The Smell of Molten Projects in the Morning
Ed Nisley's Blog: Shop notes, electronics, firmware, machinery, 3D printing, laser cuttery, and curiosities. Contents: 100% human thinking, 0% AI slop.
Prusa Mk 4 3D printer with MMU3 feeder
As fate would have it, the Step2 rolling garden seat took an untimely fall while standing in the garage and broke both its hinges:
It’s been out in the garden for maybe six years, so those chunks of plastic are fully depreciated.
Two hours after loading the solid model into PrusaSlicer:
The SiLite tray is well-weathered, but remains structurally sound: still ready for service in the D-Hall breakfast line on the morning after the Apocalypse.
Living in the future works out pretty well.
Unlike the OEM 24 V supply in the laser, the “new” supply from my heap does not have mounting flanges; it’s intended to be attached to a mounting plate from the back side. It turns out the laser does have a mounting plate with All The Things screwed onto it, but there is no way I am going to disconnect all the wiring just to drill four more holes in that plate.
So I made a pair of brackets to screw into the back of the supply and then into suitable holes in the mounting plate:
Which look like this in real life:
Those M4 rivnuts just beg for 6 mm holes in the mounting plate.
However, it turns out that their unsquished length exceeds the distance behind the panel, which means there’s no way to install them flush to the panel with the proper backside squish.
So:
The OpenSCAD code generates a simpleminded drill template:
Press a scrap of rubber firmly against the plate to dampen vibrations and thwack each hole with an automatic center punch set to stun. Deploy a succession of drills up through 6 mm, catching most of the swarf in tape strips:
Squish the rivnuts in place:
The small, vaguely tapped hole on the lower right was the “good” screw for the OEM power supply; the “bad” screw hole is invisible to the upper left, just under the raceway.
Remove the plastic spacers, snug the nuts holding the plate again, install the power supply, and it looks like it grew there:
The wires and Wago connectors scrunched underneath aren’t anything to be proud of, but longer wires didn’t seem likely to improve the outcome.
The OpenSCAD source code as a GitHub Gist:
| // Mount for 24 V laser power supply | |
| // Ed Nisley – KE4ZNU | |
| // 2025-12-07 | |
| include <BOSL2/std.scad> | |
| Layout = "Show"; // [Show,Build,Guide,Block] | |
| /* [Hidden] */ | |
| ID = 0; | |
| OD = 1; | |
| LENGTH = 2; | |
| HoleWindage = 0.2; | |
| Protrusion = 0.1; | |
| NumSides = 2*3*4; | |
| $fn=NumSides; | |
| Gap = 5.0; | |
| Rivnut = [4.0,6.0,9.0]; // body + head OD | |
| RivnutHead = [6.0,10.0,1.0]; // flat head | |
| WallThick = 6.0; // a bit more than half rivnut head OD | |
| SupplyCase = [50.0,215.0,112.0]; // power supply case size | |
| SupplyOC = [25.0,150.0,0]; // power supply mounting screw centers | |
| SupplyOffset = -1.0; // the screws are not centered on the case! | |
| SupplyScrew = [4.0,9.0,4.0]; // … LENGTH outside supply case | |
| MountOC = SupplyCase.x + 2*WallThick; | |
| MountScrewLength = 8.0; // … head-to-baseplate | |
| MountRadius = 0.5; | |
| BlockOA = [MountOC + 2*WallThick, 2*WallThick, MountScrewLength]; | |
| GuideOD = 2.0; | |
| //—– | |
| // Single mounting block | |
| module MountBlock() { | |
| difference() { | |
| cuboid(BlockOA,chamfer=MountRadius,except=BOTTOM,anchor=BOTTOM); | |
| for (i = [-1,1]) { | |
| right(i*MountOC/2) { | |
| cyl(2*RivnutHead[LENGTH],d=RivnutHead[OD],circum=true,anchor=CENTER); | |
| cyl(2*BlockOA.z,d=Rivnut[ID] + HoleWindage,circum=true,anchor=BOTTOM); | |
| } | |
| right(i*SupplyOC.x/2 + SupplyOffset) { | |
| down(SupplyScrew[LENGTH]) | |
| cyl(BlockOA.z,d=SupplyScrew[OD] + HoleWindage,circum=true,anchor=BOTTOM); | |
| cyl(2*BlockOA.z,d=SupplyScrew[ID] + HoleWindage,circum=true,anchor=BOTTOM); | |
| } | |
| } | |
| } | |
| } | |
| //—– | |
| // Guide holes in a 2D layout | |
| module DrillGuide() { | |
| difference() { | |
| square([BlockOA.x,SupplyOC.y + BlockOA.y],center=true); | |
| for (j=[-1,1]) | |
| fwd(j*SupplyOC.y/2) | |
| for (i = [-1,1]) { | |
| right(i*MountOC/2) { | |
| circle(d=GuideOD); | |
| } | |
| } | |
| } | |
| } | |
| //—– | |
| // Build things | |
| if (Layout == "Block") | |
| MountBlock(); | |
| if (Layout == "Guide") | |
| DrillGuide(); | |
| if (Layout == "Show") { | |
| for (j=[-1,1]) | |
| fwd(j*SupplyOC.y/2) | |
| MountBlock(); | |
| color("Gray",0.5) | |
| up(BlockOA.z) | |
| cuboid(SupplyCase,anchor=BOTTOM); | |
| } | |
| if (Layout == "Build") { | |
| for (j=[-1,1]) | |
| fwd(j*(BlockOA.y/2 + Gap/2)) | |
| up(BlockOA.z) zflip() | |
| MountBlock(); | |
| } | |
I have unfairly maligned the TPU snout, because the PETG snout failed the same way:
Seen with the shock cord in place, it’s obvious that combining moderately high temperature with steady compression sufficed to bend the PETG enough to pop those tabs loose from the vent.
So the OpenSCAD model now produces a stiffening ring to be laser-cut from acrylic:
The whole snout builds as a single unit in the obvious orientation:
Because the part of the snout with the tabs is 7 mm tall, I glued a 4 mm acrylic ring to a 3 mm ring, with both of them glued to the snout:
That’s “natural” PETG, which I expected to be somewhat more transparent, but it’s definitely not a dealbreaker.
Mary will sew up another cheesecloth filter and we’ll see what happens to this setup.
As the saying goes, “Experience is what you get when you don’t get what you want.”
Fortunately, living in the future makes it easy to iterate on the design & implementation until experience produces what should have been obvious at the start.
Looking at that big smashed-glass coaster from a different angle showed interesting patterns:
Although the larger fragments were still holding together when I laid them in their recesses, they apparently consist of several sub-fragments with larger continuous cracks letting the epoxy flow / ooze inside.
Now that I know what to look for, the original picture also shows them, albeit less distinctly:
They’re not obvious in the scanned image of the fragments, although I could convince myself I see some:
The many smaller fragments I’ve been turning into coasters probably separated from similar large chunks along such cracks, which is why I’ve never seen rivers of crack before.
Apologies if you arrived here expecting a tirade concerning the drug trade … :grin:
Because the BOSL2 library includes a gear generator, I can now avoid creating a gear outline in Inkscape and importing it into my stacked change gear generator.
The labels now snuggle closer to the shaft and (barely) fit on smaller gears:
The stacked B-C gears for the jack shaft work as before, with both labels on the top gear:
The admittedly flimsy motivation for all this was to make a 28 tooth gear to cut a 0.9 mm pitch, thus filling an obvious hole in the gear table.
My collection of gears could do 21-60-81-50, but the 81 T gear collides with the screw holding the 21 T gear. Rearranging it to 21-50-81-60 showed the B-C gears exceeded the space available.
Because it’s all ratios and a 28 T gear is 4/3 bigger than 21 T, reducing the rest of the train by 3/4 should work. In fact, it produced a reasonable 28-80-81-50 chain:
The fact that I do not anticipate ever needing to cut a 0.9 mm pitch has nothing whatsoever to do with it; that gear will surely come in handy for something.
While I was at it, I made a 27 T gear, because 27 = 21 × 9/7:
You can never have enough change gears. Right?
The OpenSCAD source code as a GitHub Gist:
| // LMS Mini-Lathe | |
| // Change gears with stacking | |
| // Ed Nisley – KE4ZNU | |
| // 2020-05 use Inkscape SVG gears | |
| // 2025-12 use BOSL2 gear generator | |
| include <BOSL2/std.scad> | |
| include <BOSL2/gears.scad> | |
| /* [Gears] */ | |
| TopGear = 0; // zero for single gear | |
| BottomGear = 28; | |
| /* [Hidden] */ | |
| ThreadThick = 0.20; | |
| HoleWindage = 0.2; | |
| Protrusion = 0.1; // make holes end cleanly | |
| /* [Dimensions] */ | |
| ShaftOD = 12.0; | |
| GearThick = 7.75; | |
| Keyway = [3.5,3.0,3*GearThick]; // x on radius, y on perim | |
| LegendEnable = (TopGear == 0 && BottomGear > 27) || (TopGear > 27); | |
| LegendThick = 2*ThreadThick; | |
| LegendZ = (TopGear ? 2*GearThick : GearThick) – LegendThick; | |
| LegendSize = 5; | |
| LegendRecess = [8,6,LegendThick]; | |
| LegendOffset = [0,LegendRecess.y/2 + ShaftOD/2 + HoleWindage,LegendZ + LegendRecess.z/2]; | |
| //———————– | |
| // Build it! | |
| union() { | |
| difference() { | |
| union() { | |
| spur_gear(mod=1,teeth=BottomGear,thickness=GearThick,shaft_diam=ShaftOD + HoleWindage,anchor=BOTTOM); | |
| if (TopGear) | |
| spur_gear(mod=1,teeth=TopGear,thickness=2*GearThick,shaft_diam=ShaftOD + HoleWindage,anchor=BOTTOM); | |
| } | |
| right(ShaftOD/2) | |
| down(Protrusion) | |
| cube(Keyway,anchor=CENTER+BOTTOM); | |
| if (LegendEnable) { | |
| translate(LegendOffset) | |
| cube(LegendRecess + [0,0,Protrusion],anchor=CENTER); | |
| if (TopGear) | |
| zrot(180) | |
| translate(LegendOffset) | |
| cube(LegendRecess + [0,0,Protrusion],anchor=CENTER); | |
| } | |
| } | |
| if (LegendEnable) | |
| translate([0,0,LegendZ – Protrusion]) | |
| linear_extrude(height=LegendThick + Protrusion,convexity=10) { | |
| translate([LegendOffset.x,LegendOffset.y]) | |
| text(text=str(BottomGear),size=LegendSize,font="Arial:style:Bold",halign="center",valign="center"); | |
| if (TopGear) | |
| zrot(180) | |
| translate([LegendOffset.x,LegendOffset.y]) | |
| text(text=str(TopGear),size=LegendSize,font="Arial:style:Bold",halign="center",valign="center"); | |
| } | |
| } |
In the midst of the humidification season, I spotted this while refilling one of the ancient Sears Humidifier bottles:
While it’s possible to buy replacement caps, this seemed more appropriate:
It’s PETG-CF, of course:
The shape is a ring with a simplified model of the cap removed from the middle:
It fits snugly over the cap atop a thin layer of JB PlasticBonder that should hold it in place forevermore:
The other side shows the crack over on the right:
Close inspection showed a few smaller cracks, so that cap was likely an original.
I made another ring for the other cap, only to find it was slightly larger with a black washer inside: apparently a previous owner had replaced one of the caps. The OpenSCAD program has measurements for both, not that you have either.
The OpenSCAD source code as a GitHub Gist:
| // Humidifier bottle cap reinforcement | |
| // Ed Nisley – KE4ZNU | |
| // 2025-11-29 | |
| include <BOSL2/std.scad> | |
| Layout = "Show"; // [Show,Build,Cap] | |
| /* [Hidden] */ | |
| Protrusion = 0.1; | |
| //—– | |
| // Bottle cap/valve | |
| // Collects all the magic numbers in one place | |
| Left = false; // the caps are different, of course | |
| CapODs = Left ? [43.0,42.1] : [43.1,42.9]; // [0] = base of cap | |
| CapHeight = 10.0; | |
| Notch = [0.6,2.0,8.5 + Protrusion]; // Z + hack for slight angle | |
| NumRibs = 24; | |
| RibAngle = 90 – atan(CapHeight/((CapODs[0]-CapODs[1])/2)); | |
| echo(RibAngle=RibAngle); | |
| $fn=2*NumRibs; | |
| module Cap() { | |
| difference() { | |
| cyl(CapHeight,d1=CapODs[1],d2=CapODs[0],anchor=BOTTOM); | |
| for (a=[0:NumRibs-1]) | |
| zrot(a*360/NumRibs) | |
| right(CapODs[1]/2) down(Protrusion) | |
| yrot(RibAngle) | |
| cuboid(Notch,anchor=RIGHT+BOTTOM); | |
| } | |
| } | |
| //—– | |
| // Reinforcing ring | |
| RingThick = 3.0; | |
| module Ring() { | |
| render() | |
| difference() { | |
| tube(CapHeight,od=CapODs[0] + 2*RingThick,id=CapODs[1] – 2*Notch.x,anchor=BOTTOM); | |
| Cap(); | |
| } | |
| } | |
| // Build things | |
| if (Layout == "Cap") | |
| Cap(); | |
| if (Layout == "Build" || Layout == "Show") | |
| Ring(); | |
Those of long memory will recall our vermiculture setup in the basement that turns kitchen scraps into plant food. We accumulate scraps in plastic milk jugs, which jugs get recycled after they become grody.
I finally made a decorative coaster to keep the sometimes-wet jug off the counter:
This used several of the larger smashed glass fragments from the collection:
They all fit inside a 165 mm square, with the conformal perimeter disguising the outline:
I printed the frame with the same blue PETG-CF that leaked epoxy the last time around. Using the correct filament setting (Extrusion Multiplier = 1.0) produced an epoxy-tight frame:
The overall process:
I dripped enough epoxy on each fragment to form a meniscus without having it go over the rim:
The Basement Shop temperature is just over 60 °F, so I put a heating pad in a huge ziplock bag, laid an aluminum sheet atop it as a heat spreader, put some waxed paper on the aluminum just in case, then did the filling described above:
A cardboard box on top helped the heating pad keep the coaster at a uniform 85 °F, slightly warmer than the epoxy instructions recommend, but it cured overnight with a wonderfully shiny surface.
Now that I have the process down, making glittery coasters is surprisingly easy.
The Smell of Molten Projects in the Morning 