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: Electronics Workbench

Electrical & Electronic gadgets

One-Watt LED Autopsy: No Trouble Found

Over the last several weeks the taillight on Mary’s Tour Easy became dim. I verified the LED had failed by soldering the taillight driver to a different 1 W LED stuck on a separate heatsink, whereupon the LED lit up as it should.

Testing the failed LED with a bench power supply clipped to its wires showed it passed only 15 mA, even with 6 V applied directly to the wires. Those 1 W LEDs normally drop 2 V at 400 mA, so something was definitely wrong in there.

Which meant I had to build another taillight, as described yesterday, and reuse the original electronics and PVC tube. After mounting the new taillight on Mary’s bike, I could autopsy the failed LED.

Because everything is glued into the heatsink, the first step is cutting it apart to free the LED:

1W red LED Autopsy – heatsink parting

I stopped the first cut when it just kissed the two wire holes, moved over a little more than the width of the cutoff tool, and tried again. Second time was the charm and I now have a clean aluminum tube that ought to come in handy for something, sometime.

Concerted prying under the edge of the LED PCB popped it off the heatsink:

1W red LED Autopsy – heatsink epoxy bond

The red wire sticking out on the right shows a little damage from the cutoff tool, which also chipped the epoxy while bending the aluminum LED base. The bare spot in the middle matches the screw hole in the heatsink, with the rest remaining firmly bonded, so the epoxy was doing its job.

Prying the LED off the lens holder:

1W red LED Autopsy – lens holder detached

If I didn’t know better, I’d say the red silicone insulation was sun-faded, but for sure no sun gets inside the lens holder.

An ohmmeter showed both wires had good continuity to their LED pads, which were firmly soldered to the PCB pads; the pads don’t carry any current in this setup. I then hitched up the bench supply to the wires again:

1W red LED Autopsy – no trouble found

Whereupon the LED glowed brightly and some quick testing (because it’s not on a heatsink) with the bench supply in constant-current mode showed it once again drew about 400 mA with 2 V applied and glowed brilliantly as it should.

I had previously removed the reflowed solder from the PCB thermal pad and added thermal paste before resoldering the LED:

1W LED – thermal grease

So I have no idea what caused the original failure, how the LED could survive with 6 V applied to its terminals, or what would have caused it to recover. It now has a “?” scrawled on the PCB for future reference.

The world is full of mysteries …

2026-10-23
Tour Easy Running Light: Machining & Assembly
The 1 W red LED in the taillight on Mary’s Tour Easy failed and, because the LED is epoxied into the heatsink forming the business end of the light, I had to machine a replacement. This proceeds much as before, with some improvements & simplifications.

A revised sketch of the key dimensions shows a clearance hole for the 3 mm screw (to eliminate tapping the heatsink) and a reminder about how to locate the wire holes:

Tour Easy running light – housing dimensions – M3 clear – detail

Bore a 19 mm ID recess 10 mm deep in the (1 inch OD × ¾ inch ID) aluminum tube and score the ID for improved epoxy griptivity:

Tour Easy Running Light – heatsink scoring

Knurl the ¾ inch rod to bring the OD to a slip fit in the shell:

Tour Easy Running Light – heatsink plug knurling

Take them ouside, hose down with brake cleaner to get rid of the cutting fluid, butter ’em up with epoxy, align in the lathe, and let ’em cure overnight:

Tour Easy Running Light – heatsink epoxy fixture

This time I managed to not cut off the rod, which dramatically simplifies boring the shell and facing the front end of the rod flush with the inside of the shell so the LED PCB sits flat in there:

Tour Easy Running Light – heatsink front ID

Clean up the shell OD and turn the back to 23 mm OD to fit the PVC tube from the failed taillight:

Tour Easy Running Light – heatsink shell OD

Crisp up the edge for a tidy fit:

Tour Easy Running Light – heatsink shell corner

Break / file all the corners, then bandsaw the rod and face the back end:

Tour Easy Running Light – heatsink rear facing

Set it up on the Sherline mill, find the center, and use manual jogging to poke the holes at the proper coordinates:

Tour Easy Running Light – heatsink Sherline setup

Take the finished heatsink outside again for more brake cleaner action.

The 3D printed circuit plate now sports two inserts:

Running Light Circuit Plate – two inserts – solid model

Epoxy the inserts in place, prepare two 1.6 mm = 1/16 inch steel wires, and chop the head off an M3 screw:

Tour Easy Running Light – heatsink with circuit plate

Stipulated: my center-finding hand is weak.

Aligning the LED heatsink with those wire holes is a pain:
- Clean up the 1 W LED PCB
- Solder the LED in place atop some heatsink compound
- Push the lens holder (without the lens!) over the LED
- Insert holder + LED into the heatsink
- Rotate to put one LED terminal between the wire holes in the heatsink
- Rotate to align the recesses in the PCB with the wire holes in the heatsink
- Match-mark the holes on the lens holder with a transfer punch
- Grab the holder in the Sherline vise
- Align & drill wire holes in the holder
- Solder wires to the LED PCB, aimed inward atop the LED leads
With the LED & lens holder prepared & set aside, butter the circuit plate parts with JB PlasticBonder, align the plate perpendicular to the heatsink using V-blocks / 123 blocks / whatever, and let it cure overnight.

The lens allows plenty of room to run the wires from the PCB around the LED inside the holder to the holes:

Tour Easy Running Light – 1W LED lens fit

So do that:
- Dab PlasticBonder on the bottom of the lens holder
- Stick the LED to the holder (still with no lens!)
- Thread the LED wires through the holes in the holder, the LED PCB, and into the heatsink
- Dab PlasticBonder on the bottom of the PCB
- Ease the holder + LED into the heatsink while pulling the wires through the holes
- Finally install the lens in the holder
- Clamp the whole mess lenthwise with padding atop the lens to prevent scratches
I transferred the circuitry from the failed taillight to the new one, atop refreshed double-sided foam tape:

Tour Easy Running Light – Arduino side

With hot-melt glue securing the 2 Ω current-sense resistor to the circuit plate:

Tour Easy Running Light – regulator side

Slide the PVC case over the circuit plate, secure the end cap with a 25 mm screw, and reinstall on the bike:

Tour Easy Running Light – taillight installed

It’s angled 1° up from horizontal:

Tour Easy Running Light – taillight in use

Wow, is that thing bright!
2026-10-22
One-watt LED Thermal Pad Reflow: Nope

While building up another taillight for Mary’s Tour Easy (about which, more later) and based on previous experience, I wondered if the 1 W red LED’s thermal pad was soldered to the circuit board as it should be.

Lo and behold, the solder paste was reflowed, likely while tinning the six solder pads:

Tour Easy Running Light – 1W LED PCB solder

But not with the LED in place:

Tour Easy Running Light – 1W LED base glue

What looks like melted solder is really an adhesive layer, much like the craft adhesive sheet I use to stick things together. Presumably, they just glue the LEDs to the PCB / heatsink, solder its two leads by hand, and call it done.

Checking an LED marked “Dead” on the backside of its heatsink shows the amount of solder paste varies, but the LED wasn’t there when it melted:

1W LED – base glue and PCB solder

This not being the first time, I cleaned the adhesive off the center pad and the LED, removed the solder, applied a dab of heatsink grease, and soldered the LED in place.

Which looked like this on the previous taillight:

1W LED – thermal grease

Might be a bit too much grease, but it’s certainly a better heat conductor than adhesive.

2026-10-21
Shopvac Model 500A Motor Wire Repair

The Shopvac Model 500A (1.25 HP!) inhaling dust from Tiny Bandsaw didn’t start up when I flipped the switch. It had been making horrible noises, but Shopvacs typically sound like they’re on the ragged edge of death, I wear 30 dB earmuffs when it’s running, and that wasn’t the problem anyway.

It turned out the hot (black) wire lead to the motor burned up:

Shopvac motor – charred wire

The neutral (white) wire over on the left side looked just fine:

Shopvac motor – good wire connector

But the hot wire connector was toasted:

Shopvac motor – charred connector

As the connector failed its resistance went up and so did the wire temperature, as shown by the plastic case:

Shopvac motor – melted cover

With barely an inch between the motor armature and the failed connector, this became a zero-Mulligan repair:

Shopvac motor – remaining armature wire

I clipped the failed connector off the wire, very carefully sanded off the burned insulation, then gingerly crimped the business end of a ferrule around the wire + a doubled length of 22 AWG hookup wire:

Shopvac motor – lengthened wire

Doubling the wire brought it up to 19 AWG, a bit larger than the 20 AWG armature wire, and maybe improved its durability. The ferrule surely isn’t rated for this sort of abuse, so I eased some solder into its end in the belief that any amount of solder is a better conductor than a similar amount of air.

Crimp the other end of the hookup wire into the business end of a solderless ring terminal along with the hot wire, anchor the ferrule to (and insulate it from) the armature with a wrap of silicone tape, do the same around the ring terminal, then squirt hot melt glue into the ruined connector housing to keep the wire from flopping around:

Shopvac motor – repaired wire

If it gets hot enough to melt the hot melt glue, there are other problems about to happen.

Reassemble in reverse order and the Shopvac once again makes its on-the-eve-of-destruction whine while inhaling chips.

A slip of tape marked “$5” fell off during this process, reminding me it was a garage sale find from quite a few years ago: saving it from an early grave may be an overall Good Deed.

2026-10-15
HQ Sixteen: Grip Cap Assembly

After 3D printing another set of cap parts, glob the switches into place with tan JB Weld Plastic Bonder:

HQ Sixteen grip cap – switch gluing

More Plastic Bonder joins the cap body to the plug fitting into the handlebars:

HQ Sixteen grip cap – gluing fixture

I laser-cut an acrylic disk to fit the plug stem and hold the bottom of the plug flush with the bottom of the cap, with the PTFE fabric preventing inadvertent bonding of All The Things. It’s perched atop an empty thread cone to cure: quilting requires a lot of thread and I have a good stock of cones.

The button cap body solid model now includes alignment marks to eliminate the need for the bench block & clamps:

HQ Sixteen grip cap – alignment marks

Soldering the cable directly to the switch terminals is déclassé, but the collar prevents the cable from exerting any force on them:

HQ Sixteen grip cap – wiring

This time around, dabs of hot-melt glue hold the faceplate to the cap body.

I also used hot-melt glue to stick the transparent cover onto the handlebar body, as the OEM sticky tape was losing its mojo:

HQ Sixteen grip cap – wiring closeout

The rightmost three LEDs went dark a while ago and, while I had everything apart, I tracked the failure to the third LED from the right, replaced it, and it’s all good again. Those LEDs (plus the side groups) still can’t hold a candle to the Nose Ring Lights and Under-arm Lights, but it’s good to have all the lights working.

I managed to break one tiny wire on the header (wrapped in black silicone tape next to the IC on the right) joining the two cables from the control caps, but other than that things went smoothly.

Now I can start on other long-awaited tweaks before the next quilt arrives.

2026-09-29
HQ Sixteen: New Grip Control Cap Labels

The Run/Stop (⏯) button on the HQ Sixteen’s left-hand grip control cap recently became intermittent:

HQ Sixteen – grip cap installed – left

Which is a little more than a year after I built it, so I’m not red-hot pleased with the switch reliability. On the other paw, Mary has done a lot of quilting in the last year, generally controlling the machine with her left thumb, so that particular button has gotten a lot of use.

Because I glued the caps together, rather than trying to hide screws in finger-friendly locations, the replaceable unit is the whole cap.

I originally cut the cap labels using LightBurn’s Print-and-Cut tool to align the laser to the printed label, but subsequent experience making punched cards showed it’s generally easier (at least for my simple projects) to skootch the layout to match the laser position at known features:

HQ Sixteen control caps – laser alignment

Those small magnets hold the laminated label flat against the honeycomb.

The Position Laser tool puts the laser at a selected corner of the square, with Object Snap ensuring it’s really at the corner. Firing a test pulse to verify the red-dot pointer is aligned with the focused CO₂ laser spot is good practice.

Unlike the cards, the control caps don’t have fractional-millimeter tolerances, so, after verifying the printed label alignment at all four outer corners, it’s Close Enough™:

HQ Sixteen control caps – label cut

Cutting the adhesive sheet posed no alignment problems:

HQ Sixteen control caps – adhesive cut

Then it’s just a matter of peel-n-stick.

2026-09-28
Samsung Dishwasher Drying Cycle Crash: Wax Motor

Replacing the drying fan didn’t prevent the Samsung dishwasher from crashing when the drying cycle starts, so I eventually figured out how to run the “Smart Install” test sequence:

Samsung Dishwasher – DW80K7050US – p 37 – Service Test Mode

Which ticked through everything right up to the start of the “Check Drying” step:

Samsung Dishwasher – DW80K7050US – p 39 – Check Drying

Whereupon all the control panel LEDs went dark and a little red status LED on the PCB began blinking what must be an error code: long / pause / long-short-short. Of course, I cannot find any doc for that LED’s code.

The “Auto Door Open Actuator” is a wax motor, a device I had never encountered before. It opens the door through a whippletree pushing two tabs against the door:

Samsung dishwasher – door auto-open mechanism

The meter measures the voltage across the ~~wax motor~~ Auto Door Open Actuator through improvised widowmaker jumpers jammed onto the quick-disconnect fittings. As you might expect, the meter showed 000 VAC before, during, and after the crash, suggesting something is wrong with the relay intended to activate the motor.

Because the test sequence activates the drying fan after opening the door, the fan never even twitched.

For twelve bucks (spend up to $150 if you prefer), I had a generic DD66-00089A actuator in hand before dragging the dishwasher out of its lair and, being me, had to measure what it did:

Samsung Dishwasher – Door Actuator wax motor – test layout

It’s about 1.8 kΩ across the terminals and, not being fussy about its input voltage, will dissipate anywhere from 6 to 32 W in operation. Because the “motor” works by the force generated by expanding liquid wax, the response time obviously goes down as the voltage goes up.

After a few cycles to figure things out, plotting elapsed time against piston displacement is surprisingly linear right out to the 6.5-ish mm maximum stroke:

Wax Motor – time vs. displacement

It take much longer from a cold start, which is how it’s normally used, so the 30 s intercept at 0 mm is definitely on the low side.

Samsung no longer manufactures the DD82-01337A7050 control board, the usual aftermarket suppliers report it’s out of stock, and, given the lack of doc, I am loathe to pull it out of the dishwasher for failure analysis / repair. The one-hour cycle gets the dishes clean and a 46 minute countdown timer reminds me to open the door for drying, so the failure seems survivable for the immediate future.

2026-09-16