site banner

Tinker Tuesday for May 6, 2025

This thread is for anyone working on personal projects to share their progress, and hold themselves somewhat accountable to a group of peers.

Post your project, your progress from last week, and what you hope to accomplish this week.

If you want to be pinged with a reminder asking about your project, let me know, and I'll harass you each week until you cancel the service

2
Jump in the discussion.

No email address required.

is the Tinker Tuesday bot down? where's the May 13th thread? are these done by hand?! :screaming_face:

I'm not a bot (cry_emoji, cry_emoji)!

I posted it, but the mods need to manually approve it. I think you can get there with a direct link:

https://www.themotte.org/post/1916/tinker-tuesday-for-may-13-2025

Not gonna lie, a bot would make things a lot easier for me.

It's approved. Feel free to ping me when you post these, it might help get them up faster.

Is it only Zorba that can set up the bot? I noticed that Transnational Thursday threads are automated.

huh. are all of the weekly threads done by hand?

Yup!

Edit: to clarify, I mean all the Tinker threads are.

I like making stained glass. Cutting the glass, grinding them, foiling, soldering them together. It is such a great and unique gift to give to people, and really exercises my brain in multiple ways. Precision is very important, also thinking ahead multiple steps, plus it stimulates my appreciation of aesthetics. It's a great hobby.

I've had some challenges to overcome specifically at the grinding step. During this step you grind the edges of the cut class pieces to ensure you have a consistent surface to adhere the copper foil to and ensure your pieces are the precise shape you want them to be. But I've had issues because my template pieces I've typically printed out on sticker paper using a regular printer. Problem is during the grinding process the water involved causes my templates to loosen and shift around. I've tried smearing vaseline on the pieces to make them a bit more waterproof, but it's not a great system.

So I've purchased a cutting machine akin to the popular "cricut" machines, but without their walled garden of paid software. The idea is that I can use waterproof vinyl stickers to adhere to the glass instead of paper. The precision of the machine will also help because there's some leeway when cutting using scissors that leads to weird gaps later in the process. I've been having trouble figuring out how to use the thing though. It's called a silhouette. I'm sure I'll figure it out at some point.

Anyway, I just wanted to see if anyone else has this hobby.

I've had my first significant, positive interaction with AI.

I needed a plugin for blender, bought it off an online shop, went to install it and got a big spew of error codes in Blender's console. I spent thirty minutes blind-googling error codes and various rephrasings of "why doesn't [plugin] install", to no avail. As frustration mounted, I noticed one search had popped up an abbreviated AI answer in the header, and from testimonials here and from coworkers I figured I had nothing to lose, and started feeding the error codes to the AI instead. It correctly diagnosed the problem on the first shot, and within about twenty minutes of feeding it successive results, had the problem sorted and the plugin installed. Given the nature of the problem and the solution, I'm pretty confident I would not have been able to figure it out in daily slices of a late-night hour or two after getting the kids to bed and the chores handled, before simply losing motivation and moving on to something else. AI straight-up saved my butt.

OK my stuff from mouser.com arrived w.r.t. my LED project.

I got charged a tariff!

I ordered $5.00 worth of MOSFETs. I saw a line item that said $1.50 in tariffs. If they came from China, and they were anticipating a 145%ish tariff on these items, aren't they telegraphing that their wholesale price for the MOSFETs was $1.00? Seems kind of not that big a deal. If I were a retailer I might simply hide this cost rather than give away proprietary business info like my wholesale costs this way.

Anyway!

My next biggest stumbling block with my LED lighting project for my bikes and bike accessories is coming up with adhesion. For the helmets I think I want to use double-sided tape but I'm concerned it won't hold. My experience with gluing/sticking things has historically been lousy, although I do note that Tesla Model 3 and Model Y license plate holders are stuck to the front with sticky tape so I assume it's possible. I'm eyeing 3M double-sided "very high bond" tape though not holding out a lot of hope.

If that doesn't work, or if I'm dealing with surfaces that are more porous, the next best thing might be to get LED sleeves that are more like tracks with a cover and try to rivet or staple them into the surfaces instead? Thinking of my cargo bike box here.

Thoughts?

VHB is ridiculously strong when used on properly prepared surfaces; I've bent 1/8th inch aluminum plate trying to remove two stuck together with three 8-inch strips of the stuff without solvents. The trick is figuring out what 'properly prepared' means -- 90% of the time just dunking in simple green and rinsing with water works, but heavily polished or painted and almost all rubberized materials can benefit a lot from primer, and I'd expect helmets will fall into this domain.

Most other double-sided tape is either garbage (carpet tape) or won't last very long under exposure (basically every foam mounting tape).

Rivets are a great option in most situations you can make them fit, but alignment can be a pain without kliecos. I don't recommend staples. Epoxy is a mess, but it's really strong and there exist mixtures that will secure to most everything, at the cost of never coming off period. For obvious reasons I'd keep all three but especially the staples away from helmet pieces.

The other constraint to keep in mind is what you're sticking in place; it doesn't help to just have adhesive and the PCB strip well-secured and the LEDs flopping in the wind. Those sleeves have an advantage here, though even the floppy ones might have too large a minimum bend radius. Expect to mix-and-match.

The trick is figuring out what 'properly prepared' means -- 90% of the time just dunking in simple green and rinsing with water works, but heavily polished or painted and almost all rubberized materials can benefit a lot from primer, and I'd expect helmets will fall into this domain.

What about rubbing alcohol on the shiny (not foamy) parts of a bike helmet?

Isopropyl alcohol's worth a shot first, before trying to use primer. Beyond that, depends a lot on the plastic and coating; most bicycle helmets are ABS that I'd expect it would bind fine, perhaps benefiting from a bit of heat. Higher-end ones that are using carbonfiber, or scratch-resistant polycarbonate, I'd expect you'd want the primer.

The higher ends of gorilla tape can handle those cases if you don't want to deal with (or can't get, thank you California) the primers, but I haven't gotten any experience with how weatherproof it gets.

thanks!

how do you know so much about this stuff?

VHB specifically, I've volunteered for FIRST FRC a lot, and it's one of the go-to adhesives in that realm (and most teams get free spools of it), so I've gotten a lot of hands-on experience.

WS2812s, I ran a few different STEM outreach projects using them. They're great as a way to teach and show for loops in physical space, but the constraints are very easy to run into, even with Adafruit's documentation.

Circuit assembly work in general has just been a hobby. I think it's a really important skillset, but also one that's very badly underserved by mainstream college training courses.

While I wait for mouser.com to finish delivering some stuff for my LED project I dusted off a 3D racing game I had been working on.

I originally started it after catching up on the Fast and the Furious series and Dom Toretto's Charger was top of mind. I wrote an all software rendeter in Rust but every nicely detailed free charger model I could find was way too slow without GPU acceleration. So I began rewriting it to use wgpu but it was pretty frustrating due to annoyances around Rust ownership junk but also everything in Rust that's new and exciting is unusable with LLMs because they mix old and new incompatible APIs in their answers. So I stalled out.

The last few days I decided to start over in C++ with SDL2 and bgfx and it has gone significantly better. I can get the car models fully loaded and it's pretty fast, even on crappy ancient laptop GPUs. I am getting some banding in texture maps but I'm sure I'll figure that out eventually.

Currently working on adding controls to the car and having the camera chase it as it moves around an infinite plane, and also looking into making an animated sky sphere from 360 degree YouTube videos. I suspect it'll involve Blender, at some point, which I don't know at all. Probably not too far behind you, @FCfromSSC

I am getting some banding in texture maps but I'm sure I'll figure that out eventually.

This ended up being caused by only some of the faces in the 3d model having textures, with the rest relying on material color and phong shading, but my fragment shader didn't know that so it was trying to use a junk texture for those which just led to this banding effect.

Creating another vec4 and treating the x component as a boolean to signal to the fragment shader whether it should sample from the texture is the best I could figure out how to solve this.

Anyway, looks much better now. Ride or die time with the homies is one step closer.

Still having trouble finding a 360 degree panorama I could use. I might just go to a parking lot in a scenic skyline park of town and take a photo sphere with my phone.

is the car model fully UV-mapped?

Action shot of the Charger on a "road".

/images/17467587024084995.webp

It is, yeah. I downloaded a free Dodge Charger Daytona Hellcat SRT model from sketchfab.com.

Not Dom Toretto's signature black Charger but it'll do for now.

After years of modelling with a lightweight 3d package, I'm making a burn-the-ships switch to Blender. I've spent the last week in tutorial hell in the evenings, getting the basic workspace setup finished and digging into the modifiers and shaders systems. Current goals:

  • overview of the node-based shader system - done
  • basic setup and general modelling tools and modifiers - in progress
  • rigging basics - barely started.
  • UV/Texturing - just started, probably going to need to learn a secondary software package for this as well
  • node-based geometry modifiers - no idea how deep this well is, but the goal is to use it to generate terrain and environments at first.

We'll see how it goes.

I was never that good with 3D art, but I dicked around a bit. Don't know what it is about Blender, it's obviously quite powerful, but working with it always made me want claw my eyes out. Seems to basically be the GIMP of 3D modeling.

So, my sympathies, and I hope it goes well.

thanks much!

The thing about blender is that it's a fully-general tool, and there's a straight tradeoff between power and generality and ease-of-use. Think of the difference between a dollar store calculator and a ti-85. the basic calculator is very straightforward: six or so function buttons, ten number buttons, a clear button and that's it. the graphing calculator is completely covered in obscure buttons, many with multiple functions, and these are in turn connected to nested submenus. The calculator is a physical bottleneck to a vast ocean of capabilities, and it's so complicated because the designers are trying to surface as much of the functionality as possible.

Blender is like that but possibly worse, because the functionality is broader and much more divergent. From a user standpoint, it would probably be better to be split into a suite of ten or so different programs with strong interoperability, but probably that would create other problems. And since it's such a broad generalist program, the default interface is kinda trash for specific jobs, especially if you have prior experience with other packages. I'm working through a tutorial that puts a strong emphasis on hotkeys and custom keymapping to optimize the actual modelling workflow, and blender gets a lot more usable once you've got it set up properly and build up the necessary muscle memory.

The downside was spending two nights after the kids were down working on setup, only to realize that the folder blender was saving the settings changes to wasn't write-enabled, so all my setup was lost when I restarted my computer. not a good feel.

For those interested, this is a pretty good place to start for Blender specifically. Best advice is to open a google doc or similar and take step-by-step notes as well, and certainly write down all the hotkeys he covers.

Took a small detour from optimizing the backend, and worked on the UI. I do quite like the result.

How are you doing @Southkraut?

Tinkering on my fencing, again.

Successfully helped out at the local tournament (manned the reception, which seemed 100% superfluous but apparently we really need every spectator to sign a waiver so they can't sue us if they end up on a photograph), talked to lots of people, but the first thing I'm organizing now is actually a play/grill/fence date between me and the kid on one side and another fencing parent and their kids on the other. Baby steps. Also signed up for another tournament at the end of the month. Might go to training this thursday, but it's a long, late drive and I'm unwilling to commit. Also talked to some people about organizing weekend fencing sessions, but have not made any substantial progress on that since last week.

No code.

110k words on my NaNoWriMo project. Did less work on it over the weekend than I would've liked, as I was busy with another writing project whose deadline is coming up soon.