Since I’ve been documenting failure modes in electronics, another tale of electronic woe regarding N109 type knockoff Lenovo 90W AC adapters. I have a couple T series ThinkPads, which conveniently all use the same 20V 90W supplies with the same connector. I noticed that the stress relief on one of my AC adapters was wearing, so, having had a streak of good luck buying various direct-from-China products, I bought a knockoff replacement adapter that way.
I would suggest that you don’t want do that.

It was fine for a couple weeks (sadly, long enough to give it indelible positive feedback at the vendor).
Then the case partially melted.

I blamed it on kicking a blanket over it while drawing power and didn’t worry much, figuring it was just cheaper plastics.
I was wrong.
Yesterday, while sitting with good ventilation on a desk, it cooked one of its caps off (for the uninitiated, electrolytic capacitors make a very distinctive smell and noise when they fail, and are among the most common failure modes for consumer electronics). Distressingly, it was still trying to supply power until I unplugged it.

Being me, I pulled it apart as soon as it cooled down enough to handle.

The heat deformation is centered behind what appears to be a 5N60C type N-Chanel MOSFET, although there is something that looks like a rectifier right next to it that I can’t read the label on around the also slightly melted 100µF 400V capacitor.

One of the pair of 680µF 25V capacitors on the output side was the source of the explosion. The label says it was only rated to 105°C, which I’m pretty sure was being exceeded whenever it was under load.

Also charming: note the indicator LED inside the solid black case. Clearly this circuit was custom designed for the application, and not just tossed in on the basis of being close enough.

I wish I’d thought to hook this thing up to an oscilloscope when I got it, looking at the circuit I suspect it has been emitting ripply garbage into my laptop the whole time. Hopefully it didn’t do any damage to the computer. One of these days I’ll say that enough times to give in and pick up a scope, the Rigol DS1102D scope + logic analyzers are getting ever more tempting.

Now, I’ll be getting rid of the husk so it stops stinking up my apartment. It looks like there are some possibly-authentic ones from reasonably reputable vendors for like $16-20 that are more likely to be worth owning.

The Thrunite T10 light I’ve been carrying failed to come on when I reached for it the other day. It wasn’t a dead battery, and, yesterday, I determined that the 24mo warranty offers free repair contingent on you paying for shipping to and from Shenzhen, China, which comes out similar to the purchase price. I like my little T10, so this evening, I decided to fix it myself.

tl;dr: If you have a T10 die on you, it is likely to be the negative contact spring in the bottom of the body corroding/moving. Try cleaning and/or re-seating it.

It comes down pretty easily, two spanner holes in the back of the pill to thread it out of the head, a little prying to get the two deck PCB assembly out of the pill body, and it’s as apart as it comes. I didn’t see any damage or broken connections, and all the paths that should look like a diode to a DMM looked like a diode, so I applied power with a pair of wires taped to a AA battery and… it worked. Then, I put the pill back together, pushed it on to the end of the body with a battery in place and … it didn’t work. Hooking up the assembled pill to the battery with wires did. This is bewildering, because the body is literally a spring and an aluminum tube. Probing around in the body with a DMM revealed that somehow, the aluminum-to-”copper” press fit had become non-conductive (?!).

I grabbed some forceps and pulled the spring. It still didn’t work when reassembled after I washed the spring and the tube with isopropanol. I did, however, notice that the narrow end of the spring is chrome (?) plated, while the fat end is bare coppery material. I dipped the spring in flux, and tried to plate it with some silver bearing solder, and even swimming in flux it would barely wet with solder. Whatever that spring is made of, it is not copper, and it forms a nasty, insulating oxide layer. I eventually got a few sections of the part of the spring that should be in contact with the body to take solder, re-cleaned the spring in isopropanol (which revealed that the flux did eat an oxide layer off the spring), shoved it back together, and all is well again. I’m guessing the whole spring is supposed to be plated, and it flaked (I found some gunk in the body when I washed it), causing my problem. If it dies again I may try re-cleaning the spring then coating the base in conductive glue before reassembly, I don’t think I can solder to the body or get the spring to wet reliably. Re seating the spring is simply a matter of wiggling it roughly in to place then thumping the complete light tail-down against a non-marring surface to use the battery inertia to (percussively) improve the contact between the spring and the casing.

This post is mostly up so if anyone else has a T10 die on them, they can try the easy fix. I’m curious if any other owners have looked closely at their negative terminal spring, I’d be interested in knowing if the plating defect is a widespread manufacturing problem or just my misfortune.

When I was thinking about needing to replace it, I determined that do appreciate having “soft glow” and “Blinding” as options, especially without having a stupid blink mode – I guess ideal would be a mode setting ring with detents at tactile bumps for the settings, but I haven’t seen anything (much less anything reasonably priced) with that kind of design. I probably would have just bought another T10, which is something of an endorsement even though it broke after a year on an apparent manufacturing defect.

It always makes a fun topic because it is both interestingly abstract and forces me to carefully evaluate and create specific reference cases for my preferences and positions.

TTRSS Published feed through FeedWordPress

This generates individual posts from the items in the published feed, in whatever form the interaction between the feed and the WP template generates. It gets all the desired content up, and in principle it provides a local comment scheme.

It is, however, extremely noisy compared to the other content I put up, without clearly delineating itself from original content. WordPress can be configured with alternate post types, and per-type display , but that is a level fucking about with PHP and CSS that I’m not willing to engage in unless I absolutely have to. Part of the noise problem is that it doesn’t behave terribly well with snippeting: FeedWordPress faithfully reposts all of whatever came down the feed, and what I would really like is the full note, followed by the link and a short description. While the tools are present to filter and massage the feed, it would take a nontrivial bit of parsing around inside the feed content to produce full notes and limited-length blurbs. Google Plus’s “unbecoming addiction to abbreviation,” as a friend put it, as well as the inflammatory, misleading headline issues at HN and Reddit have made me aware that clumsy truncation is not really acceptable.

A substantially larger problem is that, despite setting all the appropriate options for local comments, FeedWordPress stubbornly refuses to do anything other than pass the comment link through to the original source. Since borrowing WP’s comment system is the primary reason the news-as-posts model seemed appealing, this is not acceptable. There is a four-month-old unanswered bug report (admittedly, a crappy bug report in broken English) about the issue, but that was all I came up with.

This was the one I was most hopeful for being a drop-in solution, but it clearly would take a substantial amount of the sort of web front-end work I don’t enjoy to make it work well

TTRSS Published feed through HungryFEED

This just gives an unobtrusive feed of notes and headlines in any container your WP theme supports. Pretty nice as a “Look at these interesting things” sort of mechanism, and makes a nice way of advertising that you have a news feed available, but has absolutely no facility for local discussion. It delineates the content better, but lacks the discussion and history features I would like.

If I decide I need to get off of Google’s services, either because the services become too unsuitable or the parent company does something too horrible, I’ll look around again. Setting up a custom post type (easy), with custom display properties (straightforward but unpleasant), a segregable output feed (trivial), and working comments (??) to auto-post TTRSS’s Published feed into using a mechanism like the FeedWordPress solution above would not be horrible, and would support all the desired functionality, but isn’t something I’m currently willing to do “just because.”

For the interested: TTRSS has a publish mechanism, which creates a custom RSS feed of any article you mark published, along with whatever note you have attached to it with the built in annotation system. It even allows for non-feed content to be shared. There are various WordPress plugins that can embed an RSS feed (HungryFeed,EmbedRSS) or import an RSS feed as a post type (FeedWordPress).
Embedding as custom posts gives both distinction and a comment system, and it is a universal interfaces (can read from web, subscribe via RSS ,etc.). There is even social discovery support built in should such a thing take off.

If this experiment works really well, I might even talk myself into using it before Google gives me another reason.

I’ve been working on my Shapeoko in little fits and spurts that individually haven’t been terribly documentation worthy, but in aggregate are pretty interesting. Continuing from where I left off in Shapeoko: Part 5, I’ve iterated a bit on belt tensioners, switched to a commercial breakout board, put the spindle under computer control, attached the spindle to the machine, made some tentative test cuts, and added hall-effect endstop/homing switches to the X and Y axes.

Breakout

I picked up one of the many, many not terribly distinguished Chinese parallel breakout boards to get away from my sketchy breadboard and hand-soldered breakout cable solution. I selected this one from “Shenzhen Ele Technology Ltd.” on Aliexpress in particular, which is labelled HY-JKM5, comes with a mini-CD complete with bad documentation for this and a dozen other similar products, and what looked like a couple kilobucks worth of cracked CAD and CAM software for Windows. After rooting around in the documentation, and looking over this post that clarified a bit, (aside: the proprietor of that site apparently has exactly the same hobby projects I do, it makes good reading in general) I eventually figured out all the necessary correspondences. I has five axes worth of 4-pin headers (Enable,Step, and Direction on pins, plus a ground), four switch inputs, a built-in relay, and connections for an E-Stop, all broken out on to nice reliable easy-to-work-with screw terminals, as well as secondary headers for modular wiring, which I believe are XH2.54 connectors, but the documentation isn’t forthcoming, and that name may be specific to the vendor I matched image and dimensions with. I don’t think that the board is really properly isolated, but it is certainly less sketchy than what I had before.

To clean up the electronics a bit, I screwed standoffs into the appropriate holes of all the pieces, and pinned them into a sheet of styrofoam in roughly the layout I intend to eventually secure them in an enclosure, to reduce tugs and confusion and mess. Ideally, I’ll have the machine cut roughly the same layout as is pressed into the Styrofoam into a piece of wood or plastic or some-such to screw down and enclose with the power supplies. That will happen around the same time I do something about random bits of hookup wire plugged into ribbon connectors and the like. I’m slowly building connectors for all the serious uglyness, but it is a very manual process to build out from header strip, harvested molex connectors and such into the harness I need.

Spindle
I talked about the spindle on it’s own in detail in my part 5: the two noteworthy things I have done are attach the spindle to the machine (using the stock Shapeoko kit brackets), and get the spindle under computer control.

Parport Pin 9 is internally connected to the relay on the breakout board, which I have interrupting the ground side of the spindle/power supply connection. Pin 17 (which would be the B axis direction as marked on the breakout) is hooked up to a scaled PWM signal fed into the input of the nearly unmarked speed controller than came with the spindle. I diddled HAL settings until I had reasonable software speed control of the spindle – my current .hal and .ini are attached to the bottom of this post.

A Cut!

Once I had the breakout and spindle working, I decided I wanted to try cutting something, just to have done it. I taped a piece of EPS foam to the bed, stuck a 1/8″ diamond-cut endmill in to the collet, homed the machine by eye, hand wrote a short piece of G-Code for some partial-depth concentric circles, let it cut. Measuring the cuts, it seems to be working properly in all axes.

Switches
Running the machine with no endstops is sketchy. Running the machine with a spindle and no endstops is really sketchy. I had a bag of generic microswitches, which I wired up with some telephone cord and tried to install, but I was never happy with the physical layout, and went looking for something else. To that end, I bought a bag of 50 A3144[-ish, they're probably clone/counterfeit/etc.] sensors from this aliexpress seller (They have gone up at time of writing: I only paid $7.80/50). These are nice little self-contained unipolar hall effect switches – they ground their output pin when exposed to a sufficiently large magnetic field, and leave it floating otherwise. No moving parts, tiny packages, and incredibly sturdy (no, really, I’ve beaten the HELL out of some thermally, mechanically, and electrically – the only ones that died were due to a mishap with 12V reversed through the power and ground pins).

I cabled up with some normal hookup wire and electrical tape, but that caused problems with the wire being larger and stiffer than the part. While hunting for better cabling options, I saw this post espousing the wonder of servo cables for hooking up A3144, at which point I discovered that you can buy very inexpensive commercial Y-cables for servos that make perfect attachments for the max/min connections. The genders don’t work out as nicely, since the ends of the Y are male, but I’m not afraid of a little soldering.

I ended up using the jacketed ends of the Y cable as bases for a hot-glue-and-heatshrink protective shroud around each sensor, and simply hot gluing the packages to the X and Y motor plates for attachment. The trigger magnets (which are just some little rare earth discs I tacked on to an order from DealExtreme years ago) are unobtrusively attached to the ends of the rails with a bit of packing tape, and can be easily adjusted. The output and ground of each sensor is wired to the breakout board through an adapter made with some .1″ header and hookup wire, with the Vcc connections currently just pulled off to a small breadboard hooked to 5V.

Getting the switches working electrically exposed some quirks on the breakout board – for example, the pulled-up and grounded terminals of the X and Y switches are apparently in opposite order, which wouldn’t matter with a simple mechanical switch but makes the hall-effects that need a solid ground require some trial-and-error to hook up correctly. There is a little bit of hal and ini magic to get the homing behavior correct (which resulted in many impressive kerchunks as the carriages hit the end-plates), the settings in the attached files are apparently good.

One of my current puzzles is how to set up switches for Z – I think if I’m clever about it I can put a single sensor on the X carriage, a pair of magnets on the Z makerslide, and be good go to, but there are many moving obstructions in the area, and I can’t figure out where the low stop should be. Ideally, I suppose it would be easily adjustable for the installed bit length.

Etc.

I’m not in love with either of the Y belt anchors I have right now. One side has the aluminum plates I built as a first try, which are large and hard to establish good tension with (I currently have a wood screw jammed in to the belt loop at one end to adjust the tension), and the U channel + original bracket rigs are less than finished looking. Still not sure how to do better though, maybe it will be another “the machine cuts its own improved parts” situation down the road.

My current LinuxCNC hal and ini files are attached, I had trouble finding good examples for what I wanted to happen, so hopefully these will be useful to others. These have apparently correct descriptions for homing X and Y with the A3144s, for controlling the spindle, and for general motion, although some things like the soft travel distances and Z homing are bogus or absent.

It is now, essentially, good enough to make “real things” – I just have to get it together to do so.

Having TAed the course twice, and taken UK’s undergraduate CS curriculum, I have come to believe there are serious structural issues with the class. To swipe from writing about this previously on my G+:
We spend most of two weeks on probability, that we know every student will have to take at least one full course on, and squish relations/graphs/trees into a week and a half at the end, even though that is the part students are likely to actually use and won’t necessarily be exposed to elsewhere. We have functions at the beginning of the semester, and relations over functions at the end of the semester, which obscures the multiple representations of functions abstraction – as far as I’m concerned that is one of the primary things to get from the class. There are many similar lesser similar structural issues.

It is a very useful class in terms of skills required to be effective working with computers, but we end up spending most of it re-teaching them moderately relevant math tricks.

Taking: Taking: LIN511: Introduction to Computational Linguistics
This was a lot of fun, and I highly recommend it to anyone who likes to play with languages or computer tools in context. It wasn’t terribly difficult, it conveyed a body of useful information, and it was quite entertaining. Most of what we did used DATR and Python/NLTK to implement simple tools.

At one point near the end of the course the instructor did feel the need to apologize for it being too easy for me, but that was sort of missing why I singed up for the class: in addition to firming up some domain knowledge I was there to look at computer tools in context. The tools are quite similar to those we use for programming languages – some are more sophisticated, many are less sophisticated, but they are fundamentally similar, and contextualized very differently, since the languages are interest are generally not regular, context free, deterministic, or even particularly well specified.

It was also very interesting to observe non-computing people interacting with computers. The syntax:sematics:pragmatics relation in programming is much more tightly bound than in natural languages; many of my classmates found the degree of precision required to express things to a computer frustrating. We spent a lot of time on the command line, and there was a lot of difficulty with the idea that actions on the command line are exactly analogous to those performed in a GUI, and even the file/directory abstraction in general. I blame this largely on recent efforts to hide/ruin universal abstractions in computing in favor of apparent magic. Based on the group study sessions preparing for the final, I’m also reasonably sure I could now teach regular expressions to a potato.

For the final group project, my group built a rough prototype part of speech tagger for Persian in Python – this was exciting on several fronts: I don’t speak a word of Persian, and only one member of the group really did. The Persian writing system is very different than our familiar roman alphabet; 32 letters with four different positional forms each (a superset of Arabic), written right-to-left. The most exciting bit for me is that Unicode support in the various tools is “not always perfect,” so I spent some very informative time going over the project with differently buggy text editors, and working around Python and NLTK quirks. The code (including a modified version of one of the NLTK library components) is attached, it isn’t terribly useful as it stands, and isn’t an ideal approach, but it is interesting to look over.

Taking: GS650: Preparing Future Faculty
This turned out exactly as expected – a useful look into the conventions of greater academia, conducted mostly in the form of listening to panels of appropriate speakers, reading appropriate news streams, and writing responses.

For the final project, I went and spent a day with Dr. Jens Hannemann at Kentucky State University. I worked with Dr. Hannemann when he was at UK some years ago, and correctly expected his thoughts on the contrast to be interesting. KSU is… a bit of an overgrown highschool… but the experience was interesting in that it firmed up my growing conception that the faculty roles at regional comprehensive schools are more to my liking than I expected.

Most of the remaining PFF courses look like fun, I will likely take more (possibly enough to end up with the certificate) if my schedules work out to allow it.

We have been printing out little puzzles to test dimensional accuracy on the 3D printer in the lab, and it reminded me of a cube puzzle I had as a kid, which is now rather difficult to find information on. I’m putting this on the ‘net largely because coming up with search terms to unify all the relevant information was nearly impossible – I had to go root around at my parent’s house to find my set to connect all the dots. It was the source of many ragequits as a child, and it would be a shame to deprive future generations of the same …stimulation.

The game is a super-set of Soma Cubes. The class of puzzles in general uses a set of figures constructed from unit cubes to build a larger cube (or other target figure), and are mathematically interesting in addition to making good desk toys. The variant in question adds the two non-convex forms (1×3 and 2×2) to the traditional Soma Cube set, while still only putting seven pieces in play at any time, unlike SOMAplus, which is played with eleven pieces. This combination is apparently sometimes called a Rehm’s Cube Set, but that search term didn’t come up with much. My set was branded “Babylon The Master Builder’s Puzzle Cube” with a design credit to Helmut Huber in the instructions. The designer’s name only turns up a reference to this puzzle in Indiana University’s Jerry Slocum Mechanical Puzzle Collection. The same game has apparently been sold under some other names/brands, including Jumbo also under the Babylon title, but never been all that widely successful.

The puzzle comes with the possible two non-isomeric tricubes, and seven of the eight possible tetracubles, excluding the 1×4 form which would never fit into a 3×3 cube, and two cubic dice, one with six differently colored sides matching six of the tetracubes, and one with three black and three white faces. The game/puzzle is conducted by rolling the dice, removing the pieces corresponding to the sides from the set, and then attempting to build a 3×3 cube from the 7 remaining pieces. Detailed English directions in image and transcription below.

Babylon – The master builder’s puzzle cube
The tower of Babel was built by many peoples. In the evenings the workers spent their time with their favorite game, putting together a cube from various building components. Each group built their cube different to the other. The master Builder observed each group closely and tried endless combinations until he found the MASTER BUILDER’S PUZZLE CUBE, a game which could be played by everyone.

Contents
2 coloured dice
7 large building bricks of 4 dice, colours: lilac, yellow, red, orange, green, blue, natural
2 small building blocks of 3 dice, colours: white, black

Instructions
Each building block is a different colour. A complete dice always consist of 6 large and one small building block. The natural coloured building block is always used.
1. Roll the two coloured dice and lay the two building blocks which are the same colour as the eyes on the dice to one side.
2. Assemble the magic dice from the remaining 7 building blocks. It doesn’t matter which two building blocks are missing, the MASTER BUILDER’S PUZZLE CUBE can always be assembled from the remaining blocks!
3. Keep calm and have fun!

You can build MASTER BUILDER’S PUZZLE CUBE alone or play against one another. One player rolls the dice and everyone with a complete puzzle cube starts. The quickest “Master Builder” wins. Either points, a prize or a reward.

Note that there are 7 tetracube forms and only six sides in the removal die: the “t” shaped piece (natural wood in my set) is always in play – I am not sure if it is generally solvable if that is not the case. A solved puzzle (which happens to be a Soma Cube) looks like the following.

Some math/puzzle nerds got excited about this set a couple years ago, but the degree of specificity required to find their archive was such that I only managed to find it after I dug out mine.

At the time of this posting there is an Aliexpress seller offering a generic clone of the set, but while it is marked as being only $6.60/ea, they are only sold in pairs, and shipping to the states pushes it up past $40 to get a pair this way. It should be easy enough (as well as significantly cheaper and more satisfying) to build a set from craft cubes or sections of 1x or whatnot. Just for shits and giggles, my set can be arranged exactly as in the seller’s image, which reveals that the color choices are not completely consistent.

This summer I decided do it right. I bought the classic inexpensive LDP truck set (A Bennett Vector and a Tracker RTS, I picked the 5.0 and 129mm variants respectively), modded them for LDP, and installed them on my suitably sized deck with suitable bushings and wheels.

Distance skating attracts the best sort of crazies – It’s something of an equipment sport, so they tend to obsessively purchase and modify gear. It’s remarkably physically demanding, so there is a lot of the classic solo endurance sport mentality. It also has carryover from the “raah hardcore” skate world. Fortunately, the internet has brought together the appropriate crazies at sites like Pavedwave and skatefurther, where the discipline has been developed from its roots in the 70s.

Designed-for-LDP decks (Subsonic Pulse, LBL Walkabout, RoeRacing Mermaid, etc.) are boutique items and tend to cost in excess of $150 for the bare deck, so I wanted to ease in financially in smaller steps since I’m not sure how capable I’ll be. Most sources say 26″ to 31″ is suitable for an LDP wheelbase, and my Pakala III is around 25.6, so it is manageable if a little tight-pumping.

The mods for the trucks are pretty interesting, I implemented a number of the suggestions from this excellent thread on the pavedwave forums.

On the Bennett, I made myself a nice Polyoxymethylene(aka Delrin/Acetal) insert for the pivot. I think the technical name for this thing would be a bushing, but since that term is already in use for trucks, people have taken to calling these “hobo sphericals” or “fixed spehericals,” as the more elaborate alternative is to install a spherical bushing. I saw that Griffin sells fixed spehricals for Bennett trucks that look to be machined out of some flavor of Polyoxymethylene, looked at the sheet of 1/8″ Aceteal I had for another project, and decided to DIY. For those looking closely at the numbers, the insert is a little thinner than ideal (0.125″ sheet, gap depth measures around 0.16″) but it seems sufficient.

I hacked a small chunk off the sheet, held it under the truck pivot, and scribed the diameter into the sheet with a pin. I then did the typical chords-and-90°-angles center of a circle stunt, sanity checked it against a washer, and drilled the center to 25/64″ in a couple steps (starter drill, some intermediate size grabbed at random, 3/8″ because I wasn’t sure how sloppy I was being, then 25/64″) — Kingpins are 3/8, and it should slide freely, so a V letter drill would probably be technically correct, but I’ve never contrived an excuse to own a letter drill set, and was free-handing it with pliers and a handheld drill anyway, so 25/64. I then followed my scribe line with a coarse sanding drum chucked into a rotary tool, and fitted/finished it with a finer sanding drum. I’ve checked on it a couple times as I fiddle with the configuration, and it seems to be holding up and doing its job.

If I ever get my Shapeoko into shape for this sort of thing I should be able to punch out some nice precision parts for this on it.

On the RTS, I clipped the wings a bit with a file, as apparently the wings will chew up the pivot cup if you don’t. I then polished out the file marks and the whole pivot pin on both trucks with some tripoli compound and a felt wheel chucked up in a rotary tool.

At the moment I have the Bennett wedged to +10°, with Green and Yellow tall Reflex barrels, and the RTS at -7° with some blue Khiro barrels I had around, both stacked on top of 3/4″ of risers to keep it from biting – gives a ~5″ ride height, which is not ideal.

I built up speed by pumping the first time I stepped on this thing (and then wheelbit, got some road rash on my elbow, installed some more risers, and accelerated again without the painful sudden deceleration). I’m still clumsy and slow pumping, and I’m sure the setup can be improved, but it feels wild, and the dedicated trucks make a world of difference.

A – Belt, 1:00- Generic suitably sized phone holster, with clearance for the headphone jack cut out.
B – In the holster – T-Mobile MyTouch 4G Slide (AKA HTC Doubleshot), currently running a recent unofficial CM9 build. Detailed impressions here.
C – Wrist – My over-decade-old Fossil Blue AM-3314. It’s a dual face (analog/digital), and makes a good grounding strap.
D – Clipped in to right front pocket – Keys, on a snap-hook. RFID fob for the office.
E – Left Front Pocket – Wallet. Leather, with card slots, and an integrated coin pouch.
F – Right Front Pocket – Self-made pocket organizer, holds Black, Blue and Red 0.5mm Uniball Vision Elites, 0.7MM mechanical pencil, Flash drive, and Chap stick. Discussed below.
G – Belt, 5:00 – Leatherman Wingman. Much lighter/thinner than a Wave, particularly since it clips on directly instead of needing a sheath. Not as well made, and the screwdrivers are way worse, but the smaller/lighter/cheaper (I’ve lost/destroyed two Wingmen in the last couple years) makes up for it.
H – Right Cargo Pocket – A pair of non-isolating earbuds (safe for wearing while walking). The current pair are Sennheiser MX300s.
I – Left Cargo Pocket – ThruNite T10 Flashlight. 1xAA, three mode (Hi/Md/Lo) with NO stupid blinking modes. Detailed impression here.
J – Right Cargo Pocket – A couple of spare snagless hairbands.

The major changes are that I’ve downsized my multitool and consolidated my ubiquitous computing device. I’ve also stopped regularly ruining cheap belts and started wearing a 5.11 1.5-Inch TDU Belt, which is thus far impervious to damage or problematic wear.

The interesting bit is, as always, the custom part. The current version of my pen case is this thing:

It’s getting a little bit worn, and shows some minor design issues, but has been riding around in my pocket doing its job for several years now. The primary issue with this one is that every layer of material has a seam along the bottom, which allows the pencil point to escape through the stitching. When eventually I get around to replacing it, I think I’ll make the lower piece out of a different material (thin split leather?) and wrap it so there is a fold at the bottom and stitches up the sides, the current mid-weight upholstery fabric is otherwise adequate (light, inexpensive, and non-abrasive). I never made a proper pattern, just mocked it in paper then cut fabric to match, so there will be some work in replicating.

This would feel totally ridiculous if it weren’t so entertaining to read these from other people.