Before any of this bending happened, the flat board had already tried to kill the project.
When the clamps came off the panel glue-up from the previous session, where I had glued all the boards together into one piece, I found a gap. One end of the board, one joint that had not pulled tight enough. A millimetre or two of open glue line where there should have been none.
I cut the full metre at a point where the joint was clean, took the section without the gap, and put the rest aside for another project. You do not re-glue a metre-long panel in a shared workshop on the clock. You make a decision and move on.
M had told me before: bring more clamps than you think you need, alternate them one on top and one underneath so the pressure stays even and the assembly does not bow, and plan your clamping sequence before the glue goes on. Not during. The open time problem is real. I was spreading glue across metre-long boards while already five hours deep into the session, restarting after a long hiatus. I did not plan the sequence properly. That is how I got the gap.
Small gap between mapple and cherry.
M had told me. I just forgot.
The board after the salvage, thin strips ready to cut, and the MDF template.
That is the board after the salvage, along with the thin purpleheart and cherry strips I was about to cut, and the MDF template to the right. The plan was to cut the curve with a router following that template.
The jigsaw did not give me what I needed
The original plan was straightforward: cut the curve shape into an MDF template with the jigsaw, clamp the template to the board, and use a router with a pattern bit to follow it. The router copies the template exactly, which means a clean and repeatable curve.
The problem is that it copies the template exactly. I made two by hand, ran my finger along each curve, and something was off on both. I sanded. Still off. A router following an imprecise template does not fix the imprecision, it transfers it directly into the wood.
After a long conversation with M, who runs the workshop and knows wood the way some people know code (me), we went with CNC. Rather than fix the template and then route, the CNC cuts the curve directly into the board in one operation. Given I had already spent time on two failed templates, it was the faster path.
The jigsaw and the MDF are not money thrown away. They will find other projects. But they did not solve this one.
What that decision does not tell you is what actually happens at the machine.
M ran it, and I was glad he was there. I would not have thought to slow the feed speed through the tightest part of the curve. I would not have known how many passes the trim bit needs, or that you have to hoover constantly while keeping the piece firmly fixed for that final millimetre pass when the bit is taking almost nothing and the piece most wants to move.
CNC looks like pressing a button from the outside. It is not.
One practical note for anyone facing the same choice: with a hand-cut template, the thin strips conform exactly to the form you press them against. Every flat spot and every bump in the cut transfers directly into the lamination. You cannot fix it once the glue is dry. CNC removes that risk because the curve is mathematically continuous. It adds a dependency on someone who knows the machine, but if you have access to that person, use it.
Left to right: CNC cutting the curve / the full gap revealed / the rough edge before cleanup.
Titebond Cutting at 3mm is where the anxiety starts
The bent lamination strips are purpleheart and cherry, ripped on the table saw at 3mm.
At that thickness you are not far from the blade, and one wrong move costs you a finger or the strip, sometimes both. We had a kickback. I was standing on the correct side. The wood flew past. That is the whole story, and I am glad it is a short one. A good reminder that these machines do not negotiate.
The zero-clearance insert matters here: at 3mm the strip has almost no support right at the cut, and without it the wood can dive into the gap and shatter.
Blade choice also matters. For ripping thin strips you want a larger gullet between the teeth so the sawdust clears and does not overheat the cut. I am not certain of the exact specification we used. If you are attempting this, ask before you cut.
The purpleheart I had left was 7cm wide. No margin for a wasted strip.
On the second set, we tried hand planing the purpleheart face clean. One pass and it splintered at a third of the length, leaving me 60cm instead of a full metre. For a board that needed 40cm of curve, just enough. Only just.
Titebond 3 and how much glue to actually use
For the bent strips, the deliberate choice was Titebond 3. Food safe, but the reason I wanted it here was the longer open time and the moisture content. At 3mm the wood is already cooperative, but the extra moisture helps the fibres relax slightly around the curve. Every bit counts when you are coaxing hard exotic species around a tight arc.
How much glue: apply across the full face of each strip, then drag a rubber-toothed spreader across it. After the comb passes you will see the glue redistribute into the grooves. When you see that, you have the right amount. It sounds fussy written down. In practice you feel it immediately.
The glue-up: where it got real
Top: the full sandwich assembly. Bottom: how thin those strips actually are.
What this reference video makes look manageable with a good form and a few clamps turned into a three-person operation.
P was there. He is a fine woodworking specialist, and when I laid the purpleheart and cherry strips on the bench I watched his face change. That expression is exactly who you want standing next to you when a cut makes you nervous or a glue-up starts going sideways. He knows what the wood is going to do before you do.
When you start sandwiching thin strips between two curved halves, the whole stack wants to slide. Strips shift sideways. The top and bottom of the board stop lining up. At the tight end of the curve, where the geometry gets steep, the wood struggles to fully close.
Three things that helped, two technical and one human
Alternate your clamps. Do not run them all in the same orientation. Flip every other one and the pressure distributes across the full sandwich instead of driving the stack to one side.
Add one clamp at the very end of the curve, oriented to squeeze the tips of the strips together rather than the faces. That is the point where the geometry fights hardest and where the strips most want to spring open. It seems obvious once someone shows you and completely invisible until they do.
The third thing was M and P. This is not a one-person operation. While two people manage the clamps, someone needs to hold the stack flat and stop the strips from sliding. The moment you let go to reach for another clamp, something moves. Having them there was not a convenience. It was the reason the glue-up worked at all.
Clamps on. Three days will tell.
The gap at the tight end looks no thicker than a sheet of paper. My working theory is that the faces may actually be touching and just not bonding yet rather than a true open void. Three days will tell.
The next step is to cut that end to shape the board anyway. Even a small void at the very edge may disappear entirely at the saw.
Clamps are on.
Leave a comment. I read everything.
