![[Pasted image 20250508102712.png]] # Lab Update *2025-04-07 -> 2025-05-08 | [[#Pretty Pictures|Skip to Pics]]* `(THERE ARE A LOT OF PICTURES THIS TIME!) ` ## Highlights I got approved to exhibit at Open Sauce! That was a major goal for the year & for the Joule Thief project, so we made it. Now, there's bigger fish to fry. I also got sponsored by PCBWay! (https://www.pcbway.com/?from=Fish) I thought this was a slow month, but now it was actually quite a lot. Maybe the busiest yet tbh. Because in between all this, I'm still building all these damn data centers... I spent ~2 weeks in the depth of 3D printer fine-tuning madness wasting hundreds of dollars chasing a big boat that took 2 days and $50 to *attempt* to print. In the bits of downtime I had between this obsessive run, I kept trying to work on the Joule Thief and kept failing to have the right parts for the job. Eventually though I DID manage to solder/assemble my first SMD board, and I designed it myself, and it actually works :) Once the would-be final benchy was in the works, I took a look around the lab and decided it had been claustrophobic for too long, and that I would need more space to grow into at the rate things are going. So I took over the loft and the new setup looks and feels fantastic. Along the way, the last few weeks have also involved me doing a ton of research into what it would take to produce an integrated machine that can fully produce 2-layer boards with one click, as well as some dives into all sorts of circumferential ideas like building a PCB gigafactory in TX and the development of a novel plasma sputtering toolhead for copper-coating PCB vias as a means of shortcutting the electroplating wet processes. All things considered, pretty banger month. I appreciate everyone's support, enthusiasm, and technical guidance. Things are picking up and I'm looking forward to seeing what the next few months have in store. New Loft Lab Arc ! **New Lab Setup:** https://x.com/fishPointer/status/1918007244693881067 ## Lasers & Plasma - PCB Fabrication **(new!)** ### New Notes ### Overview on the Plasma Thing This started when I made my timeline sketch in early March, when I was mapping out what the rest of this year looks like for Lab Work. The original idea was something like the sketch below: `Let's Get the Rat Rig 1-click printing reliably at full range while I work on designing a few PCBs. Then, put together a portfolio of performant prints and boards, and then use those to start fishing for print orders and design & build contracts. Build a 500mm Acetone Vapor Chamber to offer superior print performance & post-processing to serve a high-value niche. Build a brand/following around that, and then scale it to the point I can do it full-time. Then open a PCB Fab/Assembly shop with a friend or two.` ![[Pasted image 20250508065905.png]] So naturally, the next step was to figure out how much I could shortcut that and skip to the end by pricing out all of the equipment I would need to start making decent quality boards at home. I already knew about [Stephen Hawes' Fiber Laser Video](https://youtu.be/wAiGCyZZq6w) as well as [Levi Janssen's CoreXY Electroplating Toolhead](https://youtu.be/0vtCyUHz1Mo), so I was, and still am under the impression that once the Via Problem gets solved, we are set, and that overall we are probably ~2 years away from a desktop machine that can produce full 2-layer boards automatically. As I looked into it more, it seemed like I only needed around ~$10k to get drop-in machines for each process: ![[Asparagus PCB Maker.png]] The biggest thing stopping enthusiasts from making good PCBs at home is the fact that 2-layer boards aren't easily doable yet. Single Layer Boards are wholly insufficient for general purpose design due to topological constraints. The minimum is 2. So your traces can jump over eachother. In the long journey to get from copper board to furnished PCB, the only step you can't just go buy a machine for is the electroplating. I'm sure there's some $20,000 solution from LPKF, but that's a bit out of range for home fab. I've never set up an electroless or electroplating chemical bath like that before but considering the chemical kit is only $100-300, I would've expected there to be a drop-in solution for maybe ~$1,000. But I couldn't find something like that. This whole thing is mainly held up because of the Via Problem. Currently, the canonical way to connect the 2 layers is to electroplate them together in a scary chemical bath not unlike the stuff that kills you in Portal. ![[Pasted image 20250508075801.png]] Here is my current list of Methods for Creating Vias in 2-Layer PCBs: 1. Insert a premade copper rivet and crimp it, then flow solder or use another method to bind copper layers to the rivet. 2. Insert a copper wire or rivet and spot weld it on both sides using high-frequency heating, tuned to melt copper only to a specified depth. 3. Use pre-via’d board stock and design around its templates. (Tim Smith) 4. Electroplate the boards using a traditional two-step chemical bath (copperless 5μm deposition followed by electroplating to 25μm via thickness). 5. Design a CoreXY toolhead to individually and rapidly electroplate vias under controlled conditions. (Levi Janssen) I continued asking the question of "what does the Bambu of PCB Fabrication look like?" And probing around for what its requirements might be and after a good handful of technical discussions with Larry and about a dozen others, I've got a pretty good idea of what the architecture of the machine might look like. It's pretty complicated, and would be decently expensive, but I don't see any reason why we can't start execute and start building this today. Other than the Via Problem. After probing around, `x@spin_resonant` gave me a tip to consider Plasma Sputtering, and that sent me down a very deep rabbithole that I am still currently in at time of writing. ![[Pasted image 20250508071108.png]] Basically, plasma sputtering is the process of using an energized inert gas, typically Argon plasma, and introducing it to a target material (in our case, copper). It will stochastically and ballistically knock copper atoms loose via direct momentum transfer. The target and their destination are a cathode/anode pair, and and the free copper atoms require a line-of-sight trajectory to the destination in order to make contact and begin forming a growing film on the destination surface. There are some vacuum considerations for evacuating the atmosphere to prevent collisions/scattering and oxidation during travel, but there are flavors of this process that can be performed without vacuum. So now the big idea here is: how can we get this mechanism into a CNC toolhead? Basically, it should be possible to finely control the deposition of copper atoms through a nozzle and form a kind of "spray-paint" brush, where the toolhead can now spray nanolayers of copper at an arbitrary location. With something like that on the table, the Via Problem can be solved. The only thing left is `1-2 years of really intense plasma toolhead R&D`! So I've been doing some research (harassing o3) and napkin math (harassing o3) and consulting technical advisors (DM'ing PhD's on twitter) and haven't found any critical refutation of the idea just yet. It's just a hard solution to a solved problem. I think there's actually huge potential value in taking this route for a couple reasons: - **Dry Process:** Plasma deposition completely cuts out the hazardous wet process that requires toxic chemicals and produces chemical waste the end-user would necessarily concern themselves with. This means the integrated machine should be no more hazardous than a microwave or a 3D printer (which it is essentially the lovechild of) - **No/Minimal Consumables:** Given that the typical via only demands ~0.250mg of Copper, a 1kg spool of copper wire will last even the strongest power user a year. Compared to the cost and hassle of hazardous chemical cartridges, this is trivial. - **Advanced Manufacturing IP**: Of all the potential solutions to the Via Problem, this one is the most radical, most novel, and has the greatest moat. Assuming the machine is first to market and can deliver the "Bambu for PCB" moment, it's likely the competition would have no viable immediate reaction. It also lays the groundwork for further expansion into advanced manufacturing and more "plasma printer" tech and presumably some lithography applications. - **Space Manufacturing:** As I understand it, this process actually benefits from vacuum, and shouldn't necessarily rely on gravity. After some extensive production runs on Earth, I believe this tech can be easily retooled to function in space. You can't run the electroless/electroplating wet process in space, the Chemicals That Kill You will explode and get everywhere. I believe this Plasma Deposition tech will be perfectly suitable for producing PCBs off-planet. It should also be a very strong if not foundational tech vertical for handling all sorts of off-planet advanced manufacturing processes once we have the ability to mine, process, separate, and purify base metals from asteroids. (like plasma-printing solar cells for a energy-harvesting satellites in a dyson swarm!) So anyway, taking that all into consideration there's a new goal: Go find the money to lock myself in the lab for a year and start developing this thing! If you or your loved ones have a big bag of cash laying around collecting dust please consider Suneater Labs. There's bigger picture here involving the trending narrative of American Reindustrialization, reshoring advanced manufacturing, a renaissance in hardware engineering, a Shenzhen-style SEZ in Texas, a bid for a resurgence in US technocapital hegemony through massive investment is Cislunar Infrastructure, and a roadmap to K2, but I'll save that for another time. The relevant gist here is - with this Plasma PCB machine it should be possible to put a safe, zero-consumables, full-service, 1-click same-day PCB Maker next to every 3D printer in every startup, makerspace, university, and high school. The knock-on effects of something like this will be tremendous. I've found about a dozen people with relevant expertise and various levels of interest in either the tech or business ends of both the plasma toolhead R&D or the TX PCB gigafactory buildout, and am currently working on putting together a proper deck. So if this interests you, just hang tight for a week or so. There is currently no tech demo. Also, just for the record, this sort of thing HAS been done before, to a limited extent. As far as I know, this tech has not been streamlined for implementation in a consumer device. ![[GYu1zcCawAAgKek.webp]] Here's a rough sketch of the pitch - it is all GPT slop, but it's directionally accurate: ![[Suneater Labs Venture Sketch 2.png]] ## Joule Thief ### New Notes 1. [[Joule Thief Controlled Input Voltage]] 2. [[Joule Thief Weekly 1.canvas]] 3. [[SMD Soldering Guide for Joule Thief R0]] 4. [[Joule Thief R0 Minimal Firmware]] 5. [[Joule Thief R0 Status, R1 Direction]] ### Project Status I took a detour on Joule Thief for a bit to handle the and that went great, since I now have a giant benchy! When I sat down to finally solder the board, I realized I was missing chips. It turns out I never ordered them and they were still in my cart. Then, a week later when they finally showed up, I realized they were the wrong chips and I should've double checked the cart! So much for fast-turn PCB! I finally got the board working actually and demonstrated its core functions after ripping off some traces while desoldering and swapping a wrong chip. Nothing a little magnet wire can't fix. After that, I spun up some very basic firmware to interface an Arduino with the two INA219's and the AD5280. I've got a Serial input to manually adjust the pot, and Serial outputs telling me the power through both sensors. All that's left is a little control algorithm and I'm ready to work on the checklist in [[Joule Thief R0 Status, R1 Direction]]. As far as technical problems go, it seems like the INA219 has a direct power reading function in its library, but I'm skeptical about how it's getting used. Based on what I read in the datasheet, I should be able to expect a LOT more precision out of this chip. It's currently rounding to the 10mW, or sometimes the 1mW. My concern now is that I'm going to have to dig into the library and analyze their code, or worse, interface with the INA219 more directly at a lower level to make the most out of the chip. Or even worse still, write my *own* library for this thing. I wasn't really planning on writing my own firmware libraries like that for this project, but if that's where it takes me, that's fine, since I have no experience doing something like that and it's absolutely something I'll run into again as time goes on. Just feels like a bit of a detour. I mean, either way, if I took the time to carefully read over and annotate the datasheet several times, and then go buy 20 of them, and then use them as core in a major project design, it would only make sense that I would also invest the time in making sure when I actually use the chip I'm actually getting as much utility out of it as it can give me, so. As excited as I am to dig into the Plasma thing I already got into Open Sauce on this project so I can't just drop it and chase the next shiny thing. The last few months have been a whirlwind as my x account has grown and I continue meeting so many amazing people. I can't help but re-evaluate my timelines a little bit :p - Soldering the Board - https://x.com/fishPointer/status/1913888852164682166 - https://x.com/fishPointer/status/1914025862560928105 - https://x.com/fishPointer/status/1914024152165683649 - https://x.com/fishPointer/status/1911955207820595539 - Validating AD5280 - https://x.com/fishPointer/status/1914695216839450766 ## Rat Rig ### New Notes ### Big Benchy **Big Benchy Autopsy Writeup:** https://x.com/fishPointer/status/1912529661650956784 I finally got the damn big benchy! This is my completion criteria for the whole Rat Rig Project. I can take orders for big prints now. If you need something big printed, or batch prints, I do ASA. It's a very strong filament. Let me know. Big benchy took 5 tries and about 5kg of filament. There were a handful of problems you can't really discover until you FAFO. - **Benchy 1** - died due to thermal warp, bed adhesion, layer adhesion, corner curling - **Benchy 2** - died due to MCU panic shutdown - extruder temp read >2,000,000 C - **Benchy 3** - died due to setting infill% too low, unable to form solid top surfaces - **Benchy 4** - died due to MCU panic shutdown - beacon got snagged and disconnected by stringy bridges wait what? The wire broke. The wire connecting the extruder thermistor to the Toolboard MCU broke internally and needed to be swapped. It was creating an issue where the temperature was liable to increase dramatically for a split second at random while the toolhead moved around and flexed that wire ever so slightly. After that rewire, everything worked fine. - **Benchy 5** - finished! hooray It also floats. You had to be there. Only issue is that I accidentally printed it with 2 perimeter shells instead of the typical 4. This caused a single layer disadhesion along the keel, since the overhangs create a condition of minimal overlap between the 2 layers. 4 layers would have increased the contact area and prevented this. Ultimately, this means big benchy 5 is unsuitable for acetone vapor treatment and will need to be reprinted. If I put it in the AVC, the acetone will breach the shell and dissolve the infill structure. Who knows what will happen then.... - Quality Check: https://x.com/fishPointer/status/1917599052184838555 - Benchy Floating: https://x.com/fishPointer/status/1917642999338786995 ### Getting to 1-Click - Thermal Challenges Over the last month, I continued putting a ton of time, energy, and money into the Rat Rig. Paying tribute to the local machine god shrine, etc. The biggest remaining problem was thermals. Specifically equalizing and stabilizing the temperature inside the giant ~1/2m3 chamber. The new heater is great and after some tuning, it can do a great job of getting the entire chamber up to 40C. But that's not the stable operating temperature. Once the bed turns on and starts heating to 110C, the top half the chamber gets much hotter vs. the bottom, eventually saturating around 55-65C depending on airflow and some other conditions. The exterior of the chamber, but still inside the grow tent, saturates around 100F. The big frog I had to swallow this time around was actually sitting down and figuring out how to write a klipper macro, and figuring out what kind of preheat routine would be necessary to run at Print Start in order to get the whole system to be 1-click. And it turned out to be as awful an experience as I thought. Documentation and syntax sucks, LLMs can't write functional code, etc. Probably skill issue, but it took me like 4 hours to get the preheat routine out. Here's what I ended up with by the end of it. The gist of it is: - Home XYZ - Set Chamber Heater to 40C - Wait for Saturation - Set Bed Heater to 110C - Set Chamber heater to 55C - Activate Part Cooling Fan for additional heat circulation topside - Wait for Saturation - Start Print #### Preheat Klipper Macro ## Pretty Pictures [[#Lab Update|Back to Top]] ![[Pasted image 20250508101409.png]] ![[Pasted image 20250508101415.png]] ![[Pasted image 20250508101425.png]] ![[Pasted image 20250508101432.png]] ![[Pasted image 20250508101447.png]] ![[Pasted image 20250508101455.png]] ![[Pasted image 20250508101504.png]] ![[Pasted image 20250508101512.png]] ![[Pasted image 20250508101532.png]] ![[Pasted image 20250508101542.png]] ![[Pasted image 20250508101550.png]] ![[Pasted image 20250508101557.png]] ![[Pasted image 20250508101604.png]] ![[Pasted image 20250508101611.png]] ![[Pasted image 20250508101623.png]] ![[Pasted image 20250508101631.png]] ![[Pasted image 20250508101638.png]] ![[Pasted image 20250508101648.png]] ![[Pasted image 20250508101700.png]] ![[Pasted image 20250508101707.png]] ![[Pasted image 20250508101714.png]] ![[Pasted image 20250508101721.png]] ![[Pasted image 20250508101733.png]] ![[Pasted image 20250508101739.png]] ![[Pasted image 20250508101746.png]] ![[Pasted image 20250508101756.png]] ![[Pasted image 20250508101804.png]] ![[Pasted image 20250508101811.png]] ![[Pasted image 20250508101818.png]] ![[Pasted image 20250508101828.png]]