Lessons Learned during 2D cutout

We have a hexagonal window that gets a lot of direct sunlight in the summer. I wanted to make a custom design shade for it. Due to the size, I’m planning on making six identical panels. They’ll still be too small by a bit, but I’m hoping that that will work into the design.

So here’s what I’m starting with:

As you can see, it’s really a 2D design. Probably a project I should take over to the Makerspace fifteen miles from here, and cut on their laser cutter. But I have a Nomad! No way I’m driving all that distance :wink:

Since I need the full 8x8 inches, I’m bolting my plywood directly to the Nomad table. I use two sheets at a time, so the next one to be cut serves as the wasteboard for the sheet being cut.

I loaded the design into Meshcam, and made sure to make two adjustments (per other discussions here): I set exclusion areas around the bolt heads, and then set the retraction height to 0.5 inches so it would easily clear the bolts. Then I ran the default Carbide auto toolpath for soft wood and the 0.125 inch end-mill.

I generated the gcode, and it looked pretty good.

What I didn’t pay attention to was that detail on the upper edge. You can see that for the 3D roughing pass, that very top edge of the curve is too close to the edge of the stock for the settings, so it is not roughed.
When I ran the carve, that detail became an issue. Everything was going reasonably well. I could see that the feed speed was a little too fast for this plywood based on the splintering, but that’s something I could adjust for the next one. But when it came to the waterlining, things went off the rails.
My guess is that the waterlining gcode acts as if the roughing has taken place, even in areas where it didn’t. In this case, remember that top edge where the roughing didn’t happen because it was too close to the edge? Well, when waterlining started, the system bombed into that area much too fast and much too deep. Belts squealed, steppers juddered, and I had to dive for the shutoff. You can see the damage up top in this picture.

I’d done some previous experiments in harder woods where I roughed with a larger cutter and waterlined with a 1mm cutter. The 1mm cutter snapped. I didn’t understand what had happened. Now, I suspect it was similar circumstances - an area where the cutter doing the roughing was too big to clear the material, but the waterlining pass acting as though it had.

Now I started thinking. For a 2D design, the 3D roughing is only important to clear out material so that the cutter can access where it needs for the waterline cutting. If I’m essentially doing a simple outline through plywood that’s less than 1/5th of an inch thick, I could just do everything with waterline and pencil cleanup. But it was clear that the feed rate was too high, and maybe the depth per cut was also too high?

So I tried again, cutting feed rate by 30% and cutting plunge rate in half. The plunge rate change was an error - what I should have been adjusting was step down for the waterline. It turns out that the step down was fine; lowering the plunge rate just slowed down the process rather than helping (or causing trouble).

These settings were much more successful. It took longer than needed, but the results were pretty good. I’m still getting some splintering and minor delaminations of the plywood. It may be that this is just not the right material to work with.

Lessons learned:

  • make the retraction height as low as you safely can
  • don’t confuse plunge rate with step-down distance
  • for some cases of outline cutting, you can omit the roughing pass
  • in cases where you are doing roughing, look at Meshcam’s preview carefully. Notice if there are omitted areas, and make sure they won’t cause trouble in subsequent passes.



That looks like a nice design, it would be interesting to see the final product when it is installed.

I also saw the machine stall out and ‘jump the rails’ on a cut. It turned out that there were two factors. The auto generated toolpath parameters are too fast, about half those speeds is closer to what is needed for a conservative cut. The bigger issue was not having a roughing pass before the parallel finish. I now always carefully check the simulation to make sure I see a step down in each area of the part, and that there are no areas where the parallel finish occurs without a roughing pass to get down to the correct level in the cut. The Meshcam simulator is worth every nickel, and is now a standard part of my toolchain.

UnionNine has provided some great info on how to set tollbooth parameters. Check here: First 3D Cut and Lessons Learned and here: First Experiments with new Nomad 883

Hi Samuel,

First, that’s a sharp-lookin’ project, so I’m also looking forward to the final product!

Beyond Steve’s reference to my prior comments on feeds/speed settings, the only thing I’d add here would be about the material, and picking cutters for certain tasks.

That looks like Lauan board, which is great stuff because it’s relatively cheap, and with light sanding and paint can finish up with a smooth surface. However, because the fibers are so long in it, you’re definitely going to be fighting against splintering.

You’re probably going to want to use climb cutting for roughing, because it will do a better job of chopping the fibers without tearing out as much as “conventional cutting” will in soft-woods. This assumes your tools are sharp though!

Also, Onsrud and a few other vendors offer cutters that have different flute designs to optimize their behavior in plywood to minimize splintering. “Normal” tools have flutes that are designed to cut progressively upward along the shaft, and they lead with the cutting point. However, this means they’re also exerting upward force on the material as they cut, somewhat similar to a drill-bit wanting to “pull” itself into the material.

You can find down-cutting and compression-cutting tools that work to reduce the tear-out on the top and bottom, but they may be a bit spendy by comparison to regular upcutting tools. You can also find straight-flute tools out there, but the disadvantage there is that they’re not constantly and progressively engaging the material, so they can introduce a lot of vibration, especially at slower spindle speeds.

Looking forward to seeing round 2 of the project!

Work keeps interrupting my free time, but I’ve made some progress on this.

I switched to a 1mm mill-end cutter, and based on a few tries, found some speeds that seem to work pretty well.

For this design, I’m only doing waterlining and pencil cleanup. Calculations suggested for soft wood and this 1mm two-flute cutter, a feed rate of 29inches and a step-down of 0.008 inches. Experimentally, I’ve found I can work at a feed rate of 32inches and a step-down of 0.013.

Here’s what that looks like:

In the meantime, I’ve created a new iteration of the design, and when work allows, will try cutting that. I’ll update this thread again with any progress.

EDIT: forgot to mention that this iteration, the photo is without any sanding, whereas the previous iteration, the photo is after sanding and x-acto cleanup!

This project is finally complete!

After viewing the cutouts, I revised the design a bit during production. Using a miter saw, I cut molding for the outer frame. This is where I discover that my ability to do precise work is sadly inadequate. My “exact” 60 degree cuts are off by up to several degrees apiece. My frame is not really symmetric, nor (I discovered after the glue dried) even planar.

Then again, I decided to cut myself a bit of slack. When I carefully measure the window where this goes, it turns out to be asymmetric as well. Fortunately, my errors end up working with the errors of the people who built my house.

I ended up selecting six of the best panels I’d cut on the Nomad. Sanded them, and then glued the whole assembly.

After that, it was just a matter of staining the assembly.