, so it's not so obvious at that point. And, the pedestal's not an open structure, so the solid core isn't there. (There's no need to sculpt anything, and we needed room for the clock's power supply, anyway.)

The pedestal parts aren't solid... they're all 3/4" thick pieces that are mitered into panels, and those panels are, in turn, mitered on the edges into a box. But, adjacent pieces along the edge from side to top were contiguous in the original stock, so that had to be taken into account.

Put simply, there was a lot to keep track of when I was breaking down the 12/4 stock... what pieces were book matched, or continuous, and with what. I went through (and broke) many crayons in the process, making notes on all pieces to indicate what went with what. Front pieces for the pedestal, orientation to keep grain continuity with the movement case pillars, or to keep continuity with the upper faces of the pedestal... Not exactly a Rubik's cube, but it felt a little like that sometimes.

-----
Note on the photos: All of the pictures of the finished piece were taken by David Schonbrun.







The Auto-Regulator


From January to July I worked ridiculously hard to develop this case prototype, and the associated process, to house the clock known as the Auto-Regulator. The creator of this particular clock has designed it to be the first atomic clock for home or office, so it's kind of a big deal.

This project was a monster on many fronts. There were many jigs to work out, many issues I hadn't foreseen, many, many late nights, and a lot going on at home, to boot. I haven't blogged since April because, quite plainly, I was fried, and had no time.

That's been getting better, so I'm back. In coming weeks/ months I'll go through the project from start to finish, and explain how it all came together, and how it's led me to where I am now.

Sunday, April 20, 2014

Circular Logic, Part II

Project X has been an utter monster in recent weeks. I hate calling it that. I'd rather come out with the name and details, but it's not my product, so I'm reluctant to say too much until it makes a debut somewhere. But this piece has required so much problem solving that it's eaten up a lot of time. Every time I solve one problem, another one pops up. It's resulted in a lot of sleepless nights, staring at the ceiling in the dark, trying to wrap my head around changing curves and angles, so I have a better understanding of what's going on. Then I have to figure out how I'm going to make all of that come together in the physical world. In theory, this piece will go into production, so I have to get it all figured out. It's forgivable to fudge a few details on a one-off piece, but once you get into multiples, the time spent correcting for errors gets magnified.

I can't wait to do a real write up on the project. I've been wrestling with some really interesting stuff, and came up with some cool solutions. But I hate reading... and writing... about abstract solutions without a meaningful context. I've had to re-invent the wheel on radius cutting, which has resulted in this series on dealing with cutting arcs and circles. I've made a lot of patterns to shape individual parts. There are some 3-way miters that are square, and some that transition into compound curves, that will be visible from all sides, so there's no room for error. I've had to re-examine accurate miter cutting several times over, as well as calibration of angle measurements. (Hence the review of the Shinwa bevel gauge.) And even once I could cut accurately, the first time I cut a test joint, it didn't come together at all. Normal 3-way miters are 45 degree angles, cut at 90 degrees to the surface. But once you move one of them away from 45 degrees, everything changes, and you get compound angles. And again, because it'll be visible from all sides, everything needs to be perfectly cut. Some of it starts to feel like a mathematical proof sometimes, because the ground work needs to be fully developed before it can be referenced in a larger work, and then there's still problem solving to be done on that higher level.

There's also going to be some work on veneered panels, possibly with marquetry in the future. There's going to be a bent lamination... and possibly a bent tapered lamination in future iterations. I'm still working out how I'm going to cleanly mount a piece of curved glass with radiused corners. And all of it has to be streamlined...

As always, the devil is in the details. So much effort goes into making everything look clean, so that the supporting elements can fade quietly into the background. But once you get gaps in the joinery, kinks in your straight lines, tearout, etc, the mistakes all stand out like a squeaky clarinet in an otherwise harmonious symphony. You'll notice nothing else.
 
Back to cutting curves...


----


So, this is where we were last time:


Sliding center point base for the band saw, and for the router table. Dog-leg scissors jig with pins to mount to a blank, for cutting inside or outside radii, with a center pin that will transfer neatly from band saw to router table. For my next trick, I wanted to use it to make a pile of identical parts to use for making a bending form. I figured that would test the system, to see how robust it was.

I'll confess here to being a little too cerebral. Mark wandered over, asked what I was up to, and pointedly remarked that I was doing things in 'long-hand.' And he was right. There are many ways to skin this particular cat, and almost all of them are much more efficient. (Make one master curve, and pattern-rout the rest from that, would be the fastest.) But I wanted to see this experiment through, and see just what I could learn from it, and how much it would do that the regular screw-through-the-center jig wouldn't.


Using the jig for cutting radii in either direction (inside or outside) is pretty simple.  Because the jig has so many holes, it's easy to find a setup that will work. But I realized pretty quickly that mounting pin placement was an X-factor. The holes for the pins are drilled at identical distances from the center, but the distance between pins is also relevant. Once the arc is laid out to locate the mounting pins, you can drill anywhere along that arc. But the chord length- the distance between the two pins- will determine how far into the blank that curve gets cut. Two different chord lengths will result in two cuts with identical radii, but different placement of that cut in the blank. It was one of those details that's obvious in hindsight, but still made me scratch my head for a minute. Since the object is to create a bending form, all of the layers must be identical, so pin placement needs to be the same on all of them.



I laid out the first blank, and set up the pin holes to be exactly the same distance from each side, and from the front edge, and drilled them using a fence and a stop block. Drill, flip, drill, and I have two holes with identical spacing from the ends and edge.


To the band saw, and then to the router table...


Initially, I'd used a smooth pin to hold the center. I switched to a threaded bolt that ended in a smooth pin to hold the pivot point/center because there was slop with just the loose pin. It only made for a difference of maybe 1/64"- 1/32" from one blank to the next, but for a bending form, everything has to be exactly the same. This was the part when Mark made the comment about doing things longhand, and flush trimming being faster. Obviously, he's right. But I wouldn't have learned about just how sloppy the pin was if I hadn't gone this way. 

With the slop issue ironed out, the final stack was just about perfect. There were inconsistencies that I could feel, but they were small enough to fix with a plane. It felt a little bit like cheating, since I was trying so hard to make the the jig accurate enough to not need to smooth anything out, but any play in the pivot point makes inconsistency unavoidable. All things considered, it's still a very accurate system. The fact that I can re-adjust the center point and take a second pass, means I can creep up on a very accurate radius for a master pattern, or on a wooden part. And with the router table,  I can make a finished curved surface that's ready for sanding.


-----

Part 3 will go into a little more theory on dealing with radii. The form is a 2 part form, so it will have a mating piece. But cutting that means taking the convex off-cuts from the concave form, with identical but unknown radii, and finding a way to locate the mounting pins to change the radius.