Ed Nisley's Blog: Shop notes, electronics, firmware, machinery, 3D printing, laser cuttery, and curiosities. Contents: 100% human thinking, 0% AI slop.
This being the end of the humidifcation season, I tried to set the longsuffering Sears Humidifier’s water level float to dry the thing out. After a few days, it became obvious that wasn’t working and I eventually found the clip intended to hold the float at the top of its travel had broken:
Humidifier float clips – on float
Building the retina-burn orange replacement started with a scan of the original:
Humidifier float clip
The black segment at the bottom is a shadow due to the scanner’s light bar being offset from the imaging sensor.
Using GIMP, duplicate the remaining part of the latch, flip it left-to-right, then align it at the proper position:
Humidifier float clip – repaired
The latch is the only tricky part and the ID of the ring is easy to locate, so (still in GIMP):
Trace the edge of the whole shape
Using Quick Mask mode, remove all but the latch
Convert the selection to a path
Export it as an SVG file
Then import it into OpenSCAD and eyeballometrically translate the shape to put the ring ID at the origin:
There being no obvious affordance to get the ring over the two bumps in the float, I applied Channellock pliers to the float while easing the ring into place.
Six years ago I replaced the W5W incandescent front side marker bulbs in our 2015 Subaru Forester with amber LED bulbs:
Side Marker bulbs – failed adhesive
The adhesive holding the LED PCB to the aluminum “heatsink” has fossilized and the strip on the right is peeling off (with the left one not far behind), which likely accounted for its loss of light output and flickering.
Tearing it apart reveals the LED layout and what looks like a bridge rectifier or a big resistor (to fool the CAN bus?) on a tiny PCB jammed inside the shell:
Side Marker bulbs – rectifier
The other side of the PCB could be a buck converter:
Side Marker bulbs – buck converter
In round numbers, we’ve driven 18000 miles at an average of maybe 40 mph over those years; call it 450 hours. However, the side marker lights aren’t on unless the headlights are on; we do very little night driving, which means those LED bulbs are the usual crap.
Their rubbery port covers work best with 6 mm OD PTFE tubes, but let the MMU3’s 4 mm tubes slide into / out of the boxes under normal filament extrusion / retraction forces, so I conjured an adapter for PC4-M10 pneumatic fittings:
PolyDryer PC4 Fitting – installed
A pair of M3 screws hold the adapter plate in place, with an EVA foam gasket sealing against the cover:
PolyDryer PC4 Fitting – interior view
The PC4-M10 fittings let the 4 mm tubing slide right through, so the adapter has a 0.5 mm bottom sheet to block the tube, with a small hole for the filament:
PC4 Fitting Plates – bottom – solid model
You could use PC4-M6 fittings to block the tubing, but the 2 mm lumen on the fittings I have barely pass 1.75 mm nominal filament. Comments found elsewhere suggest identical PC4-M6 fittings have smaller lumens that snag the filament as it moves in one direction or the other.
The two blind holes get heat-staked 4×4mm M3 brass inserts.
The top has a threaded hole for the fitting:
PC4 Fitting Plates – top – solid model
Despite what the description says, the thread is not an M10 metric straight thread: it is a tapered pipe thread used for gas- and liquid-tight fittings. Considerable measurement & searching suggested a ⅛BSP-28 thread, because:
British Standard Pipe threads are used everywhere in the world except the USA
Both my metric tap sets have a ⅛BSP-28 tap along with all their hard-metric straight taps
The thread is painfully close to ⅛NPT-27, which would probably work in a pinch if it was the only tap you had.
Those PC4-M6 fittings might sport 1/16BSP-28 threads, but you’re on your own.
Further searching suggests nobody uses the corresponding tapered female pipe threads and everybody goes with a straight internal thread, so I conjured a stumpy threaded rod using the BOSL2 library and removed it from the adapter plate:
The 9.7 mm diameter is the ⅛BSP-28 “major diameter”, rather than its “gauge diameter”, simply because it produced a good fit. The beveled top guides the fitting into the hole, but I still managed to cross-thread one.
The OpenSCAD code also produces SVG files to laser-cut the foam gasket and a drill template:
PolyDryer PC4 Fitting – drill template
The holes were step-drilled to ⅛ inch (which has a historic relation to the ⅛BSP-28 size, because iron pipe) for a generous fit around the M3 screws.
That was way more complicated than I expected and I’m really glad to live in the future where this is a 3D printer project, not a metalworking project involving an actual tap in, say, steel.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The switch on the Anker LC-40 flashlight serving as a running light on my Tour Easy became slightly intermittent before I replaced it with a 1 W amber LED, but it was still good enough to become the troubleshooting flashlight in the tray next to the Prusa Mk 4 printer. Eventually, of course, it failed completely and Something Had To Be Done.
Although I knew an exact replacement switch had to be available from the usual sources, I could not come up with a set of keywords capable of pulling them out of the chaff.
Which turned into a multi-dimensional search over cap geometry, TPU extrusion speeds & feeds, and various impossible-to-directly-measure sizes:
Anker LC-40 Flashlight – TPU cap iterations
The squarish block over on the left is PrusaSlicer’s version of a support structure wrapped around the first cap version; if human lives depended on it, I could surely extract the cap, but it would take a while.
The remaining debris samples occured while discovering:
An extruder temperature of 230 °C, not 250 °C, works well
A conical shape of the lip around the open end to eliminate the support structure
TPU doesn’t bridge well, so the closed end must be down
Length of the central pillar to barely touch the switch stem when released
Cap length and wall thickness so the TPU shell can collapse enough to actuate and release the switch stem
Because I expected this would be an easy job, I used snap ring pliers to unscrew and rescrew the threaded retaining ring holding the switch PCB in place. Because the pliers didn’t have a stable grip on the ring, the threads eventually became just a bit goobered.
This was not a problem, because I have a(nother) 3D printer:
Anker LC-40 Flashlight Retainer – show view
The gray thing on the right is a simple pin wrench fitting both the original and the replacement retaining rings, so I can orient the rings properly while unscrewing & rescrewing:
Anker LC-40 Flashlight – pin wrench in place
The threads have a 0.75 mm pitch and, while it’s possible to print screw threads, even a tedious 0.1 mm layer height would define each turn of the thread with only 7-½ layers.
This was not a problem, because I have a mini-lathe:
Anker LC-40 Flashlight – thread cutting
The yellow & green things on the left of those solid models are the fixture holding a retaining ring for threading and the washer applying pressure to keep the ring in place:
Anker LC-40 Flashlight – lathe fixture – detail
The alert reader will note that washer lacks holes for the alignment pins I added after seeing the washer sit not quite concentric on the fixture. I could call it continuous product improvement, although I doubt I’ll print another one.
Setting up the lathe involved finding the proper set of change gears, including the vital 42-50 stacked gear I made a while ago to print metric threads on a hard-inch lathe:
Anker LC-40 Flashlight – lathe change gear train
Although you’re supposed to measure the thread spacing on a skim pass, I find it’s easier to just measure the carriage movement for one spindle rotation:
Anker LC-40 Flashlight – lathe gear check
A few passes produced a fine retaining ring:
Anker LC-40 Flashlight – OEM vs lathe-cut threads
Sporting much nicer looking threads than the goobered original:
Anker LC-40 Flashlight – OEM vs lathe-cut threads
The original switch had a stabilizing ring around the body to prevent it from wobbling under the original rubber cap.
This was not a problem, because I have a laser cutter:
Anker LC-40 Flashlight – new switch in stabilizer
Those came from a scrap of fluorescent acrylic.
The wave washer behind the acrylic stabilizer improves the contact between the PCB trace around the rim and the flashlight tailcap, with the current passing through the body to the “light engine” up front. The retaining ring provides enough pressure to compress the wave washer, which is why it’s so easily goobered without a close-fitting pin wrench.
With everything assembled in reverse order, the flashlight worked pretty much as it did back when it was new:
Anker LC-40 Flashlight – TPU cap installed
However, after describing this during a recent SquidWrench meeting, I discovered that adding “latching” to my keywords surfaced a bodacious assortment of flashlight switches, so (a few days later) I removed the not-quite-right switch and replaced it with an identical twin of the OEM switch requiring just a little lead forming to fit the PCB.
Even better, using the 3D printed pin wrench to screw the original retaining ring into the flashlight’s aluminum threads a few times re-formed (unrelated to recent electrolytic capacitor reforming) its goobered threads well enough to fit and work perfectly again.
So I have:
… reassembled the flashlight with more-or-less original components
… a repair tool kit ready when another LC-40 fails
… re-learned the lesson that any time spent making a fixture or a special tool is not deducted from one’s allotment
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Yes, it’s built into a recycled modem case. No, they don’t make modems like they used to, either. Regrettably, the five status indicators on the left were not set up as Der Blinkenlichten.
The inside view:
S-100 Bus Cap Reforming – inside view
The multi-winding transformer in the back feeds bridge rectifiers (out of sight behind the caps) producing bulk DC:
S-100 Bus Cap Reforming – bulk supply caps
The gray cap is 52 mF = 52000 µF 15 V for the +5 V regulators supplying the TTL logic on each board.
Two of the three blue caps (each 9 mF = 9000 µF 50 V) are for the +12 V and -12 V supplies. I think the third cap is a separate supply for a different purpose, but I did not trace out the wiring.
The on-board regulators seem to use solid electrolyte caps that should still be in fine shape you should replace on principle, per ericlscott’s experience. You’d want to bring up each board separately while probing the voltages; the box of stuff accompanying the system has an extender card that should make probing easier.
I hoped to boot the thing after restoring the caps, but a casual inspection showed wire corrosion:
S-100 Bus backplane – jumper wire corrosion
You’d want to pull the backplane out and replace those jumpers, as well as clean the bus contacts, before applying power.
The system has two 8 inch floppy drives in a separate case with its own power supply:
S-100 Bus floppy drives – overview
There was some corrosion in there, too:
S-100 Bus Floppy Drive – optical sensor corrosion
So I confined myself to reforming the caps and must let someone with more powerful motivation restore the rest of the system before trying to connect everything and booting CP/M.
The general idea behind “reforming” an electrolytic capacitor is to regrow the oxide layer separating the anode and cathode electrodes, which involves passing a current of about 1 mA for as long as it takes to bring the terminal voltage up to the cap’s maximum rated voltage:
S-100 Bus Cap Reforming – 52mF 15V
That setup consists of an absurd number of PowerPole adapters putting the meter in series with a fuseholder repurposed to hold resistors to limit the current, with leads eventually ending up on the capacitor:
S-100 Bus Cap Reforming – 52 mF 15 V cap connection
The red dot is the overpressure vent, not a polarity marker.
Apparently the Greek mu symbol wasn’t in the font available for the labels, as all the capacitors use m in its place: that capacitor is 52 mF = 52000 µF.
The white plastic ejection handle belongs on the right end of the CPU board seen in the second picture, which was not plugged into its slot when I opened the case. I snapped the handle in place and plugged the board in just to keep it out of trouble. The case does not have board guide slots along the edges that would let the handle eject the board, but all that was definitely in the nature of fine tuning back then.
I started with +15 V through a 16.9 kΩ resistor and swapped in 3.3 kΩ, 1 kΩ, and 220 Ω resistors as the cap voltage crept upward over the course of two days and eventually settled to a steady state:
S-100 Bus Cap Reforming – 52mF 15V final voltage
After discharging, the cap measured 59.5 mF with a 0.3 Ω ESR, which definitely seemed Good Enough.
I reformed the three 9 mF 50 V caps at the same time by applying 50 V to three resistors captured on their screw terminals, changing the resistors as the voltages rose:
S-100 Bus Cap Reforming – 50 V caps
Those three caps eventually measured (clockwise from upper right):
9.66 mF, 1.0 Ω ESR
9.76 mF, 2.6 Ω ESR
10.46 mF, 3.4 Ω ESR
The ESRs suggest they’re somewhat dried out, but I’d be tempted to run them anyway, because the on-board regulators should knock down the ripple.
All of the reformed caps had leakage currents of a few hundred microamps. They’re not new capacitors and never will be, but they may be Good Enough.
Getting the caps out of the diskette drive power supply required easing the entire supply frame / heatsink out of the case before unscrewing the capacitor clamps:
S-100 Bus Cap Reforming – 16 mF 50V
That one eventually measured 22.1 mF with 0.14 Ω ESR. Its sibling, nominally 38 mF at 15 V, came in at 48.9 mF with 0.95 Ω ESR.
The power supply PCB carries a handful of smaller aluminum electrolytic caps that are impossible to remove without unsoldering all the TO-3 transistor leads coming through the aluminum heatsink / frame, then completely dismantling the power supply:
S-100 Bus floppy drives – power supply PCB
Although I reformed the big caps, I think a better plan would be to replace the whole thing with a contemporary switching supply. AFAICT it has 24 V and 5 V outputs; because we live in the future, dual-output switchers are cheap & readily available.
And then I closed the cases to get them ready for the next part of their adventure …
After about 7.5 years (!) the 64 GB card in my Sony HDR-AS30V helmet camera breathed its last:
SanDisk 64 GB MicroSD card – end of life
Over the course of several rides I noticed many video files ended prematurely or would not play. I gave up attempting to reformat the card in overwrite mode using the Official SD Card formatter after four hours, which says the wear leveler in the card has no spare capacity.
In round numbers, I ride 1700 miles a year at 12 mph, so the card recorded 1000 hours of 1920×1080 video at 60 frame/s, storing one 4.3 GB file every 22.75 minutes for a grand total of 12 TB of data.
A new Sandisk 128 GB High Endurance card cost a third of what the 64 GB card did and, after setting the partition label to AS30V, it’s off to a good start:
Street Lamp Pole – Rombout House Ln – 2025-05-07
That’s the street lamp pole installed on the replaced base at the corner of Rt 376 and Rombout House Lane, with the barrels gradually being pushed closer and closer to the pole by turning traffic on the newly paved lane.
That pole is not going to see the end of this year.
A postcard arrived last week telling me to call a special number for special deals on Medicare Advantage plans. Being that type of guy, I managed to read the microscopic Fine Print and found this amusing blooper amid the disclaimers weasel wording:
Medicare Advantage mail spam
Inserting insurance carrier names should have happened before printing the card, so [CarrierA] and [CarrierB] are either placeholders or mail-merge variables.
Also, you’re seeing the contrast-blown and magnified version of the postcard. The original Fine Print had faint orange ink on light green cardstock: colors having different hues with the same saturation and value to minimize legibility. In general, folks eligible for Medicare Advantage plans have trouble reading Fine Print, so the choice was not accidental.
The Smell of Molten Projects in the Morning 