One of the projects from a weekend in August 2026 was moving the big HVAC unit in the robotics area. Due to a miscommunication, it had been plunked down in the middle of a space we need. So I went in and helped knock a big dirty hole in the wall. Here's one of the other coaches on a ladder, drilling the hole that the vent goes through.

The prior hole just had a flexible tube jammed through it, but I figured I could do slightly better, so I designed an inside and an outside plate.

I fabbed up the inside plate on Sunday (plasma cutter FTW), then spot welded it together and powder coated it. It's a square steel plate with a rolled, slotted flange around a big round opening, powder coated white.

The outside was the next job: a laser cut grate with our logo on it, which I would also powder coat later in the week. That's the rest of this page.

The big round field is a honeycomb of small holes so the air can get through, with the gear and bulb left solid. That's a lot of holes to cut. Here's the laser head at work, close up and then from farther back.

The sheet on the bed of the laser with the pattern cut, and the bed afterward, with the little cut-out slugs scattered over the bed.

The pattern burned into the firebricks shows how serious a 1200W laser is: it did that even when out of focus by almost two feet!

And the finished cut, bare metal, before any coating went on.

I powder coated it in two stages. First a zinc primer, which should give it even more corrosion resistance than the topcoat alone.

Then the white topcoat.

And here it is finished, lying on the workbench.

It's on the outside wall, in an alley that few people go down, which is too bad. I still enjoy seeing it whenever I do.

It looks especially cool when it's backlit at night. I'll have to get a picture of that one of these days.

It took a few tries, but in December 2024 my experiments started to yield better results. This coin has a gear border around an open notebook and pen.

And the same coin from the top.

Some more recent examples: a couple of coins with engraved portraits. I darken the recesses with brass black and then run them in a rock tumbler for a bit to remove the black in the raised portions.

It fit.

This was basically the first "real" job on my new Gweike M3 Pro fiber laser. The kerf on it is so thin it's ridiculous. The outline of this first 'e' looks rough because the DXF I imported had a bunch of bumpiness in it, which I didn't realize until after the cut; I thought it was the laser, but it was actually faithfully reproducing flaws in the font. The other letters were much better, but even this isn't bad.

The letters stand off the wall on aluminum pins. I needed to make about 25 of these. Here's one, with a grooved and chamfered head at the top and a threaded end at the bottom. Making the other end is three operations (groove, chamfer, deburr) plus chucking and unchucking, and I was pretty happy with how fluid I got it. The threads came from a shop-made tailstock die holder, which lets me knock them out quickly on a bunch of pins.

Here's the full set of letters, with a small box of mounting hardware.

My brother tried them out on his wall.

And here they are up at the curling club, one above each sheet's scoreboard. A closer look at the E, on its standoffs:

June 6: phase one complete. No casualties. Pushing it across plywood to the deck stairs. A brother and a friend are very helpful, especially as they are both younger than me.

A line across the lawn, with the laser at the far end.

The other end of that line: a puller strapped to a tree at the far side of the yard.

The machine wasn't going to fit through the door as delivered, so before it went inside I stripped off everything that added height or weight. The CO2 laser tube, which is glass, came out first.

The control panel's buttons, disconnected from behind.

The lid came off too, which meant unhooking its safety interlock switches.

June 9: inside! And one room away from where it needs to be. Better yet, the remainder of the work can be done in air conditioning. Note that we had to pull off the top and tip it on its side to get through the door.

June 11: in the final room.

Still not a ton of clearance, even after cutting into the wall.

The view from the other side.

The gantry sliding under the cut header, with about a finger's width to spare.

June 13: final orientation.

June 14: back together, with the lid and panels on, and the control console unpacked and waiting for its cables.

The sample cut board that comes with the machine. I believe it was cut on this very one.

I think I already figured out the first upgrade I want.

June 14: success! The kerf is so tight that the piece sits in the cutout on its own. It didn't fall out when the cutout was complete.

June 16: I tried a quarter-inch steel square, target size 25 mm. The measured dimension was 24.870 mm.

And the other direction, 24.941 mm. I didn't adjust for kerf, so the error is roughly the thickness of a piece of paper.

Ventilation. It's also a CO2 laser, which can cut wood and acrylic, much of which smokes a lot more, since the material is being burned and not just melted. It exits out the back, hence the long run, because that's where it needs to be in order not to get blocked by snow in the winter.

The exhaust opening in the wall.

My other lasers are on Fiber Laser Setup and CO2 Laser Setup.

A successful test cut after eliminating an earlier problem that turned out to be an assembly error. There's still a lot more tweaking that can be done, but it's pretty usable at this point.

On February 5 I cut steel on it for the first time. It's reversed because this is meant to be the crusher head for the CanBot, a can-crushing device the robotics team made. The (overly ambitious) idea was that it would stamp the team number into the crushed cans.

My first try was a shop drawer full of green printed bins, labeled for roll pins and the various metric and fractional screw sizes.

By July 2025 I'd moved on to whole storage cases. Gridfinity is built on a 42 mm grid, and this one is 5x5, so about 200 mm square. They're printable, stackable, and parametric, and the full model is available. I think I'm impressed. "RBOC" is an abbreviation for "Random Box Of Crap".

A couple of weeks later I'd been working on them in the background and wound up re-implementing, and then augmenting, my favorite design. The end result is pleasing: stackable, parametric, and fully 3D printed except for the window.

In December 2025 I made custom trays for the tool drawer under the CNC mill. The cutouts for the tools at the left were done with tooltrace.ai, which was pretty easy to use.

I don't know what to say. This might be the high point of the rest of my life.

The clock print has started. Here's the dial ring, with its Roman numerals.

And a build plate of red parts: three large spoked gears, a threaded post on a tripod base, and some small spacers and bushings.

The clock is assembled! And does not work at all. Debugging ahead.

It's hanging on a rough wooden board, with the white dial, red gear train, black hands, and a black pendulum with a gold bob. The weight is a gold cylinder hanging from a string.

For debugging, I just pushed on the various gears to see whether the escapement was moving. It stops moving at one point in the gear train, so that's where I'll go look for friction. I guess that's my printf. It will take me an hour to disassemble it, clean out the bearings, and put it all back together again. If I'm lucky, an hour.

That's one of the things I like best about mechanical stuff: to me, the answer is often super obvious. In robotics, we have a joke. If it's software, it takes eight hours to find and five minutes to fix. If it's mechanical, it takes five minutes to find and eight hours to fix. If it's electrical, it takes eight hours to find and eight hours to fix.

I did eventually get the clock working, although I had to put a lot more weight on it than I would have liked. But it ran great: it was within a minute a day after I tuned it. For a while. Then it started stopping every few days. Then every few hours. Now it just hangs on my wall and doesn't move. I can get it to tick for a bit, but somewhere inside, some sort of friction has built up. With gear ratios as high as a clock needs to have, any little thing can be a problem.

Hopefully I'll debug it again at some point, because it really is a cool project. I don't regret building it. I just wish it was ticking away rather than just being wall art.

Drawers are mounted with standard drawer slides, and the CNC setup drills the screw holes for attaching them in both the drawer sides and the case sides, to make installation simple. The extra space below the lowest drawer is so I can fill it with concrete or something to keep the center of gravity low, and to provide stability if I park something heavy on top.

The drawer construction is nonstandard, but I've used something similar for years now without any trouble. Normally you'd make a box with a separate front rather than rabbeting the sides into the front, and the bottom would sit in a groove rather than just being attached flush. But I'm optimizing for CNC construction, and glue is pretty damn strong. I used to use pocket screws for attaching drawer sides to ends, and I might still, but I'm going to see if glue and nails will work. These don't need to be super heavy-duty, and it's hard to beat the strength-to-time ratio of glue and a pneumatic nailer.

Full sheet CNC router for the win: the cabinet parts are cut. On November 3 I was about to find out whether it would go together correctly and as easily as I was hoping.

By November 15 the first iteration was nearly complete.

An assembly fixture helped a lot. It's a sheet of OSB on sawhorses with small wood blocks screwed down at intervals to locate the parts.

This might be the first time I've put drawers into something within a few months of assembling the case.

There were issues galore, mostly tolerancing, which is typical for a CNC project. But I think they're all addressable, and the end result is completely usable.

The next day: now to actually put things away. Here's one of the drawers with a 3D-printed organizer in it.

On October 22 I started fabbing up a little catch basin. Sheet metal is fun. Here's the folded tray on the bench. I just cut the shape with some metal shears by hand, and spot welded the corners.

The drain needs a conical recess, so I 3D printed a one-off punch and die to make one. This is the design, in cross-section.

I already have a Harbor Freight 20-ton hydraulic press, like everyone has in their garage, but this is only 26 gauge steel, so a clamp or a bolt and nut is plenty.

It worked even better than I expected. Here's the clamp squeezing the printed die into the tray.

And the formed drain: a nice smooth conical recess.

The drain parts are 3D printed too. This is the drain fitting seated in the recess.

From underneath, the drain passes through the table via a printed hex nut fitting, with a clear braided hose attached below.

Here's the finished product (the photo at the top of the page), just waiting for the silicone to cure. I tilted it up ever so slightly so everything runs to the drain. I might need to add some bumpers at the edges to make sure the mill doesn't vibrate the whole thing loose. For obvious reasons I didn't want to fasten the pan down, although I could; a neoprene washer under the screw would probably work fine.

Here's the mill running with the pan in place. Working like a champ.

Fast, cheap… The laser cutter is still shrink-wrapped in the background, and the table is upside down, 2x4 legs sticking up from a plywood top, with a caster waiting to go on.

And good enough. Here's the laser on its finished rolling table, lid open, with the packing foam and exhaust fittings on the floor below.

I needed a 150 mm to 4" vacuum hose adapter. Oh noes, wherever will I get one? I printed it.

This is the bench at the end of an evening in late September. Good progress, but lots left in this build. The kit parts are laid out, with the instructions up on the laptop.

And the frame, partly assembled, surrounded by bags of hardware.

A regular socket wouldn't fit into one tight spot, so I ground down the outside wall of a 7 mm deep socket. Having a grinder means never having to say you're sorry.

By October 2 it was probably about two thirds done: you can see the frame, the table, the ballscrews and linear rails, and a stepper motor on top of the Z axis. Yes, it has ballscrews. Fancy, but necessary.

On October 3 the mechanical assembly was complete, so next up was electrical. That's the finished machine at the top of the page.

I chipped away at the electrical work when I had time. Here's the back of the machine with the electronics bay open, showing the controller boards and wiring, with the coolant reservoir on the side. The power-on smoke test didn't reveal any immediate glaring issues.

Next up was installing the software and making it move. The machine also needs a home: I'm planning to use the big CNC at the makerspace to make a rolling cart for it to sit on, with room for the power cables, coolant lines and so on.

On October 10, finally, all three axes working.

I also had to build a coolant basin for it to sit in.

I always enjoy casting, and on my birthday in November 2024 I decided to try making some aluminum bronze for an espresso funnel. This is the sand mold, with a 3d-printed ring-shaped pattern pressed into it.

And the pour, at night in the backyard. You can see the glowing crucible, the furnace and the molten metal in the mold's pour opening.

It looks like it was a success: here's the casting freshly shaken out of the sand.

A couple of days later I started cleaning it up on the lathe. The aluminum bronze looks amazing as it cleans up. It's just 90% copper and 10% aluminum, and one application is propellers. It's really pretty.

A few days after that, I'd gotten in a bit of machining over lunch. Plenty left to do. The top is faced and the inside is bored, with some casting pits still remaining.

In January 2025 I finally finished turning the espresso funnel. It's totally ridiculous, but I wanted to make something out of aluminum bronze, and the 3D printed one I'd been using was unsatisfying. This one weighs like three pounds, which is very satisfying.

Here it is next to the purple 3D-printed one it replaces. I really should thin it down some more, but it's amusing me as it is.

A quick 3D print project: an add-on steam handle for the espresso machine, clamped to the steam knob with a small socket screw.

I'd been making (slow) progress on the basement shop remodel, and in early June 2025 I was edging closer to when I'd have to move things into their new space. I tried modeling and working with the layout virtually, but I find that kind of annoying. I've done it before by cutting out paper outlines, and that works really well for me: very quick to produce, tactile, visceral. But this time I already had all the models, so I figured, why not print them out?

The biggest thing to place is the Bridgeport mill, a 2J Series 1 from around 1976. It weighs about a ton, and it was fun to get it into the basement. It shouldn't be too bad to move from one room to the other, but we'll see.

A week later it was on to the actual moving. The printed model, with its little labeled machines and furniture, has turned out to be a really useful way to visualize arrangements.

Here's what it looks like these days. I still need to add more storage to it. And clean my desk.

First, the pattern for the casting I hoped to do that weekend: a tree of a dozen small open-frame pieces on a central sprue.

And a single piece, which is what each keychain will be.

A partially successful casting on the morning of April 27th. This is the robot part in the crucible, melting.

Breaking the investment flask open to see what came out.

And the aluminum tree itself. I only got 8 out of 12 pieces to come out, so I need to do another. I think maybe the graphite gasket I made didn't give a good seal, and I didn't get as much vacuum as I could have.

After fettling, the pieces got powder coated, and here they are curing in a toaster oven.

I finished the senior gifts on May 15th. I still have a largish brick of the aluminum left over from the melted-down part, since I didn't use much of it. It's a part we removed because we were over weight, and I plan to bring it in to show people just how much aluminum it was.

After lasering, the coin goes into a rock polisher for an hour or so. It does a good job of cleaning out the soot from engraving.

But the final look is much improved by an application of "brass black", followed by about twenty more minutes in the polisher. Gives it a bit of definition and age.

Need some ductwork for your laser? Just design and print some! Here's a tall round duct with mounting flanges, and a 45-degree elbow.

I did consider PVC pipe. In the end I didn't go with it because I already had a hole saw of a given size, the fittings are fairly expensive, and I didn't feel like driving over to Home Depot to buy another one. The slicer tells me the cost of the plastic, too: one part I'd just printed was $0.73 worth of PETG.

In April 2025 I finally got the laser set up in a cabinet that other people can be in the room with. It's a plywood enclosure with a green laser-safe viewing window and an LED strip, sitting on a 2x4 stand. The laser glows away inside.

A purple-lacquered body with a turned spalted-wood cap and finial.

And three from earlier years, dyed red, purple and green and marked 2011, 2012 and 2013.

An ornament cut from a piece of our Christmas tree, in this case from 2015. My wife's Chinese astrological sign is the ox, hence the cow shape.

In December 2024 I caught up by two years. Laser cut and engraved, of course.

The rectangle on the left is engraved with a family selfie from 2018. The ring-shaped ornament on the right is dated 2019.

On September 3, 2023, I milled a lathe tool holder, mostly because I just wanted to mill something. Here it is on the mill with a long step cut along it.

A few days later I test-fitted it onto the dovetail of my 250-222 quick-change tool post. I'm pretty happy with this fit.

And here is the finished holder on the lathe, clamping a boring/turning tool with a carbide insert. There's a brass knurled adjuster nut on a threaded stud, and four hex bolts. It came out okay. I think I'll make another one and see if I can do better.

The tailstock die holder was progressing in early October. Unfortunately, I bored the end out too big. Facepalm. It was fixable without starting over, fortunately.

To make up for my mistake, I turned an adapter for the end. I also made it dual size, so it can accommodate larger dies. There was still a bit left to do on it at this point.

Having a multicolor printer is great for stuff like this. I probably should have slowed the print down a bit, and I accidentally used silver rather than white for the sizes, but it's more than good enough for its intended purpose.

Here it is installed next to the lathe's DRO, a dial indicator on a magnetic base, and a bottle of way oil.

Then there's stuff like this. It's a support for holding long, thin things stable in the bore of my lathe. It prints in 20 minutes and cost $0.29 worth of filament. Ordinarily you'd use something called a lathe spider for this, which is a whole project of its own and potentially involves drilling holes in the lathe. This is so easy that I'll just make one whenever I need a different diameter.

I'm also digging just leaving a roll of white in the printer all the time so I can embed text in stuff, like the diameter on this one.

A little lathe upgrade I'd been working on: a linear-scale digital readout for the tailstock, on 3D-printed mounts with a printed clamp collar on the quill. One of those screws is a decoy, hiding the broken tap underneath it.

In December 2024 I printed soft jaws for the lathe chuck out of TPU, a flexible filament that's like a fairly firm rubber. Here's the first version fitted on a chuck jaw.

I added a giant fillet in that corner for the final versions, since that corner was obviously going to fail. Here's the final version.

I wound up not using these because - surprise! - metal chips are hot, and plastic doesn't like that. I'll have to do some in aluminum or copper one of these days.

I picked up a Vevor 6-inch six-jaw self-centering chuck with a D1-4 cam-lock backing plate for just north of $200. It arrived covered in grinding dust and with the burrs left on, so it's getting a few hours of deburring with files and blue shop towels to take it from "done" to "operates smoothly". It's nearly finished; next up is mounting it on the lathe and seeing how accurate it actually is. Updates to follow.

Here it is torn down on the bench: the body, the six jaws, the scroll ring, and the backing plate, with a set of needle files standing by.

She also made a few more of these for my siblings. They are about 6" square, and inspired by a Chinese "chop", . The one pictured says "Shu shu", which is "father's younger brother". My sister got the equivalent one that means "father's younger sister". For these, she hand-painted the cavities after routing.

We built it in place in the apartment, starting with lumber laid out over some storage totes.

Then a 2x4 frame on the floor, which took some measuring.

Once the frame was done, a fit test in the bay window.

The slat boards got cut to length with a pull saw.

And then screwed down onto the frame.

The plan was to melt down some of the broken parts from that year's robot and cast keychains in the shape of the game piece, and maybe laser something onto them if there was time. This project was way more of a pain than I thought it would be, but I learned a fair amount, so I'll call it a win.

here's the finished keychain, presented to the seniors at the wrap-up banquet on May 19.

The Charlottesville office's version came first. It started on the computer in June 2022, where I was playing around with the design and wondering about proportions. Here's a render of one of the versions I tried (this one has Durham, NC in the corner).

In July I kept playing with the nesting software, which figures out how to pack the letters into the smallest space on a plate. It's kind of fun to see what it comes up with.

Then I cut the logo out of quarter-inch aluminum on the plasma table. The first cut failed, so I recut over the failure. It wasn't perfectly aligned, hence the slight irregularities, but it's a very good cut given that.

Here's the lowercase k after powder coating. I obviously put a bit too much powder on, but generally I like the way it came out.

I came back to the Charlottesville sign in the fall.

The Kevel logo, plus "Charlottesville, VA", on a navy panel: white aluminum letters, and a K made of two folded orange aluminum ribbons. I built it in the last days of September and the first days of October 2022, and finished it with a few minutes of sneaking away to the shop on the morning of October 3, since I was going to Charlottesville the next day. It still needed hanging hardware, and perhaps a few touches of paint to fix the flaws that probably only I can see.

More than a year later I made the one for James. In September 2023 I cut the parts on the plasma cutter, and here they are fresh out of the water table.

In October I put together a full-size layout out of taped-together sheets of printed paper on the floor.

Later that month I was at the makerspace and used the laser, the one tool I didn't have at home (I fixed that later), to make a hardboard template to help with aligning the pieces.

In November came the bending, and having done it before helped, though there were still a couple of things I had to remind myself of. The test bend went fine, shown here next to one of the laser-cut template pieces.

A second attempt a bit later that day went fine too.

The next morning it was coming along, with a single bent and polished k mounted on the navy board and still a bunch left to do.

I got the mounting hardware sorted out, so there are no more fragile 3D printed standoffs like the ones on the Charlottesville sign. Next up was powder coating.

And on November 26, 2023, it was done. Sorry it took so long, James.

I tried to DIY the container that holds the pressurized coolant and wasn't having much luck, so I went with a purchased container and some custom fittings. I had a fun morning on November 18 knocking these out. Machinist glitter included.

The system was complete by November 20. A flexible hose with a brass nozzle drips coolant toward the chuck.

The coolant lives in a wall-mounted water filter housing, plumbed with brass fittings and an air regulator with a gauge.

The term for this is MQL, Minimum Quantity Lubrication. The advantage over just a mister (which is what I started with) is that forcing the lubricant into the mix with air pressure keeps the droplet size smaller, so less of it winds up in the air. Plus I can run a straight drip system.

Even running with mist alone, it takes me from "part is too hot to touch when I'm done" to "part is completely cool when I'm done". That increases accuracy, too, since a part gets bigger when it gets hot.

To be fair, I'm abusing the commercial system. It's meant as a mist system, and I'm overloading it by using air pressure to push fluid through it rather than relying on the Venturi effect, which is the design intent. It's a bit of a hack job, really, but I think it's still an upgrade.

Step one was to figure out how to fixture it so I can mill some precision into it. Here's the base upside-down on the mill vise, with a dial indicator reading its surface. I think this will work.

Here it is on the router. It works pretty well - we align it with the side rail using 123 blocks, clamp down the vises, and we have something we can repeatably secure extrusion in.

First I had to peel off the insulation. The lathe made a convenient way to hold the spool.

Then I realized I could wind the bare wire on a mandrel on the lathe and get a tighter winding for less effort.

The coils went into a graphite crucible, and were poured into a sand mold that was not, unfortunately, packed tightly enough.

A little turning cleaned it up a lot. I have been using one of the two as a drift, for tapping things down into the vise. The other one will probably get melted down again to make something at some point.

Here's my casting setup under the deck on the night of August 16, 2023: the propane furnace with its lid glowing, the tongs, and a wooden sand flask ready for the pour.

And the result, still sitting in its sand mold.

It's about 2" long. It's okay, with a few minor defects. I may try again.

By July 13, 2023, the printer had gotten to the gears. They're herringbone gears, along with a worm and a shaft.

A week later I had it put together and winding a spool.

Here it is in action.

It works pretty well, but it was kind of a pain to assemble. Lots of the pieces didn't fit without sanding, and I bent or broke one or two of them trying to get them together. Still, I'm pretty happy with the result. The gears are pretty good. The worm gear kind of sucks, but it works nonetheless.

Better than PLA, but I wound up liking these better for removing things from the build plate.

Assuming I didn't screw anything up, the enclosure end of things was done. The odds of a screwup are high, of course; I have yet to do this and not have a loose connection or a wrong wire somewhere. This is the variable frequency drive, a Hitachi WJ200, with its terminal blocks, breaker and wiring.

A quick lunch-break pair of upgrades: I replaced the belt with a poly link belt, and the socket head cap screw with a quick-release handled screw. The new belt is slightly longer, which makes belt changes easier and puts the motor closer to the center of travel in the adjustment bracket.

And the DRO is done, with scales on both axes. I am soooo glad that's a one-time activity.

Here's the flower itself, on a board among the shavings it was made from, a couple of days before.

And the plaque with the flower attached, before I trimmed it down. I'm not thrilled with the look of the poplar, but I was out of time to do better.

To address the issue with the above unit wearing out, in October 2023 I made a couple of 3D-printed versions. First up was a prototype. It shows one of the drill bushings that I made to guide the drill, since if the mild steel I was using on the original version didn't stand up, the plastic definitely wouldn't.

My 3D printer has more than enough accuracy to place the holes. I measured the prototype and found it had placed that bushing to within 0.001" of the correct location, and since the bushing is reamed, the whole thing is probably within about 0.005" as far as putting the hole in the right place.

Of course, jigs get abused. The kids broke one and I broke another. In early November, I think this was at least version 5. Visible in the fifth hole from the left is the locating pin I made, so the jig can be used to make series of holes that are longer than the jig.

Here's one of the chains early on, with lumpy welds and the bottle hanging from a hook at the end.

Then a quick run to the makerspace so she could sandblast it.

I happened to snap this photo of her welding links together.

And a finished one, painted grey. This was the best I could do for a photo; it's too big for my lightbox.

I didn't get any video of the pour itself, because I was busy managing a crucible full of very, very hot, very heavy liquid metal.

Here's the crucible glowing hot, sitting on firebricks in a box of sand. The muffin tin next to it has a glowing ingot in it.

A few days later I decided to machine the cast iron blob I'd cast, just to see how it went. I didn't take it all the way past the defects, just enough to make it more or less square. It machined very nicely!

A few casting voids remain, and the edges are a little rough. I will definitely have to try this again.

Here's the design in CAD: two split-clamp blocks, each with a bore and clamp screws, joined by a bar.

And here's the aluminum bracket complete and installed on the Z-axis slide of the plasma table, holding the new machine torch. It was super, super nice using the Bridgeport for this.

Boring the bracket on the mill.

Here are the panel pieces coming off the printer: the front panel with the speed dial, the forward/stop/back rocker switch, and the four-digit LED display, plus a few blank panels.

I was down to my last roll of PETG when I printed this, which was a relief, because I will never use it again. It gave me stringing, voids, and blobbing. (2026 update: PETG is now my preferred material. The problems were likely due to not drying it. Having a better printer now helps, too.)

Holding the finished control box next to the mill and its DRO.

My daughter was about to turn 18, and she really likes Infinity Train, which is a bit obscure: a surreal animated series. I'd wanted to do more jewelry, and I'd done a few rings, so I designed these earrings based on One-One.

On August 27, 2022 I cleared out the area under the deck so I could move mill parts through it. As you can see, it was a bit of a job.

Attachment point for the chain hoist, installed. I never should have used that eye bolt: a forged eye bolt would have been much better from a load standpoint.

A trivial test of the chain hoist: a bucket.

Several pieces I moved myself, with various hoists. On September 3 I built a gantry (it has its own page) so I could take the top of the mill apart with a chain hoist. Here it is over the mill.

The same gantry from another angle, lifting the ram off the base.

Two stills from a video of the head coming down, in its lifting straps with a wooden bar through them.

The head, in lifting straps and hanging from an engine crane over a plywood sled.

The head being lowered under the deck with the chain hoist.

The table on the lift cart in the driveway, having just come out of the car.

And the scissor-lift cart carrying the table, hanging from the same hoist. It was easier just to lower the whole thing rather than unstrap it from the cart.

And done. And I mean done. I'm paying someone else to move the rest.

In September 2022 I started to clean up the table on my Bridgeport mill. It'll probably take a few sessions, but here's the "before".

The table is lying on the workbench, long and rusty, with its T-slots running down its length.

I was still in "get the mill fully restored" mode, so the next jobs were the upgrades that make it pleasant to use. The first was a digital readout. This is the DRO, still in its wrapper, with its first screen.

And mounted on its arm.

The scales' cables run down the side of the mill.

Next, a power feed for the Z axis, so I don't have to crank a 600 pound knee up and down by hand. This is the mounting ring for it, plasma cut and then drilled and tapped.

The Z power feed, installed and running.

In the summer of 2025 I rearranged the whole basement shop and the mill had to move to a new room. The hydraulic lift cart that made it easy to take the table off appeared to have given up the ghost, but I could still lift things with the shop crane.

The Bridgeport, still in its old spot, with a tub on the table.

The table, lifted with the shop crane.

The table hanging from the crane.

The table on its cart, going through the doorway.

The table in the new room.

The base was the hard part. I started out using a trick where you lift the Bridgeport by cranking the knee down with a board wedged under it, but switched to using a toe jack, which just felt a bit more controlled. With one side tipped up, I put some electrical conduit that I'd cut up underneath it, and I could actually roll it just by pushing on it.

First attempt, using the knee.

Closer.

The base up on the rollers. I can move it easily by myself using a crowbar when it's on these.

The base coming down the hall. A little help from my friend Andrew.

Things went well with Andrew's help, but honestly the whole thing went so well all that did was speed me up a bit. Once it's up on pipes it's easy to move: just stick a crowbar under it and you can lever it around, an inch at a time. Which was great, because I wasn't sure I liked where it was, and it turned out I could move it a few feet by myself.

In the new room, more or less.

And then a second, shorter move, into the corner, on July 13.

Once it was assembled in the basement, I thought it would work to move the Bridgeport I'd bought, and that it should fit in my car in pieces, too. So I took it apart and checked.

It fit easily, and it was pretty easy to carry, too. Three trips on foot to carry the pieces.

Then the real test, the same day: lifting the heavy ram of a milling machine with the chain hoist. It worked perfectly.

A couple of weeks later, back in my basement shop, lifting the ram into place.

Years later, I finally found a photo of it in action, straddling a machine in a garage shop.

One piece of advice about the hardware: use a forged eye bolt rated for the load you want to lift, not an eye screw or a drawn eye bolt. One of our robots unwound a 1/4" wire eye bolt lifting itself not long ago, which is not how you want to find out.

You want the ones that look like this:

Not this:

Here I'm placing the sprue in the middle of the 3D printed pattern. I wish I'd remembered to make some marks to help me center it, so that I could use it to grip the casting on the lathe.

With the sand packed, I separated the two halves of the flask so I could remove the pattern. The bottom half still has the pattern half-embedded in it, and the top half has the matching impression, with the hole for the sprue.

After the pour, I dug the casting out of the sand. It was looking good so far: you can see the disc, the pillar, and the spoon-shaped sprue coming out of the sand.

Success! It's super rough, though. I think I should have sieved my sand. But every surface is going to get machined, so it should be fine.

The next morning I started cleaning it up on the lathe. One face took a cut nicely and came up bright, with the rough sand-cast rim still around it. There were going to be some voids and inclusions in the finished part. Tim said that was fine, as long as it held the thing and conducted heat.

Flipping it around to do the other side meant gripping the faced rim with the chuck's outside jaws, with the sprue stub sticking out toward the tailstock.

It was loads easier to cut most of the sprue off on the bandsaw than to machine it away.

And done. The front face, up top, has a raised center step. This is the back: a shallow recess, with a bit of rough casting skin left on the rim. That's as far as I could take it down without running out of material. I should have allowed more room in the casting, so I'm calling the rough edge artistic. A few of the pits made it through, too.

It shipped off to Tim shortly afterward.

The logo was cut on my Mostly Printed CNC router, which I set up under the deck for the job.

Here it is at work: a wider view first, then a close-up of the bit going through the wood.

Not every shape could be routed, though. For the rest, I used the CNC as a big plotter to draw the outlines on the boards, and then cut them out on either the band saw or the scroll saw. Or both.

A few days later, the cut-out pieces of the logo laid out on the round galvanized disc on the plasma table. It was coming along nicely.

Here's the sign up on the wall at the store, with the installation still underway.

And Toby and me in front of it.

I love looking at it every time I go into his Vienna store.

On New Year's Day I cut the steel tubing for the frame. On the right are the pieces I needed; on the left is what was left over from what I started with.

And this is what they were getting welded into: an octagonal frame.

A few days later the kiln's controller was plugged in and reading. Working!

By early February the kiln was sitting on its finished base, with the controller on its stand next to it. It works, and now I can get it out of the way.

This is an earlier version of the control box, before I decided that a metal enclosure would be a good idea with 25 amps running through it.

Here's the wax pattern cast from the 3D print, in blue. Oh hey, it worked.

Then I tried again, with what I thought was a better sprue design. This one is printed in grey resin.

That earring is slightly convex, like a contact lens. I failed many times to cast it in brass and aluminum bronze, but I learned a lot doing it. In the end I just printed it and spray-painted it.

The next year, my younger daughter had another Pi Day earring concept in mind: a sphere made up of the digits of pi. I worked through the math for it in OpenSCAD.

This was the prototype, printed in white resin and coming together a few days before Pi Day. The design is hers.

The finished earrings were done just under the wire on March 14th. I'm not sure why she wanted black, but they are as ordered.

Here's a better picture, after she highlighted the numbers.

Another look at the prototype.

This is where it started: my first attempt at a sheet metal project, in January 2022. Meh. But it's a start.

Achievement unlocked: spot welding. The spot welder itself is in the background. These units aren't too expensive, and versions that run off of 120 are available, so this capability is pretty accessible even to a modest home shop.

Here's the design iteration from late November: dark walnut body, a granite-looking curling stone on top, and a brass band and handle.

By the third week of December the top was made and waiting on paint.

Here it is partway through the build, before the belt went on.

At some point during the build, I thought I had misplaced a piece of tubing that came with the kit. I looked for it absolutely everywhere, multiple times, and in the end just couldn't find it. So I decided to weld up a replacement piece.

Of course, as soon as I did that, even though the replacement was working just fine, I found the missing piece sitting right where I'd set it, right next to where I was welding.

A few days after finishing it, I tried a leather strop on the grinder. Razor sharp in seconds.

Later that same morning I got it off the bench and onto a new plywood rolling cabinet stand. Well, it's off the bench at least. The shop is getting crowded.

The rubber seal in the picture is from a replacement I bought that doesn't fit, again because of that non-standard size.

And one on the trailer in May 2022. I had to fudge it a bit, but it works.

The face was stained light and the body and wings dark. On October 24, 2021, I had the stained piece clamped to the bench and was off to find a stain color somewhere between those two.

The first route was a silicone mold. Here's the 3D-printed resin bust of Leonardo da Vinci, the blue silicone mold, and a yellow wax copy made from it. Next stop was going to be a plaster cast of the wax figure.

You can buy two-part casting silicones, but they're expensive and go bad quickly once opened, which makes them a poor choice for occasional home use. So I tried hardware store caulk as a casting medium. Straight out of the tube it's too thick to work well, but I finally hit on a recipe that is thin enough to flow around a model, cures quickly, and uses only easily available, relatively safe stuff.

The cutaway block shows that I could get it to set without voids by using a vacuum chamber. The complete mold shows that it takes detail well.

Amazing Remelt is a good product for this, but I think a homemade version costs about a fifth as much. It's just glycerin and gelatin. Here's the first experiment: the printed mold halves, with a translucent red gelatin casting between them.

And a demold success, with a teal gelatin block holding the impression of the bust next to the resin bust it came from.

The mold parted cleanly along the layer lines. There are some voids from bubbles, so I might need to refine my technique there.

Pouring the wax in stages did the trick. There is one very small area, down at the bottom, where the heat of the wax melted the mold where it was thin, but otherwise it came out great.

I think I'm done with this line of experimentation.

Here's another one, which came out looking pleasingly like an acorn, given that it's made of oak. It's some random piece of oak from I have no idea where. The finial is made out of something else, and is my least favorite part of the piece.