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.

Category: Machine Shop

Mechanical widgetry

Window Latch Cam Rebuild

One of the living room window latches stopped working and, although the correct response would be to simply replace it, the window sash already had six holes surrounded by paint scars suggesting this wasn’t the first failure.

The latch as built:

Window latch cam – as built

The stamped steel spring holder was loose on the handle arm’s square pot-metal post, due to obviously poor factory swaging during assembly. Unseen below the plate is a cam molded into the handle intended to slide the white latch plate with the stamped post (over on the left) guiding the plate around the curved track and into the recess.

I don’t have a “before” picture of the cam, but it had completely worn away against the various steel parts. The task boiled down to replacing it with something more durable, which turned out to be a reshaped M3 brass screw:

Window latch cam – test fit

The Sherline setup to remove the worn cam, with the screw in place for a test fit:

Window latch cam – Sherline setup

The vise loosely clamps the handle, mostly for alignment, with the step block clamp holding it firmly in place.

The lever arm looked like this after poking the endmill into it a few times to clear the rubble, then using a smaller endmill as a drill to get a flat-bottomed recess:

Window latch cam – milled hole

Obviously, that was accomplished with eyeballometric alignment and manual jogging.

Reshaping and trimming the screw with Tiny Lathe produced a suitable cam-like object:

Window latch cam – screw installed

I planned to epoxy it in place, but AFAICT there is no way for it to go anywhere, so I just reassembled everything in reverse order:

Window latch cam – reassembled

I brutally swaged the square post with an automatic center punch cranked to maximum throttle while forcing the spring holder firmly atop the stack. The latch now moves smoothly and everything stays properly aligned.

While that was going on, the window sash had wood pins pounded into the errant screw holes with wood epoxy filling the gaps. Aligning the latch, drilling holes in the proper spots, and screwing the latch in place proceeded as expected.

That wasn’t easy, but now I can install plastic film over those windows for the winter.

2026-10-14
Samsung French Door Refrigerator Mullion Flipper

The vertical bar between the doors of the Samsung French Door refrigerator that Came With The House™ is called a “mullion” and the curvy doodad on top that moves it out of the way when you open the door is called a “mullion flipper”, which I learned while attempting to repair it:

Samsung Refrigerator Mullion Flipper – regluing

The entire top of the mullion broke off and, while a solvent bond can (ideally) be as strong as the original plastic, if the part broke once it will probably break again. Which happened in short order, despite all the clamping.

The official replaceable part is the entire mullion, but a more practical alternative is a cap fitting the top of the broken mullion, apparently either molded or 3D printed.

Having a 3D printer of my very own, a quick search brought forth a solid model of a mullion flipper that fit exactly:

Samsung Refrigerator Mullion Flipper – 3D print top view

It’s printed in clear PETG, which seemed a better match than white, and is held in place with a wrap of double-sided tape around the top, which may not be strictly necessary.

It flips the mullion just like the original:

Samsung Refrigerator Mullion Flipper – 3D print outside view

The yellow strip is 2 mil Kapton tape applied to both sides of the torn flexible seal running the length of the door, which we agreed was less awful than attempting to replace the entire seal on both doors. It must be trimmed slightly narrower than shown to avoid jamming on the other door.

Perusing previous posts will explain my overall low opinion of Samsung appliances.

2026-10-13
Layered Paper: Whimsical Perching Bird

Start by scanning a page from the (©2013 paper!) Whimsical Designs Coloring Book and opening it in GIMP:

Whimsical Perching Bird – outline

Mary says the book is intended for appliqué quilting, which seems close enough to layered paper for my simple purposes.

Select the black lines and grow the selection to about 1.5 mm wide:

Whimsical Perching Bird – mask selection

After some editing, convert the selection to a path, then save the path as SVG to create the Black Mask layer.

Clear the selection and bucket-fill the outlines with cheerful colors:

Whimsical Perching Bird – colored

If you’re fussy, you might devote a sheet of black paper to the eye, but I just dabbed a dot of black marker.

Load into LightBurn and knock out paths on successive layers:

Whimsical Perching Bird – LightBurn layout

Stack sheets of colored paper in the appropriate order, Fire The Laser, paste the layers together, and frame:

Whimsical Perching Bird – framed

After working through the Cornell Painted Warblers, it really was that easy!

Stipulated: I cannot draw even simple Art, but I enjoy my time in the engine room.

2026-10-12
Countersunk Screw Recess Fillers

For reasons not relevant here, a friend must fill several pre-drilled countersunk flat-head screw holes with brass plugs suitable for M5 screws. These proof-of-concept samples give the general idea, with the red blotch on the left marking the first attempt:

Shelf flush screw fillers – samples

He gets to make as many as needed …

The flat-head screws have an 82° included angle, so the lathe compound angle must be 41°. Tiny Lathe lacks sufficient reach to accomplish that on the front side, but cutting the back of the brass rod with the spindle running in reverse will work:

Shelf flush screw fillers – lathe setup

The quick-change tool post is aligned parallel with the lathe bed to put the tools at the proper angle.

It turns out 41° is about as far as the compound will go:

Shelf flush screw fillers – compound angle setting

This, readers of long memory may recall, is the mini-lathe with the cross-slide DRO interfering with the compound enough to prevent it from cutting ordinary 60° screw threads, which is why the angle tops out at 42° instead of 60° in both directions.

The filler needs a 5.1 mm hole for EZ clearance, which is straightforward:

Shelf flush screw fillers – drilling

That was the first attempt after I got the compound angle worked out, which is why it eventually got the red mark. You want to drill on a flat end after countersinking with a center drill. It’s also good practice to drill only a few diameters deep, rather than trying to gun-drill sufficient hole for half a dozen fillers with one poke.

Run the spindle in reverse to cut the 41° angle on the back side:

Shelf flush screw fillers – backside cutting – detail

Run the spindle forward to separate the finished filler with a cutoff tool:

Shelf flush screw fillers – parting off

The length isn’t critical, so holding a millimeter scale against the right side of the tool and eyeballing the distance from the end of the taper should suffice.

Filing to polish, deburring to finger-frendliness, and similar activities are in the nature of fine tuning.

Now for some Quality Shop Time!

2026-10-09
Furnace Water Fill Valve Refurbishment

The water fill valve (a.k.a. water shutoff valve) for the gas furnace is inconveniently located behind the furnace and water heater:

Furnace fill valve location

After a few days of dripping Kroil penetrating oil on the valve body threads, the Greater Adjustable Wrench eased it just enough to convince me it would come apart, whereupon a search through the Big Box o’ Tubing disgorged a length of flexy silicone tube snugly fitting the valve drain port.

With the water turned off downstream of the house filter and the pipes depressurized, the tube directed most of the water into a bucket perched atop the furnace, with a scrap towel catching most of the remainder while unscrewing the valve. I clamped a rubber disk over the body while replacing the washer:

Furnace fill valve – ancient washer

Maybe that washer isn’t 60-some years old, but it’s definitely well beyond its Best Used By date. For the record, the fill valve uses the same 3/8L washer as the zone drain valves.

While I had it apart, I replaced the gasket, which was of course different from the drain valves:

Furnace fill valve – gasket

That was 25.4 mm OD, 22.7 ID, for a measured 1.3 mm width. The recess machined into the valve body limits the OD and the ID could go down to 22.0 mm. The gasket sheet is 0.8 mm thick, which seems to work well enough.

Reassemble in reverse order, wipe up various spills, turn the fill valve on to repressurize the furnace, close the valve, and the water pressure now remains at 12 psi as it should.

To the best of my knowledge & belief, every hot water baseboard furnace in all of our houses has operated with the water fill valve open: the pressure reducing valve sees normal water pressure upstream and regulates the boiler pressure, with a reverse flow check valve preventing backflow into the house plumbing as needed.

The Bell & Gossett instructions for the pressure reducing valve say, clearly and specifically, without asterisks or evasions, that you fill the furnace / boiler to the 12 psi regulated pressure, then close the fill valve. The furnace runs as a closed system, with the expansion tank keeping the pressure more-or-less constant at 12 psi.

It’s possible the pressure reducing valve is gradually failing, but it doesn’t matter with the water fill valve closed.

2026-10-08
HQ Sixteen: Improved Table Shimming

Although adding table shims / leveling blocks dramatically improved the HQ Sixteen’s travel along the table, it still had a tendency to roll away when parked at some locations. The padded shims had compressed enough to throw off the alignment between adjacent sections, so returning to the original rigid blocks seemed prudent.

I removed all the shims to let the middle three sections of the tabletop droop as it did when we got it, put a ten foot saw guide across the entire table, and added 5 mm (!) plates to the stacks already under the high (!) end of the table feet to level the two ends. The floor is not that far out of level, so the table structure itself must be mismatched from one end to the other.

Adding 3 mm (!) plates under the front pair of legs put the table far enough out of level to prevent the machine from rolling front-to-back on its carriage. I think the carriage is slightly bowed downward in the middle, but now the machine is more inclined (so to speak) to remain in one place.

The saw guide showed the table sections on each end, despite being screwed firmly to the table structure, were slightly sloped downward toward the center. Shimming the sections with thin cardboard atop their steel mounting brackets leveled them so the machine no longer wanted to roll away when parked at either end, which was a good start.

Adjoining table sections have peg-in-hole alignment, with their accuracy leaving something to be desired. Shims under the machine tracks level it across the mismatched joints:

HQ Sixteen – table track shims – detail

The brown chipboard is 0.7 mm thick and the white cardboard is 0.4 mm, with lengths eyeballometrically matching the misaligned part of the table section. They’re not quite so conspicuous in action:

HQ Sixteen – table track shims

I then inserted wood shims under the tabletop to level the three middle sections. Unlike my previous efforts, each section got four shims to level it both front-to-back and side-to-side. After considerable iteration, the HQ Sixteen would remain where it was parked at (almost) every location across the entire table span.

Then I used a set of small hole gauges to measure the distance between the table’s metal structure and the bottom of the tabletop section near each wood shim. I had previously used a taper gauge that measured the clearance on only the near side and it turns out the metal has plenty of irregularities across its bent shapes.

Then feed the measured values into the OpenSCAD shim block generator mentioned earlier to get more shims:

HQ Sixteen – table shims – array – solid model

Slide them in place and remove the wood shims:

HQ Sixteen – table shims – adjacent sections

Yes, that’s nearly a 1 mm difference, despite the bottoms of the two table sections being obviously aligned.

Each block now has a narrow strip of craft adhesive sheet holding it to the table top, in the ~~expectation~~ hope the machine’s normal vibration won’t shake them loose.

Initial user response is positive: the improved anchor may remain on the shelf.

2026-10-07
Layered Paper: Cutting and Assembling the Cornell Painted Warblers

With all the layers laid out, duplicating and plunking one in the middle of the LightBurn workspace looks like this:

Page 4 – Layer 1 – Lightburn centered layout

In Absolute Coordinates mode the laser will cut it at the center of the platform, so put a sheet right there:

Print vs cut – 100.3pct plain paper – setup

The Position Laser tool puts the laser head at a selected location in the design, with the picture showing it at the center of the upper right target. The trick is to slide the paper on the platform to put that target directly under the red-dot pointer:

Print vs cut – 100.3pct plain paper – UR align

I occasionally fire a test pulse to verify the CO₂ laser hits the red dot:

Red dot vs printed target vs laser spot alignment

Do the same for the target in the lower left corner, rotating the paper around the upper right target, and iterating until both targets match.

Because the printer does not produce accurately sized output, this can happen:

Print vs cut – 100.3pct plain paper – LR

You’ll get a hint of the problem when it’s impossible to put the targets under the the laser head at the lower targets, although it may look close enough to work.

That offset was on plain paper with the output scaled by 100.3% vertically. Applying the same scaling (to the Happy Bird picture, but with the same page template) and printing on Art Cover paper produces this result:

Print vs cut – 100.3pct Art Cover – LR

Eventually, I can find a suitable scale factor to make the printed image come out right and the cuts align properly:

Page 5 – Test piece – as cut

Those tenuous paper strands show why I couldn’t use the Letter page fixture and had to cut the sheets directly on the honeycomb: the weight of the uncut sections distorted the remaining sheet beyond belief. The squat silver cylinders in the pictures are neodymium magnets holding the paper flat on the honeycomb.

All the other layers have trivial positioning, because the cuts need not line up with anything on the sheet:

Layered paper alignment

Just frame the first layout, center the sheet around it, set the MDF stops to kiss the edges of the paper, and all the other sheets will come out just fine.

One side effect of all this cutting is the happiest scrap bin ever:

Happy Laser Scrap Bin

With all the layers cut out, assemble them face-down in the gluing fixture, align the black mask on the front by hand, put the stack into a suitable frame, and it’s ready to show off:

Cornell Warblers – page 4

That wasn’t exactly easy, but next time I can make different mistakes.

2026-10-06