I finished the rodback chair. I used mustard Real Milk Paint with the anti-foaming agent, and I’m still very happy with the results. I didn’t use an under coat of the Butternut Squash, as I did on the perch, but this chair surprisingly turned out a bit more orange than the perch.

This is a really comfortable chair. I really like the feel of the deeply carved seat. Overall I’m very happy with the way it turned out, despite the mishaps.

I used epoxy to fill the crack in the seat as well as I could. Unfortunately it didn’t fill completely, but I’m confident it’ll be strong enough. The sawdust-filled epoxy seems to hold paint well enough, but it expands and contracts differently than the surrounding wood, so it does stand out a bit.

There was another unforseeable circumstance that set things back slightly. I used two different woods when making spindles, and one of the spindles I chose for this seat was made from the yellow/green wood I haven’t identified. When the first coat of paint dried, that spindle stuck out like a sore thumb: it turned dark brown, almost black!

So that spindle has several more coats of paint than the others, just to turn it yellow. It’s not a perfect match for the rest, but it’s close enough. I have a bunch more of this wood, and it seems to be turning brown on the outer surface as it ages. Probably something in the milk paint just accelerated that process. Now I’m not sure what I want to do: I’d like to use the spindles, but I’d rather not have to add 5 coats of paint to all of them. Maybe I should use a darker color, but then I wouldn’t have matching chairs.

On all of the pine seats I’ve painted recently, there are tiny surface cracks in the paint. I think this may be related to tiny pockets of pine pitch in the wood. I think it looks basically fine, the paint hasn’t chipped on any of them. Peter Galbert heats his seats and washes them with naptha to avoid problems with pitch.

This one is marked “Alan Ferrency 2009 4” on the bottom of the seat. I must say, I didn’t expect to make 4 chairs, even stools, in the same year! And there may be more to come.

In the last few days, I figured out how my angle calculation tool was incorrect, and fixed it. Now it’s correct, just not very convenient to use. The reference axes are a bit “here and there” so you really need a sense of what makes sense, in order to extract the right answer.

I then went over all of my previous calculations, and made corrections to the seat plan.

Luckily, the amount of error in an angle calculation was directly related to how far away from perpendicular the hole is. Because of this, the errors on my spindle angles were so small that they almost all rounded off in the same direction as before. The exception was with the outside posts, which had to shift their sight angle by half a degree.

Today, I used my new plan to drill test holes for the back spindles in a scrap of dimensional lumber, to see if the back matched my expectations. It was basically a complete success: I have nothing to change with these angles. The tops of the spindles form a good curve (when viewed from above), and they’re evenly spaced. As shown here, I built up the test back with some spindles I’ve already shaved, along with some sections of 1/2″ pipe. The pipe is a lot straighter, and allows better measurements. (I picked this tip up on Peter Galbert’s blog as well, unsurprisingly.)

This gave me better information than I had about the shape of the seat back. The top of the end posts here are close to where the horizontal rod will end up. It’s wider in relation to its height than I expected. Obviously, my spindles are too long; I’ll need to trim them down on both ends and/or make some new ones. I think it’ll look fine when it’s finished, but it will be a significantly smaller chair than the bow back I made previously. This is good, since that’s what I was aiming for.

Back to the plans: My original leg lines were a complete mess. I couldn’t even figure out the values I used to I calculate them, let alone the error introduced by the tool. Always show your work! They were reasonable angles, and would have made an adequate chair, but they wouldn’t have matched my original intent. From a design perspective, I don’t know whether my plans or the old angles would result in a better chair, but I’m going to try my original plans and work from there.

I’ve already shaved some spindles, and they need to dry a lot before I can use them. I’m not using a kiln right now, so I really need let them sit around for a while: hurry up and wait. Next I need to make the top rod, bend it, and let it dry as well.

This requires a steam box, so I started building one. It is small, 30″ long, becuase I plan to use it in the kitchen. We’ll see how well it ends up working in the next week or so. I also need to accurately measure the location for the rod, so I can figure out how long to make it and build a bending form.

“I love it when a plan comes together.”

On Monday, I drew up some plans for my next project: a rod back chair.

Some woodworkers look at windsor chairs, and think they’re hard to build. Building windsor chairs isn’t difficult, but it is different from other “normal” woodworking. Non-woodworkers don’t necessarily think Windsor chairs look hard or easy to build; they have nothing to compare it to. I think the different way of working which windsor chairmaking requires fits well with my longing to extract order from chaos, and my generally intuitive way of solving problems.

The way windsor chair plans are laid out and used is also different from joined furniture plans. The plans tend to be very compressed. A rule of thumb in Information Theory says that “information (as opposed to pure data) is any difference which makes a difference.” By that measure, windsor chair plans are nearly pure information: they record only the aspects of a chair which are important, and only those which differ from other chairs. Anything not included in the plans is either not a critical dimension, or it’s derived from a partially finished chair in the same way it’s done for similar chairs.

In most woodworking, the order of the day is “square corners and flat surfaces.” The first step in a project is producing boards with all surfaces perfectly flat, and all corners exactly 90 degrees. Throughout construction, these flat surfaces and right angles act as references for various measurements. If you’ve done your job, most of these reference surfaces are interchangeable. Many finished projects retain these square angles in the finished construction.

On the other hand, a windsor chair has only one flat surface: the bottom of the seat. You usually can’t even see it, which is probably why it’s left flat. Most of the parts are never transformed into a flat board: they’re split out of logs and turned directly into legs, stretchers, spindles, posts, rods, or bows.

Since the seat is the only flat board used in a chair, it is the only natural reference surface. Because of this, windsor chair plans are centered around the seat. A seat pattern defines the outline of the seat, and the location and angle of every stick which protrudes from it.

Usually, none of the drilling angles for the legs and stretchers are included in a plan, or they are included only as a sanity check. When building a chair, you derive the required drilling angles and stretcher lengths from the partially finished chair instead of using predetermined dimensions, so your chair will fit together even if the legs aren’t installed perfectly.

This is another difference between windsor chair building and “other woodworking.” In most woodworking projects, you (ideally) construct a number of identical interchangeable parts whose dimensions are all fully specified, and then you assemble them. If you mess up a part, make a new one. You almost never take measurements from a half-finished project and build a part which fits. That would be like hand carving the missing piece from a jigsaw puzzle after it’s almost finished.

So, how do you make a chair plan? I don’t know. I mean… I don’t know how other people do it, and no one ever taught me how, so I figured it out. I may have gotten it wrong.

I started with measured drawings, which I cut and pasted together to show the chair I wanted to make. I know what a finished plan should look like. Making the plan is just translating the drawings into a usable plan. The problem with measured drawings is that many of the measurements are irrelevant, and the rest are not measured in the way a chair maker uses them. To create a plan from a drawing, I had to convert the measurements on the drawing into the measurements used to build the chair.

The biggest conversion task is angles. Measured drawings generally show 3 views: top, front, and side. You can easily measure the rake and splay angles of the legs and spindles from the front and side views of the chair, but these are not the angles used when drilling holes for a chair. Instead, you use an incident angle, which is the angle at which the drill bit enters the seat, and a sight line which tells you which direction to point the drill (the bit is always perpendicular to the sight line).

Apparently Drew Langsner has charts in the back of one of his books, which convert rake and splay angles into incident angles and sight lines. Unfortunately, I don’t have a copy of that book; it’s out of print and expensive. But I do have a bachelor’s degree in math/computer science, which was even more expensive. So, in lieu of a set of tables, I wrote a script.

Unfortunately, now that I’m “done” with my plan and I’m sitting down to write about it, I see that my script is not giving entirely consistent results. Doh! I’ll have to go revisit that before I post a link or build the chair.

I ran into another complication with the drawings I’m working from. This chair has its seat at a 3.5 degree angle in relation to the floor, but the drawings depict everything with the horizon set to the floor, and not to the seat. There was no way to measure the “front” angles properly with relation to the seat. So my script had to take that angle into account as well.

The result of these labors is the plan shown above. As you can see, it only depicts one half of the seat; the other half is a mirror image. I follow Brian Cunfer’s practice of using half-patterns, to avoid screwing up half of a pattern and building asymmetrical chairs.

With only a seat plan, there are clearly some undefined aspects of the chair, but only some of them are determined in advance. I’ll have to calculate the length of the legs to achieve the proper seat height, but the shape of the turnings isn’t very important from a structural standpoint as long as the leg is thick enough at the joints. The stretcher lengths and angles are measured from the position of the legs in the chair. The post lengths are chosen from a design perspective, and the spindles are made long enough to fit the bent rod. The rod itself needs to have the proper length and curve to meet the spindles and bows properly, but if I’ve calculated the spindle angles correctly they’ll match the measured bow from my drawings.

There’s a saying that “In carpentry, you work to the nearest 1/16th of an inch. In woodworking, you work to the nearest 1/64th of an inch. In boatmaking, you work to the nearest boat.” In other words, it’s more important to keep water out of the boat, than it is to get the measurements correct. Stated another way, this applies to windsor chairmaking as well: It’s more important for your parts to fit together properly, than for the measurements to be correct.