Getting Fractal Patterns Into Wood Without Losing Your Mind
I've been doing fractal pyrography for about twelve years now, mostly because someone on a woodcarving forum suggested I try Mandelbrot sets on a cutting board and I never figured out how to stop. The main headache is translating those infinite recursive equations into something your wood-burning pen can actually follow. Most people try to trace fractal outlines freehand and end up with something that looks like a error message from 1998.
What Is the Fractal Wood Burning Solution
It's basically a workflow that converts mathematical fractal data into vector paths or grayscale height maps your burner can read. The software side handles the heavy lifting of recursion depth and color banding so you're not manually plotting coordinates for eight hours. I used to build everything in Mathematica and export SVGs. That worked fine until I started getting into high-iteration Julia sets on small boxwood blocks. The files got unwieldy and my burner's controller board choked on anything over two megapoints. The workaround was switching to Processing with an export library I wrote that thins the path data down to what the machine actually needs. Cut the file size by about eighty percent with zero visible quality loss at burn resolution.
The Actual Process
Generate your fractal first. I use a modified version of the XaoS viewer for real-time exploration, then dump the output as a high-bit-depth PNG when I find something worth burning. Depth matters here. Anything under sixty-four iterations looks muddy on hardwood. You want that fine structure. Next step is converting that raster to a burn map. This is where most people blow it. They throw it straight into their CAM software and wonder why the details turn to soup. What you need to do is pass it through a dithering routine that preserves local contrast. FLOYDSTEINBERG works for most cases but for finer detail I use a custom Bayer matrix pattern tuned to my tip size. I burn on maple primarily. Cherry works but it's forgiving in the wrong way. You'll get acceptable results everywhere instead of sharp results in the right places.
Get the Full Details
The Edge Case Nobody Talks About
Last spring I burned a two-foot panel with a Sierpinski carpet at three hundred inches per square foot. The fractal has those predictable voids, and on a surface that size the voids line up in ways that warp the wood as it cools from sustained burning. I lost the piece because I didn't account for thermal gradients across the negative space. The middle section pulled and the whole thing cracked along a symmetry axis. My fix was burning from the center outward in quadrants and letting each section cool under light clamping weight for twenty minutes before moving on. Slower but it saved the grain structure. Also helped that I was using a Kilz primer coat underneath to seal the pores. The sealant stuff is a pain to work around but it makes a noticeable difference on open-grain boards.
Software Options That Actually Work
The common tools are Pyrography Pro and a few Python scripts floating around GitHub. The free options are mostly adequate. What trips people up is the tip calibration. Your machine will read the same g-code file differently depending on whether you're using a needle point or a shader blade, and the software doesn't always adjust the burn rate automatically. I run a test pattern every time I switch tips. A simple gradient with known intervals. Takes five minutes and saves you from losing a three-hour burn because your duty cycle was off by ten percent.
Limitations Worth Knowing
Fractal burning doesn't work well on curved surfaces. The recursion assumes a flat plane and once you introduce radius the patterns distort in ugly ways. I've seen people try it on bowls and the results look like the fractal was melting, which is accurate but not what most buyers want. Also, the recursive nature means some areas will always burn hotter than others simply because there's more line density. A Mandelbrot set at iteration level one hundred and twenty-eight will char through thin sections of basswood unless you modulate your speed dynamically. If your machine doesn't support variable feed rates you're limited to simpler patterns or you're doing a lot of manual passes. The biggest bottleneck is file generation time. A decent fractal at burn-quality resolution can take forty-five minutes to render on a consumer CPU. I keep a machine dedicated to just this because it's annoying to wait on a computer you need for something else.
If you're just starting out I'd suggest a smaller commitment. A six-inch panel on maple with a medium-complexity Julia set at about eighty iterations. It'll teach you the calibration and the material behavior without eating a week of your life. The learning curve is steep but not because the math is hard. It's because the intersection of software, hardware, and wood physics creates a lot of failure modes you won't find documented anywhere.
