Making Kumiko on the Tablesaw: Description and Build Notes
This is my spin on making Kumiko on the table saw. It is my variant of the approach Mike Farrington wrote about in Fine Woodworking called "Create Kumiko at the Tablesaw". In some areas, I think I improved upon Mike's approach, and in other areas I used power tools whereas Mike used hand tools.
Overall, I found this whole project was an exercise in precision.
The first step is to cut and mill strips that are exactly the same thickness as your saw kerf. In my case, I started with boards that were ½ inch thick, and I trimmed off 1/8 inch slices. That is too thin to push the wood through between the blade and the fence, so I had to put the bulk of the board between the fence and the blade, and leave the 1/8 inch hanging out on the other side of the blade. The downside of this approach is that you have to readjust the fence after every slice. This can be ameliorated a bit by cutting multiple board at the same time. Then you can adjust the fence, rip each of the boards, and then readjust the fence.
At first I tried to be really careful with how I adjusted the fence, but as I expected, there was too much variance from slice to slice. Because of this, one had to rip the slices slightly thick, and then trim them to the desired thickness.
These strips were clearly too thin to go through the surface planer. Mike's approach was a jig and a hand plane. I tried this, but either my plane or my technique was not up to it. Eventually, I ended up configuring my router table as a sideways “surface plane”.
My first attempt was feeding the strip in from the left for a conventional cut. When the spiral bit hit the end grain of the slice, bad things happened. That didn't work. My second attempt was to try to feed from the right (Danger! Danger!) for a climb cut. I knew it was dangerous and I tried to be really careful and keep a firm grip on the slice. That attempt was worse than the first one.
Eventually the approach I settled on was to make a custom featherboard that would hold the wood against the fence before and after the blade, and then to bevel the leading edge of the slice with a block plane. This resulted in the end grain not being impacted by the bit. Instead, the end of the strip slid behind the bit, and then the bit gradually bit into the surface. That worked really well.
You'll also note that I clamped a board against the fence at the bit to use as a fence. This served two purposes: I avoided a bump where one fence ended and the other began. More importantly, I could grab the strips before and after the temporary fence without pulling the strips away from the fence—it gave me some space behind the strip for my fingers.
This worked amazingly well. I would feed a strip in from the left, pushing with both hands. Partway through I would transition my right hand to the outfeed side, so that I was pushing with my left hand and pulling with my right, and then I would transition my left hand to the outfeed side, so that I was pulling with both hands.
One thing that I learned was that I needed to have the featherboard loose enough that I could pull the board through with one hand. On one occasion, I had the featherboard too tight, figuring that tighter was better, but then I found I needed both hands to move the strip. This resulted in momentary pauses when I was moving one hand or the other, and these left little burn marks on the surface of the wood.
I found that if I wanted to have a snug fit in the kerf, a few thousands of an inch matter. On the other hand, I found it was fairly easy to position a dial indicator against the fence, loosen one end of it, tap it gently, and move it by a thousandth at a time.
The key feature of the triangular lattice (Kumiko) is the three-way joint. You might think of the strips going in a given direction as a "layer", so that you have a bottom, middle, and top layer. You cut the appropriate notches in the strips for each layer, and then they go together like this:
Notice that the top and bottom layers are the same, just one is turned upside down. So there is really two types of strips to make: middles and top/bottoms.
The middle strip has a pair of kerf-width notches, one on either side, at an angle of 60-degrees. The notches go in opposite directions, but this is automatically achieved when the strip is flipped over. So if one cuts a series of notches on one side, flips the strip upside down, and then makes the same set of notches on the other side, they will line up correctly. Note that the depth of these notches should be marginally greater than 1/3 the width of the strip.
The notches on the top/bottom strips are formed by making kerf-width cuts at +60° and -60°. This results in pointed sides to the resulting notch, where the point should be centered in the strip. The depth of these notches should be marginally greater than 2/3 the width of the strip.
The key to making these notches is a pair of jigs.
The first jig (Jig-1) is a variation of a standard box-joint jig. There is a fence at 60-degrees, and a key (the width of the cut) located some distance along the fence. In operation, one slides the strip against the key, makes the first cut, puts that cut over the key, makes the second cut, and repeats.
To make the middle strips, the blade is adjusted to the 1/3 height, a series of notches made on one side, and then strip is flipped. Started with the same end, an identical series of notches are made. Note that in order for this to work, the end that was initially against the key must have been cut square to the strip.
In my work, I used ½ inch strips, and spaced the key to kerf (as measured along the fence) at about 2½ inches.
The top/bottom strips start out in a similar fashion. One pass is made on Jig-1 in the same manner as for the middle strips, but with the blade height adjusted to 2/3 the strip width. In order to make the second cut for these notches, we need a second jig, which has the fence angled in the opposite direction.
This is an area where I departed from the author of the article. His approach was to make essentially the same jig as the first one, but with the fence angled in the opposite direction. The problem with this is that each cut is referenced against the previous cut made on the same jig. This means that any error is cumulative.
I found that the tolerances on this second jig to be quite tight, because the cumulative nature of the errors. The distance from the key to the saw path had to be almost exactly the same on both jigs. If they were not, then gradually the location of the second cut would drift with respect to the first cut. It doesn't take much of a drift to move the "points" of the notch noticeably off center. There was no good way to adjust the second fence other than making a test piece, seeing how much drift there was (and in what direction), tweak the fence slightly, and try again. Furthermore, I was concerned that if I spent a lot of time and the got the fence adjusted just perfectly, then over time it would shift or get bumped, and then I would start ruining strips.
What I really wanted was a jig that would reference the cut off the same notch as made in the first step. That would mean that the errors would no longer be cumulative. This resulted in Jig-2.
If you look at this jig, you'll see that it essentially mimics the joint itself. The three strips of the joint are represented by the strip being cut, the key, and the path of the blade. The obvious problem with this approach is that the first cut would cut off the key! The solution to this is that the key is movable.
The jig is formed by cutting a notch in the fence. Then the fence is attached to the base so that it is at the proper angle (60° in the other direction) and the tip of the fence notch is centered in the saw kerf. Then a bit of strip is cut so that it slides through the fence notch, and a piece thinner than 2/3 of the strip width sticks out. A simple screw in the base controls how far back the key can slide.
In operation, the key is slid out, the notch from the first operation is dropped over the key, the strip is held to the fence so that it doesn't slide, the key is slid back, and the cut is made.
An additional advantage of this approach is that it will work regardless of the spacing of the key and cut in Jig-1. An alternative that I thought of later was to make essentially this same jig, but with the key moved down the fence by the grid spacing. This means that each notch would be indexed off of the previous notch, but the difference between this approach and the original one is that it would be indexing off of the previous notch cut in step-1, not the previous notch as cut in step-2. This means that once again, the errors would not be cumulative. The advantage of this alternate approach is that the key could be fixed. The disadvantage is that they distance from the key to the kerf would have to match that of Jig-1 (although not as precisely because the errors are not cumulative).
I think that Jig-2 is the better approach, and using it is almost as fast as using Jig-1. In honesty, however, I have to admit to once forgetting to pull back the key, cutting it off, and then having to make a new key. One other time, I pulled it back but not far enough, so I cut off the end of the key. In this case, however, it was still long enough to be useful, so I didn't need to make another key.
One of the keys to using these jigs is consistency. It helps to always push the wood consistently in the same direction, e.g. towards the key. You also need to avoid as much as possible slop in the jigs, e.g. either in how the runner on the jig fits into the miter slot, or how the key fits into the kerf cuts.
I found that to adjust the blade height it was best to adjust it as best one could, erring on the side of being slightly low, making a test cut, measuring the depth of the cut, and raising the blade slightly until you were happy with the depth of cut.
One of my improvements on the jigs was that I glued some small blocks next to where the saw blade goes. This was to make sure that my fingers, which were somewhat near the blade, didn't wander into the path of the blade. As long as my fingers were on one side of the block, and the blade was on the other side, my fingers were safe.
One issue that I ran into was getting some tear-out at the near end of the slot (where the blade leaves the slot). In theory this shouldn't have happened because I was essentially cutting into a zero-clearance backer, but it happened—not always, but sometimes. Initially I thought that this was happening when I backed the jig back across the blade, so I started pushing the jig through blade, removing the strip, and then pulling it back, but this didn't help. Eventually, I decided that this was at least mostly occurring when the strip passed through the back half of the blade. So I clamped a stop onto the table saw, so that the jig would stop when the strip was just past the center of the blade. Then I could back it out and move on to the next slot. This seemed to work better for me.
After making one set of middle strips and two sets of top/bottom strips (one for the top and one for the bottom), you get a series of strips that look like:
These are then assembled in the obvious way to make the triangular lattice. You could do this with no glue, gluing all of the joints, or just gluing the joints at the edges. In any event, if you use glue, you should just a tiny bit on the end grain, so that there is little to no squeeze out. I used the point of a toothpick to apply the glue.
One issue that I ran into was that after assembling the middle and bottom layers, when I went to put in the top layer, I found that the spacing of the assembly was slightly larger than the spacing of the notches on the top layer. This meant that I had to compress the assembly slightly to get the top strips to fit. Initially I thought that this was a good thing, as it tightened up the joint, but then I found that on larger panels, it tended to warp the panel. If you picture the triangular grid, and then apply compression on one diagonal at the top and on the other diagonal on the bottom, you end up with something sort of resembling a potato chip: two opposite corners are high and the other two corners are low.
I'm not exactly sure what is causing this or how to fix it. I tried making the strips slightly (by a few thou) thinner, so the joints wouldn't be as tight. This helped but didn't eliminate the problem.
There are a number of different patterns that one can use for infill. I chose to implement two of them.
The first was the Kasane-rindo pattern. This is formed by two strips, which cross each other at 90-degrees. The long end is beveled on one side to form a 15-degree angle, and the short end is beveled to form a 75-degree angle. Each piece is notched halfway through at 90°, so that they can interlock.
The author made these with jigs and hand plane. My skills and/or tools were not up to it, so I devised ways to do this with the table saw. Not only was this more neat, accurate, and repeatable, but it was also significantly faster. This matters when you are trying to make a lot of them.
For these jigs, I put several fences (jigs) onto one base, so that I wouldn't need to make many bases.
These pieces used three gigs. The first one cut the 15° bevel on the end of a long strip. It is the bottom half of this jig:
The jig itself is trivial—just a fence at the appropriate angle. The tricky part was how to hold the strip during the cut. Due to the angle of the cut, there was no way that I was going to hold this with my fingers. Particularly as it needed to be held fairly firmly, as the force of the blade wanted to make the strip slide along the fence.
I first tried using an existing clamp, perhaps a c-clamp, but these didn't work. They were fiddly to set up, they didn't hold the strip near the point where it was cut, and it put a piece of steel too close to the blade for comfort.
Eventually, I ended up making a wooden clamp. I thought about fancy mechanisms for applying the clamping pressure, but I settled on something really quick and easy. I just got a large cap screw, drilled a hole through one end of the clamp, and used a screw with some slots cut in it as a crude tap and tapped the hole. I put an Allen wrench into the cap, and I had a lever that could tighten or loosen the clamp easily. I didn't want it gouging up the fence, so I screwed a small strip of metal to the fence for the screw to bear against. This worked pretty well. Occasionally the Allen wrench would fall out of the screw head, not too often. I could have glued the Allen wrench in, but I wanted to be able to keep using the Allen wrench for other purposes.
This jig made the 15° bevel. The second step was to use another jig to make the 75° bevel. This one had a movable stop, because it was determining the length of the piece. This stop had to be tweaked to make the pieces fit well. The strips could be held by hand or by a spring clamp as the cut was made.
Since these two steps need to be alternated, then these two jigs really need to be mounted on the same base, so that you can alternate their use without having to constantly swap jigs.
The third step was cutting the notch. This was done with a third jig. In this case, the stop was adjusted to put the notch in the correct spot. I referenced the slot off of the 75° bevel, because it was generally a lot cleaner than the 15° bevel. That one tended to be a lot more feathery and not as crisply defined.
The second style of infill that I made was the asa-no-ha style. This is basically a strip of wood with one end pointed by two 30° bevels and the other end pointed by two 60° bevels. Three of these pieces go like the spokes of a wheel into the middle of one of the triangular lattice cells.
The pointed ends are made by cutting a bevel, flipping the strip, and cutting a second bevel. If the location of the strip is the same for both cuts, the two cuts will intersect at a point along the center of the strip. Like the previous infill, these pieces are made with three different jigs.
The first step is to cut them to length. This length is not critical, but it needs to be slightly longer than the finished length. This is done with a trivial jig composed of a 90° fence and a stop. This results in a bunch of rectangular pieces.
The second step is to cut the 60° bevels. This uses another jig. The process is to put the piece in the jig, make one cut, flip the piece over, and make the second cut. As long as the piece registers against the stop properly in both passes, the point will be centered.
Again, the length in this jig is not critical as long as it is slightly too long. Here, one can choose to hold the piece by hand or with a spring clamp.
The third step is to cut the 30-degree bevels. This is the tricky one. I again had to make a wooden clamp to hold the strip during the cut. However this time, the length of the clamp and the amount of wood available was less than for the previous clamp. My first attempt failed. I had made it similar to the previous clamp, but when I tried tightening the bolt, I ended up cracking the end of the clamp off.
I ended up making a pretty kludgy-looking but effective clamp. For this one, I used a piece of threaded rod laid in a channel to handle the tension, and some wooden blocks to deal with the compression. It looks ugly, but it worked.
The length of this cut is absolutely critical. I was amazed at the level of precision that was needed. I again ended up breaking out my dial indicator when adjusting the stop on this jig. You want the piece to slide into the lattice with a little resistance. I found that if the length was a few thousands short, the fit would be sloppy, and in particular the center junction would skew so that there was tiny triangular hole in the center rather than the three points being coincident. On the other hand, if the length was a few thousands long, the third piece wouldn't fit in the available space. I found that when I was making a new batch of these pieces, I would first cut three of them, check them for size, and then tweak the stop on the third jig accordingly.
My first real "products", were what I call "snowflakes". These feature each joint have all six spokes, with the outside edges being cut to a consistent length. This was done with yet another jig (just a 90° fence and an angled key the appropriate distance from the kerf). On the larger snowflakes, this required short pieces to complete each point.
As one can see from the pictures above, my strategy was to make the lattice out of maple, for a pale white-ish "background", and then use a variety of colored woods for the infill, in various patterns. Even though I only made two types of infill (there are a lot more possible), by changing the colors, a large number of patters was possible. For the infill, I used cherry, black walnut, purpleheart, and wenge.
The largest pieces that I made called for a rectangular frame. I couldn't think of any larger project in which to embed these frames, so I made the frames as the entire project. It is a little tricky turning a triangular lattice into a rectangle.
I made the lattice so that the middle layer ran horizontally from side to side. I glued up the joints at the edges of the lattice, but not in the middle. Then I cut the lattice vertically through the center of the joints so that exactly a half triangle was left. To do this, I first marked out the cut line, then trimmed the pieces to just longer than this with a scroll saw. I would have used a band saw, but I don't have one of those. Then I carefully clamped it onto a sliding table jig, so that the apex of the triangles lined up with the cut line, and make the final cut with the table saw. (Originally, I didn't use the scroll saw, but I found that as I was making the cut with the table saw, the cut off pieces would sometimes fall into the blade and become projectiles.) Then I glued a strip to each side.
When the glue had dried, I then trimmed the top and bottom with the scroll saw and table saw. Originally, I tried trimming the top and bottom flush to the top and bottom strips. But I decided that this looked ugly where the diagonal strips came through. This wouldn't be an issue if the frames were being mounted into a piece of furniture, but it mattered if the frame was an object by itself. So I ended up trimming the top and bottom so as to take 1/32 of an inch off of the top and bottom pieces. This left them 3/32 in thickness. Then I trimmed 1/32 off of a maple board, and glued that on as a really thin piece of veneer. Originally, I cut the veneer strips from the same boards that I used to make the lattice strips, so the veneer was ½ inch wide. I tried to align it very carefully against the top and bottom, but inevitably, I would get some misalignment. Eventually, I started cutting the veneer from ¾ inch stock. This made the glue-up much easier, and then I trimmed them veneer edges with a block plane. (See, I don't use power tools for *everything*!)
After that, it was lots of sanding. The primary goal was to level the surface, particularly between the infill and the lattice. I didn't trust using a random-orbit sander. I was afraid that it might catch on some edge (particularly of the infill) and rip some wood off. So I did all of the sanding by hand. I finished the pieces with Watco oil. This was a bit of a pain, as there are a zillion small nooks, but I can't imagine trying to do this with a film finish like polyurethane. I ended up using a pipe cleaner to apply the oil, and then paper towels to remove the excess. The paper towels weren't the best, but I didn't know what would be better.
My final results were:
Here are some closing hints and notes:
I had some irregularities in the grid—not all of the triangular sections were identical. When I tried putting the asa-no-ha pieces in, they would be tight in one section but loose in another. This points out the need for consistency in making the lattice.
I had some issues matching colors between pieces for the infill. For example, some pieces of wenge were almost black, but other pieces were just gray.
When I was cutting the sharp bevel on the Kasane-rindo pieces, sometimes the wood strip wouldn't be completely vertical, so that the sharp end of the cut ended up being angled. I'm not sure what was causing this, but my guess is that perhaps those strips were a bit twisted.
I also found that on some of the types of wood, when making this same cut, the thin sharp corner on top would tend to chip off, so that the finished piece was effectively missing a bit of one corner. Sometimes I found that orienting the strip the other way would help. Some of the woods were more susceptible to this than others.
The asa-no-ha pieces tended to be much more fiddly regarding length, but if the length was right, they tended to fit well. The pieces for kasane-rindo tended to be more forgiving of length. If they were too long, one could easily trim a tiny bit off of the sharp end, and it wouldn't be obvious. On the other hand, some of the time, the long cut face wouldn't seat closely with the lattice--there would be a small gap there. I "cheated" and used glue and a clamp made from a modified clothes pin to close the gap.