Ed Nisley's Blog: Shop notes, electronics, firmware, machinery, 3D printing, laser cuttery, and curiosities. Contents: 100% human thinking, 0% AI slop.
Mary’s feet are exquisitely sensitive to irregularities in the insoles of her shoes, which poses a real problem with her bike shoes: those SPD cleat recesses are no good at all.
This is a view down into one shoe, with the SPD cleats adjusted all the way to the rear. That leaves a large recess in the front, which was painfully obvious to her sole. The white shape is the gap filler…
I pressed a sheet of paper across the gap to get the general shape, traced it twice onto a slab of 0.060-inch aluminum with a nice pebbly paint job, and cut the two pieces out. A few conversations with Mr Belt Sander, a few licks with a rat-tail file, and they dropped right onto place. The recess is slightly curved, but I didn’t have to bend the pieces to fit.
I laid duct tape across the whole affair, put the insoles back in place, and it was all good.
The backing plate is 0.072 inch thick and she was content with the difference.
In previous shoes, with the cleat near the middle of the adjustment range, I’ve stuffed epoxy putty into the gaps. That works, but it doesn’t bond to the (miracle engineering plastic) soles and tends to crumble. This is Not A Good Thing…
The discussion following that post prompted me to take a closer look at the corroded spider. I planned to pull the spider off the back of the drum and examine the pieces, but a week of dribbling thread lube around the bolts left two of them firmly affixed.
While I don’t have it completely apart yet, some observations are in order…
Spider mounting bolts through drum
The bolts are stainless steel and utterly immovable with the usual screwdriver-handle-mounted Torx bit. I got the first two bolts out by putting a T30 bit in a 1/4-inch socket in a ratchet wrench and applying brute force.
A few days of thread lube (the incomparable PB B’Laster) persuaded two more out of their lairs. The remaining bolts may require even more brute force, but I’ll give the lube a few more days to work its magic.
Despite that, the bolts and holes are not corroded. They may have some thread locker down in there, but I see little evidence of that. I think it’s just a case of being torqued down hard, then set adrift in ionic water for half a decade.
The outer third of each arm has a covering of corrosion products, but the metal below that (now dried and flaking) gunk seems undamaged. The arms have severe corrosion and cracking throughout the inner two-thirds of their length.
Spider corrosion
If this were chemical corrosion, I’d expect it to apply evenly throughout the length of the arm, because the presence of corrosion products over the entire arm indicates pretty good distribution.
However, galvanic corrosion should follow the same pattern, so I’m not sure what to make of this.
The fact that an oxidation layer on the stainless steel tends to passivate it may not really matter. Compare the surface areas of the drum and the spider: there’s a whole lot more drum than spider, so even a passivated drum could provide enough current to rot the spider.
The ends of the spider spend their lives whipping through the water inside the tub at a pretty good clip. That could dislodge most of the crud and leave them reasonably clean, at least compared to the hub that moves more slowly (same rotational speed, smaller radius). It’s also true that the water level never reaches the hub, remaining below the level of the door seal.
Thus, the hub probably gets splashed, but never immersed, and thus has no way to remove any contaminants. The corrosion products simply build up there, keeping it wet throughout its life.
I maintain there’s little drying going on, even with the door open, in the relatively short intervals between washings. The hub region would be least likely to dry, however, because there’s absolutely no ventilation back there.
All that notwithstanding, this corrosion should not happen.
I’d very much like to see some measurements: we’re all obviously guessing at the conditions. The plastic tub surrounding the drum has a port for the rear vent near the perimeter, so it’s possible to get a (cramped, inconvenient) look in there without tearing the washer apart.
We spent the night aboard BB62 in Camden NJ, with our bikes lashed to a post on the dock. Follow the light-color brick track from the upper-left GPS point across the dock to the black dot marking a memorial stone: we tied up just to the left of that spot.
Position Jitter – ZNU at NJ2BB-15
NJ2BB-15 is the APRS digipeater aboard BB62 with an antenna high in the superstructure. While I didn’t have any trouble with RF reception, packet collisions pose a problem in a dense urban environment. For what it’s worth, essentially everything in the superstructure is an antenna; the NJ2BB ham shack is a wonder to behold.
BB62 starboard side
The point-to-point jitter is about 20 meters (18.52, says the GPS info dump), so you’re looking at the un-augmented GPS accuracy of a long-term stationary object. I’m sure there’s a slight registration mismatch between the satellite imagery coordinates and the GPS coordinates, enough to put the upper-left point across the dock.
If you get the chance, take the tour. The guides are retired Navy, some served aboard BB62, and they take their storytelling duties very seriously. The bunk space, even with air conditioning, is claustrophobic at best; a tip of the bike helmet to you folks who live in these machines!
[Update: Our daughter discovered three itchy bites in a line across her tummy the morning after spending a night in the bunks. That means BB62 has bedbugs, which you do not want to bring home in your luggage. As a result, I cannot recommend an overnight on BB62, alas. We wish she’d mentioned that before we got home…]
[Further update: when I reported this to the folks at BB62, they had an exterminator check out the berthing spaces and conclude they have no bedbugs on board. That’s encouraging, but I still heartily recommend that you follow the same decontamination procedures that you should use after all trips.]
It turned out one rider in our group was an active-duty Rear Admiral who, evidently, could (or should) have supervised the signal gun firing after Colors and Taps. She was traveling incognito, though, and didn’t stand on ceremony.
I was running my GPS-to-APRStracker while on a bicycling vacation along rail-trail paths around southeast PA. I expected good coverage in urban areas and not much in the woods, which is pretty much how it worked out.
Here’s a plot of my track (from aprs.fi), with superimposed half-size PHG (Power-Height-Gain) “circles” for the digipeaters that caught my signal. Clicky for many more dots.
The first part of the ride, from BB62 in Camden NJ to Pottstown PA, had good coverage.
A bus jaunt from Pottstown to White Haven, just north of I80 along the Lehigh River gorge, accounts for the abrupt jump. I dropped off the face of the earth at White Haven, riding south along the Lehigh River to Jim Thorpe, then along some undeveloped trails to resurface just north of Allentown.
Strangely, there are no points east from Allentown to the Delaware, then south along the river & canal to Trenton. We stayed overnight on Bull’s Head Island where, as nearly as I could tell, there were no other APRS signals at all.
A plot of all the APRS activity (and, thus, all the active digis) for a different 24 hours shows the gaps in coverage match up fairly well with where I wasn’t heard. These are also half-size circles, but don’t take topography into account. Notice that the trails along the Delaware run right through the no-coverage zone!
PHG Allentown to Camden – 24 hours
I’m not sure why the digis caught me going into Allentown and not going out, but the vagaries of RF propagation remain inscrutable. Even if the digipeater could receive a clear signal, a collision between two transmitters can kill both packets stone cold dead. In addition, I’m using 100 Hz tone squelch and some receivers may not decode packets with tones.
Another possibility is a path (WIDE1-1, WIDE2-2) that allows only two hops to an Internet gateway. In those remote regions, it may well be that I should have had a path allowing three or four hops. However, I wasn’t hauling along all the programming gear to tweak the TinyTrak3+ on my bike. If I lived around there, I’d have a better appreciation of what’s needed to get out of the valleys.
The amount of ink dumped into the external waste ink tank is staggering. A single head cleaning results in a stream of ink pouring into the tank. After a few weeks of watching that, I stood the tank on end: to my astonishment, the ink pretty much fills the black endcap.
Waste ink collection
In round numbers, the cylinder is 40 mm ID and the cap is 20 mm tall. Volume of a cylinder is πr2h, so you’re looking at 25×103 mm3 of waste ink.
Seeing as how 1 mm3 = 0.001 ml, the tank currently holds about 25 ml of ink!
The printer has six cartridges. Assuming head cleanings drain an equal amount from each cartridge, that’s 4 ml apiece. Given that the large OEM ink cartridges come with 11 ml of ink, you can do the math: a third of a cartridge of each color just for head cleanings so far.
I do not object to head cleanings; that’s how they keep all those teeny little nozzles free of gunk. However, coupling that ink usage with minuscule ink tanks is robbery, plain and simple.
The next time you hear a printer manufacture tout their greenness, you can spit right into their shadow for me.
This is a tweak to the previous design, based on some road testing.
An attenuator on the output of the MAX4467 voice amp allows gains below unity. Right now, the MAX4467 has Av=5 and the attenuator cuts it back by about 1/5, so the overall gain is about unity. I have a bunch of surplus electret mic capsules and some have come through really hot; this allows backing the gain way down with the mic amp set to Av=1.
That requires stiffening the Vcc/2 supply by swapping in a 33 µF cap for the original 1 µF unit. If you don’t do that, the amp turns into a oscillator: the attenuator jerks the Vcc/2 supply around, which feeds back to the non-inverting input of the MAX4467. In principle, the gain should be less than unity, but I wouldn’t bet on it.
The MOSFET relay sometimes didn’t quite turn on from the piddly 4 mA available through the ICOM IC-Z1A’s mic power supply; it was vaguely temperature dependent. I returned to an ordinary optocoupler with a CTR of about 100% driving a 2N2907 PNP transistor, as in the first-pass design that you never saw.
The two 2N2907 devices allow either a through-hole TO-92 or SMD SOT-3 package, depending on what you have and the power dissipation you need. In my situation, the SMD version suffices, with less than 100 mV of VCE saturation.
Let me know if you need the Eagle PCB files or PCB layouts.
[Update: I’m not convinced the Vcc/2 supply is stiff enough. I ripped out the attenuator and cut the amp gain to 1.0. If I get some really hot capsules, I’ll think it over a bit more.]
Somehow I managed to shred the silicone cushion of the earbud on my bike radio. As nearly as I can tell, it got caught between the seat and the back; the missing part certainly isn’t inside my ear.
Anyhow, I have a bag of spare cushions from all the other earbuds, so this isn’t a showstopper.
The adhesive snot holding the earwax filter in place also failed, so I figured I should fix that while I had the hood up. The old filter was all ooky with earwax & oil & dried sweat, which meant that any new adhesive wouldn’t stick. I chopped a disk from a random foam earbud cover with a 7/32-inch hollow punch and glued it in place with some acrylic sealant.
Earbud cushion and wax filter replacement
While I had the sealant out, I replaced the tape sealing the vent hole (on the other end of the earbud) with a dot of glop, much as I should have done originally.
The Smell of Molten Projects in the Morning 